1

RED TEMÁTICA CONACYT: ÁREAS

NATURALES PROTEGIDAS

PROYECTO: 269540

PRODUCTO H

PUBLICACIONES GENERADAS

2016

Coordinador Dr. Joaquín Sosa Ramírez Profesor-Investigador Titular

Universidad Autónoma de Aguascalientes

1

PUBLICACIONES 2016

Coordinador

Dr. Joaquín Sosa Ramírez Profesor-Investigador Titular

Universidad Autónoma de Aguascalientes

A continuación se enlistan las Publicaciones que fueron concluidas, en este año

de 2016, bajo la coordinación de miembros activos de la RENANP a inicios de 2016, no

incluyendo desde luego en este listado a aquellos quienes ingresaron a partir del mes

de Junio de 2016, que es la fecha cuando se recibieron los recursos económicos para

apoyar este proyecto.

En cada una de estas Publicaciones se dan los agradecimientos

correspondientes a la Dirección de Redes Temáticas de CONACYT así como al

Proyecto Red Áreas Naturales Protegidas RENANP-CONACYT.

Los resultados son muy satisfactorios a pesar de ser una red relativamente

pequeña, que inicio este 2016 con tan solo 21 miembros, en tan solo un año hemos

generado, ya publicados 4 artículos científicos 2 Libros y 23 Capítulos de Libros.

2

A resaltar es que uno de los libros (y los 23 capítulos del mismo) habían sido

ya publicados digitalmente en Diciembre de 2015, pero no es hasta Noviembre de

2016 y solo gracias al apoyo de este Proyecto CONACYT, que aparece publicada la

1er Versión impresa de este Libro.

Muchas Gracias a la Dirección de Redes Temáticas de CONACYT

3

ARTÍCULOS CIENTÍFICOS PUBLICADOS

Almendarez-Hernández, M.A., G. Avilés-Polanco, V. Hernández-Trejo, Alfredo Ortega-

Rubio and L. F. Beltrán Morales. 2016. Residential Water Demand in a

Mexican Biosphere Reserve: evidence of the effects of perceived price.

Water 2016 (8) 428: 1-14.

Ortíz-Alcaraz, A., Y. Maya-Delgado, P. Cortés-Calva, A. Aguirre-Muñoz, E. Rojas-

Mayoral, M. Cordoba-Matson and Alfredo Ortega-Rubio. 2016. Recovery of

Vegetation Cover and Soil after the Removal of Sheep in Socorro Island,

Mexico. Forests. 2016, 7 (91): 1-13.

Lagunes-Díaz E., M. E. González-Ávila, S. Díaz-Castro, F. Poujol and Alfredo Ortega-Rubio.

2016. Energy consumption drivers: demand side management vs device

penetration, which is stronger? Implications in climate change in arid

Mexico. Fresenius Environmental Bulletin. 25 (8): 3023-3029.

Ortiz-Alcaraz, A., Aguirre-Muñoz, A., Méndez-Sánchez, F. and Ortega-Rubio, Alfredo.

2016. Feral sheep eradication at Socorro Island, Mexico: a mandatory step

to ensure Ecological Restoration. Interciencia, 41 (3): 184-189.

Esparza-Carlos Juan Pablo, John W. Laundré, Lucina Hernández, Luis Ignacio Íniguez-

Dávalos. 2016. Apprehension affecting foraging patterns and landscape use

of mule deer in arid environments. Mammalian Biology 81, 543–550.

4

LIBROS PUBLICADOS

Geiger Villalpando A. y Alfredo Ortega Rubio. 2016. La reglamentación de la pesca

deportiva en México y en Baja California Sur. Universidad Autónoma Baja

California Sur. 130 pp. ISBN: 978-607-7777-67-0

Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia-Moreno (Editores). 2016. Las Áreas

Naturales Protegidas y la Investigación Científica en México. Centro de

Investigaciones Biológicas del Noroeste S. C., La Paz B. C. S., Universidad

Autónoma de Yucatán, Mérida, Yucatán y Universidad Michoacana de

San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN: 978-607-

424- 558-5. (Este libro fue publicado Digitalmente en 2015 y es hasta este

2016 que gracias al apoyo del Proyecto CONACYT de REDES TEMÁTICAS se

publica la 1er Edición Impresa).

El cual consta de 23 capítulos, distribuidos en 7 secciones:

Halffter, G., C. Tinoco-Ojanguren, L. I. Iñiguez-Dávalos y Alfredo Ortega-Rubio. 2016. La investigación científica y las Áreas Naturales Protegidas en México: una relación exitosa. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.3-18). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

5

Ortega-Rubio, Alfredo, C. Jiménez-Sierra, L. Jiménez-Badillo, M. J. Pinkus-Rendón, V. J. Arriola-Padilla, J. Sosa-Ramírez, F. Valadez-Cruz, G. Rodríguez-Quiroz, E. Barba-Macías, C. Espitia-Moreno, P. Cortés-Calva, A. Breceda Solís-Cámara, L. I. Iñiguez-Dávalos, H. A.González–Ocampo y M. A. Pinkus-Rendón. 2016. Prioridades de investigación para las Áreas Naturales Protegidas de México. Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp. 493-512). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Pinkus Rendón, M.Á., M. J. Pinkus Rendón, G. Rodríguez Quiroz y Alfredo Ortega-Rubio. 2016. La sustentabilidad y las Áreas Naturales Protegidas de México: perspectivas de país. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.513-536). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

González-Ocampo, H. P., G. Rodríguez-Quiroz y Alfredo Ortega-Rubio. 2016. Una revisión Panorámica de las Áreas Naturales Protegidas de México. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.19-40). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Torres-Orozco, D., C. L. Jiménez-Sierra, J. Sosa-Ramírez, P. Cortés-Calva, A. Breceda Solís-Cámara, L. I. Iñiguez Dávalos y Alfredo Ortega-Rubio. 2016. La importancia de las Áreas Naturales Protegidas en nuestro país. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.41-64). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

6

Iñiguez-Dávalos, L.I., C. Jiménez-Sierra, J. Sosa-Ramírez y Alfredo Ortega-Rubio. 2016. Valoración de las diferentes categorías de las Áreas Naturales Protegidas en México. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.67-84). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Ortiz-Gallarza, S. M. y Alfredo Ortega-Rubio. 2016. Índice para la toma de decisiones sobre recursos bióticos en Áreas Naturales Protegidas. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.147-176). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Espitia-Moreno, I.C., C. Jiménez-Sierra y Alfredo Ortega-Rubio. 2016. Informes de Responsabilidad Social en las Áreas Naturales Protegidas de México. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.179-193). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Rodríguez Quiroz, G., H. A. González-Ocampo, E. Barba-Macías, L. Jiménez-Badillo, M. J. Pinkus-Rendón, M. Pinkus-Rendón y Alfredo Ortega-Rubio. 2016. Aspectos Socioeconómicos de la Pesca Artesanal en las Áreas Naturales Protegidas. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp. 195-2019). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

7

Lagunas-Vázques, M., A. Gerardo Sosa y Silva, L. F. Beltrán-Morales y Alfredo Ortega-Rubio. 2016. La perspectiva de género en los estudios sociales en las ANPs de nuestro país: una propuesta conceptual y metodológica. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.211-248). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Olmos-Martínez, E., Rodríguez Rodríguez, G., Salas S. y Alfredo Ortega-Rubio. 2016. Efecto de la implementación de una Área Protegida sobre el bienestar de comunidades rurales de Baja California Sur. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.249-282). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Valadez–Cruz, F., G. Rosiles–González y Alfredo Ortega–Rubio. 2016. Potenciales Bioindicadores del Elenco Ficológico de la Reserva De La Biosfera Sian Ka’an. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.285-314). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Arriola-Padilla, V. J., E. Estrada-Martínez, R. Medellín-Jiménez, A. R. Gijón Hernández, L. A. Pichardo-Segura, R. Pérez-Miranda y Alfredo Ortega-Rubio. 2016. Áreas Naturales Protegidas del Centro de México: degradación y recomendaciones. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.337-374). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

8

Cortés-Calva, P., I. González López, B. Bermúdez Almada, C. L. Jiménez-Sierra y Alfredo Ortega-Rubio. 2016. Reserva de la Biosfera el Vizcaíno: 25 años de manejo y conservación. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.375-392). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Ortega-Rubio, Alfredo, F. Santiago-León, M. Lagunas-Vázques, E. Olmos-Martínez, E. Rubio Fierro-Bandala y L. F. Beltrán-Morales. 2016. Vulnerabilidad ecológica, económica y social del sitio RAMSAR Bahía Magdalena, ante el cambio climático: una aproximación espacial. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.411-428). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Sosa Ramírez, J., A. Breceda Solís, C. L. Jiménez-Sierra, L. I. Iñiguez Dávalos y Alfredo Ortega-Rubio. 2016. Los ecosistemas de la Sierra Fría en Aguascalientes y su conservación. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp. 447-472). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Breceda Solís Cámara, A., J. Sosa-Ramírez, C. L. Jiménez-Sierra y Alfredo Ortega-Rubio. 2016. La Sierra La Laguna en Baja California Sur: importancia de su conservación. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.473-490). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN: 978-607-424-558-5

9

Soto-Galera Ernesto y Myriam Adela Amezcua-Allieri. 2016. Uso de la Geomática para el análisis espacio-temporal de las Áreas Naturales. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.411-428). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

Sonia Gallina y Alberto González-Romero. 2016. Los estudios sobre vertebrados y su aplicación en recomendaciones de manejo. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp. 447-472). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5.

Everardo Barba-Macías, Francisco Valadez–Cruz, Miguel Angel Pinkus- Rendón, Manuel Jesús Pinkus-Rendón, y Juan Juárez Flores. 2016. Reserva de la Biosfera Pantanos de Centla: aspectos socio- ambientales prioritarios. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.473-490). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN: 978-607-424-558-5

Manuel Jesús Pinkus-Rendón, Miguel Angel Pinkus-Rendón y Everardo Barba- Macías. 2016. Desarrollo y conservación de la Reserva de la Biósfera Ría Celestún: ¿el ecoturismo es la respuesta? En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp. 447-472). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN 978-607-424-558-5

1

Luis Fueyo Mac Donald y Vladimir Pliego Moreno. 2016. Requerimientos de la Comisión Nacional de Áreas Naturales Protegidas a los investigadores de México para el adecuado manejo de las Áreas Naturales Protegidas. En: Ortega–Rubio, A., M. J. Pinkus-Rendón e I. C. Espitia- Moreno (Editores). Las Áreas Naturales Protegidas y la Investigación Científica en México. (pp.473-490). Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, México. 572 pp. ISBN: 978-607-424-558-5

water

Article

Residential Water Demand in a Mexican BiosphereReserve: Evidence of the Effects of Perceived PriceMarco Antonio Almendarez-Hernández 1, Gerzaín Avilés Polanco 2, Víctor Hernández Trejo 2,Alfredo Ortega-Rubio 1 and Luis Felipe Beltrán Morales 1,*

1 Centro de Investigaciones Biológicas del Noroeste, Instituto Politecnico Nacional 195,Colonia Playa Palo de Santa Rita, 23096 La Paz, B.C.S., Mexico; malmendarez@cibnor.mx (M.A.A.-H.);aortega@cibnor.mx (A.O.-R.)

2 Department of Economics, Universidad Autónoma de Baja California Sur, Carretera al Sur Km 5.5,23080 La Paz, B.C.S., Mexico; gaviles@uabcs.mx (G.A.P.); victorhtrejo@gmail.com (V.H.T.)

* Correspondence: lbeltran04@cibnor.mx; Tel.: +52-612-123-8484

Academic Editor: Jay R. LundReceived: 11 August 2016; Accepted: 21 September 2016; Published: 29 September 2016

Abstract: The purpose of this paper is to provide empirical evidence for policy-makers of watermanagement, evaluate the applicability of economic variables such as price and other factors thataffect demand, and determine the impact thereof on decision-making surrounding water managementin the El Vizcaino Biosphere Reserve in Mexico. We estimated a dynamic function with an averageprice specification, as well as price perception specification. Findings demonstrated that consumerstend to react to perceived average price but not to the marginal price. Furthermore, long-termprice elasticity was found to be higher than short-term elasticity, and both elasticities were foundto be inelastic. Inelastic elasticities, coupled with rising prices, generate substantial revenues withwhich to improve water planning and supply quality and to expand service coverage. The resultssuggest that users’ level of knowledge surrounding price is a key factor to take into account whenrestructuring rates, especially in situations where consumers do not readily possess the necessaryinformation about their rate structure and usage within a given billing period. Furthermore,the results can help water management policy-makers to achieve goals of economic efficiency, socialequity, and environmental sustainability.

Keywords: lagged consumption; dynamic function; water management; average price; marginal price

1. Introduction

The principal sources of water for human use are rivers, lakes, and aquifers, which togetherrepresent approximately 10 million km3, less than 1% of the total volume of water that exists inthe hydrosphere. Each year, approximately 505,000 km3 of water evaporates from the ocean, ofwhich 90% returns to the sea in the form of precipitation, with the remaining 10% falling on thecontinents. Together with local precipitation, the volume of which is approximately 68,500 km3

per year, a total of about 119,000 km3 falls on the Earth’s landmasses each year. Asia and SouthAmerica are the continental zones where the largest volume of water runoff occurs, with 14,100and 12,200 km3, respectively [1]. Latin America is the region with the largest volume of water perinhabitant, with 48,000 m3 [2]. Sixty-five percent of water consumption in Central America is sourcedin subterranean water, and in South America this number ranges from 40% to 60% [3].

In Mexico, water availability per inhabitant is 4547 m3 [4]. The principal water-related problemsin Mexico are linked to inefficiencies in use. The agricultural sector utilizes 77% of water allocatednationally, of which 67.34% is extracted from surface sources and 32.65% originates underground.Of total water extraction, transport efficiency is 63.8%, while the remaining quantity evaporates,

Water 2016, 8, 428; doi:10.3390/w8100428 www.mdpi.com/journal/water

Water 2016, 8, 428 2 of 14

is filtered, or is lost in the process. This indicates that although the Mexican agricultural sectorconsumes 56.1 km3 of water annually, it only actually utilizes 35.8 km3. Meanwhile, the industrialsector uses 10% of national water allocations (6.9 km3 annually), 76.8% of which is extracted fromsurface sources and 23.2% from aquifers. The primary problem related to water use in this sector is thecontamination of watersheds and aquifers by residual waste, given that Mexican industry generates5.62 km3 of residual water, of which only 0.85 km3 is recovered for treatment, while 4.77 km3 isdischarged directly into large bodies of water [5].

The urban public sector in Mexico uses 13% of national water allocations (9.6 km3). The majorityis extracted from aquifers (65.62% or 6.3 km3 annually), and only 34.37% (3.3 km3) is taken from thesurface. Eighty of the 188 largest aquifers in Mexico, which together supply 66% of the water used inthe country and in which 79% of groundwater recharge is captured, are being overused. Other factorsrelevant to inefficient use by the urban public sector are the deficient coverage of potable water andsewerage (10.2% and 23.8%, respectively); inappropriate pricing structures; lack of information forusers; deficiencies in micro level water metering; and externalities related to watershed and aquifercontamination, given that 70% of the largest watersheds in the countries are contaminated by residualwater discharge of up to 8.05 km3 per year. Of this only 80.24% (6.46 km3) is collected, and only 35%(2.26 km3) is treated [5].

Another issue relevant to water use and management in the urban public sector is the lack ofaccurate water metering. According to the Comisión Nacional de agua (CNA) [5], in 39 cities withpopulations greater than 50,000 residents, only 46% of water taps have an installed meter. The lackof water metering infrastructure is the cause and consequence of budgetary insufficiencies in themajority of the Potable Water and Sewage Operating Units in Mexico. Given this situation, these Unitsare obliged to bill users based on approximate consumption, known as “Averaged Consumption”.This estimate tends to be undervalued, and as a consequence users who pay an averaged, fixed feehave no need to match the cost of consuming an additional unit of water to a marginal benefit; instead,users take advantage of the situation by consuming until the marginal benefit is equivalent to zero.The result is inefficiency in use and economic infeasibility.

Barkin and Klooster [6] argue that problems exist at federal, state, and municipal levels forthe implementation of appropriate water management, including the following: (a) institutionalbarriers; (b) administrative shortcomings; (c) environmental impacts such as overexploitation ofaquifers and contamination thereof; (d) impossibility of quantifying with certainty the hydrologicalbalance; (e) poor quality service provision for consumption; (f) unclear prospecting in thehydrological sector; (g) insufficient technical and administrative capacities of relevant personnel;and (h) information shortage.

In the case of the Vizcaino Biosphere Reserve, 50% of users have a water meter installed, and theconsumption volume reported by Organismo Operador Municipal del Servicio de Agua potable yAlcantarillado (OOMSAPA) can reach up to 2,879,461 m3 per year, of which the residential sectorabsorbs 90% [7]. To date, only one study has been conducted in this region surrounding the impactof improved water consumption metering, and this study focused on the commercial sector. Resultsindicated that in the short-term, improved water metering positively impacts financial revenue,and over the long-term, measurement reduces water consumption, and price elasticity of demand ishighly inelastic [8].

Other studies have indicated that pricing systems for water consumption should be designedbased on legal and environmental aspects, and also that pricing structure represents one of the mostimportant management instruments through which to achieve economic efficiency, improve equity,and maintain the sustainability of hydrological resources [9–11]. Pricing policies can also have theeffect of incentivizing changes in the behavioral patterns of the individual’s water consumption,promoting responsible use and thereby controlling water demand, which is especially important inregions where water resources are limited. Furthermore, revenues generated through higher prices cangenerate increased financial resources for water management, for instance for the agencies responsible

Water 2016, 8, 428 3 of 14

for water supply planning. It is important to recognize that in order to improve access to water acrosspopulation sectors and guarantee that quality is consistent with international standards, technologiesthat use water efficiently and enable water recycling and reuse are required. Similarly, pricing increaseshave had the effect of fomenting water reallocation between sectors (e.g., from agricultural irrigationto domestic and industrial uses).

According to Rogers et al. [9] water usage rates should meet the following objectives: (1) maximizethe efficient allocation of resources; (2) be perceived as fair by water users; (3) be equitable betweencustomer classes; (4) generate sufficient income; (5) provide net income stability; (6) involve a processof rate setting that is understood by the public; (7) promote resource conservation; (8) avoid shocks inrates; (9) be easily implemented; (10) entail water accessibility; (11) take future changes into account;(12) reduce administrative costs; (13) include environmental costs; (14) not be in conflict with othergovernmental policies; (15) reflect the characteristics of water supply and quality, as well as reliabilityand frequency of supply; (16) vary depending on measurability and consumption; (17) take into accountdaily peaks and seasonal variations in water demand (for more sophisticated pricing schemes).

Given these various considerations, the desire to measure the potential impact of pricing policieson water demand management has in turn motivated the proliferation of methods to more preciselyestimate price elasticity of demand and income [12,13]. The variables most frequently used in thedevelopment of this research are marginal price (MP), average price (AP), and the combination ofboth. However, in the existing literature surrounding water demand, controversy exists on whichprice variable is the most appropriate for use in obtaining precise estimates of elasticities when userspay for water consumption based on a pricing structure block. This debate centers on the fact thatinformation about consumer price blocks is imperfect. Consumers themselves are not typically familiarwith blocked pricing structures, and therefore they are not aware of the marginal price for usage.Users adjust their consumption behaviors to variations in average prices, because they do not havesufficient incentives, including time, to learn about marginal prices [14,15].

Therefore, it is difficult to develop hypotheses assuming that residential water consumers havecomplete knowledge about the relevant rate scheme [16,17]. Arbués et al. [12], and Worthingtonand Hoffman [13], in reviews of empirical studies of water demand, demonstrate that in mostcases there are no significant differences in the results surrounding elasticities derived from thetwo aforementioned price variables, and results are inconclusive about which alternative is preferred.Although studies reveal price elasticities of demand to be inelastic, long-term elasticity is greater thanthat of the short-term by a considerable amount, producing substantial effects on consumer reactions.Furthermore, elasticities have been shown to be higher under tiered pricing schemes whose rateincreases are non-linear, as compared to those associated with decreasing blocks and uniform prices,given that the first structure tends to send stronger signals to users to reduce water consumption [18,19].Increasing blocks represent the rate structure typically employed by developing countries [20].

Researchers conducting empirical work generally face challenges surrounding the availabilityof information where new variables are frequently added and micro level data is scarcely available.Given the lack of empirical information surrounding the effectiveness of water pricing structures andthe benefits associated with efficacious pricing methodologies, the objective of the present study wasto estimate a dynamic function that would incorporate average price, employing the methodology ofShin [21] in seeking to verify the hypothesis that consumers react to perceived changes in prices ratherthan actual prices. The results of Shin [21] demonstrated that residential electricity consumers did notpossess the necessary information to optimize usage in a situation of imperfect information availability,in turn indicating that consumers would not recognize the difference between average price andmarginal price or the impacts thereof on consumption. Thus, the authors utilized average price asa heuristic. In applying the same method to the context of residential water demand in a MexicanBiosphere Reserve, it should be understood that consumers have imperfect information about thepricing structure as well as concerning their water usage during the relevant billing period.

Water 2016, 8, 428 4 of 14

2. Materials and Methods

2.1. Study Area

The geographical boundaries of the Vizcaino Reserve are: from the west beginning at the 28thparallel extending from the Laguna Guerrero Negro to the estuary El Datil, and on the east extendingfrom the 28th parallel to Santa Rosalia, and from here to the southeast through the transpeninsularhighway, through the Laguna San Ignacio and Barra of San Juan (Figure 1).

The Vizcaino Reserve is one of the largest in Latin America with about 2.55× 106 ha. It is classifiedas an arid region, with a dry climate and low annual rainfall [22]. Due to weather patterns andgeological conditions, water use by the local population involves pumping from subterranean aquifers.

Water 2016, 8, 428  4 of 14 

2. Materials and Methods 

2.1. Study Area 

The geographical boundaries of the Vizcaino Reserve are: from the west beginning at the 28th 

parallel  extending  from  the  Laguna  Guerrero  Negro  to  the  estuary  El  Datil,  and  on  the  east 

extending  from  the  28th  parallel  to  Santa  Rosalia,  and  from  here  to  the  southeast  through  the 

transpeninsular highway, through the Laguna San Ignacio and Barra of San Juan (Figure 1). 

The Vizcaino Reserve  is  one  of  the  largest  in Latin America with  about  2.55  ×  106  ha.  It  is 

classified as an arid region, with a dry climate and low annual rainfall [22]. Due to weather patterns 

and geological conditions, water use by the local population involves pumping from subterranean 

aquifers. 

 

Figure 1. Area of Biosphere Vizcaino Reserve. 

2.2. Specification of Water Demand Dynamics 

Relationships  in  a  model  whose  character  is  dynamic  by  nature  can  be  studied  by 

incorporating  the  lagged dependent  variable  among  the  regressors,  known  as  an  autoregressive 

model. The equation has a bilogarithmic form, 

ln δln , ln β μ   (1) 

where  ln   is the natural logarithm of the average community water consumption over the course 

of time t;  ln ,   is the natural logarithm of consumption lagged by one month; z is a matrix of size 

(n  ×  m)  containing  m  independent  variables  (price,  income,  maximum  temperature  and  total 

monthly  rainfall); β is a vector of  size parameters (m × 1); μi constitutes  the discrepancies among 

consumers that correspond to each individual community, which is generally not designed by the 

researcher and  is  taken  into account as a  fixed effect. The stochastic component   represents  the 

distance  between  the  estimated  consumption  and  consumption  frequently  observed,  and  is  not 

inferred by the researcher. 

Figure 1. Area of Biosphere Vizcaino Reserve.

2.2. Specification of Water Demand Dynamics

Relationships in a model whose character is dynamic by nature can be studied by incorporatingthe lagged dependent variable among the regressors, known as an autoregressive model. The equationhas a bilogarithmic form,

lnwit = δlnwi,t−1 + lnz′itβ+ µi + vit (1)

where lnwit is the natural logarithm of the average community water consumption over the courseof time t; lnwi,t−1 is the natural logarithm of consumption lagged by one month; z is a matrix of size(n × m) containing m independent variables (price, income, maximum temperature and total monthlyrainfall); β is a vector of size parameters (m × 1); µi constitutes the discrepancies among consumersthat correspond to each individual community, which is generally not designed by the researcher andis taken into account as a fixed effect. The stochastic component vit represents the distance between theestimated consumption and consumption frequently observed, and is not inferred by the researcher.

Studies by Nauges and Thomas [23], García [24], and Chang and House-Peters [25] indicatethat the use of dynamic models that include a lagged dependent variable generate greater precision

Water 2016, 8, 428 5 of 14

in the estimation of parameters, leading to improved predictions surrounding water consumption.Furthermore, it is assumed that demand responds instantaneously to changes in prices, which is themost common means to econometrically evaluate the relationship between demand and consumption.The reason that the dependent variable is lagged is because current consumption is strongly associatedwith past consumption habits, given that users do not typically change consumption patters forpsychological, technological, or institutional reasons. Current consumption estimates therefore dependon data surrounding usage in the previous month, including by household fixtures and appliancessuch as showers, toilets, washing machines, etc. The frequency of use of such equipment changesgradually over the short-term but may be substantially impacted by replacements or upgrades in thelong-term. However, consumers frequently resist upgrading household appliances that consume lesswater due to the high costs associated with replacements.

Such factors lead to slow reactions by users to changes in prices, as well as gradual adjustmentsto their consumption. As such, larger displacements are expected to occur in the demand curve in thelong-term. Accordingly, it is desirable to formulate a methodology utilizing a partial adjustment modelbecause water is a basic necessity. Therefore, existing inertias in consumption are highly relevant,and these inertias are taken into account by incorporating the lagged dependent variable among theregressors. Furthermore, the average price methodology was selected because users lack informationsurrounding the pricing structure block, in addition to the marginal price. In the Vizcaino Reserve,consumers receive invoices almost immediately after water is consumed, and payment must be madequickly due to the short time limits imposed by the municipal system. For these reasons, in theeconomic specification the price was not lagged by one month, based on the recommendations of otherstudies for situations in which the water bill is delivered to the home between one and two monthsafter consumption has occurred [11,24,26].

Additionally, few previous studies consider inertia in consumption resulting from consumerhabits, or from the fact that individuals tend not to respond instantaneously to variations in rates.Thus, current consumption has been assumed to adjust quickly (i.e., during the same period) to thedesired level. Such an approach may not be realistic, as illustrated in arguments explained in thepartial adjustment model. To not consider real consumption patterns could generate partial andinconsistent estimates. Additionally, the methodology employed by Shin [21] is used in the presentstudy to test their hypothesis, which argues that consumers do not adjust consumption to variations inreal prices, but rather to changes in perceived prices. There are several reasons for which it would betoo costly for the consumer to determine the actual rate paid for water consumption: (1) It is difficultfor consumers to know the difference between the average and the marginal price and the impactsthereof on consumption, given that consumers are typically unaware of the pricing structure block;(2) Even if the consumer was aware of the pricing structure block, it would be difficult to respondimmediately to changes in prices and he would only adjust consumption after receiving an invoice forthe current billing period; (3) It is very unlikely that the consumer would differentiate water pricesfrom other charges on the invoice, such as sewer service.

In summary, the methodological approach of Shin [21] assumes that consumers consider themarginal costs and marginal benefits in calculating a marginal price without solid foundations, therebyreacting to variations in perceived prices and not to any estimated amount that would reflect actualprices. Depending on how the consumer weighs relevant factors, three responses are possible: (1) If themarginal benefit is less than the marginal cost, the consumer will not respond to the marginal priceand will instead determine consumption based on other price information; (2) If the marginal expectedbenefit is greater than the marginal cost, the consumer will likely determine the real marginal priceand the perceived price will equal the marginal price; (3) If the price structure equals the marginal cost,the consumer will stop searching for information and the perceived price will oscillate between themarginal price and the average price.

According to Shin [21] the perceived price is formulated as follows,

P∗ = MP(AP/MP)k (2)

Water 2016, 8, 428 6 of 14

where MP is the marginal price, AP is the average price, and k is a parameter representing priceperception. The ratio between AP and MP captures the effect of the difference variable on priceperception. It is expected that the parameter k will not be negative. The possible results of k are:(1) When k = 0, the consumer reacts to changes in MP; (2) When k = 1, the consumer reacts to AP;(3) Assuming a structure of prices with increasing blocks (i.e., the rate is progressive), when P∗

varies between AP and MP, we get the result 0 < k < 1. When k > 1, P∗ < AP < MP and when k < 0,P∗ > AP > MP.

The econometric specification is expressed as follows,

lnwit = β0 + αlnPMPit + βlnz′it + µi + vit= β0 + α [(1− k) lnMP + klnAP] + βlnz′it + µi + vit= β0 + β1lnMP + β2lnAP/MP + βlnz′it + µi + vit

(3)

where β1 = α (1− k), β2 = αk, being k = β2/β1. z is a matrix of size (n × m) containing m independentvariables (income, maximum temperature, and total monthly precipitation), β is a vector of sizeparameters (m × 1). Recent studies by Ito [27] on electric consumption and Wichman [28] onwater consumption employed quasi-experimental methods, and their results have demonstratedthat behavior is best explained by average price.

2.3. Method of Estimating Water Demand Dynamics

The dynamic panel with fixed effects is the econometric strategy used to estimate the function ofwater demand, which includes past consumption as a control variable,

yit = δyi,t−1 + x′itβ+ µit, i = 1, . . . , N; t = 1, . . . , T (4)

with uit = µi + vit, following the approach of Baltagi [29] µi is fixed and constant for each individualsuch that vit ∼ IID

(0,σ2

v). (I.I.D. means that errors are independent and identically distributed.)

The model also assumes that the explanatory variables are uncorrelated with random error,but may be correlated with individual effects. However, the presence of the lagged dependentvariable in the model causes problems of endogeneity because of correlation with the error term [29].As an alternative solution, Kiviet [30] suggested the use of a Least Squares Dummy Variable (LSDV)as an estimator suitable for finite samples. The correction of bias within the transformation estimatoris known as the method of Least Squares Dummy Variable Corrected (LSDVC).

In their Monte Carlo simulations, Judson and Owen [31] demonstrated evidence that whenthe period is 30, the bias of the fixed effects estimator is considerable. These authors recommendusing the LSDVC estimator when the lapse is ≤10 and the Anderson and Cheng estimator [32]when the lapse is significant. Other alternatives exist to correct the problem of endogeneity, such asInstrumental Variables estimators (IV) and the Generalized Method of Moments (GMM). However,these are designed for N → ∞ with a fixed T, that is, they demonstrate consistency for a large numberof cross-sectional units (N), even where the length of the time series is short. Meanwhile, the Arellanoand Bond [33] estimator has a significant downward bias in small samples. This is because one ofthe disadvantages inherent in IV estimators such as Anderson–Hsiao (AH) and Generalised Methodof Moments (GMM) estimators such as Arellano–Bond (AB) and Blundell-Bond (BB), is that theirasymptotic properties depend on having a large N, which is characteristic of micro panel data.

Recently, Bruno [34] developed a method for utilizing the LSDVC estimator for unbalanced panels.In this method, bias is corrected through a consistent estimator such as AH, AB and BB, where thethree alternatives used to initialize the bias correction are asymptotically equivalent. In our case T isrelatively large, that is instances of T → ∞ and N are either few or N is fixed. Cermeño [35], throughan empirical study, demonstrated that LSDV estimator bias is lower compared to the estimates thatconsider T to be small.

Water 2016, 8, 428 7 of 14

2.4. Description of the Database

The present study considered data from the 2010 to 2014 period to inform the econometricmodel, including information from seven communities: San Ignacio, Bahia Tortugas, Bahia Asuncion,Villa Alberto Alvarado, Guerrero Negro, Mulege, and Santa Rosalia. The description of each isprovided in Table 1.

Table 1. Description of variables used in the regression analysis.

Variable Description Source

w Symbolizes average water consumption per capita inresidential use. The variable is measured in cubic meters (m3).

System Operator Agency WaterSupply and Sewerage

(OOMSAPA).

AP

Average price, obtained by dividing the water bill paid by theconsumer living in one housing unit and the volume of waterconsumed. Additionally, the measurement of price wasdeflated using the National Consumer Price Index (NCPI),base 2010 = 100, where 2010 is the bases year for theestimation, obtained from the Bank of Mexico (BM).

System Operator Agency WaterSupply and Sewerage

(OOMSAPA).

MP

Marginal price, representing the amount that the consumermust pay, according to the fee structure for final consumptionunits associated with the average amount. The price wasdeflated using the CPI, base 2010 = 100.

System Operator Agency WaterSupply and Sewerage

(OOMSAPA).

Income

Defined as the average daily wage by state according to theMexican Social Security Institute (IMSS). In the regressionanalysis it is used as a proxy for income, representing anindicator of household income. For purposes of inclusion inthe dynamic equations, we calculated monthly wage.This variable was deflated with CPI base 2010 = 100 andweighted with the working population.

National Commission forMinimum Wage in the State of

Baja California Sur.

t Monthly maximum temperature, measured in degreesCelsius (◦C).

National Water Commission(CONAGUA).

P Total monthly precipitation, measured in millimetres (mm). National Water Commission(CONAGUA).

Table 2 shows the descriptive statistics used in the econometric specifications.

Table 2. Descriptive statistics of variables selected for the regression analysis.

Variable Mean Standar Devation Minimun Maximun

Natural Logarithm of Water Consumption 3.08 0.39 1.27 3.92Natural Logarithm of Average Price 1.59 0.37 1.22 3.22

Natural Logarithm of Income 8.44 0.04 8.33 8.5Natural Logarithm of Temperature 3.48 0.18 2.89 3.78Natural Logarithm of Precipitation 9.5 25.15 0 218

Natural Logarithm of Marginal Price 1.49 0.19 1.12 2.88

3. Results and Discussion

The first step, prior to conducting the econometric analysis of the demand functions, was theanalysis of each one of the series represented in the variables. For this analysis, unit root tests wereperformed on the panel data environment, proposed by: (1) Breitung [36]; (2) Levin et al. [37]; (3) Harrisand Tzavalis [38]; (4) Im et al. [39], known as IPS (Im, Pesaran and Shin); and (5) Fisher-type tests [40]known as the Dickey-Fuller (ADF) and Phillips-Perron (PP). The results in Table 3 demonstrate thatit was not necessary to apply cointegration vectors, providing support for a methodology involving

Water 2016, 8, 428 8 of 14

stationary methods, given that the null hypothesis, which would assume nonstationarity at commonlevels of significance, is rejected.

Starting with the dynamic panel, the regression results presented in Table 4 show the variation ofthe demand for residential water attributed to the independent variables considered in the analysis.A significance level of 1%, with a value between 0 < wi,t−1 < 1, resulted from the analysis of averagelagged consumption. The speed of adjustment was obtained by subtracting 1 from 0.62 (its coefficient),the difference being 0.38. The interpretation of this result is that that the gap of 38% separating actualand desired demand for water is closed within a period of one month. With respect to the findingssurrounding price elasticity of demand, the expected results were obtained and demand was foundto be inelastic, consistent with the general economic theory that postulates an inverse relationshipbetween quantity demanded and price. This result indicates that for user response to the percentagechange in price, the percentage change in quantity demanded is less than the percentage changein price.

The short-term elasticity for estimating water demand for domestic use is approximately −0.27and long-term elasticity is −0.71. The value of the first number is lower than that of the second,suggesting that consumers react primarily to continued increases in rates and not to variations of onemonth in duration. The permanence of higher prices enables individuals to adapt after one month,thereby adjusting their consumption patterns. The explanation of this result is based on the observationthat consumption habits of domestic users tend to remain stable, caused in part by typically minimalvariations in water prices. Furthermore, psychological and technological factors may result in gradualrather than immediate adaptation to increases in water rates [23,24]. Schleich and Hillenbrand [41]suggest that the results of the elasticities found in more recent studies could demonstrate a downwardtrend, possibly because the rates represent a small proportion of household income.

One method for verifying that the estimates of panel data with a lagged dependent variable donot contain inconsistencies and demonstrate that the specification is correct is to compare the estimateof this function with the estimate of a static form such as Equation (4), that is, to identify an alternativeassumption in accordance with the suggestion of Angrist and Pischke [42]. The results of the staticequation (model 3) of the coefficient MP and AP/MP are similar to those from model 2 from Table 3,and as such the findings are robust.

An important aspect of the analysis is to highlight that endogeneity may originate because theprice is considered as an exogenous variable related to water consumption, and if not treated withappropriate econometric techniques it will lead to partial and inefficient estimators. However, we useda database compiled by a representative community, and Shin [43] argues that the endogeneity problemis not very serious in equations that use aggregated information as the original source, compared withequations that use microdata for the synchronization effect created by the existence of a correlationbetween price and the error term.

Additionally, a test of the error term from Hausman [44] was calculated to check the problemof endogeneity. This was done in order to compare two estimators, one determined to be consistentunder the null hypothesis and another an estimator for the instrumental variables (IV). If the nullhypothesis is not rejected, the estimator that is considered to be consistent will produce parameterestimates that are unbiased and efficient, and if the null hypothesis is rejected it will produce biasedand inefficient estimates. The Hausman statistic was 1.09, indicating that at the 10% significance levelthe null hypothesis is not rejected, suggesting that the difference in the coefficients is not systematic,and as such there is no evidence for the problem of endogeneity. Of additional relevance is the factthat the water rate is fixed by the operating organism, and does not change based on demand.

Water 2016, 8, 428 9 of 14

Table 3. Results of unit root tests of the dynamic function variables.

TestVariable

Natural LogarithmWater Consumption

Natural LogarithmAverage Price

Natural Logarithmof Income

Natural LogarithmTemperature

Natural LogarithmMarginal Price Precipitation

Levin, Lin and Chu t-stat 1 No trend −3.5058 * −2.9591 * −4.7032 * −3.3096 * −5.3825 * −14.8757 *Trend −3.7844 * −3.6753 * −4.5454 * −2.9802 * −7.1843 * −16.4279 *

Breitung t-stat 1 No trend −2.016 ** 0.8364 −2.8989 * −3.8967 * −3.8093 * −11.1172 *Trend −2.9742 * −1.7696 ** −1.6902 ** −3.1495 * −1.5693 −11.1841 *

Harris-Tzavalis 1 No trend −12.4917 * −9.9615 * −39.8499 * −14.5436 * −29.5214 * −39.8661 *Trend −8.4543 * −8.6226 * −30.2714 * −7.3701 * −24.2629 * −24.5517 *

Im, Pesaran and Shin W-stat 2 No trend −4.7914 * −5.1862 * −6.9979 * −9.5987 * −7.1840 * −13.172 *Trend −4.9233 * −3.7272 * −13.1415 * −8.9286 * −9.8964 * −13.7871 *

ADF-Fisher Chi-square 2 No trend 61.273 * 24.9921 ** 76.8509 * 117.714 * 87.2638 * 170.723 *Trend 55.8855 * 36.2216 * 158.721* 98.8576 * 112.370 * 161.536 *

PP-Fisher Chi-square 2 No trend 61.0525 * 42.2184 * 209.668 * 81.1587 * 83.0181 * 183.702 *Trend 54.2991 * 59.9753 * 234.038 * 57.8068 * 105.175 * 156.455 *

Notes: * The null hypothesis of unit root is rejected at 1%; ** The null hypothesis of unit root is rejected at 5%; 1 Represents a common unit root process; 2 Represents an individual unitroot process.

Table 4. Estimates of the dynamic function.

VariableModel 1 Model 2 Model 3

Coefficient t-Ratio Probability Coefficient t-Ratio Probability Coefficient t-Ratio Probability

Constant 0.6426 1.4173 0.1572 0.5768 1.2110 0.2940 1.0790 1.3621 0.1739Lagged consumption 0.6171 13.2977 * 0.0000 0.6116 12.2336 * 0.0000 - - -

AP −0.2735 −5.9151 * 0.0000 - - - - - -MP - - - −0.2587 −4.3048 * 0.0000 −0.2830 −5.5984 * 0.0000

AP/MP - - - −0.2803 −5.6858 * 0.0000 −0.3123 −7.5952 * 0.0000Income 0.1047 1.7747 *** 0.0767 0.1127 1.8093 *** 0.0712 0.1329 1.8579 *** 0.0639

Temperature 0.0259 1.3592 0.1748 0.0245 1.2785 0.2018 0.0557 1.1124 0.2666Precipitation −0.00015 −2.5679 ** 0.0106 −0.00015 −2.4170 ** 0.0161 −0.00017 −2.8669 * 0.0043

k - - - 1.0832 4.8567 * 0.0000 - - -R2 0.9253 0.9251 -

F Test of fixed effects 13.1292 *, 0.0000 13.0892 *, 0.0000 -DURBIN-WATSON 2.1497 2.1507 -

Notes: To compute the ratios t, heterostedasticity robust standard errors were used. * Significant at 1%; ** Significant at 5%; *** Significant at 10%.

Water 2016, 8, 428 10 of 14

Meanwhile, income elasticity of demand demonstrated the positive sign expected according toextant economic theory, with a coefficient of 0.10. This result indicated that for every 1% increase inthe income of residential water users, changes in water consumption experienced a correspondingincrease of 0.10%. In other words, change manifested as an increase in consumption patterns ratherthan a change in consumption itself, suggesting that water is a normal good.

With respect to the environmental variables analyzed, maximum temperature was not found tobe significant. A negative correlation was found between total monthly rainfall and consumption.This result is logical, because when heavier or more frequent rains fall, water users use less stored water,for example to irrigate, thereby reducing overall consumption. This variable cannot be interpretedas involving directly proportional change, and should therefore be transformed exponentially.When converting the maximum temperature variable, the resulting calculation is 0.9998, indicatingthat for each additional mm of precipitation, demand is reduced by 0.02%, which is very low. This lowcoefficient indicates that variations in precipitation have little impact on users’ consumption decisions.

It is important to highlight the similarity between these results and the coefficients obtainedthrough the econometric analysis, which sought to test the hypothesis proposed by Shin [13], in whichit is postulated that consumers respond to perceived price. Here, the price perception parameterwas 1.08 and the null hypothesis of k = 0 with t = 4.8567 was rejected with 4.8567 at a significancelevel of 1%; meanwhile, the null hypothesis of k = 1 was not rejected with 0.3730, corroborating thefindings suggesting that consumers react to the average perceived price when making usage decisions.Effectively, users believe that the price paid is lower than the actual rate.

According to Shin [21], consumer decisions are based on a perceived price that they believe to bebelow that specified in institutional rates. However, our results suggest a marginal price lower than theperceived price, indicating that incentives for rational consumption in the short-term are low, which isworrisome given that the study took place in an arid region. For environmental policies and politicaldecision makers, this situation is not ideal because it appears that the rate structure is not achieving atleast one of the objectives for which it was designed, that is, promoting water conservation. As such,the rate structure is operating in an environment in which clients do not know with certainty themarginal price that they pay.

Nevertheless, in the long-term the results indicate that permanent increases in prices couldimprove conservation even if not of the same magnitude as that corresponding to a situation inwhich users are conscious of the marginal price. Furthermore, the present study could be consideredalongside the work of Almendarez-Hernández et al. [45] in which a contingent valuation study wasconducted in order to understand the willingness of homes located in the El Vizcaino BiosphereReserve to pay for improvements of resource availability in the future. Homes would be required toimplement conservation practices but supply would be of sufficient quantity and quality. The resultsobtained signaled a rate increase of 14%.

In the present study, short-term elasticity was −0.26 while elasticity in the long-term was −0.67.These findings are very similar to those reported in recent studies estimating water demand usingalternative average price. In a study conducted in France, Nauges and Thomas [23] estimateda dynamic demand function with annual panel data using a non-lineal GMM and a GMM basedon Blundell and Bond, with a double-difference estimator. The results obtained included a short-termelasticity of −0.26 and a long-term elasticity of −0.40. Musolesi and Nosvelli [46], with annualpanel data from Italy and also using the GMM system, obtained a short-term elasticity of −0.24and a long-term elasticity of −0.47. Similarly, in Germany, Schleich and Hillenbrand [41] estimateda static model under the Ordinary Least Squares (OLS) method and found an elasticity of −0.242.Fullerton et al. [47] in the region of Halifax in Canada calculated an elasticity of −0.31 usingcross-sectional data and an OLS estimator. Finally, with information from Tunisia organized astrimestral, non-seasonal panel data, Younes [48] used a Full Modified OLS method (FMOLS) in whichblocks of low and high consumption were segmented, and through which elasticities between −0.08and −0.46 were found.

Water 2016, 8, 428 11 of 14

Researchers who have employed the price perception methodology introduced by Shin [21]include Nieswiadomy and Molina [49] in Texas, United States of America, with information arrangedin monthly panel data and using the IV estimator. This study examined usage in increasing anddecreasing blocks, and found elasticities between −0.11 and −0.30. Meanwhile, in the communityof Windhoek in Namibia, Kavezeri-Karuaihe et al. [50] used a static model with a full informationmaximum likelihood (FIML) focus, and obtained elasticities between −0.25 and −0.60. Binet et al. [51],using cross-sectional data from France, estimated an elasticity of −0.31 through an optimal GMMmodel. In North Carolina, USA, Wichman [28] used a Difference-in-Difference-in-Difference (DDD)model for information ordered in monthly panel data and calculated elasticities between −0.43 and−1.14 and a regression discontinuity framework of −0.31. Similarly, Wichman et al. [52] used paneldata in which price and non-price policies were evaluated for periods of drought, and found elasticitiesbetween −0.15 and −1.08.

Finally, the results found in the present study are within the range of findings reported in studiesthat have econometrically evaluated demand functions for residential use in Mexico with a Nordinspecification and an average price, such as those discussed in Jaramillo-Mosqueira [53]. This latterstudy used intra-annual information structured in panel data, employing Discrete-Continuous Choicemodels and an IV estimator, and calculated elasticities between −0.22 and −0.58. Garcia-Salazarand Mora-Flores [54], also using panel data, found elasticities between −0.18 and −0.20. Salazarand Pineda [55] used IV and Generalized Least Squares (GLS) methods for panel data, and obtainedelasticities of −0.33. Avilés-Polanco et al. [56], with monthly time series data, used an IV estimator andfound a short-term elasticity of −0.51 and a long-term elasticity of −0.90.

4. Conclusions

The present study presents relevant, quantitative information that could be considered by policymakers for improving their decision-making surrounding water management policies. The resultscould provide guidance to individuals and institutions responsible for water management practices,empowering them with information with which to evaluate or redesign water rates for residential use.This could be especially useful in contexts in which consumers have imperfect information and it wouldbe too expensive for water managers to invest the requisite time and resources in consumer education.

Furthermore, the findings of the present study surrounding elasticities in water price increasesand in relation to inelastic price could lead to the generation of increased revenue, which in turn couldbe used to improve supply planning. Results could also be considered alongside of those found in otherstudies applying contingent valuation methods to understand the implementation of conservationprograms for aquifer restoration. As such, long-term policies could be designed to provide servicesto populations without access to potable water and responsible resource use [45,57,58]. For instance,previous studies have examined the willingness of households that are alternately connected or notconnected to the drinking water network to pay for improvements to the system in order to ensureresource sustainability in the future, and have found that connected households are more willing to pay.

The introduction of a discriminatory pricing policy that considers seasonality as a factor(i.e., winter and summer) would likely not be as effective for conservation purposes as alternativepricing policies, given its low coefficient. On the other hand, income elasticity of demand was smallfor this variable, and fell within the range of income elasticity frequently reported in the literature.Estimates suggest that implementing management strategies including instruments such as price areof fundamental importance. Such strategies should be understood as a basis for the evaluation ofthe functions for which these policies have been structured, and specifically for the development ofrates that would satisfy policy objectives. An adequate design of the pricing system for residentialwater use can complement other social, legal, and environmental policies designed to improve watermanagement in protected natural areas.

Acknowledgments: We would like to thank the National Council of Science and Technology of Mexico(CONACYT) for the financial support that this agency provided for this research (Grants numbers 251460,

Water 2016, 8, 428 12 of 14

269540 of Redes Temáticas and 251919 of Ciencia Básica), as well as the Northwest Biological Research Center ofMexico (CIBNOR), the Biosphere Reserve of Vizcaino, the Municipality of Mulege, and for the support provided byMC. Diana Dorantes and David Jefferson in drafting the English version of this manuscript. We thank the time andeffort that Sonia Guan and two anonymous reviewers devoted to improve an earlier version of our manuscript.

Author Contributions: Marco Antonio Almendarez-Hernández and Luis Felipe Beltrán Morales developedthe idea. Marco Antonio Almendarez-Hernández, Gerzaín Avilés Polanco, Víctor Hernández Trejo andAlfredo Ortega-Rubio analyzed the data; Marco Antonio Almendarez-Hernández, Gerzaín Avilés Polanco,Víctor Hernández Trejo, Alfredo Ortega-Rubio and Luis Felipe Beltrán Morales wrote the paper.

Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations

The following abbreviations are used in this manuscript:

AP Average pricem3 Cubic metersMP Marginal pricep Precipitationt Temperaturew Water consumption average per capita residential use

References

1. Gleick, P.H. Water in the 21st century. In Water in Crisis; Gleick, P.H., Ed.; Oxford University Press: Oxford,UK, 1993; pp. 105–113.

2. The Food and Agriculture Organization of the United Nations (FAO). El Riego en América Latina y el Caribe enCifras; FAO: Roma, Italy, 2000; p. 348.

3. The Secretariat of Environment and Natural Resources (SEMARNAT). El Medio Ambiente en México 2005;SEMARNAT: Mexico City, Mexico, 2005.

4. Foro Consultivo Científico y Tecnológico (FCCyT). Diagnóstico del Agua en las Américas; FCCyT: Mexico City,Mexico, 2012.

5. Comisión Nacional del Agua (CNA). Estadísticas del Agua en México; CNA: Metepec, Mexico, 2008.6. Barkin, D.; Klooster, D. Estrategias de la Gestión del Agua Urbana. En la Gestión del Agua Urbana en México:

Retos, Debates y Bienestar; Barkin, D., Ed.; Universidad de Guadalajara: Guadalajara, Mexico, 2006.7. Beltrán-Morales, L.F.; Borgues-Conterras, J.; Lagunas-Vazquéz, M.; Beltrán-Morales, J.; García-Rodríguez, F.

Volúmenes de consume de agua por localidad en la Reserva de La Biosfera del Vizcaíno, B.C.S. México.In Valoración Hidrosocial en la Reserva de la Biosfera del Vizcaíno; Beltrán-Morales, L.F., Chávez López, S.,Ortega-Rubio, A., Eds.; CIBNOR: La Paz, Mexico, 2010; pp. 113–128.

8. Almendarez-Hernández, M.A.; Avilés-Polanco, G.; Beltrán-Morales, L.F. Demanda de agua de uso comercialen la Reserva de la Biosfera El Vizcaíno, México: Una estimación con datos de panel. Nova Sci. 2015, 7,553–576. [CrossRef]

9. Rogers, P.; de Silva, R.; Bhatia, R. Water is an economic good: How to use prices to promote equity, efficiency,and sustainability. Water Policy 2002, 4, 1–17. [CrossRef]

10. PRI Project. Economic Instruments for Water Demand Management in an Integrated Water Resources ManagementFramework; Policy Research Initiative, The Walter and Duncan Gordon Foundation, Agriculture andAgri-Food Canada, Environment Canada the Canadian Water Network: Edmonton, AB, Canada, 2004.

11. Arbués, F.; Villanúa, I. Potential for pricing policies in water resource management: Estimation of urbanresidential water demand in Zaragoza, Spain. Urban Stud. 2006, 43, 2421–2442. [CrossRef]

12. Arbués, F.; García-Valiñas, M.; Martínez-Espiñeira, R. Estimation of residential water demand: A state of theart review. J. Socio-Econ. 2003, 32, 81–102. [CrossRef]

13. Worthington, A.C.; Hoffman, M. An empirical survey of residential water demand modelling. J. Econ. Surv.2008, 22, 842–871. [CrossRef]

14. Billings, R.; Agthe, D. Price elasticities for water: A case of increasing block rates. Land Econ. 1980, 56, 73–84.[CrossRef]

15. Bachrach, M.; Vaughan, W. Household Water Demand Estimation; Working Paper ENP 106; Inter-AmericanDevelopment Bank: Washington, DC, USA, 1994.

Water 2016, 8, 428 13 of 14

16. Charney, A.H.; Woodard, G.C. A test of consumer demand response to water prices: Comment. In LandEconomics; University of Wisconsin Press: Madison, WI, USA, 1984; Volume 60, pp. 414–416.

17. Opaluch, J. A test of consumer demand response to water prices: Reply. In Land Economics; University ofWisconsin Press: Madison, WI, USA, 1984; Volume 60, pp. 417–421.

18. Stavins, R.N. Managing Water Demand: Price vs. Non-Price Conservation Programs; A Pioneer Institute WhitePaper; Pioneer Institute: Boston, MA, USA, 2007.

19. Tracy Mehan, G., III; Kline, I. Pricing as a demand-side management tool: Implications for water policy andgovernance, Unites States: American Water Works Association. J. Am. Water Works Assoc. 2012, 104, 61–66.

20. Boland, J.; Whittington, D. The Political Economy of Water Tariff Design in Developing Countries: IncreasingBlock Tariffs versus Uniform Price with Rebate. In The Political Economy of Water Pricing Reforms; Dinar, A., Ed.;Oxford University Press: New York, NY, USA, 2000; pp. 215–235.

21. Shin, J.-S. Perception of Price when Price Information is Costly: Evidence from Residential ElectricityDemand. Rev. Econ. Stat. 1985, 67, 591–598. [CrossRef]

22. The Secretariat of Environment and Natural Resources (SEMARNAT); Comisión Nacional del Agua (CNA).Estadísticas del Agua en México; SEMARNAT and CNA: Mexico City, Mexico, 2004.

23. Nauges, C.; Thomas, A. Long-run study of residential water consumption. Environ. Resour. Econ. 2003, 26,25–43. [CrossRef]

24. García, M. Fijación de precios para el servicio municipal de suministro de agua: Un ejercicio de análisis debienestar. Hacienda Pública Española/Revista Economía Pública 2005, 172, 119–142.

25. Chang, H.; House-Peters, L. Urban water demand modeling: Review of concepts, methods, and organizingprinciples. Water Resour. Res. 2011, 47, 1–15.

26. Arbués, F.; Barberán, R. Price impact on urban residential water demand: A dynamic panel data approach.Water Resour. Res. 2004, 40, 1–9. [CrossRef]

27. Ito, K. Do Consumers Respond to Marginal or Average Price? Evidence from Nonlinear Electricity Price.Am. Econ. Rev. 2014, 104, 537–563. [CrossRef]

28. Wichman, C. Perceived price in residential water demand: Evidence from a natural experiment. J. Econ.Behav. Organ. 2014, 107, 308–323. [CrossRef]

29. Baltagi, B. Econometric Analysis of Panel Data, 3rd ed.; John Wiley and Sons: New York, NY, USA, 2005.30. Kiviet, J. On Bias, Inconsistency, and Efficiency of Various Estimators in Dynamic Panel Data Models.

J. Econom. 1995, 68, 53–78. [CrossRef]31. Judson, R.; Owen, A. Estimating Dynamic Panel Data Models: A Practical Guide for Macroeconomists.

Econ. Lett. 1999, 65, 9–15. [CrossRef]32. Anderson, T.; Cheng, H. Estimation of Dynamic Models with Error Components. J. Am. Stat. Assoc. 1981, 76,

598–606. [CrossRef]33. Arellano, M.; Bond, S. Some tests of Specification for Panel Data: Monte Carlo Evidence and Application to

Employment Equations. Rev. Econ. Stud. 1991, 58, 277–297. [CrossRef]34. Bruno, G. Approximating the bias of the LSDV estimator for dynamic unbalanced panel data models.

Econ. Lett. 2005, 87, 361–366. [CrossRef]35. Cermeño, R. Median-Unbiased Estimation in Fixed-Effects Dynamic Panels. Ann. D’Econ. Stat. 1999, 55,

351–368. [CrossRef]36. Breitung, J. The Local Power of Some Unit Root Tests for Panel Data. In Advances in Econometrics;

Baltagi, B., Ed.; Elsevier: New York, NY, USA, 2000; Volume 15.37. Levin, A.; Lin, C.-F.; Chu, C.-S.J. Unit root tests in panel data: Asymptotic and finite-sample properties.

J. Econom. 2002, 108, 1–24. [CrossRef]38. Harris, R.D.F.; Tzavalis, E. Inference for unit roots in dynamic panels where the time dimension is fixed.

J. Econom. 1999, 91, 201–226. [CrossRef]39. Im, K.; Pesaran, M.; Shin, Y. Testing for unit roots in heterogeneous panels. J. Econom. 2003, 115, 53–74.

[CrossRef]40. Baltagi, B.; Chihwa, K. Nonstationary panels, cointegration in panels and dynamic panels: A survey.

In Advances in Econometrics; Baltagi, B., Ed.; Elsevier: New York, NY, USA, 2001; Volume 15.41. Schleich, J.; Hillenbrand, T. Determinants of residential water demand in Germany. Ecol. Econ. 2008, 68,

1756–1769. [CrossRef]

Water 2016, 8, 428 14 of 14

42. Angrist, J.; Pischke, J.-S. Mostly Harmless Econometrics: An Empiricist’s Companion; Princeton University Press:Princeton, NJ, USA, 2009.

43. Shin, J.-S. Aggregation and the Endogeneity Problem. Int. Econ. J. 1987, 1, 57–65. [CrossRef]44. Hausman, J. Specification tests in econometrics. Econometrica 1978, 46, 1251–1271. [CrossRef]45. Almendarez-Hernández, M.A.; Jaramaillo-Mosqueira, L.; Áviles-Polanco, G.; Beltrán-Morales, L.F.;

Hernández-Trejo, V.; Ortega-Rubio, A. Economic Valuation of Water in a Natural Protected Area ofan Emerging Economy: Recommendations for El Vizcaino Biosphere Reserve, Mexico. Interciencia 2013, 38,245–252.

46. Musolesi, A.; Nosvelli, M. Dynamics of residential water consumption in a panel of Italian municipalities.Appl. Econ. Lett. 2007, 14, 441–444. [CrossRef]

47. Fullerton, T.M., Jr.; Katherine, C.W.; Wm, D.S.; Adam, G.W. An empirical analysis of Halifax municipal waterconsumption. Canadian Water Resources Association. Can. Water Resour. J. 2013, 38, 148–158. [CrossRef]

48. Younes, B.Z. A long-run analysis of residential water consumption. Econ. Bull. 2013, 33, 536–544.49. Nieswiadomy, M.; Molina, D. A note on Price Perception in Water Demand Models. Land Econ. 1991, 67,

352–359. [CrossRef]50. Kavezeri-Karuaihe, S.; Wandschneider, P.R.; Yoder, J.K. Perceived Water Prices and Estimated Water Demand in

the Residential Sector of Windhoek, Namibia; Food and Agriculture Organization of the United Nations: Rome,Italy, 2005.

51. Binet, M.-E.; Carlevaro, F.; Paul, M. Estimation of residential water demand with imprecise price perception.Environ. Resour. Econ. 2014, 59, 561–581. [CrossRef]

52. Wichman, C.; Taylor, L.; von Haefen, R. Conservation policies: Who responds to price and who responds toprescription? J. Environ. Econ. Manag. 2016, 79, 114–134. [CrossRef]

53. Jaramillo-Mosqueira, L. Evaluación econométrica de la demanda de agua de uso residencial en México.Trimestre Económico 2005, 72, 367–390.

54. García-Salazar, J.A.; Mora-Flores, J.S. Tarifas y consumo de agua en el sector residencial de la ComarcaLagunera. Región Sociedad 2008, 20, 119–132.

55. Adams, A.S.; Pablos, N.P. Factores que afectan la demanda de agua para uso doméstico en México.Región Sociedad 2010, 22, 3–16.

56. Avilés-Polanco, G.; Almendarez-Hernández, M.; Hernández-Trejo, V.; Beltrán-Morales, L. Elasticidad-preciode corto y largo plazos de la demanda de agua residencial de una zona árida. Caso de estudio: La Paz,B.C.S., México. Tecnología Ciencias Agua 2015, 4, 85–99.

57. Avilés-Polanco, G.; Huato, L.; Troy-Diéguez, E.; Murillo, B.; García, J.; Beltrán-Morales, L. Valoracióneconómica del servicio hidrológico del acuífero de La Paz, B.C.S.: Una valoración contingente del uso deagua municipal. Frontera Norte 2010, 22, 103–128.

58. Sánchez-Brito, I.; Almendarez-Hernández, M.A.; Morales-Zárate, M.V.; Salinas-Zavala, C.A. Valor deExistencia del Servicio Ecosistémico Hidrológico en la Reserva de la Biosfera Sierra La Laguna, Baja CaliforniaSur México. Frontera Norte 2013, 25, 97–129.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open accessarticle distributed under the terms and conditions of the Creative Commons Attribution(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).

O

Ad

JLa

b

a

ARAHA

KFPAPM

I

ti1optp

lm

j

P

h1

Mammalian Biology 81 (2016) 543–550

Contents lists available at ScienceDirect

Mammalian Biology

jou rn al hom epage: www.elsev ier .com/ locate /mambio

riginal investigation

pprehension affecting foraging patterns and landscape use of muleeer in arid environments

uan Pablo Esparza-Carlos a,1, John W. Laundré a,2, Lucina Hernández a,3,uis Ignacio Íniguez-Dávalos b,∗

Centro Regional, Instituto de Ecología, AC, Boulevard del Guadiana 123, Los Remedios, Durango CP 34100, MexicoCentro Universitario de la Costa Sur, Universidad de Guadalajara, Av. Independencia Nacional 151, Autlán, Jalisco CP48900, Mexico

r t i c l e i n f o

rticle history:eceived 7 October 2015ccepted 21 July 2016andled by Emmanuel Serranovailable online 22 July 2016

eywords:ear ecologyredatorspprehensivenessuma concolorexican Chihuahuan Desert

a b s t r a c t

Foraging animals make trade-offs between food and predation risk. To avoid predation, animals will seekout safe habitats for foraging, often at the sacrifice of food resources. Habitat structure and specific preda-tors prey face determine the risk level of specific habitat types. In northern forested areas, forest edgeshave been shown to be high risk habitat for mule deer (Odocoileus hemionius) foraging under predationrisk from pumas (Puma concolor). In more open habitats, e.g. deserts, a priori knowledge of what charac-teristics might contribute to predation risk levels for this species complex is lacking. Our objective wasto evaluate the potential predation risk faced by mule deer from pumas under different habitat charac-teristics in the open habitat of the Chihuahuan Desert. We measured apprehension levels in mule deerwith Giving Up Densities (GUDs) to evaluate predation risk levels of different habitat variables. We usedsupplemental (experimental) GUDs and natural GUDs: browse of candelilla (Euphorbia antisiphylitica).Supplemental GUDs were best modeled with macro-habitat variables: they were negatively associated

variables, with the interaction between distance to hills and bush height being the strongest predic-tor, indicating possible predation risk by stalking pumas. Also natural GUDs (candelillas) indicated thatmule deer perceived the proximity to hills as riskier. Identification of habitat characteristics that maycontribute to predation risk in the Chihuahuan desert will enable further investigation into how habitatcharacteristics in the Chihuahuan Desert contribute to predator-prey dynamics of mule deer and pumas.

© 2016 Deutsche Gesellschaft fur Saugetierkunde. Published by Elsevier GmbH. All rights reserved.

ntroduction

It is now well established that while foraging, prey species needo incorporate both the levels of food resources and predation riskn their decisions of habitat use (Brown et al., 1999; Lima and Dill,990). There is also growing evidence that prey live in a landscapef fear (Kuijper et al., 2013; Laundré et al., 2001) where the level of

redation risk varies over the landscape dependent on the predatorhey face and how the habitat influences the efficiency of predation,redator lethality (Laundré, 2010). Preys become more apprehen-

∗ Corresponding author.E-mail addresses: juan.esparza.c@gmail.com (J.P. Esparza-Carlos),

aunjohn@hotmail.com (J.W. Laundré), liniguez@cucsur.udg.mx,urcimx@hotmail.com, murcimx@gmail.com (L.I. Íniguez-Dávalos).1 Present address: Centro Universitario de la Costa Sur, Universidad de Guadala-

ara, Av. Independencia Nacional 151, Autlán de Navarro, Jalisco, 48900, Mexico.2 Present address: University of California Natural Reserve System-James Reserve,

O Box 3355, Idyllwild, CA, 92549, USA.3 L. Hernández passed away in October of 2013 after a battle with cancer.

ttp://dx.doi.org/10.1016/j.mambio.2016.07.006616-5047/© 2016 Deutsche Gesellschaft fur Saugetierkunde. Published by Elsevier Gmb

sive, fearful, when faced with increased predation risk (Altendorfet al., 2001; Hernández et al., 2005). It is well documented that preyrespond to this risk by either changes in time allocation, avoidingrisky areas, or in behavior, increased vigilance (Brown et al., 1999;Kotler et al., 1994; Laundré et al., 2001) and these responses can beused to identify risky habitats (Iribarren and Kotler, 2012; Shraderet al., 2008). Although it will vary with the particular predator,habitat characteristic and how they contribute to the lethality ofpredation are the main mediators of risk levels (Laundré et al., 2001;Laundré, 2010). For example, wolves (Canis lupus) are more adaptat running down their prey and so open habitats have a higher pre-dation risk for their prey than closed forests (Laundré et al., 2001).Conversely, pumas (Puma concolor) primarily stalk their prey andso open areas are of low predation risk while forest edges appearto be high risk areas (Altendorf et al., 2001; Hernández et al., 2005).

Identifying what habitat characteristics contribute to predation

risk becomes an important first step in characterizing the makeupof the landscape of fear for a particular prey and its predator(s).Once the link between predation risk and apprehension levels iswell established, prey species can enable us to identify risk levels

H. All rights reserved.

5 malia

tpeoioae(te

rscaeka(Gd(

dahcoblpattiksoritntllttahNomfbopga

M

S

I

44 J.P. Esparza-Carlos et al. / Mam

hough their behavior, either in changes in their spatial foragingatterns or behavioral levels of vigilance (Kotler et al., 1994; Kuijpert al., 2013; Hernández and Laundré, 2005). These changes canften be easily observed visually or measured through standard-

zed GUD techniques (Brown, 1988). The latter method involvesffering boxes with fixed limited quantities of food mixed in withn inedible substrate. This offers a diminishing return on foragingfforts and eventually the animal will stop foraging. The amountdensity) of food left is the GUD. These GUDs have been showno be directly related to apprehension levels (Brown, 1988; Kotlert al., 1994).

With these techniques, investigators have been able to identifyelative levels of risk of different habitat characteristics over a land-cape (Iribarren and Kotler, 2012; Shrader et al., 2008). This is thease for mule deer (Odocoileus hemionius) in forested habitats for-ging under predation risk from pumas (Puma concolor) (Altendorft al., 2001; Hernández et al., 2005). A majority of mule deer areilled by pumas along forest edges (Laundré and Hernández, 2003)nd pumas were found to spend more time hunting in these areasHolmes and Laundré, 2006). Subsequently, through the use ofUDs and fecal group distributions, it was demonstrated that muleeer perceived these areas as being risky and avoided using themAltendorf et al., 2001; Hernández et al., 2005).

Although we have been able to identify risky habitat for muleeer faced with predation by pumas in forested area, mule deernd pumas coexist over large expanses of relatively open desertabitat. In these areas, such as the Chihuahuan Desert, habitatharacteristics change more subtly and clear distinctions, such aspen vs. edge in forested areas, can rarely be made. Then, it cane asked what habitat characteristics in the Chihuahuan Desert

andscape might contribute to risk levels that mule deer face underuma predation? Identifying these habitat characteristics becomes

first step in understanding the spatial and behavioral aspects ofhe predator-prey relationship between pumas and mule deer inhis vast ecosystem. Given that levels of apprehension (fear) exhib-ted by prey in feeding trials (supplemental and natural GUDs) arenown to be related to risk levels, we postulate that foraging inten-ity of mule deer is strongly influenced by habitat characteristics inrder to avoid predation and, to a lesser degree, for habitat variableselated to food resources. Specifically, we postulate that foraging isnfluenced by vegetation structure associated with stalking pumas:all bushes, less visibility, high bush density, stepper slopes, andearness to hills. Also, we postulate that foraging would be posi-ively related to food abundance. We evaluated the perceived riskevels of various habitat characteristics across a Chihuahuan Desertandscape through the use of a combination of mule deer pellets dis-ribution (spatial use) and GUDs (apprehension levels). Results ofhis study will help to determine if variables known to be sources ofpprehension influence foraging decisions of mule deer in the Chi-uahuan Desert, similar to behaviors observed in northern regions.atural GUD’s measurements were based on foraging intensityf candelilla (Euphorbia antisiphylitica), a known principle food inule deer diet in the area (Guth, 1987). Based on a priori knowledge

rom other studies of what might constitute risky habitat, e.g. possi-le stalking cover, we identified a variety of habitat characteristicsn the macro and micro habitat scales as possible contributors toredation risk. We then used linear multiple regression to investi-ate the relationship between GUD measurements (apprehension)nd each identified characteristic.

aterial and methods

tudy area

The work was conducted in the Laboratorio del Desierto (LD-NECOL), located at 26◦41′10′′N and 103◦44′42′′W, in Mapimí

n Biology 81 (2016) 543–550

Biosphere Reserve (MBR), in the State of Durango, Mexico. The MBRis a closed basin, characterized by extended flat zones, small hills,and isolated mountains. The climate is arid; annual average tem-perature is 21 ◦C, winters are cool and summers are warm, with15 ◦C and 26 ◦C mean temperature respectively. Annual averageprecipitation is 287 mm, with summer rains (data from weatherstation, LD-INECOL). In the first year of our study (Jun 2004–May2005), it rained 10 months for a total of 417 mm and mean tem-perature was 20 ◦C; in the second year (Jun 2005–May 2006), itrained 10 months for total of 205 mm and mean temperature was18 ◦C. Finally, during the third year (Jun 2006–May 2007) it rained9 months for total of 374 mm, with a mean temperature of 16 ◦C.

The study was conducted at altitudes between 1150 and1200 masl, in a relatively flat area called the Bajadas, next to hillsof igneous and sedimentary origins. This landscape element com-prises 37% of the region, and is where mule deer are primarily found(Sánchez-Rojas and Gallina, 2000). Specifically, the study was con-ducted in the upper Bajadas (closer to the foothills and the base ofthe San Ignacio Mountain) because it is where mule deer principallyforage. In the lower Bajadas there are patches of dense vegetation(mesquite trees-Prosopis glandulosa- and associated bushes) calledmogotes, which provide horizontal and vertical cover, and wheremule deer do rest and start foraging every day (pers. obs.). In upperBajadas, the few sites where occasionally mule deer occasionallydid rest (under isolated big bushes), were excluded from analysis.

The vegetation in the upper Bajadas is homogeneous in compo-sition, dominated by creosote (Larrea tridentata), ocotillo (Fouqueriasplendens), prickly pear (Opuntia rastrera), yucca (Yucca rigida), andagaves (Agave scabra and A. lechuguilla) (Montana and Bremier,1988). The diet of mule deer in the area is primarily bushes (43% ofannual diet), forbs (34%, principally in summer), succulents (13%),and grasses (10%). The candelilla bush is an important food itemin the Chihuahuan Desert, especially in dry years (Urness, 1981;Esparza-Carlos et al., 2011), contributing 14% of the annual diet(range 10–24% between seasons) (Guth, 1987). Wildlife in the areaincludes the black-tailed jack rabbit (Lepus californicus), one of thefew species defecating in pellets; however, the shape and size, aswell as the number of pellets, make them clearly distinct from thoseof the mule deer. The only animals in the area that could interferewith the habitat use or activity patterns of mule deer is the collaredpeccary (Dicotyles angulatus); however, the region is in the edgeof its potential distribution. Historically peccaries did not occur inthe study area, but there are about 6–8 individuals found there,remnants of a group that was released in 2000 (approximately 30individuals). Therefore, we consider that the effect of interactionsbetween species is minimal. Cattle could potentially have impactedthe movements and distribution of mule deer. However, they do notconsume candelilla and were located outside the study area duringthe supplemental GUD experiments.

In this work we consider predation risk to mule deer only bypumas. Besides pumas, the coyote (Canis latrans) is the only otherpotential predator in the area. However, diet and habitat studiesof coyotes in the area have shown that they are not a threat formule deer (Hernández and Delibes, 1994). During our experimentsperiod, we registered puma occurrence by tracks, scats, trail camerarecords, visual observations and five kill sites in the study area.Thus, it was assumed that predation risk by pumas was sufficient toinfluence foraging behavior of mule deer. Human disturbance in thearea is low, hunting is prohibited and human activities are centeredin cattle management on water catchments (3.5–6 km far).

Mule deer habitat use during foraging and candelilla abundance

estimation

We initially plotted the distribution of mule deer fecal groups ina 74 ha study site. We used the previously established relationship

malia

batanildogbteia

(ttaduPtuta2r(l2omcd(patlGacTlaiggemigApdWdlcd(dlW

J.P. Esparza-Carlos et al. / Mam

etween fecal group distribution and use levels by deer (Hernándeznd Laundré, 2005; Loft and Kie, 1988; Månsson et al., 2011). Wehen selected a sample set of use areas and measured the perceivedpprehension levels of mule deer with the use of supplemental andatural GUDs (Olsson et al., 1999). After the rainy season, usually

n October, mule deer start to concentrate in the Bajadas and pel-et groups begin to accumulate. Mule deer stay there during thery season until the next June, when the rainy season begins (pers.bs.) They then presumably move out to other areas to feed on newrowth vegetation. We conducted the study in the upper Bajadas,ecause this is where mule deer forage more intensively and haveo rely on naturally concentrated food during the dry season. How-ver, the lack of observed bed sites indicates that they rarely restn this area. This allowed us to test the relationship between foodnd predation risk relative to habitat use while deer were foraging.

Mule deer density varies between 0.7–4.2 per km2 seasonallySánchez-Rojas and Gallina, 2000). By night counts and camera-raps, we calculated that our study area was used by 35–40 deer inhree groups at least; each group has an average of eight individu-ls, although sometimes we recorded groups up to 25 individuals,ue to temporal associations. The habitat use of mule deer was eval-ated using a deer fecal (pellet) group census for the whole area.ellet counts are a widely use technique used with deer, moose, cat-le and other ungulates, and are a good indicator of relative habitatse for deer, which does not seem to differ from estimates of habi-at use based on telemetry and GPS (Hernández et al., 1999; Loftnd Kie, 1988; Månsson et al., 2011; Sánchez-Rojas and Gallina,000). In arid lands the principal cause of pellets loss is heavyain, so the best time to use this technique is November to MarchWallmo et al., 1962). For the two sampled years, we let the pel-ets accumulate from October 2004 to March 2005 and October005 to April 2006. We could not identify any natural separationf the area into patches of vegetation because of the high unifor-ity in structure and composition. Thus, foraging areas did not

hange markedly across the landscape. Therefore, for analyses, weivided the area, an irregular polygon (180 ha), into a lattice of 1 ha100 × 100 m) squares, hereafter referred to as “quadrats” (=sam-ling unit). This size was chosen as representing a small enoughrea that mule deer could move across easily but yet large enougho detect use differences among quadrats. The locations of all pel-et groups encountered were entered as waypoints with a handheldPS (Garmin®12XL; Esparza-Carlos et al., 2011). Starting in Marchnd April, respectively, we did a complete census of pellets andandelilla plants in 74 randomly selected quadrats (out of 180 ha).o map all pellet groups and all candelilla plants, we first estab-

ished parallel lines every 20 m on 1:75,000 orthophotos. The startnd endpoints of these lines in UTM coordinates were enterednto the GPS and the map page and track functions were used touide field personnel along the lines. The locations of all pelletroups encountered within 10 m on either side of the line werentered as waypoints into the GPS. To avoid double counts, wearked a cross on the ground at each group found. Vegetation

n the area was sparse enough to facilitate location of all pelletroups and candelilla plants, which grew in varying size clumps.s the precipitation during the rainy season either washes theellets away or bleaches them to a white color, we were confi-ent we did not double count pellet groups across sample periods.ith ArcView 3.2 software, we superimposed the locations of the

eer pellet groups and candelilla plants on the quadrats and calcu-ated the density of pellet groups and candelilla plants/quadrat. Weounted the number of candelillas only once because this speciesoes not have a significant new recruitment over a two year period

Villa-Castorena et al., 2010). To test if quadrats were used by deerifferently between years, we classified quadrats in: high, medium,

ow classes, based on tertiles, according to density of pellet counts.e compared classes with a one way ANOVA of repeated measures

n Biology 81 (2016) 543–550 545

using standardized Z scores: the number of pellets per class did notdiffer between years.

GUDs evaluation

We used the GUDs technique to evaluate apprehension levelsand thus the level of predation risk. In a depletable patch, as a for-ager eats, the harvest rate decreases, until the benefit of harvestingdoes not compensate the foraging cost anymore. Foragers shouldleave a patch when the patch harvest rate is no longer greater thanthe sum of the energetic cost, predation risk and missed opportu-nity costs of foraging. If energetic cost and missed opportunity costare assumed to be constant, the value of predation risk can be esti-mated using harvest rates (Bedoya-Perez et al., 2013; Brown, 1988).Additionally, some natural foods can be treated as food patches,e.g. as more time is spent browsing off one bush, it becomes lessrewarding to stay and continue eating. At some point, the foragershould leave the bush for another and edible branches remainingcan be considered a natural GUDs estimate. Natural GUDs havebeen used successfully for branches consumed by rabbits and preylocated in branches by woodpeckers (Hodson et al., 2010; Olssonet al., 1999). Differences in the extent to which mule deer use can-delilla bushes of similar shape and size in different places of thehabitat in our study area may be attributed to risk levels.

We tested our predictions by using standard supplemental feed-ing trials to relate GUDs with different habitat variables. The feedingbox design followed Altendorf et al. (2001). Feeding boxes wereplastic rectangular trays (40 × 24 × 12 cm) in which were mixed:250 g of dried corn, 250 g of compressed alfalfa pellets, and 1000 g ofcorncobs cut into pieces as the inert nonedible substrate. The driedcorn and alfalfa are depletable food and the corncobs increase theforaging effort over the foraging bout. The trays were covered withfencing iron wire-mesh, with six spaces of 11 × 11 cm, to preventthe mule deer from pushing the cobs out of the trays. Plastic trayswere placed at 70 cm height on iron structures, which successfullyexcluded peccaries and small mammals. To minimize rodent con-sumption we used eight ShermanTM live traps at each site, trappingbefore and during the experiments. We did not expose food from10:00–17:00, which minimized the use by birds. We verified thatall feeders were visited previous to the experiment with digital trailcameras (Leaf River model DC-2BU).

Data collection for every feeder was for 12 consecutive days. Wereplenished boxes every day at 17:00 and collected the remain-ing food at 10:00 the next day. According to our observations andphotographs from the cameras, this was the time period is whenmule deer were most active. The habituation process to get muledeer to feed at food boxes took three years: in 2005 (five months)mule deer did not visit feeders; 2006 (four months) they started tofeed before the rainy season began. Finally, in 2007 (one month),at the end of the dry season deer were accustomed to eat fromboxes; the criteria to consider a deer group as habituated were (1)the animals visit all the feeders in the area, and (2) they visit regu-larly the feeders (almost daily). Due to different feeding habituationtimes between mule deer groups, the area was divided in two adja-cent zones. The supplemental GUDs evaluation was just before therainy season started: zone one was sampled from May 13–25, 2007and contiguous zone two, from June 16–27; for each zone we used14 feeders. However, mixed multiple regression model exploratoryanalyses showed that it was not necessary to nest the data by zone:a likelihood ratio test indicated the model with quadrat nested inzone was not better than the model using only quadrat as a ran-

dom effect (L = 0.000003, P = 0.999). We placed one feeder box inthe center of each sampled quadrat. The nearest distances betweenfeeders were 100–225 m; distances between first and last feedersby zone were 1.43 km and 1.10 km, respectively.

5 malia

ttdntrcgosaqbruio3tat

E

smnsobttpiccTE

omrqtet1drsrwtaq

F

hfibtm

46 J.P. Esparza-Carlos et al. / Mam

To evaluate if predation risk affected browsing behavior pat-erns, we used the candelilla bush as a model. Candelilla is suitableo estimate natural GUDs, because: (a) it has tillers consisting ofense leafless vertical cylindrical stems, which have similar thick-ess between plants, (b) mule deer only eat the annual growthip of individual stems, (c) deer experience diminishing rates ofeturns because the longer they exploit the bush, the more diffi-ult it is to find additional uneaten stems, (d) new shoots will notrow until the next two rainy seasons (pers. obs.), (e) the numberf tips of the stems browsed by deer could easily be recorded. Totandardize the method, we selected plants with a circular shapend as similar in size as possible (approx. 0.5 m radius). From the 74uadrats, we randomly selected 50 quadrats, where we recordedrowsing levels on 15 randomly selected candelilla plants, whichepresented the browse intensity of the same years when mule deerse was measured (2004–2006). Generally after two years, the eat-

ng browse-scars disappear (pers. obs.). To estimate browsing levelsn each plant we choose two random orientations between 0 and60◦ across the outer part of crown—the more accessible part ofhe plant- and examined 15 stems (30 by plant), recording thems either browsed or not. Browsing intensity was the proportion ofhe sampled stems that had been eaten.

xplanatory variables

We investigated which variables might contribute to apprehen-ion levels by measuring several variables at two spatial scales:acro-habitat and micro-habitat.We related supplemental and

atural GUDs with micro-habitat variables focused on vegetationtructure and quality as forage at two scales: (1) feeding box area,ne point using feeder as center, and (2) quadrat, six random pointsy quadrat. At each sample point we used the point quarter methodo measure: (a) bush density, (b) bush crown size (>50 cm diame-er), (c) height of bushes (>50 cm height), (d) slope at each sampleoint with a handheld Suunto® Clinometer, (e) estimated visibil-

ty to 20 m, and (f) a food quality index: for each edible bush, wealculated their crown area and multiplied by the percent theyomprised of the annual diet of mule deer for the area (Guth, 1987).he details on how these variables were obtained are explained insparza-Carlos et al. (2011).

At the macro-habitat scale, we explored the possible relationf supplemental and natural GUDs with the following measure-ents: distance to (u) permanent water source, (v) nearest dirt

oad, (x) the mountain; (y) density of candelilla, and (z) vegetationuantity (m2) by classes: dense vegetation, scarce-medium vege-ation and bare soil. We obtained the distances of the first threelements to the quadrat centroid. To determine vegetation quan-ity we utilized the Soil Adjusted Vegetation Index (SAVI; Huete,988) applied in a Quick Bird Satellite image (September 20, 2005)one in Idrisi 17.1 (“Selva Edition”; Eastman, 2012); the high spatialesolution, 2.5 m, of the multispectral imagery helped us to detectmall bushes characteristics of the Chihuahuan desert, which areelatively stable in size in three year periods (pers.obs.). The imageas classified in three greenness classes by using supervised con-

rol points: (1) bare soil, (2) scarce and intermediate vegetation,nd (3) dense vegetation. We calculated the area of each class peruadrat using ArcView.

oraging analysis

We tested if supplemental and natural GUDs were related toabitat variables at the micro- and macro-habitat scales: (1) We

rst preselected the set of variables which could be related withoth GUDs, using regression trees, because they are a good guideo which variables will be included in the model (Crawley, 2007). To

ake it more robust, we used the random forests approach (James

n Biology 81 (2016) 543–550

et al., 2013). We include all variables to generate 5000 regressionstrees, which used randomly 80% of data by chance. We used a scriptwritten by Dr. Roger Guevara (Instituto de Ecología, A.C., unpub-lished script). (2) We include in the global model the variablessupported in random forests (Table 1). Furthermore, we includedthe density of candelilla plants (C) to test if missed opportunitycosts influenced supplemental and natural GUDs. We included thedensity of pellet counts (U) to determinate if use of area by muledeer influences both GUDs. Finally we included variables and theirinteractions logically related to food, deer use and predation risk.

To determine which factors contributed most to explainingGUDs, we used multiple regression linear models. The global mod-els included: (a) for supplemental GUDs, the response variable asthe average of twelve days for the supplemental GUDs. Data werelog-transformed to meet normality and variance homogeneity. Theexplanatories variables were: Distance to hill (H), height of bushes(B), density of candelilla plants (C), density of pellet counts (U, muledeer use of area) and the interactions: H:B, H:C, H:U, C:U, whichwe thought had more biological meaning. We limited our analysisto those to avoid over-parameterization; Crawley (2007) suggestsn/3 parameters (9 in our case). For (b) natural GUDs, we useda mixed multiple regression linear model. The response variablewas the proportion of tips browsed by candelilla plant. Data weretransformed with arcsine square root, to meet normality and vari-ance homogeneity. The global model only included macro-habitatexplanatory variables for 50 quadrats: H, C, U, distance to roads (R),and all second order interactions between these variables. Quadratvariables were treated as random effects. Previously we did thesame analysis for 36 quadrats, in which we had both macro- andmicro-habitat variables (we did not have micro-habitat variablesfor the other 14 quadrats), but the micro-habitat variables werenot retained in the best model. Therefore, we ran the models withmacro-habitat variables using all the quadrats data (n = 50). Finally,quadratic terms of independent variables did not have significanteffects in neither of the two analyses.

We started with the global model and dropped variables inbackward steps using the lowest AIC value (Aikaike InformationCriterion) for Linear Models, and Loglikelihood proportion forMixed Models; then we select the simplest model if there wereno differences between models. All analyses were done with R-language (R Core Team, 2015) and associated R-packages: “MASS”(Venables and Ripley, 2002), “nlme” (Pinheiro et al., 2015), “lattice”(Sarkar, 2008), and “piecewiseSEM” (Lefcheck, 2015).

Results

Over the 12 days of exposure, feeders were visited 65% of thetime, 7.8 days on average (SD = 2.12), the range varied from 4 to12 days (30–100%). Supplemental GUDs left in feeders were vari-able between days for each feeder (mean = 290 g, SD = 132, range39–500 g).

Supplemental GUDs relationship with habitat variables

The best model for supplemental GUDs included the macro-and micro-habitat variables (F4,23 = 38.1, P < 0.001, r2adj. = 85, mul-tiple linear regression). Most of the variation (58.9%) was explainedby distance to hills, which had a negative effect on supplementalGUDs: as distance to hills increases, GUDs diminishes in feeders(F1,23 = 98.7, P < 0.001). The interaction distance to hill:plant heightwas significant (F1,23 = 26.1, P < 0.001); this interaction was second

in importance regarding explained variance (14%; Table 2). Therewere higher supplemental GUDs in areas near hills (0.5–1.5 kmfrom hills approximately) that had tall bushes (>1 m). In general,both variables had negative effects on supplemental GUDs. There-

J.P. Esparza-Carlos et al. / Mammalian Biology 81 (2016) 543–550 547

Table 1Pre-selection of variables with random forest approach. Deviance explained: 90.2% in supplemental GUDs (g), 89.2% in Natural GUDs. Distance to hills (“H”, km), Mule deeruse (“U”, density of fecal groups by ha), Bush height (“B”, cm), Distance to road (“R”, km).

Node Split point Terminal node n Deviance GUDs (g) SE

Suplemental GUDsDistance to hills 1.95 22 116302 290 1.5Bush height 92 15 23584 328 10.8

H > 1.95 7 18968 207 16.4H < 1.95–HB < 92 8 8972 313 8H < 1.95–HB > 92 7 6884 349 9.3

Natural GUDs Browsed tips (%)Distance to hills 0.93 600 47.7 88 0.02Mule deer use 94 156 10.2 60 0.08

H < 0.93–U < 94 117 7.3 56 0.57H < 0.93–U > 94 38 2.1 72 1.34

Distance to road 0.199 444 20.7 98 0.02H > 0.932–R > 199 67 2.9 93 1.03

Distance to road 0.188 274 11.5 100 0.26H > 0.932–R = 0.199–0.188 169 5.9 100 0.37H > 0.932–R < 0.188 170 7.6 91 0.51

Table 2Overall deviance analysis of the best models for explain: (a) supplemental GUDs, r2adj.=0.846, P < 0.001; (b) natural GUDs, R2 = 0.401, P < 0.001; and their relation with habitatvariables. Distance to hills (“H”, km), Mule deer use (“U”, density of fecal groups by ha), Bush height (“B”, cm).

Variable (%) Coefficients Standard Error F P Df Variance explained (%)

(a) Supplemental GUDsIntercept 101.1 1.46 149.9 <0.001 1, 23Distance to hills 2.0 1.27 98.7 <0.001 1, 23 58.9Bush height 4.0 1.08 3.9 0.059 1, 23 1.8Mule deer use 1.0 1.46 15.9 <0.001 1, 23 12.33Distance to hillsaBush height −1.5 1.03 26.1 <0.001 1, 23 13.8(b) Natural GUDsIntercept 20.6 0.14 164.7 <0.001 1, 700Distance to hills 3.4 0.02 152.8 <0.001 1, 47 35.0

fnetFapeaht

N

qtsvvff(adfoatm

Mule deer use 0.000077 0.0000035

a Coefficients were re-transformed to grams for (a) and percentage for (b).

ore, in this condition the relation of GUDs with distance to hills isegative whereas the relation with bush height is positive. How-ver, for bush height, the effect was negative farther to hills, whereallest bushes are associated with less supplemental GUDs (Fig. 1a).inally, the density of pellet groups had a positive effect on GUDs:s the density of fecal groups increases, mule deer left more sup-lemental GUDs in feeders boxes (Fig. 1b); this effect is minor,xplaining only 12% of variance (F1,23 = 3.1, P < 0.001). Other vari-bles and their interactions, including density of candelilla plants,ad no effect in explaining supplemental GUDs’ response; thereforehey were not included in the best model.

atural GUDs relationship with habitat variables

We checked 22,500 stems, on 750 candelilla plants in 50uadrats, of which 13,222 were browsed. The average of browsedips by plant was 58.8% (SD = 24.39, range 0–100%). Proportions oftems foraged by plant in quadrats were related with macro-habitatariables and pellet groups, and not related with micro-habitatariables. Distance to hill had a positive effect in proportions oforaged stems (R2 = 0.40): as distance to hills increases, mule deeroraged more on candelilla plants and left lower unbrowsed tipsF1,47 = 152.8, P < 0.001; Fig. 2a). The intensity of mule deer use of anrea (mean of pellet groups) also had a positive effect on foraging; aseer use more of an area, they ate more on candelilla plants and left

ewer unbrowsed stems tips (F1,47 = 152.8, P < 0.001; Fig. 2b). None

f variables density of candelilla (plants/ha), distance to roads (m),s well as interactions between variables, had significant effects inhe number of browsed tips, so they were not retained in the best

odel.

22.3 <0.001 1, 47 5.1

Discussion

We identified which habitat characteristics might contribute torisk levels that mule deer face from puma predation in the Chihua-han Desert. At the macro-habitat scale the risky habitat is closer tothe hills, as is shown both by natural and natural GUDs. The mostinteresting result found with supplemental GUDs was that the levelof apprehension depends on a mix of macro habitat variables, tallbushes promote apprehension when close to hills (area frequentlyused by pumas), whereas there is decreased apprehension in areasfurther away from hills. Our results indicate that mule deer for-aged less in risky habitats in Chihuahuan Desert, as they do in otherecosystems (Altendorf et al., 2001; Hernández et al., 2005).

Supplemental GUDs relationship with habitat variables

Apprehension affected foraging strategy of mule deer in the Chi-huahuan Desert of northern Mexico, according to the results ofsupplemented GUDs. While mule deer avoided tall shrubs in theriskiest habitat close to hills, tall shrubs did not negatively influ-ence mule deer foraging in less riskier areas. We propose that ata macro-habitat scale, the nearness to the hills represents moreprobability of encounters with pumas. Similar behavior has beendocumented for domestic goats, with the hillsides as risky areas,because they provides potential ambush or stalking sites for cara-cal (Felis caracal), a feline predator (Shrader et al., 2008). Althoughour study area has little slope to it (0–9◦), we identified the areasclose to the hills as riskier, due to higher number of puma occur-

rences recorded there. We found many signs of pumas: tracks, scats,visual observations, in the hills or close to them during our exper-iments (2005–2007). Furthermore, two kill sites were located at0.5 and 1.5 km from hill base. Rough terrain increases pumas hunt-

548 J.P. Esparza-Carlos et al. / Mammalian Biology 81 (2016) 543–550

F :heighz < 0.00

impbdu

ipdp

ig. 1. Relation of supplemental GUDs with habitat variables: (a) Distance to hillsones show more GUDs. (b) Area used by mule deer (mean of fecal groups by ha) (P

ng success. For example, uneven terrain, creeks and big rocks inountains reduce visibility, therefore making stalking easier for

umas; for example, these features were found to be avoided byighorn sheep (Laundré et al., 2002). This could explain why muleeer females during the breeding season in our study area avoidneven slope terrains (Sánchez-Rojas, 2000).

At the micro-habitat scale, while mule deer avoided tall shrubs

n the riskiest habitat close to hills, probably the height of bushesrovides cover to ambush deer in the risky areas, while tall shrubsid not negatively influence mule deer foraging in less riskier areas,robably because the lower risk of puma occurrence. This pattern is

t of bushes interaction (P < 0.001). Dark zones show lower GUDs in feeders, light1).

similar in Europe, where ungulates browsed less inside core areasof wolf pack activity than in areas far from them. At the micro-habitat scale, if habitat characteristics favored wolf predation insideof the core areas, the foraging intensity decreased, but they have noeffect on foraging outside of the core areas (Kuijper et al., 2013). Thenegative effect of plant height on predation risk for mule deer wasfound in fragmented forests, where bushes (averaging 73 cm) were

higher in the forest edge than in less risky open areas (Altendorfet al., 2001). In our study area, bushes >90 cm should provide goodstalking cover. Felids need a minimal cover provided by bushes,grasses or rocks to successfully hunt. Pumas and lions (Panthera

J.P. Esparza-Carlos et al. / Mammalia

Fig. 2. Natural GUDs: browsing intensity in candelilla plants by quadrats. The plotsshow the general relationships between proportions of browsed stems tips of can-delilla plants and variables. Tendency line and dots were predicted with the bestmvb

ltli

tgpaapeipatTit

mts1rIob

odel, using average of proportions of browsed tips by quadrat, when the otherariables were set to their mean values: (a) distance to hills (“H”, km), (b) area usedy mule deer (“U”, mean of fecal groups by ha).

eo) need to approach their prey undetected to at least 20 m beforehe attack (Robinette et al., 1959; Van Orsdol, 1984). For Africanions (substantially larger than pumas), the minimal cover heights 40–80 cm (Van Orsdol, 1984).

We found an unexpected result in that there is a positive rela-ionship between intensity of use by deer (as measured with pelletroups) and GUDs, which increases slightly. Two explanations areossible: (a) mule deer use of area has no relation with GUDs,nd our result is an spurious outcome that could be attributedt two or three points (see Fig. 1b, down left corner), this is sup-orted because this variable is less influential in the model (9.5% ofxplained deviance), or (b) the use of the area by deer was differentn the year when we did GUDs evaluation, compared with the tworevious years when we measured the pellet groups density. Weveraged both years, because the level of use was similar even withwo different precipitation regimes (Esparza-Carlos et al., 2011).hus we assumed the areas in the experimentation year were used

n a similar way. We therefore consider the first explanation to behe more probable.

The fact that candelilla density was not retained in the bestodel indicates that GUDs were not affected by missed oppor-

unity costs (MOC). One prerequisite to avoid MOC effects is thateveral feeders should be available for the same animal (Brown,988). For white-tailed deer, O. virginianus, feeder distances in the

ange of few meters did not affect the GUDs (Rieucau et al., 2007).n our study the feeders were apart of each other in the rangef 100–200 m; we consider these distances can be easily traveledy mule deer. In domestic goats, higher amount of available food

n Biology 81 (2016) 543–550 549

seems to promote the leaving of lower GUDs (Shrader et al., 2008);this seems not be the case in mule deer.

Natural GUDs with candelilla browsing

We identified variables that can be sources of apprehensionand affect the area use by deer through evaluating natural GUDsfrom candelilla browsing. Again, mule deer perceived sites closerto hills as risky habitat, as in the supplemental GUDs experiment.The proportion of browsed tips in candelilla plants diminishes withdistance to hills. In the study of the response of ungulates to coreactivity areas of wolves, the proportions of branches browsed werelower in the core areas, where ungulates are more exposed to pre-dation (Kuijper et al., 2013). Snowshoe hare (Lepus americanus)browsed less on bushes, leaving more natural GUDs, and spent lesstime foraging in forest gaps, which was assumed to be riskier, thanin closed forest (Hodson et al., 2010). The browsing intensity ishigher where deer increased the area use; however, the effect ofthis variable is minimal (4.5% of the explained variance). In addi-tion, the effect was more evident in heavily used areas (with morethan 140 pellet groups), while in areas with less than 100 pelletgroups there was not a clear pattern (see Fig. 2).

One advantage of using natural food as natural GUDs is that wedo not distort the normal behavior of foragers; this has been identi-fied as a potential problem with supplemental GUDs (Brown, 1988).In natural vegetation (e.g. candelilla), once one tip is browsed, it willnot be available again until the next growing season; in supple-mental GUDs experiments, feeders are replenished rapidly (everyday), which does not occur naturally for most of systems. Moreover,natural GUDs can provide us additional information. For example,candelilla is a succulent plant, which means that additionally it isa water source and can be very important in arid lands; mule deerincrease the proportion of succulents in their diets in dry years(Urness, 1981). Probably the foraging intensity could increase intwo sequential dry years; this will require further investigation, asour results represented the combination of one wet and one dryyears.

Consequently, both techniques (experimental and naturalGUDs) can be used to detect effects of apprehension on foragingactivities, as in this work they support the idea that distance tohills reduces apprehension and increases mule deer foraging activ-ity. Candelilla foraging reflect the effect in a larger time scale (twoyears), obviously this is affected by the intensity of use of the area,but not by availability of candelilla. Whilst GUDs reflect forag-ing decisions in a diel basis, the interaction distance to hill:bushheight represents the baseline of predation risk in the habitat. Wehighlight the importance of using supplemental and natural GUDsexperiments at the same time; if both are consistent, we could usenatural foraging as proxy for predation risk evaluation in long termmonitoring processes.

We conclude that mule deer incorporate variables at differentscales associated with apprehension in their foraging decisions. Thesites near of hills at the macro-habitat scale, and taller bushes atmicro-habitat scale were perceived as risky habitats. Finally, nat-ural plants such as candelilla or other bushes, which representdiminishing returns, could be used as rapid measures of possibleapprehension sources, when calibrated with simultaneous GUDsexperiments.

Acknowledgements

We give thanks to Consejo Nacional de Ciencia y Tecnología(CONACyT) for scholarship number 179205 to JPEC. Instituto deEcología A.C. provides research funds for JWL and LH, and grantedaccess to Laboratorio del Desierto. We thank Earthwatch and

5 malia

iSMrRicetbw

R

A

B

B

B

CEE

G

H

H

H

H

H

H

H

I

J

K

K

M.L., Arreola-Ávila, J.G., 2010. Production of candelilla seedlings (Euphorbiaantisiphyllitica Zucc.) by cuttings. Revista Chapingo Ciencias Forestales y del

50 J.P. Esparza-Carlos et al. / Mam

ts volunteers. Thanks to Vinicio Sosa-Fernández and Gerardoánchez-Rojas for comments. Thanks for fieldwork: Aracely andanuela Terrazas, and Itzel Arias. Thanks to the Herrera and Ter-

azas families. This research is part of the CONACyT Thematicesearch Network on Natural Protected Areas (RENANP). This work

s dedicated to Dr. Lucina Hernández who, after a long battle withancer, passed away in October 2013. It is through her tirelessfforts and many hours both in and out of the field that helpedhe senior author and many other students realize their dreams ofecoming wildlife scientists. She was an excellent researcher and aarm and giving individual. She is missed by all that have met her.

eferences

ltendorf, K.B., Laundré, J.W., López, C., Brown, J.S., 2001. Assessing effects ofpredation risk on foraging behavior of mule deer. J. Mammal 82, 430–439.

edoya-Perez, M.A., Carthey, A.J., Mella, V.S., McArthur, C., Banks, P.B., 2013. Apractical guide to avoid giving up on giving-up densities. Behav. Ecol. Socbiol.67, 1541–1553.

rown, J.S., Laundre, J.W., Gurung, M., 1999. The ecology of fear: optimal foraging,game theory, and trophic interactions. J. Mammal 80, 385–399.

rown, J.S., 1988. Patch use as an indicator of habitat preference, predation risk,and competition. Behav. Ecol. Sociobiol. 22, 37–47.

rawley, M.J., 2007. The R Book. John Wiley & Sons West Sussex, England.astman, J.R., 2012. IDRISI Selva 17.1 Software. Clark University, Worcester.sparza-Carlos, J.P., Laundré, J.W., Sosa, V.J., 2011. Precipitation impacts on mule

deer habitat use in the Chihuahuan Desert of Mexico. J. Arid Environ. 75,1008–1015.

uth, M.C.G.A., 1987. Hábitos Alimenticios del Venado Bura (Odocoileus Hemionus,Rafinesque 1817) en la Reserva de la Biosfera de Mapimí, Dgo. BSc. Thesis.Universidad Nacional Autónoma de México, México.

ernández, L., Delibes, M., 1994. The Bolsón de Mapimí, Southern ChihuhuanDesert, Mexico. Z. Saugetierkd Seasonal Food Habits of Coyotes Canis Latrans,59, pp. 82–86.

ernández, L., Laundré, J.W., 2005. Foraging in the landscape of fear and itsimplications for habitat use and diet quality of elk Cervus elaphus and bisonBison bison. Wildl. Biol. 11, 215–220.

ernández, L., Barral, H., Halffter, G., Colon, S., 1999. A note on the behavior of feralcattle in the Chihuahuan Desert of Mexico. Appl. Anim. Behav. Sci. 63, 259–267.

ernández, L., Laundré, J.W., Gurung, M., 2005. From the field: use of camera trapsto measure predation risk in a puma-mule deer system. Wildl. Soc. Bull. 33,353–358.

odson, J., Fortin, D., Bélanger, L., 2010. Fine-scale disturbances shape space-usepatterns of a boreal forest herbivore. J. Mammal 91, 607–619.

olmes, B.R., Laundré, J.W., 2006. Use of open, edge and forest areas by pumasPuma concolor in winter: are pumas foraging optimally? Wildl. Biol. 12,201–209.

uete, A.R., 1988. A soil-adjusted vegetation index (SAVI). Remote Sens. Environ.25, 295–309.

ribarren, C., Kotler, B.P., 2012. Foraging patterns of habitat use reveal landscape offear of Nubian ibex Capra nubiana. Wildl. Biol. 18, 194–201.

ames, G., Witten, D., Hastie, T., Tibshirani, R., 2013. An Introduction to StatisticalLearning. Springer, New York.

otler, B.P., Gross, J.E., Mitchell, W.A., 1994. Applying patch use in Nubian ibex to

measure resource assessment ability, diet selection, indirect interactionsbetween food plants, and predatory risk. J. Wildl. Manage. 58, 300–308.

uijper, D.P.J., de Kleine, C., Churski, M., van Hooft, P., Bubnicki, J., Jedrzejewska, B.,2013. Landscape of fear in Europe: wolves affect spatial patterns of ungulatebrowsing in Białowieza Primeval Forest, Poland. Ecography 36, 1263–1275.

n Biology 81 (2016) 543–550

Laundré, J.W., Hernández, L., 2003. Winter hunting habitat of pumas Puma concolorin northwestern Utah and southern Idaho, USA. Wildl. Biol. 9, 123–129.

Laundré, J.W., Hernández, L., Altendorf, K.B., 2001. Wolves, elk, and bison:reestablishing the landscape of fear in Yellowstone National Park, USA. Can. J.Zool. 79, 1401–1409.

Laundré, J.W., Hernández, L., Arias-Del Razo, I., 2002. The landscape of fear and itsimplications to sheep reintroductions. Biennial Symposium Northern WildSheep and Goat Council 13, 103–108.

Laundré, J.W., 2010. Behavioral response races, predator-prey shell games, ecologyof fear, and patch use of pumas and their ungulate prey. Ecology 91,2995–3007.

Lefcheck, J.S., 2015. PiecewiseSEM Piecewise structural equation modeling in R forecology, evolution, and systematics. Methods Ecol. Evol. 7, 573–579.

Lima, S.L., Dill, L.M., 1990. Behavioral decisions made under the risk of predation: areview and prospectus. Can. J. Zool. 68, 619–640.

Loft, E., Kie, J., 1988. Comparison of pellet-group and radio triangulation methodsfor assessing deer habitat use. J. Wildl. Manage. 52, 524–527.

Månsson, J., Andrén, H., Sand, H., 2011. Can pellet counts be used to accuratelydescribe winter habitat selection by moose Alces alces? Eur J. Wildl. Res. 57,1017–1023.

Montana, C., Bremier, R.F., 1988. Major vegetation and environments units. In:Montana, C. (Ed.), Estudio Integrado de los Recursos Vegetación, Suelo y Aguaen la Reserva de la Biosfera de Mapimí. Instituto de Ecología A.C, México, pp.99–114.

Olsson, O., Wiktander, U., Holmgren, N.M., Nilsson, S.G., 1999. Gaining ecologicalinformation about bayesian foragers through their behaviour: II. A field testwith woodpeckers. Oikos 87, 264–276.

Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., R Core Team, 2015. nlme: Linear andNonlinear Mixed Effects Models R Package Version 3. 1–122, URL: http://CRAN.R-project.org/package=nlme(accessed 28.08.15).

R Core Team, 2015. R: A Language and Environment for Statistical Computing RFoundation for Statistical Computing. R Core Team, Vienna, Austria, URLhttps://www.R-project.org/.

Rieucau, G., Vickery, W.L., Doucet, G.J., Laquerre, B., 2007. An innovative use ofwhite-tailed deer (Odocoileus virginianus) foraging behaviour in impactstudies. Can. J. Zool. 85, 839–846.

Robinette, W., Gashwiler, J., Morris, O., 1959. Food habits of the cougar in Utah andNevada. J. Wildl. Manage. 23, 261–273.

Sánchez-Rojas, G., Gallina, S., 2000. Mule deer (Odocoileus hemionus) density in alandscape element of the Chihuahuan Desert, Mexico. J. Arid Environ. 44,357–368.

Sánchez-Rojas, G., 2000. Conservación y manejo del venado bura en la reserva de labiosfera de Mapimi. In: Doctoral Dissertation. Instituto de Ecología AC, México.

Sarkar, D., 2008. Lattice: Multivariate Data Visualization with R. Springer, NewYork.

Shrader, A.M., Brown, J.S., Kerley, G.I.H., Kotler, B.P., 2008. Do free-rangingdomestic goats show ‘landscapes of fear’? Patch use in response to habitatfeatures and predator cues. J. Arid Environ. 72, 1811–1819.

Urness, P., 1981. Desert and chaparral habitats. Part 1. Food habits and nutrition.In: Wallmo, O.C. (Ed.), Mule and Black-Tailed Deer of North America.University of Nebraska Press, Lincoln, pp. 347–365.

Van Orsdol, K., 1984. Foraging behaviour and hunting success of lions in QueenElizabeth National Park, Uganda. Afr. J. Ecol. 22, 79–99.

Venables, W.N., Ripley, B.D., 2002. Modern Applied Statistics with S. Springer, NewYork.

Villa-Castorena, M., Catalán-Valencia, E.A., Inzunza-Ibarra, M.A., González-López,

Ambiente 16, 37–34.Wallmo, O., Jackson, A., Hailey, T., Carlisle, R., 1962. Influence of rain on the count

of deer pellet groups. J. Wildl. Manage. 26, 50–55.

© by PSP Volume 25 – No. 8/2016, pages 3023-3029 Fresenius Environmental Bulletin

3023

ENERGY CONSUMPTION DRIVERS: DEMAND SIDE

MANAGEMENT VS DEVICE PENETRATION, WHICH IS

STRONGER? IMPLICATIONS IN CLIMATE CHANGE IN

ARID MEXICO

Elio Lagunes-Diaz1, Maria Eugenia Gonzalez-Avila 2, Sara Diaz-Castro1,

Federico Poujol3, Alfredo Ortega-Rubio1

1 Centro de Investigaciones Biológicas del Noroeste. Instituto Politécnico Nacional 195, C.P. 23096; La Paz, B.C.S. México.

2 El Colegio de la Frontera Norte. Calle Técnicos 277, C.P. 64700, Monterrey NL.

3 Universidad Autónoma de Baja California Sur. Carretera al Sur Km. 5.5, C.P. 23080, La Paz, B.C.S.

ABSTRACT

Baja California Sur (BCS) is an isolated and

arid state in Northwestern Mexico. Departing from

a global and national framework, the state's trends

and characteristics of electricity consumption,

generation and policies and factors modeling these

are depicted. The trends of parameters such as

energy intensity of GDP, GHG emissions by energy

unit produced, and share of renewables are first

analyzed. Then, a test for detecting a decoupling of

electricity consumption from weather is carried on,

using correlation in subsets of time series together

with statistical trend detection, for assessing an

overall state-level change, instead of punctual

savings or estimations, which is attributable to

building and appliance efficiency standards. The

work focuses on the residential and tourism sectors,

which demand above 70% of the electricity in the

State, intending to develop an approach to regional

scale change assessments, instead of estimations or

punctual saving calculation in warm regions, with a

comprehensive data input requirement for

developing countries. The findings show that GDP

has grown slightly more energy intensive, the share

of renewables in the mix has diminished, and, in

spite of punctual savings from demand side

management measures, due to an overall increase in

air conditioning and other devices penetration, there

is no significant decoupling in energy consumption

from temperature, a main driver of consumption in

BCS, where total summer demand is about 50%

greater than that of winter and the state which ranks

second in non industrial GHG emissions per capita

in Mexico. Implications in GHG emissions and

development of policies seeking to mitigate climate

change are discussed.

KEYWORDS: Greenhouse gas emissions; Climate change; Energy

efficiency; demand side management; arid regions.

INTRODUCTION

In some regions, such as the one depicted in

this work, conditions such as isolation, high

temperatures in summer, a large share of residential

demand, little availability for wind and none for

hydro power, as well as nightly peaks of

consumption which cannot be covered with solar

power make demand side management, i. e. energy

efficiency, the most feasible option to better the

sustainability of the energy system. Policies on

demand side management seek to decrease energy

intensity by means of technical and legislative

policies and encouraging the rational use of energy,

while, on the opposite side, there is a steadily

growing electrification of the everyday life [1], as

new appliances that use electricity are developed

and adopted by the people, which cause increases in

consumption.

As an approach to ponder the results of

demand side management policies enforced in Baja

California Sur (BCS) during the last two decades, a

time series analysis for detecting a decoupling of

energy use from temperature is carried out, at the

state scale, instead of estimations or punctual

reckoning of savings, which are more conventional

in the literature. The method has a comprehensive

data disaggregation requirement, for warm regions

where consumption data is offered on a monthly

basis.

The present work is intended to increase

understanding about the challenges and barriers of

using electricity in a sustainable fashion in arid

regions of developing countries, with high weather-

sensitive residential demand. The final purpose of

the work is to provide a different and medium scale

insight for addressing climate change and

sustainability issues in future planning.

© by PSP Volume 25 – No. 8/2016, pages 3023-3029 Fresenius Environmental Bulletin

3024

MATERIAL AND METHODS

Study area. BCS is the most arid and isolated

State of Mexico, its population in 2010 was

637,026 inhabitants. It is among the Mexican States

with the fastest economic and demographic growth,

around 7% in the last decade, due to the recent

development of the tourism industry and to an ever

increasing number of American and Canadian

citizens who look for a place to retire [2].

The State Energy Grid relies on fossil fuel

based facilities, which make up 98.2% of the total.

Besides the fossil-fuel-based, BCS has geothermal

and solar power plants; since there are no perennial

rivers in the State, hydropower is absent in BCS [3].

The resulting hourly cost of production of this

technology mix ranges from US$0.094 before dawn

in winter to US$0.371 per kWh in the afternoon

(15:00 to 17:00) and before midnight (21:00 to

23:00) [4]; power is sold for an annual average of

US$0.098 per kWh to residential customers and

US$0.12 per kWh average to all other sectors [5].

FIGURE 1

Power consumption by sectors of activity, 1988-

2014. Data source: [7].

Electric consumption in BCS. In BCS, large

Industries demand 10.3% of the total power

consumption, compared with a 27.5% at a national

level; meanwhile, in the state, residential demand is

33.14%, high above the 26% average for the

Country. Medium sized industries (MSI) are the top

consumers (37.8%)(Fig. 1); MSI Tariff is used

mainly by the tourism sector, as well as in other

economical activities such as welding shops,

bakeries, automotive shops, etc. This sector has a

20% weather sensitive consumption, a figure

obtained by subtracting the winter consumption as a

baseline demand, when air conditioning is not used,

an approach used in warm regions to determine air

conditioning loads [6]. Residential sector has a 35%

weather dependent consumption.

Due to isolation, the predominantly fossil

fueled energy mix and the warm weather, the State

ranks second place nationally in non-industrial

energy related GHG emissions per capita, with 0.7

Tons CO2Eq per capita yearly, only after Quintana

Roo, well ahead of states with large industries, such

as Coahuila, Nuevo León, Veracruz, San Luis

Potosí and Tamaulipas, which in consequence of its

industrial activity have larger GHG emissions (Fig.

2). Overall emissions from the eight greatest

facilities added 1.6 million tons of CO2EQ in 2010,

yielding 752 tons of CO2EQ per GWh produced,

from a former value of 740 in 2007 [8][9]. A

comparison of per capita energy consumption,

GHG emissions and GDP among BCS, Mexico and

the OECD is presented in Table 1.

FIGURE 2

CO2 Emissions from energy production per

capita, excluding industrial consumption. Size of

bubble represents the total CO2 emissions in

tons.

Energy intensity has grown in the recent

years: in the 2005-2009 period, energy consumption

from all conventional sources (gasoline, gas, diesel,

fuel oil, kerosene and electricity) grew at an

average yearly 1.07 rate [7], exceeding the pace of

GDP, which only grew at an average 1.03 rate in the

period. Elasticity from energy use is a desirable and

non-controversial sign of a sustainable economy

[10].

In BCS, the highest consumption loads, with

its consequent highest costs of production, occur in

summer, peaking around 15:00 hrs, when outdoor

temperature is highest, and around sleeping time at

midnight, when comfort is needed the most, as it

happens in warm regions in the world [3]. Energy

consumption is lowest in winter, since gas or

electrical heating are scarce, present in less than

0.6% homes [13].

© by PSP Volume 25 – No. 8/2016, pages 3023-3029 Fresenius Environmental Bulletin

3025

TABLE 1

Comparison of energy, emission and GDP indicators for BCS, Mexico and the OECD

GDP per capita USD

constant

prices, 2005 PPP

GHG Emissions, kg,

per capita Electricity

consumption, MWh

per capita

Emission

intensity, kg

GHG/1000 USD

2000 2010 2000 2010 2000 2010 2000 2010

BCS 17,717 6.33 (2005) 2.53 2.83 0.59 (2005)

Mexico 11,946 12,741 5.6 6.14 1.70 1.92 0.46 0.48

OECD 28,252 30,574 14.03 12.92 7.93 8.27 0.50 0.42

Sources: [11][2], [11], [12]; values for BCS were calculated from [2], [7], [9]

A net growth in consumption per household

has occurred in the State in the last two decades

(Fig. 3), although some pulses have driven energy

consumption to a temporarily decrease, such as a

Tariff re-structure in 2002 [14], a Tariff change for

Los Cabos to a cheaper Tariff in 2006 and the

World Financial Crisis in 2009, when the trend in

tourism GDP in BCS, which rose from US$389.629

million to US$752.93 million in the 2003-2007

period, dropped dramatically to US$560 million in

2010 [2]. The trend in consumption per household

is less steep than the per capita, due to the long

term household size reduction in the country.

FIGURE 3

Residential consumption average for NW region,

BCS and Nationwide. Source: [7].

The perceivable overall rise in energy

consumption per household is related to

technological penetration -as more households

acquire newly developed appliances, economic

growth, as well as changes in housing type [15]. In

BCS, electric appliances have spread widely in the

past decade. Table 2 summarizes the increase of

availability per household for the most widespread

and most energy consuming devices in the 2000-

2014 period.

TABLE 2 Device penetration in BCS

Device 2000(%) 2010 (%) 2014(%)

Refrigerators 82 89 91

Air conditioners 23 30 36

Laundry

machines 54 68 72

TV sets 89 98 99

Computers 11 39 42

Source: [13].

Other factor influencing the rising trend in

consumption is the increase in temperatures in the

main cities, which, for La Paz, the city with the

longest records, adds up to 2.5° C for the period

1923-2013. Considering the need for air

conditioning, government subsidizes up to 80% of

consumption during summer. Meteorological

phenomena, such as hurricanes and heavy rainfall

temporarily affect energy consumption due to

interruptions of the supply.

Energy efficiency measures in Mexico. First

Mexican Official Standards (NOM, initials in

Spanish) regarding power consumption appeared in

the mid nineties, these focused primarily on

industrial devices; the first standard directed to a

domestic device was published in 2002,

establishing efficiency limits for refrigerators; it

was followed by standards limiting energy

consumption for air conditioners (2006) laundry

machines (2010), domestic fluorescent lamps

(2010) and refrigerators (2012), complemented by

standards for passive efficiency measures for office

buildings (2001) and residential buildings

(2011)[16]. The results of most of these recent

controls are yet to be analyzed, hence, an approach

is presented in this work for doing so.

© by PSP Volume 25 – No. 8/2016, pages 3023-3029 Fresenius Environmental Bulletin

3026

The National Commission for the Efficient

Use of Energy (CONUEE, initials in Spanish)

operates programs for substituting inefficient

domestic appliances, giving credits for the

acquisition of refrigerators and air conditioners,

since 2009. The replacement of both cooling

devices by this initiative in the first year reached

has reached 3,773,000 replacements in the country

[17]. Along with the cooling device substitution, the

Fund for the Energy Transition and the Sustainable

Use of Energy (FOTEASE, initials in Spanish)

program substituted eight incandescent light bulbs

for fluorescent lamps in every household in the

country. Previous to this programs, ASI, a trust for

energy efficiency, replaced 10,600 devices from

2004 to 2008 in BCS, giving benefits to a

considerable fraction among 198,500 households,

whose power savings may reach 30% by switching

to new devices [18]

Time series analysis of the coupling of

energy consumption and temperature.

Considering all the aforementioned regulatory and

financial measures to improve energy efficiency on

the cooling devices -which are the highest in power

consumption, from standards for electronic devices

and building designs and financial support schemes,

a perceivable decoupling on the relation between

the consumption and the temperature should be

expected, if the effects of the undertaken measures

were exceeding the rising trends on electric

consumption, which are due mainly to device

penetration and population growth.

Energy savings from efficiency measures have

been regarded as difficult to assess [19], [20].

Testing for a decoupling has to overcome factors

that may mislead the findings, to find whether

occurrence of savings is truly related to

optimization, or whether it is masked by changes in

the demand shares, economic recession, migration

or milder summers, for instance. Among the scarce

literature about decoupling, [21] reported a

decoupling of energy use from temperature in

specific buildings constructed according to the

bioclimatic architecture in the US and UK.

In the present work, a test was made for

detecting decoupling of energy use from

temperature, as an approach to assess the

achievements of energy conservation policies by

using the Monthly State Electricity Index (IEME,

initials in Spanish) [2], as the electricity

consumption variable, together with data from

meteorological observatories [22], under the

assumption that if a decoupling occurred during the

evaluated period, it should be noticeable among

subsets of the time series.

The IEME synthesizes generation and demand

volume and cost information for all activity sectors,

presented as a Physical Volume Index, fixed at a

base year, so it takes into account the effects of

variability in electricity pricing. It is readily

available information, updated every other month,

and time series depart from 2003 on, so it is useful

to detect changes in trends in the past decade.

Monthly temperature data for the

meteorological stations in BCS were kindly

supplied by the observatories of the CONAGUA

[22], from 2003 to 2012. For each year in the

period, Pearson linear correlation was obtained for

annual, warm (April to October) and temperate

(November to March) seasons subsets of data. As

should be expected, correlation IEME-temperature

was highest for the warm (0.91) and lowest for the

temperate (0.61) seasons. Annual correlation was

high, averaging 0.88. Standard deviation was 0.06,

0.05 and 0.3, for the annual, warm and temperate

groups, respectively.

Finally, for statistically detecting any

monotonic trend in the temperature-consumption

correlation within the period 2003-2012 (Fig. 4),

the non-parametric test Mann-Kendall Z was

carried on using R Studio [23].

Z values were 0.36, 1.07 and -0.18 for annual,

warm and temperate subsets, respectively, meaning

that no trend in any of the subsets was significant,

not even at α=0.1. Warm season had the highest

positive value, but not enough to be considered a

trend from the test outcome.

FIGURE 4

Energy consumption (IEME)-mean temperature

correlations for BCS, annual and seasonal, 2004-

2013.

On Figure 5 a Holt Winters [23] forecast of

the current trends and conditions for the climate

sensitive consumption (med tension and residential)

is plotted; if these persist, by 2020 the peak of

consumption in summer and the winter baseline

would be 50,000 MWh and 60,000 MWh higher,

respectively.

© by PSP Volume 25 – No. 8/2016, pages 3023-3029 Fresenius Environmental Bulletin

3027

FIGURE 5

Climate sensitive consumption 2009-2015 and 5 year forecast with 95% confidence prediction intervals.

RESULTS

The present attempt to study the effect of

policies regarding air conditioning efficiency and

passive temperature control in BCS does not

suggest that they have been unworthy, but that they

yet do not seem to curb the same opposing forces

that policies on refrigerators did overcome on the

national scale, which related emissions decreased

from 9.9 Tg to 8.6 Tg CO2 in the 1996-2006 period,

outpacing even a 13% growth in population and an

increase from 9% to 16% in penetration, while

those of air conditioning increased 3.9 Tg to 5.5 Tg

in the same period, according to [24].

Large correlations, as those occurring in

summer, difficult the detection of growing trends,

although the absence of a positive trend is to be

remarked, since a number of studies around the

world have found increasing dependence, mostly

within temperate climates, where warmer summers

are occurring, such as the regions of Maryland in

the USA [25] or the Netherlands [26].

Demand side management policies have not

matched the speed of growth in economy and

population in the state. The first passive control

standard, the NOM for office buildings, was

published in 2001; ten years later, the NOM for

residential buildings appeared, in 2011. Air

conditioning efficiency standards are also relatively

young, without yet having reached a decade since

their appearance: the first was published in 2006,

for central air conditioning; the standard for

regulating room systems appeared in 2008, when

air conditioning had reached almost 30% of

penetration in BCS dwellings. The challenge of

mitigating GHG emissions from the electric sector

can only be overcome including more stringent

demand side management measures.

As can also be observed from the records,

winter demand has increased by 215%, exceeding

the rate of summer, which grew by 207%. This is

linked to an increase in tourist arrival during winter,

from a total of 593,000 in 2003 to 726,000 in 2011

in the March season, which attracts 24% more

tourism than the monthly average throughout the

year in the period of study.

CONCLUSION

This work has shown that parameters related

to sustainability of the electric system have had

little progress or receded at the state level; GDP

grew more energy intensive, renewable generation

decreased in the electricity generation mix, GHG

emissions per energy unit increased due to fuel mix

and device penetration and consumption remains

tightly coupled to temperature. Institutional, market

and behavioral barriers have to be broken in the

state to counter these trends and to achieve

sustainability.

Policies in BCS must be tailored differently to

those of the rest of Mexico, taking into account the

very different conditions the state has, seeking to

mitigate its GHG emissions and to optimize the

energy use. This should be solved through the

creation of a state energy commission, an approach

that has been already taken in the states of Baja

California, Sinaloa, Jalisco and Hidalgo. An

outstanding case of success in state planning is that

of the California Energy Commission, which has

been at the vanguard of energy efficiency in the US

ever since the publication of its first standard -the

first in the US as well, in 1974 [27]. Dependence

from centralized planning is an institutional barrier

that has to be overcome.

Education has remained overlooked within

policy development: neither the Energy Reform, the

LAERFTE nor the ENE 2013 take into account any

educational measure; the LASE does mention

including rational use of energy in educational

programs in Mexico, as well as fostering research.

Previous literature has pointed out that education

can outperform other measures for conserving

energy, through breaking behavioral barriers among

end users [28] hence, since BCS residential users

are a significant share of the total users, education

can be a valuable tool for breaking behavioral

© by PSP Volume 25 – No. 8/2016, pages 3023-3029 Fresenius Environmental Bulletin

3028

barriers thus saving energy and achieving climate

change mitigation goals.

A more rational use of energy can be achieved

through tariffs, as was pointed from the overall

changes in consumption when electricity became

more expensive or when economy receded, but it

should not be the first neither the only policy

approach to use.

GHG emission savings from efficiency

standards need to be assessed and the pace of its

publication should not lag behind that of the

appearance of new devices: in BCS, energy related

emissions per capita increased from 1.7 tons CO2EQ

2005 to 2.4 tons CO2EQ in 2013 [2, 9] and electricity

generation makes up 35.6% of the total emissions,

after transport, with 52%, so promoting the rational

use of energy through all its different forms, i.e.

efficiency standards, education, tariffs and

regulation should be a critical strategy for

mitigating GHG emissions.

The path towards energy sustainability

through demand side management has been steep in

these decades. Although some true achievements

have been made, and more changes are to be

expected from the new measures and policies,

population growth and the electrification of the

everyday life still seem to be outpacing the

accomplished advances. In the isolated and arid

BCS, fostering a change in the currently intensive

end-user behavior is crucial for achieving the real

sustainability of the energy system.

ACKNOWLEDGEMENTS

To the Comisión Nacional del Agua for

providing meteorological data and to the Comisión

Federal de Electricidad, División de Estadísticas for

their advice on electrical sales information. This

study is the product of the work of CONACYT

Thematic Research National Network of Protected

Natural Areas (RENANP, by its acronym in

Spanish).

REFERENCES

[1] Sánchez-Ramos, I. and Pérez-Rebolledo, H.

(2011) Consumo de energía por potencia en

espera en casas y oficinas, Bol. Inst. Investig.

Eléctricas, 2011 143–149.

[2] INEGI, Banco de Información Económica

(BIE) (2015). [Online]. Available:

http://www.inegi.org.mx/sistemas/bie/.

[Accessed: 10-Sep-2015].

[3] SENER (2012), Prospectiva del Sector

Eléctrico 2012-2026. Secretaría de Energía,

México, D.F.

[4] CFE (2014) Costos totales de corto plazo

incurridos. 2014. Comisión Federal de

Electricidad, México, D.F.

[5] CFE (2015) Tarifas autorizadas por la

Comisión Federal de Electricidad. [Online].

Available:

http://app.cfe.gob.mx/Aplicaciones/CCFE/Tarif

as/Tarifas/tarifas_casa.asp. [Accessed: 10-Sep-

2015]. Comisión Federal de Electricidad,

México, D.F.

[6] Pereira, I. M. and Assis, E. S. D. (2013) Urban

energy consumption mapping for energy

management, Energy Policy, vol. 59, 257–269.

[7] SENER (2015) Sistema de Información

Energética. [Online]. Available:

http://sie.energia.gob.mx/. [Accessed: 10-Sep-

2015].

[8] Lagunes-Díaz, E., Beltran-Morales, L. Stoyan,

S. and Ortega-Rubio, A. (2011) Pasado,

presente y futuro del consumo y generación de

energía eléctrica en Baja California Sur:

planeación y optimización para el estado más

árido y aislado de la República., in Hacia la

sustentabilidad ambiental de la producción de

energía en México., M. E. González-Ávila, G.

Muñoz-Meléndez, and A. Ortega-Rubio, Eds.

Tijuana: El Colegio de la Frontera Norte y

CIBNOR, 139–172.

[9] SEMARNAT (2015) Registro de Emisiones y

Transferencia de Contaminantes. Secretaría de

Medio Ambiente y Recursos Naturales,

México, D.F.

[10] Manzini, F., Islas, J. and Macías, P. (2011)

Model for evaluating the environmental

sustainability of energy projects, Technol.

Forecast. Soc. Change, vol. 78, no. 6, 931–944.

[11] World Bank (2015), GDP per unit of energy

use (constant 2011 PPP $ per kg of oil

equivalent) | Data | Table, 2015. [Online].

Available:

http://data.worldbank.org/indicator/EG.GDP.P

USE.KO.PP.KD. [Accessed: 10-Sep-2015].

[12] OECD (2015) Greenhouse gas emissions.

[Online]. Available:

https://stats.oecd.org/Index.aspx?DataSetCode

=AIR_GHG. [Accessed: 10-Sep-2015].

[13] INEGI (2015) Encuesta Nacional de Ingresos y

Gastos de los Hogares (ENIGH). [Online].

Available:

http://www.inegi.org.mx/est/contenidos/proyec

tos/encuestas/hogares/regulares/enigh/default.a

spx. [Accessed: 10-Sep-2015].

[14] Cámara de Diputados (2003), Impacto en los

hogares Mexicanos por el incremento de las

tarifas eléctricas residenciales. Análisis por

deciles de ingresos.

[15] Rosas-Flores J., and Gálvez, D. (2010) What

goes up: Recent trends in Mexican residential

energy use, Energy, vol. 35, no. 6, 2596–2602.

© by PSP Volume 25 – No. 8/2016, pages 3023-3029 Fresenius Environmental Bulletin

3029

[16] SENER (2010), Evaluación y plan de manejo

ambiental del programa de eficiencia

energética. Secretaría de Energía, México, D.F.

[17] Auditoría Superior de la Federación (2012),

Secretaría de Energía. Administración y

Desarrollo de Programas de Ahorro de Energía

por Fideicomisos Públicos. [Online]. Available:

http://www.asf.gob.mx/Trans/Informes/IR2012

i/Documentos/Auditorias/2012_0443_a.pdf.

[Accessed: 14-Sep-2015].

[18] ASI (2008), Acciones financiadas por el

fideicomiso para el aislamiento térmico de la

vivienda en BCS, Ahorro Sistemático Integral.

[19] Rosenow, J. and Galvin, R. (2013) Evaluating

the evaluations: Evidence from energy

efficiency programmes in Germany and the

UK, Energy Build., vol. 62, 450–458.

[20] Escrivá-Escrivá, G. Álvarez-Bel, C., and

Peñalvo-López, E. (2011) New indices to

assess building energy efficiency at the use

stage, Energy Build., vol. 43, no. 2–3, 476–

484.

[21] Mohareb, E. a. Kennedy, C., Harvey, D. and

Pressnail, K. (2011) Decoupling of building

energy use and climate, Energy Build., vol. 43,

no. 10, 2961–2963.

[22] CONAGUA, Registro estación meteorológica

WMO 76405, 2003-2012, 2012. Comisión

Nacional del Agua, La Paz, BCS, México.

[23] R Core team (2014) R: A language and

environment for statistical computing. R

foundation for statistical computing.

[24] Rosas, J. Sheinbaum, C. and Morillon, D.

(2010) The structure of household energy

consumption and related CO2 emissions by

income group in Mexico, Energy Sustain. Dev.,

vol. 14, no. 2, 127–133.

[25] Ruth, M. and Lin, A (2006) Regional energy

demand and adaptations to climate change:

Methodology and application to the state of

Maryland, USA, Energy Policy, vol. 34, no. 17,

2820–2833.

[26] Hekkenberg, M., Benders, R., Moll, R. and

Schoot-Uiterkamp, A. (2009). Indications for a

changing electricity demand pattern: The

temperature dependence of electricity demand

in the Netherlands, Energy Policy, vol. 37, no.

4, 1542–1551.

[27] Alliance to Save Energy (2013) The history of

energy efficiency. Washington, D.C..

[28] Dias, R., Mattos, C. and Balestieri, P. (2004)

Energy education: breaking up the rational

energy use barriers, Energy Policy, vol. 32, no.

11, 1339–1347.

Received: 30.10.2015

Accepted: 15.03.2016

CORRESPONDING AUTHOR

Alfredo Ortega Rubio. Centro de Investigaciones Biológicas del Noroeste,

Av. IPN #195, Col. Playa Palo de Santa Rita,

23200. MEXICO

E-mail: aortega@cibnor.mx

Interciencia

ISSN: 0378-1844

interciencia@ivic.ve

Asociación Interciencia

Venezuela

Ortiz-Alcaraz, Antonio; Aguirre-Muñoz, Alfonso; Méndez-Sánchez, Federico; Ortega-

Rubio, Alfredo

FERAL SHEEP ERADICATION AT SOCORRO ISLAND, MEXICO: A MANDATORY

STEP TO ENSURE ECOLOGICAL RESTORATION

Interciencia, vol. 41, núm. 3, marzo, 2016, pp. 184-189

Asociación Interciencia

Caracas, Venezuela

Available in: http://www.redalyc.org/articulo.oa?id=33944256007

How to cite

Complete issue

More information about this article

Journal's homepage in redalyc.org

Scientific Information System

Network of Scientific Journals from Latin America, the Caribbean, Spain and Portugal

Non-profit academic project, developed under the open access initiative

184 MARCH 2016, VOL. 41 Nº 30378-1844/14/07/468-08 $ 3.00/0

KEYWORDS / Environmental Restoration / Eradication / Exotic Fauna / Feral Sheep / Ovis aries / Socorro Island /Received: 11/24/15. Modified: 02/08/2016. Accepted: 02/09/2016.

Introduction

Insular ecosystems have un-usually high rates of endemic species present in their envi-ronments. Islands are import-ant breeding grounds for a variety of birds and marine mammals. Mexican islands are known for their high biodiver-sity richness (Aguirre-Muñoz et al., 2005, 2011). Despite their ecological significance, many islands around the world receive serious impacts derived from human activity (Johnson and Stattersf ield, 1990). Approximately 75% of all ani-mal extinctions have occurred on islands, the majority being caused by introduced species (Diamond, 1989; Groombridge, 1992). Feral species are among the main causes of biodiversity loss and ecological disequilib-rium in many environments (Courchamp et al., 2003; Blackburn et al., 2004, Aguirre-Muñoz et al., 2008),

even the primary cause of ex-tinctions (Baillie et al., 2004; Reaser et al., 2007). In partic-ular, the settlement of herbi-vores like feral sheep (Ovis aries) has caused important ecological impacts on insular ecosystems. In 1869, 100 sheep were introduced to Socorro Island for ranching (they supplied food to boaters and f ishermen); over time, without humane care, sheep became feral, successfully adapting to island conditions (Levin and Moran, 1989; Álvarez-Cárdenas et al., 1994; Brattstrom, 2015). Since its introduction, feral sheep has caused great modification to the natural habitat. Erosion rates and loss of vegetation coverage derived from the sheep presence were docu-mented along the south-ern-central region of the island (León de la Luz et al., 1994; Maya-Delgado et al., 1994). Particularly, a shrub land

environment with Psidium so-corrense and Guettarda insu-laris, both endemic species, was severely disturbed. Constant overgrazing resulted in the abundance of paths formed by the trampling of sheep on the soil, which trig-gered different erosive pro-cesses and put native flora and fauna at risk (Figure 1). Among the most significant changes on the original floral composition has been an in-crease in the presence of graminacea species and scrub species as well as a reduction of the area covered by the na-tive flora. Sheep are particu-larly efficient in the propaga-tion of introduced plant spe-cies (they carry seeds in coat and excret). The change in native vegetation has been ob-served in every habitat that sheep occupied (SEMARNAT, 2004). Nearly 30% of the orig-inal soil and vegetation on Socorro Island was lost due to

erosive processes caused by feral sheep (Ortega-Rubio et al., 1992). This degradation problem was alarming since the erosion rate for deforested areas on Socorro Island was estimated at 20 to 90 tons/ha/year (Arnaud et al., 1993; Maya-Delgado et al., 1994; Ochoa-López et al., 1998).

Similarly, areas covered by introduced vegetation have a negative effect on local natu-ral processes, since native fauna species depend on na-tive plants for food, shelter or nesting. Arthropods are rare on disturbed areas (Jimenez et al., 1994). Vertebrates were negatively affected due to habitat loss; for example, the Socorro Blue Lizard (Urosaurus auriculatus) and the endemic birds Socorro Mockingbird (Mimus gray-soni) and Socorro Wren (Troglodytes sissonii) were less abundant in eroded areas (Martínez-Gómez et al.,

SUMMARY

Domestic sheep have been introduced to numerous islands worldwide; they became feral and caused indiscriminate im-pacts on ecosystems. On Socorro, a tropical Mexican island in the Pacific Ocean, feral sheep caused soil compaction and erosion, loss of vegetation, habitat destruction, landscape al-terations, and species extinction. In response to these biodiver-sity threats, feral sheep were removed from Socorro Island. In 2009, 1257 sheep were dispatched using aerial hunting over 35 hours of flight time. To locate remaining herds, twelve sheep

were captured and equipped with radio-collars (Judas sheep). From March 2010 to April 2012 eradication continued using terrestrial hunting; 505 sheep were dispatched at this stage. The use of Judas sheep and detection dogs were critical in lo-cating the last remaining sheep. A systematic monitoring pro-gram conducted during two years confirmed eradication suc-cess. In June 2014, Socorro Island was declared free of feral sheep. To date (November 2015), a remarkable ecosystem re-covery has been recorded.

FERAL SHEEP ERADICATION AT SOCORRO ISLAND, MEXICO: A MANDATORY STEP TO ENSURE ECOLOGICAL RESTORATIONAntonio Ortiz-Alcaraz, Alfonso Aguirre-Muñoz, Federico Méndez-Sánchez and Alfredo Ortega-Rubio

Antonio Ortiz-Alcaraz. Biologist, Universidad Autónoma de Baja California (UABC), Mexico. Masters in Use, Management and Conservation of Natural Resources, and Ph.D. candidate in Ecology, Centro de Investigaciones Biológicas del Noroeste (CIBNOR), Mexico. Director, Gulf of California

Islands Project, Grupo de Ecología y Conservación de Islas, (GECI), Mexico.

Alfonso Aguirre-Muñoz. B.Sc. in Oceanography, UABC, Mexico. Ph.D. in Sustainable Development and Regional Studies, El Colegio de la Fron-tera Norte, Mexico. Executive Director, GECI, Mexico.

Federico Méndez-Sánchez. B.Sc. in Oceanography, UABC, Mexico). M.Sc. in Environmental Management, University of Auckland, New Zealand. Direc-tor for Development, GECI, Mexico.

Alfredo Ortega-Rubio. Doctor of Science in Ecology, Instituto Politécnico Nacional,

Mexico. Researcher, CIBNOR, Mexico. Address: CIBNOR. Av. Instituto Politécnico Nacional 195, Playa Palo de Santa Rita Sur, La Paz, Baja California Sur, C.P. 23096, México e-mail: aortega@cib-nor.mx

185MARCH 2016, VOL. 41 Nº 3

ERRADICACIÓN DE BORREGO FERAL EN ISLA SOCORRO, MÉXICO: ACCIÓN REQUERIDA PARA LOGRAR SU RESTAURACIÓN ECOLÓGICAAntonio Ortiz-Alcaraz, Alfonso Aguirre-Muñoz, Federico Méndez-Sánchez y Alfredo Ortega-Rubio

RESUMEN

equipados con radio-collares (borregos Judas); una vez libe-rados sirvieron para localizar las manadas remanentes. De marzo 2010 a abril 2012 se continuó con la erradicación me-diante cacería terrestre; 505 borregos fueron sacrificados en esta etapa. El uso de borregos Judas y perros de muestra en-trenados fue fundamental para la localización de los últimos borregos. Con el fin de confirmar la erradicación se llevó a cabo por dos años un programa de monitoreo sistemático. En junio 2014, Isla Socorro fue declarada libre de borregos fera-les. A la fecha (noviembre 2015) se ha registrado una mejoría notable del ecosistema.

Borregos domésticos han sido introducidos en numerosas islas alrededor del mundo; éstos se vuelven ferales y pro-vocan impactos indiscriminados en los ecosistemas. En Isla Socorro, una isla tropical mexicana en el Océano Pacífico, los borregos ferales causaron la erosión y compactación del suelo, pérdida de vegetación y destrucción de hábitat, alte-raciones del paisaje, y la extinción de especies. En respues-ta a estas amenazas a la biodiversidad, los borregos fueron erradicados de Isla Socorro. En 2009, 1257 borregos ferales fueron sacrificados durante la fase de cacería aérea, con 35 horas de vuelo. Un total de 12 borregos fueron capturados y

ERRADICAÇÃO DE CARNEIROS FERAIS NA ILHA SOCORRO, MÉXICO: UM PASSO OBRIGATÓRIO PARA GARANTIR A RESTAURAÇÃO ECOLÓGICAAntonio Ortiz-Alcaraz, Alfonso Aguirre-Muñoz, Federico Méndez-Sánchez e Alfredo Ortega-Rubio

RESUMO

leiras (carneiros Judas) para localizar os rebanhos restantes. Entre março 2010 a abril 2012, continuo a erradicação com caça terrestre e 505 carneiros foram sacrificados nesta fase. O uso de “carneiros Judas” e cães de detecção foram funda-mentais para localizar os últimos carneiros. Para confirmar o sucesso da erradicação um programa de monitoração sistemá-tica foi realizado nos dois anos seguintes. Em junho 2014, Ilha Socorro foi declarada livre de carneiros selvagens. Até a pre-sente data (novembro 2015) se registra uma recuperação notá-vel do ecossistema.

Carneiros domésticos foram introduzidos em numerosas ilhas em todo o mundo; tornaram-se ferais e tiveram impactos amplamente indiscriminados sobre os ecossistemas. Em Socor-ro, uma ilha tropical mexicana no Oceano Pacífico, os carnei-ros ferais causaram a compactação e erosão do solo, perda de vegetação e habitat, mudança da paisagem, e extinção de es-pécies. Em resposta a estas ameaças à biodiversidade os car-neiros ferais foram retirados da Ilha Socorro. Em 2009, 1257 ovinos foram sacrificados durante 35 horas de vôo de caça aé-rea. Doze ovinos foram capturados e equipados com rádio co-

2001; Martínez-Gómez and Jacobsen, 2004), but highly abundant in areas where the f loral composition remained unaltered (Galina et al., 1994; Rodríguez-Estrella et al., 1994). Feral sheep contribut-ed, along with feral cats (Felis catus), to the extinction from the wild of the endemic Socorro Dove (Zenaida

graysoni), with the last re-maining specimens now kept only at zoos in Mexico, USA and Germany.

The removal of feral spe-cies has showed to be an achievable task in many is-lands of the world (e.g., Veitch and Clout, 2002; Veitch et al., 2011). With funds from both Mexican

(CONANP, CONABIO and INECC) and USA (USFWS) governments, and national (Alianza WWF-Fundación Carlos Slim) and international (American Bird Conser- vancy, Packard and Marisla Foundation) donors, and with the invaluable logistic support from the Mexican Navy (Secretaría de Marina;

Armada de México), the Mexican civil society organi-zation Grupo de Ecología y Conservación de Islas, A.C. (GECI) completed the feral sheep eradication on Socorro Island by 2012. In this paper we report the methods used to achieve this significant resto-ration project.

Methods

Study site

Socorro Island (13,200ha; 1,040masl) is located in the Revillagigedo Archipelago (Figure 2), 460km west from Los Cabos, Baja California Sur and 694km west from Manzanillo, Colima (18º47´N, 110º58´W). It harbors 118 na-tive vascular plants, 30 of which are endemic (León de la Luz et al., 1994). Socorro’s fauna includes 103 bird species (terrestrial, marine, wading and

Figure 1. Loss of vegetation (left) and soil (right).

186 MARCH 2016, VOL. 41 Nº 3

migratory); eight of the terrestri-al birds are endemic to the is-land. Also, two terrestrial rep-tiles are present, one being en-demic (SEMARNAT, 2004). The island is permanently in-habited by military personnel (40-50 people) stationed at a Naval Base located in the south-ernmost part of the island.

Based on the experienced gained around the world (e.g., Campbell and Donlan, 2005) and in Mexico (Aguirre-Muñoz et al., 2011) in the eradication of ungulates (mainly goats and sheep), we selected aerial hunt-ing aided with global position-ing systems, geographic infor-mation systems (GIS), terrestri-al hunting and trapping as main eradication methods. Judas animals and specialized hunting dogs were also used for localizing remaining sheep.

Feral sheep monitoring

In 2002, GECI began with a feral sheep monitoring program on Socorro Island to assess eradication feasibility; herbs lo-cation and major distribution areas were determined. It was concluded that eradication was

possible and that aerial hunting was the best method. During October 2005, several f lights aboard a Beechcraft Bonanza aircraft were made over the is-land. This allowed to identify the main areas where sheep were distributed. At the same time, land surveys were con-ducted by foot and in motorized vehicles (ATV s). Mexican Navy officers guided these surveys. The locations of the sheep herds were recorded into a GIS.

Aerial hunting

The aerial hunting cam-paign, using a single turbine helicopter (model MD369D), took place between April 20 and 29, 2009. Twelve gauge semi-automatic shotguns (Winchester SX3) equipped with near-range scopes (Sightmark SM13003), and semi-automatic .308 rif les (Benelli R1) were used. Guns permits from the Secretariat of National Defense were ob-tained and hunting was super-vised by Mexican Navy per-sonnel on the island. Two hunters were hunting simulta-neously during the flights, one

behind the pilot and another next to the pilot. Manual counting devices were used to register the number shot. Also, a GPS was used in order to record the f light trajectories and the areas where the high-est numbers of sheep were found. All f lights were done between 07:00 and 11:00, and between 16:00 and 19:00, when sheep have a higher ac-tivity. During the f irst two f lights we covered the whole island in order to identify the main herds of sheep. For the subsequent 25 flights, hunting efforts were concentrated in areas with the higher densities of sheep (east zone of the is-land). Flights were done at an average speed of 42km/h and average height of 35m.

Judas sheep

During the hunting flights 12 live animals were captured, to be used as Judas sheep (Taylor and Katahira, 1988). These an-imals were neutered and fitted with radio-telemetry collars. The captured animals were taken to Socorro’s Naval Base and kept in a barn prior to

their sterilization. The latter was done by GECI’s personnel, previously trained on Guada-lupe Island by experts from the Galápagos Islands National Park. Sheep were kept for ob-servation three days prior to their liberation. The animals were equipped with radio-te-lemetry collars (Telenax Mexico), with frequencies be-tween 164 and 168Mhz. These Judas sheep were deployed back to the sites where they were captured.

Terrestrial hunting and trapping

From February 2010 to April 2012, 4 to 7 experienced hunters (with gun permits) carried out this stage of the project. Every hunter had a handheld GPS to record their hunting tracks. Rif les with telescopic sights (Leupold VX-I), caliber .222, .243 and .308, as well as 12 gauge shotgun with cartridges 00 buck were used. Periodically, Judas sheep were located to locate herds. Simultaneously, leg hold traps (Oneida Victor Soft Catch # 3) and snare traps were used on identified trails; both types of t raps were checked daily. Traps were used to capture remaining sheep in places where previously hunted; thus, hunting efforts focused on new areas of the island.

Hunting dogs

As the last stage of the erad-ication, we used two hunting dogs (Beagle and Foxhound) to track down the last sheep herds. Dogs were fitted with GPS collars to record their lo-cations and movements. The dogs were rented; they were trained for localizing of sheep and goat. Two dogs worked si-multaneously during 2h periods with 30min rest.

Results and Discussion

Feral sheep monitoring

Both aerial and ground sur-veys confirmed the description made in 1992 by Álvarez-Cárdenas et al. (1994) that

Figure 2. Location of Socorro Island.

187MARCH 2016, VOL. 41 Nº 3

sheep were present in the east zone of the island (in an area of ~6000ha). Patches of Bumelia, Ficus, Psidium and Guettarda are abundant in this area. However, due to the re-building of the island’s airstrip and to ground hunting done by the Mexican Navy for local human consumption, sheep were congregating in higher areas of the island. The areas frequented by sheep were wide open spaces, with scarce vege-tation, but further away from human presence.

Aerial hunting

A total of 35 hours of flight were done in one week for the aerial hunting stage, in which the majority of the is-land was covered (Figure 3). During this period, 1,257 in-dividuals were dispatched. The total amount of f light hours and the sheep dis-patched are detailed in Figure 4. The average f light t ime was 1h 20min. We stopped hunting when sheep were difficult to localize, and

few animals were dispatched in an hour of f light. Aerial hunting is a highly effective method to use in invasive species eradications (Campbell and Donlan, 2005). The tech-nique allows hunters to access animals that otherwise would have been completely inacces-sible by foot, considering that their location was confined to cliffs or canyons. The effec-tiveness of this technique is backed up by the results, since it only took 35h of f ly-ing time to dispatch a large par t (~70%) of the sheep population.

Sheep were mainly present at least 1km away from the main trails and the airstrip. The rest of the hunting sites were located >2km away from the Navy Base. The distance from the trails made it very diff icult to carry back the hunted sheep for human con-sumption. Nevertheless, certain areas with little or no vegeta-tion allowed the use of the helicopter to carry back some sheep. Approximately 75 sheep were recovered and given to the Mexican Navy personnel. The Mexican Navy carries food to Socorro Island every two weeks for its personnel, so sheep did not represent their source of meat. They were

convinced of the ecological impacts caused by sheep and cooperated with this project.

The tropical climate of the island allowed the carcasses to decompose rapidly. Land inver tebrates, aided by the microbial array present in the soil, were capable of fully decomposing one adult sheep in two weeks. The high irra-diance rates of the tropical sun avoid for the carcasses to become a source of infection on the island. Furthermore, the decays enrich the soil of the area. GECI personnel car-ried out an on-foot survey, the day after the last hunting ef-for t, in order to verify the state of the carcasses; they were beginning to dry out.

Ground hunting and trapping

Ground hunting was concen-trated on sites identified by air with the greatest abundance of sheep traces (Figure 5). A summary of hunting and trap-ping effor ts is shown in Table I; 505 sheep were dis-patched through ground hunt-ing, in 9 months over a period of two years (March 2010 to April 2012). Judas sheep were most effective when there was a greater amount of remaining sheep, because it increases

Figure 5. Hunting tracks (black line) and use of dogs (white line), from 2010 to 2012.

Figure 3. Helicopter flight trajectories on Socorro Island.

Figure 4. Amount of sheep dispatched by hours of f light (total accumulated).

188 MARCH 2016, VOL. 41 Nº 3

possibility of aggregation. Hunting dogs were only used to locate the last ten remaining animals, which were difficult to locate for hunters.

A total of 1762 sheep were dispatched from Socorro Island in a 3-year eradication cam-paign (April 2009 to April 2012). After aerial hunting, it took two years to dispatch 30% of the sheep population by terrestrial hunting, due to difficult logistics because of the distances, topography and size of the island. During the following two years (May 2012 to December 2014), a to-tal of 14 expeditions of 21 days each were carried out to find traces of sheep (footprints and feces) on the island; no traces of sheep were recorded. Conventionally, islands are declared free of invasive spe-cies after two or three years without evidence of exotics (Rejmanek and Pitcairn, 2002; Regan et al., 2006). Socorro Island was declared free of feral sheep in June 2014. This project had an approximate cost of US$ 500,000. This amount does not include the contribution in-kind from the Mexican Navy: maritime transportation, accommodation, use of heliport and vehicles; and the costs of the two con-secutive years for confirmation of the absence of sheep (US$ ~200,000). Compared with oth-er islands where goat or sheep have been eradicated (Van Vuren, 1992), feral sheep erad-ication on Socorro Island can be considered highly effective as it was completed in three years, when similar projects take 3-5 years, and even de-cades, to conclude (Campbell et al., 2005). Moreover, the

methods used reduced project cost, which was of US$ 38/ha, while other projects of sheep eradication, such as that on Santa Cruz Island, California, had a cost of US$ 80/ha, due to capture and transport of sheep to the continent (Faulkner and Kessler, 2001).

Since the sheep eradication project was concluded vegeta-tion began to recover passive-ly. Comparison of satellite im-ages from 2008 and 2013 show that vegetation has recovered in 1450ha. Results from a field assessment of vegetation showed an average of 21% of vegetation cover in 2009, at the most impacted sites by sheep, while in 2014 average vegetation cover reached 79%.

Conclusions

The aerial hunting method showed to be an ideal tech-nique for the eradication of sheep from Socorro Island. It enabled the eradication team to dispatch a high number of animals in few days of work, while allowing the hunters to complete difficult access ar-eas. The use of Judas sheep and hunting dogs were cru-cial for concluding eradica-tion. There-fore, these tech-niques can be used on other islands in Mexico and other countries.

ACKNOWLEDGMENTS

This work would have not been possible without the suppor t of the Mexican Navy. SEMAR’s invaluable contribution to this eradica-t ion campaign included al-lowing researchers into the Socorro Island Naval Base,

t ranspor tation of staff and equipment from the Manza-nillo Harbor in Colima to the island, accommodation and use of the heliport and other facil-it ies. The support of the Mexican federal government agencies SEMARNAT, DGVS, CONANP, SEGOB, SEDENA and the granting of all needed permits were key for the suc-cess of the program. Donors were: Mexican (CONANP, CONABIO, INECC) and United States (USFWS) gov-ernments, Alianza WWF-Fundación Carlos Slim, American Bird Conservancy, Packard and Marisla Foun-dation. We thank support from CONACYT Natural Protected Areas Network (RENANP).

REFERENCES

Aguirre-Muñoz A, Croll D, Donlan J, Henry RW, Hermosillo MA, Howald G, Keit t B, Luna-Mendoza L, Rodríguez-Malagón M, Salas-Flores LM, Sama-niego-Herrera A, Sánchez-Pacheco JA, Sheppard J, Tershy B, Toro-Benito J, Wolf S, Wood B (2008) High-impact conserva-tion action: A case study from the islands of Western Mexico. Ambio 37:101-107.

Aguirre-Muñoz A, Samaniego-Herrera A, Luna-Mendoza L, Ortiz-Alcaraz A, Rodríguez-Malagón M, Méndez-Sánchez F, Félix-Lizárraga M, Hernández-Montoya JC, González-Gómez R, Torres-García F, Barredo-Barberena JM, Latofski-Robles M (2011) Island restoration in Mexico: Ecological outcomes after systematic eradications of invasive mammals. In Veitch CR, Clout MN, Towns DR (Eds.) Island Invasives: Eradication and Management. Proc. Int. Conf. on Island Invasives. IUCN Species

Survival Commission Nº 42. IUCN-CBB. Gland, Switzerland. pp. 250-258.

Aguirre-Muñoz A, Samaniego-Herrera A, García-Gutiérrez C, Luna-Mendoza L, Rodríguez-Malagón M, Casillas-Figueroa F (2005) El Control y la erradica-ción de Fauna Introducida como Instrumento de Restauración Ambiental: Historia, Retos y Avances en México. Temas so-bre Restauración Ecológica. Instituto Nacional de Ecología. México. 254 pp.

Álvarez-Cárdenas S, Castellanos-Vera A, Galina-Tessaro P, Arnaud G (1994) Aspectos de la población y el hábitat del borrego doméstico (Ovis aries). In Ortega-Rubio A, Castellanos V (Eds.) La Isla Socorro, Reserva de la Biosfera Archipiélago de Revillagigedo, México. Centro de Investi-gaciones Biológicas del Noroeste. México. pp. 301-317.

Arnaud G, Rodríguez A, Ortega-Rubio A, Álvarez CS (1993) Predation by cats on the unique endemic lizard of Socorro Island (Urosaurus auriculatus), Revillagigedo, México. Ohio J. Sci. 93(4): 101-104.

Baillie JEM, Hilton-Taylor C, Stuart SN, Eds. (2004) IUCN Red List of Threatened Species. A Global Species Assessment. IUCN. Gland, Switzerland. 191 pp.

Blackburn TM, Cassey P, Duncan RP, Evans KL, Gaston KJ (2004) Avian extinction and mammalian introductions on oceanic islands. Science 305: 1955-1958.

Brattstrom B (2015) Food Webs and Feeding Habits on the Revillagigedo Islands, Mexico. Pacif. Sci. 69: 181-195.

Campbell K, Donlan J (2005) A review of feral goat eradication on islands. Conserv. Biol . 198: 1362-1374.

Courchamp F, Chapuis JL, Pascal M (2003) Mammal invaders on

TABLE IGROUND HUNTING EFFORT

Year Months Personnel Hours/hunter Distance (km) Sheep with Judas Traps/Night Captured

sheep Hours/dogs Dispatched sheep

2010 Mar-Apr 7 1323 815 53 900 41 - 355Jul 6 588 460 18 - - - 48

2011 Apr 5 512 433 11 650 8 - 67Ago-Sep 4 728 644 4 - - - 25Nov-Dec 4 420 385 - - - 49 8

2012 Apr 4 240 216 - - - - 2Total 3811 2953 86 1550 49 49 505

189MARCH 2016, VOL. 41 Nº 3

islands: impact, control, and control impact. Biol. Rev. 78: 347-383.

Diamond JM (1989) Overview of re-cent extinctions. In Western D, Pearl MC (Eds.) Conservation for the Twenty-First Century. Oxford University Press. New York, USA. pp. 37-41.

Faulkner KR, Kessler CC (2011) Live capture and removal of feral sheep from eastern Santa Cruz Island, California. In Veitch CR, Clout MN, Towns DR (Eds.) Island Invasives: Eradication and Management. IUCN. Gland, Switzerland. pp. 295-299.

Galina P, Álvarez S, Ortega-Rubio A (1994) Aspectos ecológicos de la herpetofauna. In Ortega-Rubio A, Castellanos V (Eds.) La Isla Socorro, Reserva de la Biosfera Archipiélago de Revillagigedo, México. Centro de Investiga-ciones Biológicas del Noroeste. México. pp. 185-197.

Groombridge BE (1992) Global Diversity - Status of the Earth’s Living Resources. Compiled by the World Conservation Monitoring Center. Chapman & Hall, London, UK. 585 pp.

Jiménez M, Tejeda A, Aguilar R (1994) Los artrópodos terres-t res. In Or tega-Rubio A, Castellanos V (Eds.) La Isla Socorro, Reserva de la Biosfera Archipiélago de Revillagigedo, México. Centro

de Investigaciones Biológicas del Noroeste. México. pp. 153-183.

Johnson TH, Stattersfield AJ (1990) A global review of island ende-mic birds. Ibis 132: 167-180.

León de la Luz JL, Breceda A, Coria-Benet R, Cancino-Hernández J (1994) Asocia-ciones vegetales. In Ortega-Rubio A, Castellanos V (Eds.) La Isla Socorro, Reserva de la Biosfera Archipiélago de Revillagigedo, México. Centro de Investigaciones Biológicas del Noroeste. México. pp. 115-141.

Levin JA, Moran R (1989) The Vascular Flora of Isla Socorro, México. San Diego Society of Natural History, University of Michigan, USA. 71 pp.

Martínez-Gómez JE, Flores-Palacios A, Curry RL (2001) Habitat re-quirements of the Socorro Mockingbird Mimodes graysoni. Ibis 143: 456-467.

Martínez-Gómez JE, Jacobsen J (2004) The conservation status of Townsend’s S hearwater Puffinus auricularis auricularis. Biol. Conserv. 116: 35-47.

Maya-Delgado Y, Salinas-Zavala F, Troyo-Diéguez E (1994) Estado actual del suelo y propuestas para su conservación. In Ortega-Rubio A, Castellanos V (Eds.) La Isla Socorro, Reserva de la Biosfera Archipiélago de Revillagigedo, México. Centro de Investigaciones Biológicas del Noroeste. México. pp. 63-75.

Ochoa-López E, Reyes-Bonilla H, Ketchum-Mejía J (1998) Daños por sedimentación a las comuni-dades coralinas al sur de la isla Socorro, Archipiélago de Revillagigedo, México. Cs. Mar. 24: 233-240.

Ortega-Rubio A, Castellanos-Vera A, Arnaud G, Maya-Delgado Y, Rodríguez R, León de la Luz JL, Cancino J, Jiménez C, Llianas J, Álvarez S, Galina P, Breceda A, Troyo E, Salinas F, Díaz S, Servín R, Romero H, Rodríguez A, Coria R (1992) Estudio de los recursos naturales de la Isla Socorro, Revillagigedo. Ciencia 43: 175-184.

Reaser JK, Meyerson LA, Cronk Q, De Poorter M, Eldrege LG, Green E, Kairo M, Latasi P, Mack RN, Mauremootoo J, O’dowd D, Orapa W, Sastroutomo S, Saunders A, Shine C, Thrainsson S, Vaiutu L (2007) Ecological and socioeco-nomic impacts of invasive alien species in island ecosystems. Environ. Conserv. 34: 98-111.

Regan TJ, McCarthy MA, Baxter PWJ, Panetta FD, Possingham HP (2006) Optimal eradication: when to stop looking for an invasive plant. Ecol. Let t . 9: 759-766.

Rejmanek M, Pitcairn MJ (2002) When is eradication of exotic pest plants a realist ic goal? In Veitch CR, Clout MN (Eds.) Turning the Tide: The

Eradication of Invasive Species. Invasive Species Special ist Group. IUCN. Gland, Switzerland. pp. 249-253.

Rodríguez-Estrella R, Rivera-Rodríguez , Mata E (1994) Avifauna terrestre. In Ortega-Rubio A, Castellanos V (Eds.) La Isla Socorro, Reserva de la Biosfera Archipiélago de Revillagigedo, México. Centro de Investigaciones Biológicas del Noroeste. México. pp. 199-224.

SEMARNAT (2004) Programa de Conservación y Manejo, Reserva de la Biósfera Archipiélago de Revillagigedo. Secretaría de Medio Ambiente y Recursos Naturales México. 220 pp.

Taylor D, Katahira L (1988) Radio telemetry as an aid in eradica-ting remnant feral goats. Wildl. Soc. Bull. 16: 297-299.

Van Vuren D (1992) Eradication of feral goats and sheep from is-land ecosystems. In Borrecco JE, Marsh RE (Eds.) Proc. 15th Vertebrate Pest Conference. University of California, Davis, USA. pp. 377-381.

Veitch CR, Clout MN (2002) Turning the Tide: The Eradication of Invasive Species. Invasive Species Specialist Group. IUCN. Gland, Switzerland. pp. 374-380.

Veitch CR, Clout MN, Towns DR, Eds. (2011) Island Invasives: Eradication and Management. Invasive Species Specialist Group, IUCN. Gland, Swit-zerland. 542 pp.

Article

Recovery of Vegetation Cover and Soil after theRemoval of Sheep in Socorro Island, Mexico

Antonio Ortíz-Alcaraz 1,2, Yolanda Maya-Delgado 1,*, Patricia Cortés-Calva 1,Alfonso Aguirre-Muñoz 2, Evaristo Rojas-Mayoral 2, Miguel V. Cordoba-Matson 1 andAlfredo Ortega-Rubio 1

1 Centro de Investigaciones Biológicas del Noroeste, S.C. Av. Instituto Politécnico Nacional 195, La Paz,B.C.S. 23096, Mexico; antonio.ortiz@islas.org.mx (A.O.-A.); pcortes04@cibnor.mx (P.C.-C.);mcordoba@cibnor.mx (M.V.C.-M.); aortega@cibnor.mx (A.O.-R.)

2 Grupo de Ecología y Conservación de Islas, A.C. Moctezuma 836, Ensenada, B.C., 22800, Mexico;alfonso.aguirre@islas.org.mx (A.A.-M.); evaristo.rojas@islas.org.mx (E.R.-M.)

* Correspondence: ymaya04@cibnor.mx; Tel.: +52-612-123-8480

Academic Editors: Sadanandan Nambiar and Timothy A. MartinReceived: 4 February 2016; Accepted: 20 April 2016; Published: 22 April 2016

Abstract: For over 140 years, the habitat of Socorro Island in the Mexican Pacific has been altered bythe presence of exotic sheep. Overgrazing, jointly with tropical storms, has caused soil erosion, andmore than 2000 hectares of native vegetation have been lost. Sheep eradication was conducted from2009 to 2012. Since then, the vegetation has begun to recover passively, modifying soil properties.The objective of our study was to verify that this island was resilient enough to be recovered and in arelatively short time scale. To confirm our hypothesis, we analyzed changes in the physical-chemicalproperties of the soil and vegetation cover, the last one in different times and habitats after sheeperadication. The change in vegetation cover was estimated by comparing the normalized differencevegetation index (NDVI) between 2008 and 2013. In sites altered by feral sheep, soil compactionwas assessed, and soil samples were taken, analyzing pH, electrical conductivity, organic carbon,total nitrogen, phosphorus, calcium, and magnesium. After a year of total sheep eradication, clearindications in the recovery of vegetation cover and improvement of soil quality parameters wereobserved and confirmed, specifically compaction and nitrogen, organic carbon, phosphorus, andcalcium. The results seem to support our hypothesis.

Keywords: feral sheep; soil erosion; soil recovery; Socorro Island

1. Introduction

Habitat degradation caused by feral herbivores is one of the most serious ecological problems seenin island ecosystems worldwide [1–14]. Invasive exotic herbivores often have evolutionary strategiesthat favor their adaptation to the insular environment [15] and usually have negative effects on nativeplants and animal species, including soil erosion and compaction, vegetation cover loss, competitionwith native herbivores, and extirpation and extinction of endemic species [8,16–18].

Islands have higher concentrations of endemic species than the mainland; the proportion ofendemics increases with isolation, island size, and topography. These centers of endemism arefrequently threatened by human activity [19–21]. In 1869, 100 sheep (Ovis aries Linnaeus, 1758) fromAustralia were introduced to Socorro Island in the Revillagigedo Archipelago. They became feraland adapted to the insular environment [22]. In the absence of natural enemies, sheep populationgrew to be about 5000 individuals in 1960 [23], ca. 2000 in 1988 as a result of increased hunting effortby the Mexican Navy [24], and they became the main disturbance agent of the island’s ecological

Forests 2016, 7, 91; doi:10.3390/f7040091 www.mdpi.com/journal/forests

Forests 2016, 7, 91 2 of 13

conditions [25,26]. Sheep are credited with the loss of soil and vegetation cover in 30% of the islandsurface [27,28], habitat quality deterioration, and population decline of endemic vertebrates [24,29–31].

The island has different erosion degrees by removing vegetation cover due to overgrazing, which,in addition to sheep bed sites and rest areas, produced soil compaction, removal of the litter layer,and destruction of seedlings [24,32]. The erosion was exacerbated due to the geographical position ofthe island, located in the path of hurricanes in the Pacific, causing torrential rains with a consequenteroded soil landslide. Because runoff and erosion rates increase in an inverse relationship with plantcover [33], water was mobilized in large runoff through the bare areas and concentrated in naturalhollows or in the trails used by the sheep to move from one grazing area to another, resulting indeep erosion gullies. Its formation was favored by the soil texture since the island’s eroded areas arecharacterized by a high percentage of clay.

Some ecosystems may be sufficiently resilient to recover if the degree of damage has not reachedthe ecological irreversible threshold [34]. Other possible scenarios are limited recovery, unintendedconsequences [35], or that it may take long geological time scales [36]. The degree of recoverywill depend on the magnitude and duration of impacts, as well as on the nature of the disturbedecosystem [37]. Our hypothesis was that Socorro Island is resilient enough to recover in a relativelysmall time scale, after the removal of the pressures caused by a large exotic herbivore. Thus, theobjectives of this study were to (1) perform a temporal analysis of the normalized difference vegetationindex (NDVI) in images taken in 2008 and 2013; (2) assess the changes in vegetation cover and thenumber of plant species in different habitats of the island by field sampling; (3) identify pioneerspecies in eroded soils; and (4) evaluate the physical and chemical properties of soil under the differentpercentages of plant cover that derived from the elimination of the main source of disturbance.

2. Materials and Methods

2.1. Study Area

The Socorro Island is a tropical volcanic island located in the Revillagigedo Archipelago (18˝42’28”N, 111˝02’49” W), east of the Pacific Ocean (Figure 1). It has an area of 132 km2 and a maximumelevation of 1040 m. It is located in the path of tropical cyclones that occur from June to Novemberwith an average of 2.75 events per year, with precipitation of more than 400 mm during the season [38].There are no freshwater streams, and approximately 27% of 117 vascular plant species are endemic tothe island [39]. Except for the disturbed areas, a dense deciduous and evergreen tropical scrub withpatches of low tropical forest covers Socorro Island. Deciduous trees are dominant in the lowlands,with endemic evergreens in the higher parts of the island. On the southern slopes (500–800 m), morewoodland is observed than on the steeper northern side of Socorro due to deeper soils and flatterterrain [24,40,41]. No records of native herbivore mammals are available [42].

Since their introduction, sheep had become feral and the population varied from 2000 to 5000individuals [22–24]. Sheep were mainly concentrated in the central-southern portion of Socorro Island.The dense and tangled island scrub, as well as a dangerously sharp and brittle surface of severalhigh-ridged rocky lava flows and a scarcity of freshwater sources, appeared to be the reason whythey did not disperse and impact the northern portion of the island [24]. The sheep caused erosionon much of the central-southern portion of the island [24,32], particularly in the areas correspondingto the guayabillo (Psidium socorrense I.M.Johnst.) scrub. The eroded surfaces were transformed insavannah-like and prairie-like open spaces, with a mix of native and exotic vegetation [24] besides asmall number of individuals and species [39] with a consequent reduction in vegetation cover. As aconservation measurement of natural resources in Mexico and particularly in Socorro Island, severalMexican agencies started eradication programs of feral sheep in 2008 ending in 2012. Along this study,sheep population decreased drastically due to eradication campaigns. Aerial sightings have recordedan estimation of 1800 individuals in 2009, 500 in 2010, and 100 in 2011 [43], which have been eradicatedby 2012. Since then, passive vegetation recovery has been documented.

Forests 2016, 7, 91 3 of 13

From an edaphological point of view, Socorro Island has three areas: (1) The northern area hasthin soils overlying recent lava, and pyroclastic spills dominate the soils most recently formed on theisland; it has dense vegetation that does not require deep soil to thrive; (2) The central area has soilsderiving from volcanic glass, which cover older soils and rocks. Toward its southern region, clay andred and deep soils have formed from basalt with a high iron content that provides the characteristicred color of the island landscapes [32]; (3) The central-southern portion of the island, the most clearlyeroded and denuded, was our sampling area.Forests 2016, 7, 91  4 of 14 

 

Figure 1. Location of Socorro Island and location of the sampling transects of vegetation and soils. (a) 

Mexico; (b) Revillagigedo Archipelago; (c) sampling area in the central‐southeast of Socorro Island. 

2.4. Soil Compaction 

In  2013,  soil penetration  resistance measurements were  taken by using a penetrometer  (Soil 

Compaction Tester Dickey‐john®) within the 20 sampling transects (Figure 1). Eroded soil sites and 

50% and 100% recovered vegetation cover were chosen; 50 readings in total were obtained in each 

condition. In addition, 50 readings out of transects were taken at sites that were not disturbed by the 

sheep  (ND = non‐disturbed), with 100% vegetation coverage  [47], making a  total of 200  readings 

among all conditions. Walter and Levin [24] observed this condition (sites without sheep), which we 

corroborated when performing the aerial monitoring of sheep. An analysis of variance and Tukey’s 

honest  significance  test were performed  to determine  significant differences  among  the different 

degrees of vegetation cover. 

2.5. Physical‐Chemical Soil Parameters 

Soil samples, approximately 1 kg each, were collected from each transect at a depth of 0–10 cm: 

16  of  them  on  eroded  soils,  16  on  soils with  50%,  and  16  on  soils with  100%  vegetation  cover 

recovered. Samples were also obtained outside transects in places that were not disturbed by sheep 

with  100%  vegetation  cover  (100%  ND).  The  following  physical‐chemical  parameters  were 

determined:  pH  and  electrical  conductivity  (for  reading  potentiometer  and  conductivity, 

respectively); total nitrogen by the Dumas method in a LECO nitrogen analyzer; organic matter by 

the method  of Walkley‐Black;  phosphorus  by  colorimetric  reading  of  a  spectrophotometer,  and 

calcium and magnesium by  the EDTA method  [48–50]. The results were compared by analysis of 

variance, and comparisons of means were performed by Tukey’s  test  to determine similarity and 

differences among sites. 

3. Results 

3.1. NDVI 

The  results obtained by  the NDVI  indicated  that,  in 2013, healthier vegetation was observed 

compared to 2008. The multi‐temporal analysis performed by NDVI is shown in Figure 2, where red 

Figure 1. Location of Socorro Island and location of the sampling transects of vegetation and soils.(a) Mexico; (b) Revillagigedo Archipelago; (c) sampling area in the central-southeast of Socorro Island.

2.2. Normalized Difference Vegetation Index (NDVI)

To assess the changes in vegetation cover, we evaluated the photosynthetic vegetation vigor onthe island by obtaining the normalized difference vegetation index (NDVI) [44] to distinguish areaswithout vegetation. The NDVI was obtained using the equation:

NDVI “ pNIR´VISqˆ pNIR ` VISq´1, (1)

where VIS and NIR stand for the spectral reflectance measurements acquired in the visible andnear-infrared regions, respectively. For this purpose, the QGIS 2.12.2 V Lyon software (Boston MA,USA) was used, generating two maps. The first (pre-eradication) was created using a QuickBird satelliteimage with a resolution of 60 cm per pixel, dated 11 May 2008. The second map (post-eradication)was generated using an image WorldView2, 50 cm per pixel resolution dated 9 May 2013. The imageswere geometrically corrected and aligned to a cartographic map projection (Universal TransverseMercator UTM, Datum WGS84, Zone 12 Q). Due to the use of sensors with different spatial and spectralresolutions, a nearest neighbor resampling method was applied. The final resolution was 60 cm. Boththe algorithm and resolution were chosen to keep as much of the original numerical information of theWorkdView2 image as possible. The change between the two dates was determined by subtractingthe 2008 image NDVI raster pixel image values from the image of 2013. A Boolean map was createdconsidering only differences exceeding 0.2 [45].

Forests 2016, 7, 91 4 of 13

2.3. Field Assessment of Vegetation

To assess the changes in vegetation cover in different habitats of the island, we selected20 transects strategically distributed in the eastern portion of the island (the area with the mostsheep abundance, determined by aerial monitoring in 2009, Figure 1). There were seven transects inforest, six in mixed scrub, and seven in eroded surface (type of habitat based on the classification ofLeón de la Luz et al. [39]). Monitoring of the island was performed in two periods (2009 and 2013). Dueto the environment heterogeneity, transects of 10 m ˆ 100 m were established in the more disturbedarea. This method was useful for including all the plant types from the emergent herbs to trees [46].In 2009, it was possible to record the whole coverage of the plant species and the total number ofindividuals per unit area in each transect. Nevertheless, in 2013, the number of individuals andcoverage was so extensive that it was necessary to make 40 subsamples of 2 m ˆ 2 m within eachtransect. To identify pioneer species in eroded soils, all plants were identified and counted, forming thedatabase of online information. Results were compared by analysis of variance with repeated measures(rANOVA) to determine significant differences in vegetation cover and the number of species betweenthe years of study.

2.4. Soil Compaction

In 2013, soil penetration resistance measurements were taken by using a penetrometer (SoilCompaction Tester Dickey-john®) within the 20 sampling transects (Figure 1). Eroded soil sites and50% and 100% recovered vegetation cover were chosen; 50 readings in total were obtained in eachcondition. In addition, 50 readings out of transects were taken at sites that were not disturbed bythe sheep (ND = non-disturbed), with 100% vegetation coverage [47], making a total of 200 readingsamong all conditions. Walter and Levin [24] observed this condition (sites without sheep), which wecorroborated when performing the aerial monitoring of sheep. An analysis of variance and Tukey’shonest significance test were performed to determine significant differences among the differentdegrees of vegetation cover.

2.5. Physical-Chemical Soil Parameters

Soil samples, approximately 1 kg each, were collected from each transect at a depth of 0–10 cm:16 of them on eroded soils, 16 on soils with 50%, and 16 on soils with 100% vegetation cover recovered.Samples were also obtained outside transects in places that were not disturbed by sheep with 100%vegetation cover (100% ND). The following physical-chemical parameters were determined: pH andelectrical conductivity (for reading potentiometer and conductivity, respectively); total nitrogen bythe Dumas method in a LECO nitrogen analyzer; organic matter by the method of Walkley-Black;phosphorus by colorimetric reading of a spectrophotometer, and calcium and magnesium by the EDTAmethod [48–50]. The results were compared by analysis of variance, and comparisons of means wereperformed by Tukey’s test to determine similarity and differences among sites.

3. Results

3.1. NDVI

The results obtained by the NDVI indicated that, in 2013, healthier vegetation was observedcompared to 2008. The multi-temporal analysis performed by NDVI is shown in Figure 2, where redindicates bare ground (values close to zero), orange-yellow identifies senescent vegetation (NDVIvalues from 0.1 to 0.4), while bluish-green indicates vigorous vegetation (values above 0.5).

In order to perform a detailed analysis of the results, we divided the island in four portions(Figure 3). The results of the average values obtained in 2008 and 2013 (Table 1) indicated major valuesof NDVI in the eastern portion of the island, while the values in the south and southwest portionswere lower. However, the biggest difference or vegetation recovery seems to have occurred preciselyin the eastern portion.

Forests 2016, 7, 91 5 of 13

Forests 2016, 7, 91  5 of 14 

indicates bare ground  (values close  to zero), orange‐yellow  identifies senescent vegetation  (NDVI 

values from 0.1 to 0.4), while bluish‐green indicates vigorous vegetation (values above 0.5).   

In order  to perform a detailed analysis of  the results, we divided  the  island  in  four portions 

(Figure 3). The  results of  the average values obtained  in 2008 and 2013  (Table 1)  indicated major 

values of NDVI  in  the eastern portion of  the  island, while  the values  in  the south and southwest 

portions were lower. However, the biggest difference or vegetation recovery seems to have occurred 

precisely in the eastern portion. 

 

Figure  2. Maps  of  the  normalized  difference  vegetation  index  (NDVI)  obtained  from  images  of 

Socorro Island. Left: QuickBird, 11 May 2008; Right: WorldView 2, 9 May 2013.   

Table 1. Average values of NDVI in 2008 and 2013. 

Portions of the island    2008 2013 Difference 

Northwest (NW)  0.443  0.501  0.058 

Northeast (NE)  0.461  0.612  0.150 

Southwest (SW)  0.394  0.499  0.105 

Southeast (SE)  0.346  0.533  0.187 

To view the change in vegetation, a Boolean map (Figure 3) was generated by the subtraction of 

the images of 2008 from 2013. Red indicates the sites where there was a positive difference (NDVI > 

0.2), i.e. an increase in vegetation cover in 2013. Calculations showed a difference of 1452 ha, which 

is equivalent to the vegetation recovery in 11% of the island surface. The eastern part of the island 

was the area with the highest habitat disturbance [22], and where the greatest vegetation recovery 

seemed to have occurred in the analyzed period. 

Figure 2. Maps of the normalized difference vegetation index (NDVI) obtained from images of SocorroIsland. Left: QuickBird, 11 May 2008; Right: WorldView 2, 9 May 2013.

Forests 2016, 7, 91  6 of 14 

 

Figure 3. Map of recovery and increase of vegetation cover in 2013 (red). Dashed lines indicate the 

portions of the island to which reference is made in Table 1. 

3.2.Field Assessment of Vegetation 

When doing field verification, we found that the outcome of the satellite images corresponded 

to  the actual habitat condition  (Figure 4). Due  to  the presence of sheep, most studied sites  lacked 

vegetation, and few species were present in 2009 (Table 2). Additionally, trails made by the sheep 

were observed with compacted soils. 

Statistical tests showed significant differences from 2009 to 2013 in the number of species present 

in the eroded sites as well as in coverage percentage. As shown in Table 2, it was possible to record 

an  evident  recovery  in  all of  the habitats  in  2013,  i.e.,  the  forest habitat with  the highest values, 

followed by the mixed scrub, and then the eroded surface. 

Figure 3. Map of recovery and increase of vegetation cover in 2013 (red). Dashed lines indicate theportions of the island to which reference is made in Table 1.

To view the change in vegetation, a Boolean map (Figure 3) was generated by the subtraction ofthe images of 2008 from 2013. Red indicates the sites where there was a positive difference (NDVI > 0.2),i.e., an increase in vegetation cover in 2013. Calculations showed a difference of 1452 ha, which isequivalent to the vegetation recovery in 11% of the island surface. The eastern part of the island wasthe area with the highest habitat disturbance [22], and where the greatest vegetation recovery seemedto have occurred in the analyzed period.

Forests 2016, 7, 91 6 of 13

Table 1. Average values of NDVI in 2008 and 2013.

Portions of the Island 2008 2013 Difference

Northwest (NW) 0.443 0.501 0.058Northeast (NE) 0.461 0.612 0.150Southwest (SW) 0.394 0.499 0.105Southeast (SE) 0.346 0.533 0.187

3.2. Field Assessment of Vegetation

When doing field verification, we found that the outcome of the satellite images correspondedto the actual habitat condition (Figure 4). Due to the presence of sheep, most studied sites lackedvegetation, and few species were present in 2009 (Table 2). Additionally, trails made by the sheep wereobserved with compacted soils.

Statistical tests showed significant differences from 2009 to 2013 in the number of species presentin the eroded sites as well as in coverage percentage. As shown in Table 2, it was possible to record anevident recovery in all of the habitats in 2013, i.e., the forest habitat with the highest values, followedby the mixed scrub, and then the eroded surface.

Forests 2016, 7, 91  7 of 14 

Figure 4. Forest recovery. Left: Site eroded in 2009. Right: Recovery of vegetation in 2013. 

Table 2. Vegetation cover in the 20 transects studied in 2009 and 2013.   

Type of habitat No. of plant species

p Vegetation cover (%) 

p 2009  2013  2009  2013 

Forest  8.86 (± 4.41)  15.00 (± 2.31)  *  29.29 (± 8.38)  85.00 (± 9.24)  ** 

Mixed scrub  6.67 (± 2.73)  13.67 (± 4.23)  *  19.17 (± 2.99)  72.50 (± 11.54)  ** 

Eroded surface  4.71 (3.45)  8.29 (± 2.56)  **  14.29(± 4.27)  50.29 (± 16.13)  ** 

Type of habitat based on the classification of León de la Luz et al. [39]. *, p < 0.005; **, p < 0.001 

After sheep eradication from the island, among the pioneer plants that began the colonization 

of eroded sites, not only were native and endemic species present, but the exotic ones were as well. 

Shrubs  and  herbs  were  the  most  successful  life  forms.  The  results  indicated  that  there  were 

differences  among  the  northeast‐southeast  zones  of  the  sampling  area. Native  species  such  as 

Pteridium caudatum (L.) Maxon and Dodonaea viscosa Jacq. were those with higher densities and the 

most dominant species in the sampling area. Hyptis pectinata (L.) Poit. was one of the most prolific 

exotic species (Table 3).   

Table 3. Maximum density and life forms of the plant species of major occurrence in the colonization 

of eroded soil in transects in 2009 and 2013. A p < 0.001 was observed between 2009 and 2013 for all 

the species. (ind ha−1) = individuals per hectare. 

    Zones within the sampling area 

Species  Life form 

Northeast Southeast 

ind ha−1Increase(%) 

ind ha−1 Increase(%) 

2009 2013 2009  2013 

Hyptis pectinata (L.) Poit. (x)  Shrub  60  210  350  330  4860  1473 

Dodonaea viscosa Jacq. (n)  Shrub  910  6190  680  140  1370  979 

Pteridium caudatum (L.) Maxon (n)*  Herbs  6%  58%  967  11%  43%  391 

Nicotiana stocktonii Brandegee (e)  Herbs  30  190  633  40  260  650 

Erigeron socorrensis Brandegee (e)  Herbs  240  1420  592  80  350  437 

Mitracarpus hirtus (L.) DC. (x)  Herbs  80  440  550  20  90  450 

Aristida spp. (x)^*  Herbs  1%  3%  300  6%  31%  517 

Amaranthus palmeri S.Watson (n)  Sub‐shrub  220  900  409  100  180  180 

Psidium socorrense I.M.Johnst. (e)  Tree  50  190  380  30  100  333 

Chamaesyce spp. (x)^  Herbs  70  160  229  100  370  370 

Cenchrus spp. (x)^  Herbs  40  110  275  160  530  331 

Figure 4. Forest recovery. Left: Site eroded in 2009. Right: Recovery of vegetation in 2013.

Table 2. Vegetation cover in the 20 transects studied in 2009 and 2013.

Type of Habitat No. of Plant Species p Vegetation Cover (%) p2009 2013 2009 2013

Forest 8.86 (˘ 4.41) 15.00 (˘ 2.31) * 29.29 (˘ 8.38) 85.00 (˘ 9.24) **Mixed scrub 6.67 (˘ 2.73) 13.67 (˘ 4.23) * 19.17 (˘ 2.99) 72.50 (˘ 11.54) **

Eroded surface 4.71 (3.45) 8.29 (˘ 2.56) ** 14.29(˘ 4.27) 50.29 (˘ 16.13) **

Type of habitat based on the classification of León de la Luz et al. [39]. *, p < 0.005; **, p < 0.001.

After sheep eradication from the island, among the pioneer plants that began the colonizationof eroded sites, not only were native and endemic species present, but the exotic ones were as well.Shrubs and herbs were the most successful life forms. The results indicated that there were differencesamong the northeast-southeast zones of the sampling area. Native species such as Pteridium caudatum

Forests 2016, 7, 91 7 of 13

(L.) Maxon and Dodonaea viscosa Jacq. were those with higher densities and the most dominant speciesin the sampling area. Hyptis pectinata (L.) Poit. was one of the most prolific exotic species (Table 3).

Table 3. Maximum density and life forms of the plant species of major occurrence in the colonizationof eroded soil in transects in 2009 and 2013. A p < 0.001 was observed between 2009 and 2013 for all thespecies. (ind ha´1) = individuals per hectare.

Species Life Form

Zones within the Sampling Area

Northeast Southeast

ind ha´1Increase(%)

ind ha´1Increase(%)

2009 2013 2009 2013

Hyptis pectinata (L.) Poit. (x) Shrub 60 210 350 330 4860 1473Dodonaea viscosa Jacq. (n) Shrub 910 6190 680 140 1370 979

Pteridium caudatum (L.) Maxon (n)* Herbs 6% 58% 967 11% 43% 391Nicotiana stocktonii Brandegee (e) Herbs 30 190 633 40 260 650Erigeron socorrensis Brandegee (e) Herbs 240 1420 592 80 350 437

Mitracarpus hirtus (L.) DC. (x) Herbs 80 440 550 20 90 450Aristida spp. (x)ˆ* Herbs 1% 3% 300 6% 31% 517

Amaranthus palmeri S.Watson (n) Sub-shrub 220 900 409 100 180 180Psidium socorrense I.M.Johnst. (e) Tree 50 190 380 30 100 333

Chamaesyce spp. (x)ˆ Herbs 70 160 229 100 370 370Cenchrus spp. (x)ˆ Herbs 40 110 275 160 530 331

Guettarda insularis Brandegee (e) Tree 80 220 275 170 320 188Perityle socorrosensis Rose (e) Herbs 390 780 200 310 620 200

Type of life-form based on Levin and Moran [41]; * Due to the impossibility of accounting for individualplants, only % coverage per hectare was considered; only ˆ are located close to the paths on the island transects;n = native; e = endemic; x = exotic.

3.3. Soil Compaction

The results of the study of soil compaction showed that eroded soils were the most compactedand had become less compacted as vegetation began to recover (50% and 100% recovered vegetationcover). Transects that retained eroded soils (0% vegetation) showed greater penetration resistance(>300 pounds-force per square inch, or psi, to 12 inches deep), resulting from the rams [51–53]. Inrelation to sites with 50% and 100% recovered vegetation cover, soils were very compacted and shallow(100–120 psi to three inches deep) and became more compacted at greater depths (300 psi to 24 inchesdeep). At sites with 100% coverage without disturbance (ND), the soil was kept with little variation(230–300 psi until 21 inches depth), which was in the range of values conducive for the growth of mostplants from 200–400 psi to 24 in [54], which could be due to the constant stable condition. Significantdifferences (p < 0.001) were observed among sites with 0% and those with 50% and 100% recoveredvegetation cover. Significant differences were found between sites with 50% and 100% recoveredcoverage compared to undisturbed sites (100% ND, Figure 5).

Forests 2016, 7, 91  8 of 14 

Guettarda insularis Brandegee (e)  Tree  80  220  275  170  320  188 

Perityle socorrosensis Rose (e)  Herbs  390  780  200  310  620  200 

Type  of  life‐form  based  on  Levin  and Moran  [41];  * Due  to  the  impossibility  of  accounting  for 

individual plants, only % coverage per hectare was considered; only ^ are located close to the paths 

on the island transects; n = native; e = endemic; x = exotic 

3.3. Soil Compaction 

The results of the study of soil compaction showed that eroded soils were the most compacted 

and had become less compacted as vegetation began to recover (50% and 100% recovered vegetation 

cover). Transects that retained eroded soils (0% vegetation) showed greater penetration resistance 

(>300 pounds‐force per square  inch, or psi, to 12  inches deep), resulting from the rams [51–53]. In 

relation  to  sites with  50%  and  100%  recovered vegetation  cover,  soils were very  compacted  and 

shallow (100–120 psi to three inches deep) and became more compacted at greater depths (300 psi to 

24 inches deep). At sites with 100% coverage without disturbance (ND), the soil was kept with little 

variation  (230–300 psi until 21  inches depth), which was  in  the range of values conducive  for  the 

growth of most plants  from 200–400 psi  to 24  in  [54], which  could be due  to  the  constant  stable 

condition. Significant differences (p < 0.001) were observed among sites with 0% and those with 50% 

and 100% recovered vegetation cover. Significant differences were found between sites with 50% and 

100% recovered coverage compared to undisturbed sites (100% ND, Figure 5). 

 

Figure 5. Soil compaction in eroded areas and areas without disturbance (ND). 

3.4. Physical‐Chemical Soil Parameters 

The  results  of  physical‐chemical  analyses  of  soil  samples  are  shown  in  Table  4;  pH  values 

remained  close  to  neutral,  showing  a  significant  difference  (p  <  0.021)  between  sites  without 

vegetation and 50% vegetation cover. Electrical conductivity, which is an indicator of salt presence in 

soil, also had significant differences (p < 0.013) between the eroded and 100% vegetation cover sites, 

although no difference was observed between eroded soils and those that were not disturbed.   

In  the  case of  total nitrogen, organic  carbon, phosphorus, and  calcium,  sites with  recovered 

vegetation were significantly different (p < 0.001) than those with erosion. Both nitrogen and organic 

carbon  doubled, while  phosphorus  and  calcium  values  almost  tripled  in  places with  increased 

vegetation cover regarding the eroded sites.   

Meanwhile, magnesium showed significant differences among the eroded sites (0%, 50%, and 

100% recovered vegetation cover) and undisturbed sites (100% ND). The sites that were never altered 

by the presence of sheep exhibited a concentration twice that of disturbed sites.   

 

Figure 5. Soil compaction in eroded areas and areas without disturbance (ND).

Forests 2016, 7, 91 8 of 13

3.4. Physical-Chemical Soil Parameters

The results of physical-chemical analyses of soil samples are shown in Table 4; pH values remainedclose to neutral, showing a significant difference (p < 0.021) between sites without vegetation and50% vegetation cover. Electrical conductivity, which is an indicator of salt presence in soil, also hadsignificant differences (p < 0.013) between the eroded and 100% vegetation cover sites, although nodifference was observed between eroded soils and those that were not disturbed.

In the case of total nitrogen, organic carbon, phosphorus, and calcium, sites with recoveredvegetation were significantly different (p < 0.001) than those with erosion. Both nitrogen and organiccarbon doubled, while phosphorus and calcium values almost tripled in places with increasedvegetation cover regarding the eroded sites.

Meanwhile, magnesium showed significant differences among the eroded sites (0%, 50%, and100% recovered vegetation cover) and undisturbed sites (100% ND). The sites that were never alteredby the presence of sheep exhibited a concentration twice that of disturbed sites.

Table 4. Results of soil physical-chemical analyses.

Vegetation Cover

0% 50% 100% 100% ND

pH 7.20 (0.16)a 6.99 (0.28)b 7.03 (0.16)a,b 7.11 (0.22)a,b

Electrical conductivity (dS m´1) 0.09 (0.03)a 0.07 (0.03)a 0.07 (0.02)b 0.09 (0.03)a

Total nitrogen (%) 0.20 (0.07)a 0.45 (0.11)b 0.48 (0.14)b 0.51 (0.17)b

Organic carbon (%) 1.53 (0.52)a 3.60 (0.86)b 3.22 (0.78)b 3.27 (1.08)b

Phosphorus (meq 100g´1) 0.01 (0.01)a 0.03 (0.02)b 0.03 (0.02)b 0.03 (0.02)b

Calcium (meq 100g´1) 0.03 (0.01)a 0.08 (0.04)b 0.05 (0.02)b 0.09 (0.03)b

Magnesium (meq 100g´1) 0.10 (0.04)a 0.10 (0.03)a 0.11 (0.03)a 0.20 (0.10)b

ND = Non-disturbed area with little or no alteration by sheep. The numbers in parenthesis are the standarddeviation; values followed by different lowercase letter (a, b) indicate that means are significantly differentamong eroded and non-eroded conditions. The correspondent p values for each case is specified in the text.

4. Discussion

4.1. NDVI and Field Assessment of Vegetation

Remote sensing allows for easy analysis of extended and/or complex areas difficult to access. Inour study, NDVI was a useful tool to evaluate changes in photosynthetic vegetation along time,especially in an island with a very complex topography with plenty of areas difficult to reach.The results of this analysis support those obtained in field assessment in a more quantitative way,specifically the increment in vegetal coverage.

We found differences not only in the number of species and vegetation cover in the samplingarea between 2009 and 2013 but also in the habitats located along the northeast-southeast zones ofthe sampling area. The forests and mixed scrubs in the southeast zone showed the greatest recovery,probably favored by their vegetal components (Tables 2 and 3) and the permanence of seed banksbecause of a more stable landscape in addition to water availability due to precipitation patterns; theendemic tree species recovering were Guettarda insularis and Psidium socorrense. The smaller number ofplant species found in the isolated patches of mixed scrubs included in large expanses of erosion in thenortheast zone could be due to slope steepness and wind exposure in this area. Gravity would makethe permanence of naturally occurring soil seed bank difficult.

Some species of exotic grasses have also been favored with sheep eradication because they areno longer grazed on. Species of the genera Aristida and Cenchrus are now more abundant near roadswhere vehicles can travel on the island.

Herbivores can affect plant species diversity by modifying local extinction or colonization rates, orboth. The question whether herbivores increase or decrease plant diversity over a productivity gradient

Forests 2016, 7, 91 9 of 13

thus translates into the question of how herbivores impact extinction and colonization rate changeswith productivity [55]. In the case of Socorro Island, productivity can be qualitatively evaluated basedon the NDVI analysis results because it shows the increment in vegetal coverage, i.e. in biomass. Theseresults evidence the strong negative impact that sheep had on vegetation due to their permanenceon the island.

The introduction of large grazers is crucial for plant diversity worldwide. Nutritional qualityof plants growing in poor soils is low, so large herbivores require a lot of forage to feed on. Largeherbivorous should not be introduced everywhere, as they can have negative effects on diversity,especially in unproductive areas or areas that have no recent evolutionary history of abundant largeherbivore grazing [56], as in the case of Socorro Island.

The ecological resilience corresponds to the degree to which the system would have to bealtered before it begins to reorganize itself around another set of processes. Small and fast scales aredominated by biophysical processes that control plant physiology and morphology [34]. In the caseof Socorro Island, a fragile ecosystem with endemism and no naturally occurring large herbivores,sheep disturbance elimination evidences the importance of plant composition in the recovery of certainspecies. Some endemic species such as Erigeron socorrensis and Perityle socorrensis have successfullycolonized eroded soil, even taking advantage of microhabitats created by H. pectinata and D. viscosa,which could be due to the long-term evolution in isolation and adaptation to an exceedingly harshsubstrate and climate, as established by Walter and Levin [24].

At the larger and slower scale of patch dynamics, interspecific plant competition for nutrients,light, and water influences local species composition and regeneration [34]. Some plants colonizingeroded soils could become very aggressive, likely preventing biodiversity recovery. Such could be thecase of Pteridium caudatum, a successful fern species that thrives worldwide in disturbed sites and indifferent climates and soil types, which, in 2013, had covered extended surfaces in Socorro Island.

Finally, at a still larger scale of stands in a forest, mesoscale processes determine structure andsuccessional dynamics from tens of meters to kilometers, and from years to decades [34]. As Walterand Levin [24] argued, our results support the possibility that sheep removal would be followed bythe eventual return of the former island vegetation, which seems possible given that only a year afterthe eradication of feral sheep concluded, the natural ecosystems of Socorro Island showed signs ofpassively recovering.

4.2. Soil Recovery

The changes in soil physical-chemical properties in Socorro Island seem to be related to thegradual recovery of vegetation after the eradication of feral sheep. From 2009, the first plants tocolonize disturbed areas (Table 3) were detected. Prostrate Chamaesyce sp. and E. socorrensis have beenobserved to have a great capacity to retain soil. H. pectinata and P. caudatum settled in high densities;in addition to retaining soil, they have generated much organic matter. Possibly the most successfulspecies to colonize disturbed areas has been D. viscosa, which has a great ability to germinate in erodedsoil [40], generating organic matter and preventing the germination of other species [57]. In the absenceof trampling, soil aggregate stability increases, which enhances infiltration, reduces erosion, and maypromote nutrient accumulation and soil retention [58,59].

As pioneer plants began to establish, the ground became less compacted because the roots ofplants, particularly annual grasses in these stages, act as biological perforators, also incorporatingorganic matter into the soil; once the roots die and shrink, these pores are large enough to allow theroots of perennial shrubs to penetrate [60]. Greater ease of water movement in the soil matrix, coupledwith heavy rainfall, could be causing leaching and replacing cations with H+ ions, acidifying the soil.Another cause may be nutrient absorption by plant roots, decreasing the concentration of basic cations(Ca, Mg, K, and Na). In the case of electrical conductivity, the average values from 0.07 and 0.09 dS m´1

were not considered limiting for normal plant development [61].

Forests 2016, 7, 91 10 of 13

Both the results obtained with the NDVI calculation and field observations suggested thatsome pioneer plants had the ability to germinate on eroded soils and were instrumental in thesuccession process by providing the right conditions such that tree species could germinate. Theprogressive increase in vegetation cover reduces soil compaction and restores the biogeochemicalcycles of essential nutrients, such as nitrogen, phosphorus, and calcium, which are essential for therecovery of communities and the ecosystem in general, as well as the incorporation of carbon on theground, which is essential for the proper functioning of important microbiological component.

Any change in the habitat that produces changes in litter production, soil aeration, or any otherfactor affecting microorganisms will be reflected in changes in biogeochemical cycles, such as those ofcarbon and nitrogen [62]. The recovery in these important cycles on the island has been allowing theestablishment of species that need these elements available in the soil in order to establish. Althoughthe significant difference in the results of organic carbon and nitrogen seems to correspond to theobserved recovery, it is important to mention that it might be due to the high values in the standarddeviations obtained for all the vegetation cover conditions.

The magnesium content can be explained as a result of the weathering of mafic rocks, which arerich in magnesium and iron [63] and not by the litter of pioneer plants that have gained a footholdin poor soils.

5. Conclusions

Some studies have discussed the importance of herbivorous species of small or large size, havingdiscrepancy as to the negative or positive effect they can have [55]. For this reason, large herbivoresshould not be introduced in environments where environmental conditions of precipitation andhumidity are not balanced, as in the case of Socorro Island.

Removing the exotic herbivorous species from the island is a conservation tool [7], which is evidentin its recovery contribution of the natural environment. Habitat fragmentation and degradation causedby the presence of invasive species (Ovis aries) was evident on the island, whose main involvementwas on the ground and in vegetation. The resistance of native species has been transcendental not onlyin the relatively rapid recovery of the vegetation cover but also in offering the possibility of recoveringthe former island vegetation.

The results reflect the important role of vegetation in erosion control, both for establishingmechanical support due to its roots in the soil structure and in capturing water flow andnutrients, providing fresh organic matter to the soil, thus restoring biogeochemical cycles andecosystem processes.

The time after sheep removal has so far been too short to determine ecosystem behavior, but it isclear that the system had already started changing by 2013. With the habitat recovery, wildlife recoveryis expected as food availability and resources for the native species of the island gradually increase.

The results of this study seem to support our hypothesis that Socorro Island is resilient enoughto recover in a relatively small time scale, after the removal of the pressures caused by a largeexotic herbivore.

Acknowledgments: The authors thank CIBNOR staff Gil Ceseña for advice and support in the GIS Laboratory;Manuel Trasviña and Myriam Hernandez for advice and support in the Soil Science Laboratory; Diana Dorantesfor editorial services in English; the staff of Grupo de Ecología y Conservación de Islas, A.C. for support duringthe expeditions to Socorro Island; and two anonymous reviewers for their critical and helpful comments. Thiswork has been funded by various organizations, including the National Network of Protected Areas RENANP ofCONACYT, grant 259540, Grupo de Ecología y Conservación de Islas, A.C., and the WWF–Carlos Slim FoundationAlliance, grant Gca 2014-11.

Author Contributions: Alfredo Ortega-Rubio, Antonio Ortiz-Alcaraz, and Yolanda Maya-Delgado conceived anddesigned the experiment; Alfonso Aguirre-Muñoz, and Evaristo Rojas-Mayoral contributed with reagents andmaterials, statistical analysis and discussion; Antonio Ortiz-Alcaraz and Patricia Cortés-Calva analyzed the data;Alfredo Ortega-Rubio and Miguel V. Cordoba-Matson contributed with analysis tools; Antonio Ortiz-Alcaraz,Yolanda Maya-Delgado, and Patricia Cortés-Calva wrote the paper.

Conflicts of Interest: The authors declare no conflict of interest.

Forests 2016, 7, 91 11 of 13

References

1. Mason, H.L. The edaphic factor in narrow endemism. II. The Geographic occurrence of plants of highlyrestricted patterns of distribution. Madroño 1946, 8, 241–257.

2. Tilman, D.; May, R.M.; Lehman, C.L.; Nowak, M.A. Habitat destruction and the extinction debt. Nature 1994,371, 65–66. [CrossRef]

3. Coblentz, B.E. The effects of feral goat (Capra hircus) on islands ecosystems. Biol. Cons. 1978, 13, 279–286.[CrossRef]

4. Eberhard, T. Introduced birds and mammals and their ecological effects. Swed. Wildl. Res. 1988, 13, 1–107.5. Rosenberg, D.K. The impact of introduced herbivores on the Galapagos rail (Laterallus spilonotus).

Monogr. Syst. Bot. Missour.i Bot. Gard. 1990, 32, 169–178.6. Mooney, H.A.; Hobbs, R.J. Invasive Species in a Changing World; Island Press: Washington, DC, USA, 2000.7. Donlan, C.J.; Tershy, B.R.; Croll, D.A. Islands and introduced herbivores: Conservation action as ecosystem

experimentation. J. Appl. Ecol. 2002, 39, 235–246. [CrossRef]8. Primack, R.B. Essentials of Conservation Biology; Sinauer Associates Press: Sunderland, MA, USA, 2002.9. Veitch, C.R.; Clout, M.N. Turning the Tide: The Eradication of Invasive Species; IUCN, SSC Invasive Species

Specialist Group: Gland, Switzerland, 2002; pp. 374–380.10. Courchamp, F.; Chapuis, J.L.; Pascal, M. Mammal invaders on islands: Impact, control, and control impact.

Biol. Rev. 2003, 78, 347–383. [CrossRef] [PubMed]11. Nogales, M.; Martín, A.; Tershy, B.R.; Donlan, C.J.; Veitch, D.; Puerta, N.; Wood, B.; Alonso, J. A Review of

feral cat eradication on islands. Cons. Biol. 2004, 18, 310–319. [CrossRef]12. Campbell, K.; Donlan, C.J. Feral goat eradication on islands. Cons. Biol. 2005, 19, 1362–1374. [CrossRef]13. Pimentel, D.; Zúñiga, R.; Morrison, D. Update on the environmental and economic costs associated with

alien invasive species in the United States. Ecol. Econ. 2005, 52, 273–288. [CrossRef]14. Aguirre-Muñoz, A.; Samaniego-Herrera, A.; Luna-Mendoza, L.; Ortiz-Alcaraz, A.; Rodríguez-Malagón, M.;

Méndez-Sánchez, F.; Félix-Lizárraga, M.; Hernández-Montoya, J.; González-Gómez, R.; Torres-García, F.; et al.Island restoration in Mexico: Ecological outcomes after systematic eradications of invasive mammals. InIsland Invasives: Eradication and Management; Veitch, C.R., Clout, M.N., Towns, D.R., Eds.; IUCN: Gland,Switzerland, 2011; pp. 250–258.

15. MacArthur, R.H.; Wilson, E.O. The Theory of Island Biogeography. Vol 1; Princeton University Press: Princeton,NJ, USA, 1967.

16. Álvarez-Castañeda, S.T.; Ortega-Rubio, A. Current status of rodents on islands of the Gulf of California.Biol. Cons. 2003, 109, 157–163. [CrossRef]

17. Blumstein, D.T.; Daniel, J.C. The loss of anti-predator behavior following isolation on islands. Proc. R. Soc. BBiol. Sci. 2005, 272, 1663–1668. [CrossRef] [PubMed]

18. Cortés-Calva, P.; Gallo-Reynoso, J.P.; Delgadillo-Rodríguez, J.; Lorenzo, C.; Álvarez-Castañeda, S.T. Theeffect of feral dogs and other alien species on native mammals of Isla de Cedros, Mexico. Nat. Areas J. 2013,33, 466–473. [CrossRef]

19. Cincotta, P.R.; Wisnewski, J.; Engelman, R. Human population in the biodiversity hotspot. Nature 2000, 404,990–992. [CrossRef] [PubMed]

20. Whittaker, R.J.; Fernandez-Palacios, J.M. Island Biogeography; Oxford University Press: Oxford, UK, 2007.21. Donlan, C.J.; Croll, D.A.; Tershy, B.R. Islands, exotic herbivores, and invasive plants: Their roles in Coastal

California restauration. Res. Ecol. 2003, 11, 524–530. [CrossRef]22. Álvarez-Cárdenas, S.; Castellanos, A.; Galina, P.; Ortega-Rubio, A.; Arnaud, G. Aspectos de la población y el

hábitat del borrego doméstico (Ovis aries). In La Isla Socorro, Reserva de la Biosfera Archipiélago de Revillagigedo,México; Ortega Rubio, A., Castellanos, A., Eds.; Centro de Investigaciones Biológicas del Noroeste S.C.:La Paz, Mexico, 1994; pp. 301–317.

23. Villa, B. La Isla Socorro. Vertebrados terrestres. Monogr. Inst. Geofís. 1960, 2, 203–2016.24. Walter, H.S.; Levin, G.A. Feral sheep on Socorro Island: Facilitators of alien plant colonization and ecosystem

decay. Divers. Distrib. 2008, 14, 422–431. [CrossRef]25. Richards, A.F.; Brattstrom, B.H. Bibliography, cartography, discovery and exploration of the islas

Revillagigedo. Proc. Calif. Acad. Sci. 1959, 29, 315–360.

Forests 2016, 7, 91 12 of 13

26. Veitch, C.R. The Eradication of Cats and Sheep from Socorro Island: A Report to the Socorro Island Project; NorthernRegion Technical Report Series No. 11; Department of Conservation: Auckland, New Zealand, 1989.

27. Castellanos, A.; Arnaud, G.; Álvarez, S.; Rodríguez, R.; Galina, P. El borrego doméstico (Ovis aries) en islaSocorro, su situación actual e impacto en el suelo, la vegetación y la fauna nativa. In Proceedings of the XICongreso Nacional de Zoología, Mérida, Mexico, 28–31 October 1991.

28. Ortega, A.; Castellanos, A.; Arnaud, G.; Maya, Y.; Rodríguez, R.; León, J.L.; Cancino, J.; Jiménez, C.; Llinas, J.;Álvarez, S.; et al. Estudio de los Recursos Naturales de la Isla Socorro, Revillagigedo. Ciencia 1992, 43,175–184.

29. Castellanos, A.; Rodríguez-Estrella, R. Current status of the Socorro Mockingbird. Wilson Bull. 1993, 105,167–171.

30. Galina, P.; Álvarez, S.; Ortega Rubio, A. Aspectos ecológicos de la herpetofauna. In La Isla Socorro, Reserva de laBiosfera Archipiélago de Revillagigedo, México; Ortega Rubio, A., Castellanos, A., Eds.; Centro de InvestigacionesBiológicas del Noroeste S.C.: La Paz, Mexico, 1994; pp. 185–197.

31. Rodríguez-Estrella, R.; Rivera-Rodríguez, L.; Mata, E. Avifauna Terrestre. In La Isla Socorro, Reserva de laBiosfera Archipiélago de Revillagigedo, México; Ortega Rubio, A., Castellanos, A., Eds.; Centro de InvestigacionesBiológicas del Noroeste S.C.: La Paz, Mexico, 1994; pp. 199–224.

32. Maya-Delgado, Y.; Salinas-Zavala, F.; Troyo-Diéguez, E. Estado actual del suelo y propuestas para suconservación. In La Isla Socorro, Reserva de la Biosfera Archipiélago de Revillagigedo, México; Ortega Rubio, A.,Castellanos, A., Eds.; Centro de Investigaciones Biológicas del Noroeste S.C.: La Paz, Mexico, 1994; pp. 63–75.

33. Adema, E.O.; Babinec, F.J.; Peinemann, N. Pérdida de nutrientes por erosión hídrica en dos suelos delcardenal pampeano. Cienc. Suelo 2001, 19, 144–154.

34. Peterson, G.; Allen, C.R.; Holling, C.S. Ecological resilience, biodiversity, and scale. Ecosystems 1998, 1, 6–18.[CrossRef]

35. Zavaleta, E.S.; Hobbs, R.J.; Mooney, H.A. Viewing invasive species removal in a whole-ecosystem context.Trends Ecol. Evol. 2001, 16, 454–459. [CrossRef]

36. Holl, K.D.; Aide, T.M. When and where to actively restore ecosystems? Forest Ecol. Manag. 2011, 261,1558–1563. [CrossRef]

37. Jones, H.P.; Schmitz, O.J. Rapid Recovery of Damaged Ecosystems. PLoS ONE 2009, 4, 53–56. [CrossRef][PubMed]

38. CONANP-SEMARNAT. Programa de Conservación y Manejo, Reserva de la Biósfera Archipiélago de Revillagigedo;Comisión Nacional de Áreas Naturales Protegidas: Mexico, D.F., Mexico, 2004.

39. León de la Luz, J.L.; Breceda, A.; Coria, R.; Cancino, J. Asociaciones vegetales. In La Isla Socorro, Reserva de laBiosfera Archipiélago de Revillagigedo, México; Ortega Rubio, A., Castellanos, A., Eds.; Centro de InvestigacionesBiológicas del Noroeste S.C.: La Paz, Mexico, 1994; pp. 115–141.

40. Campa-Molina, M.A. Flora y potencial apícola de Isla Socorro, Archipiélago de las Revillagigedo. Master’sThesis, Universidad de Guadalajara, Jalisco, Mexico, 1989.

41. Levin, G.A.; Moran, R. The Vascular Flora of Isla Socorro, Mexico; San Diego Society of Natural History:San Diego, CA, USA, 1989.

42. Hall, E.R. The Mammals of North America; Wiley Inter-Sciences Publications: New York, NY, USA, 1981; p. 191.43. Ortiz-Alcaraz, A.; Barredo-Barberena, J.M.; Aguirre-Muñoz, A.; Santos del Prado-Gasca, K. Acciones para

la Recuperación del Ambiente en Isla Socorro, Archipiélago de Revillagigedo. Reporte Final; Grupo de Ecología yConservación de Islas, A.C.-Instituto Nacional de Ecología: Mexico City, Mexico, 2011; p. 32.

44. Valdez-Lazalde, J.; González-Guillen, M.; de los Santos-Posadas, H. Estimación de cobertura arbóreamediante imágenes satelitales multiespectrales de alta resolución. Agrociencia 2006, 40, 383–394.

45. Cabral, F. Dinámica espaciotemporal de uso del suelo en la UMAFOR 12 del Estado de México mediante eluso de sistemas de información geográfica (1990–2007). Master’s Thesis, Universidad Autónoma Chapingo,Mexico, 2011.

46. Mostacedo, B.; Fredericksen, T.S. Manual de Métodos Básicos de Muestreo y Análisis en Ecología Vegetal; Proyectode Manejo Forestal Sostenible (BOLFOR): Santa Cruz, Bolivia, 2000.

47. Ortiz-Alcaraz, A.; Aguirre-Muñoz, A.; Méndez-Sánchez, F.; Ortega-Rubio, A. Feral sheep eradication atSocorro Island, México: A mandatory step to ensure ecological restoration. Interciencia 2016, 41, 184–189.

48. Jackson, M.L. Análisis químico de suelos, 4th ed.; Ediciones Omega S.A.: Barcelona, Spain, 1982; p. 652.49. Bates, R.G. Determination of pH. Theory and Practice; John Wiley and Sons: New York, NY, USA, 1964; pp. 11–12.

Forests 2016, 7, 91 13 of 13

50. Page, A.L.; Miller, R.H.; Keeney, D.R. Methods of Soil Analysis. Part 2, Chemical and Microbiological Properties,2nd ed.; American Society of Agronomy, Inc. and Soil Science Society of America, Inc: Wisconsin, WI, USA,1982; p. 1159.

51. Pinzón, A.; Amézquita, E. Compactación de suelos por pisoteo de animales en pastoreo en el piedemonteamazónico de Colombia. Pasturas tropicales 1979, 13, 21–26.

52. Riney, T. Study and management of large mammals; John Wiley & Sons Ltd.: Chichester, UK, 1982.53. González, M.; Nogues, E. Pisoteo animal y su efecto en la densidad aparente del suelo en un Haplustol

franco arenoso bajo diferentes manejos. Rev. Divulg. Técnic. Agríc. Agroind. 2012, 31, 1–9.54. Bakker, E.S.; Ritchie, M.E.; Olff, H.; Milchunas, D.G.; Knops, J.M.H. Herbivore impact on grassland plant

diversity depends on habitat productivity and herbivore size. Ecol. Lett. 2006, 9, 780–788. [CrossRef][PubMed]

55. Milchunas, D.G.; Sala, O.E.; Lauenroth, W.K. A generalized model of the effects of grazing by large herbivoreson grassland community structure. Am. Nat. 1988, 132, 87–106. [CrossRef]

56. Vibrans, H. Taller de identificación de Malezas. In Manejo de Malezas en México. Vol. 1 Maleza terrestre;Bojórquez-Bojórquez, G., Rosales-Robles, E., Zeta-Padilla, G., Eds.; Universidad Autónoma de Sinaloa:Culiacan, Mexico, 2011; pp. 293–308.

57. Castellano, M.J.; Valone, T.J. Livestock, soil compaction and water infiltration rate: Evaluating a potentialdesertification recovery mechanism. J. Arid Environ. 2007, 71, 97–108. [CrossRef]

58. Allington, G.R.H.; Valone, T.J. Reversal of desertification: The role of physical and chemical soil properties.J. Arid Environ. 2010, 74, 973–977. [CrossRef]

59. Amézquita, E.; Chávez, L.F. La compactación del suelo y sus efectos en la productividad de los suelos.In Proceedings of the Congreso Costarricense de la Ciencia del Suelo, San José, Costa Rica, 9 April 1999.

60. Sellés, G.; Fereyra, R.; Ruíz, R.; Ferreyra, R.; Ahumada, R. Compactación de Suelos y su Control: Estudio de Casosen el Valle de Aconcagua, Boletín INIA 234; Instituto de Investigaciones Agropecuarias: Santiago, Chile, 2012.

61. USDA. Soil Survey Laboratory Methods Manual. Soil Survey Investigations Report No. 42. Version 3.0; NaturalResources Conservation Service: Washington, DC, USA, 1996.

62. Hartmann, A.; Aßmus, B.; Kirchhof, G.; Schloter, M. Direct approaches for studying soil microbes. In ModernSoil Microbiology; van Elsas, J.D., Trevors, J.T., Wellington, E.M.H., Eds.; Marcel Dekker Inc.: New York, NY,USA, 1997; p. 712.

63. Mediavilla, M.J. Meteorización química. Enseñanza de las Ciencias de la Tierra; Universidad de Girona: Girona,Spain, 2014; pp. 230–238.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open accessarticle distributed under the terms and conditions of the Creative Commons Attribution(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).