Modeling land-use change and future deforestation in two spatial scales

 

Modelaje del cambio de uso del suelo y la deforestación futura en dos escalas espaciales

 

Carmina Cruz-Huerta¹; Manuel J. González-Guillén^1*; Tomás Martínez-Trinidad¹; Miguel J. Escalona-Maurice²

 

¹ Postgrado en Ciencias Forestales, Colegio de Postgraduados. Carretera México-Texcoco km 36.5. C. P. 56230. Montecillo, Texcoco, Edo. de México.

² Desarrollo Rural, Colegio de Postgraduados. Carretera México-Texcoco km 36.5. C. P. 56230. Montecillo, Texcoco, Edo. de México. Correo-e: manuelg@colpos.mx, Tel.: 595 95 202 00 ext. 1464 (*Autor para correspondencia).

 

Received: June 5, 2014.
Accepted: April 24, 2015.

 

ABSTRACT

Land-use changes in the Chignahuapan-Zacatlán region of the state of Puebla, Mexico, and in each of the two aforementioned counties, were determined using Landsat imagery and a bitemporal analysis of the period 1986-2011. The predominant land uses at the regional level are the agricultural (49.7 %) and forest (46.1) groupings. In the case of Chignahuapan county, agricultural use is the main activity with 58.9 % of the territory, while in Zacatlán county the predominant activity is forest use with 57.3 % of the land base. At the regional level and for Zacatlán county, the probabilistic model of forest land-use change was significantly correlated (P ≤ 0.05) with 21 independent variables; however, for Chignahuapan, the model only considered 16 variables. At the regional level, the probability of forest land changing to other uses ranged from 5-90 % and at the county level from 7-99.8 %. Finally, the projection for the year 2030 estimates that the deforestation risk at the regional level and in Chignahuapan and Zacatlán counties is 13,063.8, 10,966.6 and 4,405.5 ha, respectively.

Keywords: Deforestation patterns, Chignahuapan, Zacatlán, land uses.

 

RESUMEN

En este estudio se determinó el cambio de uso de suelo en la región de Chignahuapan- Zacatlán, Puebla, mediante un análisis bitemporal entre 1986 y 2011; la evaluación se hizo también para cada municipio. El cambio de uso de suelo se detectó mediante el manejo de imágenes Landsat. Los usos predominantes a nivel regional son el agrícola (49.7 %) y el forestal (46.1 %). En el caso de Chignahuapan, el uso agrícola representa la actividad principal con 58.9 % del territorio, mientras que en Zacatlán predomina el uso forestal con 57.3 %. A nivel regional y para el municipio de Zacatlán, el cambio de uso del suelo forestal se correlacionó significativamente (P ≤ 0.05) con 21 variables independientes, mediante un modelo probabilístico; en el caso de Chignahuapan, el modelo consideró 16 variables. A nivel regional, la probabilidad de cambio de uso de suelo forestal varió de 5 a 90 % y a nivel municipal de 7 a 99.8 %. Finalmente, la proyección para el año 2030 estima que el riesgo de deforestación a nivel regional y en los municipios de Chignahuapan y Zacatlán será de 13,063.8, 10,966.6 y 4,405.5 ha, respectivamente.

Palabras clave: Patrones de deforestación, Chignahuapan, Zacatlán, usos del suelo.

 

INTRODUCTION

Deforestation is a process of change in land cover and use from natural vegetation to residential, agricultural, livestock, industrial or commercial uses (Evangelista, López, Caballero, & Martínez, 2010; Rosete, Pérez, & Bocco, 2008) at different spatial-temporal scales (Gomben, Lilieholm, & González-Guillen, 2012; Hunter et al., (2003). Deforestation has a direct impact on the reduction and fragmentation of forest ecosystems by negatively modifying their structure and functioning. The impacts of deforestation alter the natural capital (Sarukhán, Carabias, Koleff, & Urquiza-Haas, 2012) and its ability to meet human needs (Vitousek, Mooney, Lubchenco, & Melillo, 1997). Changes in land cover and use also determine, in part, the vulnerability of people and places to climatic, economic and socio-political disturbances. When these changes are aggregated globally, they affect central aspects of earth system functioning (Lambin, Geist, & Lepers, 2003).

Understanding the impact caused by changes in land cover and use means studying natural, social, cultural and institutional aspects; however, in Mexico there are few studies (Rosete et al., 2008) that provide a quantitative analysis of the relative importance of these aspects, so interpretations of how these factors interact to induce this change vary widely from one place to another (Skole, Chomentowski, Salas, & Nobre, 1994). Therefore, the impact of change in land cover and use should be analyzed under different spatial scales (O'Neill, 1989), for example at the village, town, city, county, region, state or country level. Each spatial scale represents an appropriate and unique approach to certain factors of interest (Alcaraz, Bandi, & Garbulsky, 2008; Bocco, Mendoza, & Masera, 2001). The information generated is essential for developing effective public policies related to land-use planning.

Several studies have modeled and analyzed the deforestation process at the regional level using small scales and considering socioeconomic and environmental variables (Bocco et al., 2001; Chaves & Rosero-Bixby, 2001; Pineda, Bosques, Gómez, & Plata, 2009). In general, studies at small scales are associated with large units of analysis (region, state, country) and vice versa; however, there are exceptions depending on the level of precision of the analysis. For example, the Instituto Nacional de Ecología y Cambio Climático (INECC, 2011), Mexico's National Institute of Ecology and Climate Change, determined the national deforestation risk index considering only the national economic pressure index. The study notes that the state of Puebla has an average deforestation rate of 3.3 %, equivalent to 15,000 ha·yr^-1; however, one possible downside to this study was the resolution used (9-ha cells). To avoid this, some authors like Alcaraz et al. (2008) and Trucíos, Rivera, Delgado, Estrada, and Cerano (2013) have analyzed the risk of deforestation at multiple scales, while Hunter et al. (2003) and Gomben et al. (2012) did it to determine land-use change.

Qualitative and quantitative analyzes allow us to understand and evaluate the environmental impacts of deforestation at different spatial and temporal scales. Although limited resources often reduce the opportunity to make such analyzes at multiple scales (Alcaraz et al., 2008; Krannich, Carroll, Daniels, & Walker, 1994; Romero & Fuentes, 2009), data and information gathering should be appropriate to the phenomenon or process to be studied (Gomben et al, 2012; Hunter et al., 2003). Because a phenomenon or process observed at one scale may or may not be repeated in another, it is important to know the environmental, social and economic patterns that stimulate deforestation and their impacts at different spatial scales. Recognizing these patterns and their interaction in the deforestation process is an important step in the generation of future land-use scenarios (Trucíos et al., 2013). Therefore, the aims of this study were to build and apply a land-use change and deforestation risk model for the Chignahuapan-Zacatlán region of Puebla, and use the regional model to analyze the effects of scale with models fitted to the county level, on the patterns that govern such processes in the study area. In addition, a future land-use change and deforestation risk scenario at the regional and county levels was generated to provide better decision-making in relation to the conservation of forest resources.

 

MATERIALS AND METHODS

Location of the study area

The study area was the Chignahuapan-Zacatlán region, located in the Sierra Norte of Puebla state. The region is bordered to the north by Huauchinango and Chiconcuatla, Puebla; to the south by Tlaxco, Tlaxcala; to the east by Aquixtla, Puebla; and to the west by Apan, Cuatepec de Hinojosa and Almoloya, Hidalgo (Secretaría del Medio Ambiente y Recursos Naturales [SMRN], 2007).

Detection of land-use change (dependent variable)

Land-use change was detected by managing two Landsat imagery scenes (1986 and 2011), obtained from the files of the U.S. Geological Survey (USGS, 2013). These were corrected and classified in a supervised manner through 250 training points and field trips, with the aid of PCI GEOMATICS version 9.1^® software (PCI GEOMATICS, 2004). Later a county and regional section was made from the images using the IDRISI Taiga program with the aid of the Image Calculator model, in order to later perform a bitemporal analysis of the 1986-2011 period with the CrossTab change modeler (Eastman, 2009) and obtain three land-use change maps (one regional and two at the county level). The stable forest cover and changes from forest use to agricultural, livestock and residential uses were obtained in the three maps. Subsequently, the hectare-level information was extracted using a 100 x 100 m grid with ArcGIS version 9.3 software (Environmental Systems Research Institute [ESRI], 2011) and the Create fishnet^® module. Finally, the centroid of each polygon was obtained to develop a database of points and extract their values with the raster values to point extract command. The dependent variable in the model took the following values: 1 = Forest to forest, 2 = Forest to agriculture, 3 = Forest to livestock and 4 = Forest to other uses.

Estimation of the independent variables

A total of 21 independent social, economic and environmental variables (Altamirano & Lara, 2010; Pineda et al., 2009; Pinedo, Pinedo, Quintana, & Martínez, 2007) correlated with the dependent variable, which are shown in Table 1, were considered. The values of the independent variables were calculated and extracted in a similar way to that used for the dependent variable. Subsequently, the database was purged by checking that each hectare or cell was represented by values of the dependent and independent variables through a visual inspection with the Microsoft Office Excel^® software.

Baseline scenario for deforestation risk at regional and county levels

With the independent and dependent variables, a multinomial logistic regression model was constructed (Hunter et al., 2003; Infante & Zárate, 2003; Montgomery, 2004). This type of model is a variant of the classic binary logistic model that relates the dependent variable with two or more explanatory or independent variables that are either qualitative or quantitative (Allison, 1999):

where:

Ŷ_i = Probability of a hectare of forest use changing use (where i = agricultural, livestock or residential uses)

e = Base of natural logarithm

Intercept and estimators of the independent variables (x_i)

x_i = Independent variables (x₁, x₂,...,x₂₁)

The model was constructed using the stepwise procedure of the SAS 9.0 program (Statistical Analysis System, 2004) that lets you choose the subset of independent variables or regressors that should be considered in the model. This selection is made according to the order of importance or significance that helps explain the dependent variable (Beltrán, 2011; Silva, 1995). Finally, the land-use change probability maps (regional and county) were built from the logistic regression model using the 2011 forest area, obtained from satellite imagery (USGS, 2013).

Future deforestation scenario

The future deforestation scenario was generated by considering the increase in inhabitants for the period 2010-2030 at the regional and county levels (Consejo Nacional de Población [CONAPO], 2012). For both levels, the average population density for the period 1985-2010 (Instituto Nacional de Estadística, Geografía e Informática [INEGI], 2010) was determined per land-use type. For example, the population density at the regional level in the agricultural, livestock and residential uses was 6.65, 49.14 and 100.82 inhab·ha^-1, respectively; in Chignahuapan, it was 1.10, 90.26 and 70.82 inhab·ha^-1 in the agricultural, livestock and residential uses, respectively; and in Zacatlán, it was 4.54, 11.62, and 327.47 inhab·ha-1 in the agricultural, livestock and residential uses, respectively. This "historical" density was taken as a base and then the new inhabitants were "populated," giving priority to those cells with the greatest likelihood of land-use change in the probability maps.

Comparison of scales

A comparative future analysis of the deforestation risk maps at the regional and county levels was performed qualitatively and quantitatively.

 

RESULTS AND DISCUSSION

Detection of land-use change (dependent variable)

Regional

The main land use in the Chignahuapan-Zacatlán region is agriculture, which increased by 7.9 % in the period from 1986 to 2011. This value is consistent with data reported by Franco, Regil, González, and Nava (2006) in a study conducted in Nevado de Toluca National Park, Mexico, during the period 1972-2000. In general, Evangelista et al. (2010) note that in the Sierra Norte de Puebla there is a trend towards increased agricultural and residential land use; the livestock area in 2011 decreased by 75 %, while agricultural and residential areas increased. For its part, forest use, second in importance, retained only 85 % of its land base due to losses caused by population growth and inordinate use of natural resources (SMRN, 2007).

County

Agricultural land is the most important in Chignahuapan county. Mahar and Schneider (1994) indicate that the advance of the agricultural frontier is one of the main causes of deforestation. On the other hand, forest use accounts for 53 % of the territory in Zacatlán. During the 25 years analyzed, livestock area decreased in both counties, while residential use increased. Despite the land-use changes, 57,693 ha of stable forest use were determined at the regional level, of which 29,654 ha (51.40 %) are in Chignahuapan and 28.039 ha (48.60 %) in Zacatlán.

Importance and significance of variables at the regional and county levels

The results of logistic regression and the significance of the variables that help explain the change in land use are shown in Tables 2, 3 and 4. At the regional level, the number of significant variables for land-use changes from forest to agriculture, forest to livestock and forest to residential were 17, 17 and 14, respectively (Table 2); for Chignahuapan, they were 14, 5 and 4 (Table 3); and for Zacatlán, they were 19, 13 and 11 (Table 4). The number of significant variables (17, 14 and 19) that predict land-use change from forest to agriculture is similar at the regional and county level. This was not the case for land-use changes from forest to livestock and forest to residential in Chignahuapan, for which the number of variables are 5 and 4, respectively. In the same county there were five non-significant variables for the three land-use changes (Table 3).

In the nine models obtained, the estimators of the 21 independent variables show either a positive or negative sign depending on the change of use in question and the scale analyzed. For example, at the regional level there is a positive correlation between the change in land use from forest to agriculture and the distance to sawmills (x1), distance to areas with high poverty (x13), distance to pasture areas (x19), ejidal (communal land) ownership (x20) and private ownership (x21). That is, one hectare of land with forest use is more likely to be changed to agricultural use if the land is farther away from sawmills and pasture areas and if it is ejido or private land. For the same use and region, all other significant variables show negative correlation (Table 2). The interpretation of the signs is similar in other uses for the region and for the counties under study (Tables 2, 3 and 4).

All variables have different relative importance in each model at the regional and county levels (Table 5). At the regional level, and with a negative relationship, the variables minimum distance to agricultural areas (x18), elevation (x8) and slope (x9) are the ones that most influence deforestation (Tables 2 and 5). This contrasts with Chaves and Rosero-Bixby (2001), who indicate that the minimum distance to transportation routes has little influence on the deforestation process, and with Pineda et al. (2009), who agree that elevation and slope (> 20 %) are causes of deforestation. This is probably because the studies were performed with a different spatial scale. For example, at the county level, the behavior of the three aforementioned variables was different; for Chignahuapan, they were not significant, except for the change in land use from forest to agriculture (Table 3). Some variables such as the distance to: paved roads (x3), dirt roads (x4), permanent watercourses (x5) and communities with a population of more than 100 inhabitants (x17) helped explain deforestation in both counties; however, these variables were not significant at the regional level. In short, estimators of the independent variables at the regional and county levels generate a different model for each land-use change (forest-agriculture, forest- livestock and forest-residential) (Alcaraz et al, 2008; Bocco et al., 2001). This allows us to understand that the patterns of land-cover and land-use change vary widely from one region to one county and vice versa.

Model validation

In building the deforestation risk models, there was no correlation between independent variables. In addition, Table 6 shows that the models were not affected by multicollinearity, since they had tolerances of greater than 1 and variance inflation factors of less than 10. This suggests that there is no linear relationship between the model's independent variables (Kutner, Nachtsheim, Neter, & Li, 2005). On average 80.7 % concordant observations were obtained at the regional and county levels, indicating that the independent and dependent variables are ordered in the same direction, thereby favoring the model's predictive ability. A perfect positive association occurs when all pairs are concordant. All three models have R²_adj greater than 20 %; although the value is low, it is considered appropriate due to the inclusion of biophysical, economic and social variables. This situation has already been reported in previous research (Gomben et al., 2012; Frías- Armenta, López-Escobar, & Díaz-Méndez, 2003; Hunter et al., 2003); therefore, the models were considered appropriate to predict the phenomenon of interest.

Estimating the probability of spatial deforestation in two scales

Regional: Forest-agriculture

Figure 1a shows the probability of one hectare of forest use changing to agricultural use, which is 45-90 %. Tables 2 and 5 indicate that the variables elevation and distance to agricultural limits are the most influential in the process. This is consistent with Mahar and Schneider (1994), who concur that the advance of the agricultural frontier increases the likelihood of deforestation.

Regional: Forest-livestock

The probability that one hectare changes from forest to livestock use varies from 10-50 %. This is related to 17 significant variables in the deforestation process due to livestock use (Table 2). The most important variables at the regional level were elevation and distance to pasture areas with a negative relationship (Table 2), indicating that forest land with lower elevation and less distance to pasture areas has a greater probability of deforestation. Figure 1b shows that this phenomenon occurs mainly in the cloud forest in the region's northwest area (SMRN, 2007).

Regional: Forest-residential

Figure 1c plots the probability of land-use change. Table 2 shows the 14 variables that have an impact on the deforestation process. The trend towards changing to residential use is limited because the forest areas are far from urban centers and surrounded by agricultural areas, having a greater tendency to change from agricultural to forest (Pineda et al., 2009) and from residential to agricultural use.

County: Forest-agriculture

The probability of one hectare of forest use changing to agricultural use (Figures 2a, 2b) is greater than 60 %. In Chignahuapan and Zacatlán, the distance to high-poverty areas was not significant (Tables 3 and 4) in determining the probability of change from forest to agricultural use, although agriculture is the main activity of both counties and the main driving force behind deforestation in Mexico (Deininger & Minten, 1999).

County Forest-livestock

The probability of change from forest to livestock use (Figure 2c) is a function of the significant variables for each model; for example, for Chignahuapan, five significant variables with probabilities of more than 95 % to change a small forest area to livestock use were considered (Table 3). On the other hand, for Zacatlán county, 13 significant variables (Table 4) were considered to determine the probability of change (Figure 2d). In this regard, Rojas-López, González- Guillén, Gómez-Guerrero, and Romo-Lozano (2012) obtained a probability of 1 % for the change from forest to livestock use; these significant differences are probably due to the methodology used for the calculation.

County Forest-residential

The probability of forest use changing to residential use (Figures 2e, 2f) is greater than 50 %. In both counties, the significant variables were distance to permanent watercourses, slopes, high-poverty areas and a community of more than 100 inhabitants. The slope variable was significant with a negative sign, indicating that areas with gentle slopes are at increased risk of deforestation, while the variables permanent watercourses and a population of more than 100 inhabitants are considered indicators for establishing urban areas.

Scalar comparison of the risk of future deforestation

Figure 3 shows a qualitative comparison of the deforestation risk in Zacatlán and Chignahuapan. The regional model estimated a smaller area (2,308.2 ha) at risk of deforestation than the county-level models; however, this area is located in a different space, since the variables and estimators that determine this risk are different for each level. However, the risk of deforestation in both scales occurs mainly in the forest- agriculture frontier (Mahar & Schneider, 1994).

Figure 4 quantitatively compares the deforestation risk trend at the regional and county levels; it shows that the risk decreases as the area increases. At the regional level, an estimated 13,063 ha will be required to accommodate new inhabitants in 2030. Of these, 1,774.2 ha (13.6 %) are considered at high risk, 5,790.0 ha (44.3 %) at medium risk, and 5,499.6 (42.1 %) at low risk of deforestation (Figure 4a). The Chignahuapan model shows the same deforestation trend as the regional model; of the total area (29,654 ha), 7 % (2,097 ha) is at high risk, 14 % (4,395 ha) at medium risk and 15 % (2,097.2 ha) at low risk of deforestation (Figure 4b). These areas are located near agricultural and livestock limits. For its part, Figure 4c shows the deforestation risk in Zacatlán, with 1,797.7, 2,271 and 336 ha being at high, medium and low risk, respectively. The deforestation probability trend in Zacatlán differs from that shown at the regional level in Chignahuapan. Overall, the forest area shows a pattern fragmented by agriculture and livestock, increasing the risk of deforestation (SMRN, 2007).

Deforestation risk models at the county and regional levels integrated biophysical and socio-economic variables, highlighting that elevation, slope and distances to agricultural and livestock areas have a greater influence on land-use change and deforestation risk; however, distance to permanent watercourses is the most influential variable. Finally, Saab (1999) and Cueto (2006) agree that scale is a very important factor in delineating plans and strategies, so the selection of the scale to use depends on the objectives and characteristics of the study (García, Teich, & Balzarini, 2011); models built at one spatial scale and applied to another would generate biased interpretations. For example, if the regional model of land-use changes were applied to make estimates at the county level, an "ecological fallacy" would be generated, whereas if the county model were applied at the regional level, an "individualistic fallacy" would be committed. On the other hand, if the model generated for Chignahuapan county were applied to Zacatlán county or vice versa, a "cross fallacy or irregularity" would be committed (Brenner, 2001). Therefore, each scale must be appropriate to the phenomenon or process under study and must be used at the spatial level at which the policies and strategies for natural resource management, conservation and use are designed and applied.

 

CONCLUSIONS

Through identifying patterns that govern land-use changes and deforestation risk, probabilistic models were constructed and applied for the region and the counties of Chignahuapan-Zacatlán, Puebla. Considering both study levels and the land-use change model, it was observed that residential use increased substantially, whereas livestock use dropped dramatically. For its part, the county-level probabilistic model estimated probabilities of more than 90 % of forest use changing to agricultural or residential use, while probabilities at the regional level are less than 90 %. The generation of future scenarios for 2030 yielded information for the design of natural resource management strategies at the regional and county levels. Since an overview of the areas susceptible to deforestation due to land-use change was successfully obtained, we conclude that the constructed models are flexible, powerful and useful tools depending on the scale at which one wishes to generate strategies for use of natural resources.

 

ACKNOWLEDGMENTS

The authors are grateful to the Colegio de Postgraduados for providing funding for this study under Priority Research Line 1 (known by the Spanish acronym LPI1): Sustainable Management of Natural Resources.

 

REFERENCES

Alcaraz, S. D., Baldi, P. G., & Garbulsky, M. F. (2008). Análisis de la dinámica temporal del NDVI en áreas protegidas: Tres casos de estudio a distintas escalas espaciales, temporales y de gestión. Ecosistemas, 17, 108–117. Obtenido de http://www.redalyc.org/articulo.oa?id=54017106010        [ Links ]

Allison, P. D. (1999). Logistic regression using the SAS system: Theory and application. Cary, NC, USA: John Wiley & Sons-SAS Institute Inc.         [ Links ]

Altamirano, A., & Lara, A. (2010). Deforestación en ecosistemas templados de la precordillera andina del centro-sur de Chile. Bosque, 31, 53–64. doi: 10.4067/ S0717-92002010000100007        [ Links ]

Beltrán, C. (2011). Aplicación del análisis de regresión logística multinomial en la clasificación de textos académicos: Biometría, filosofía y lingüística informática. INFOSUR, 5, 53–64. Obtenido de http://www.infosurrevista.com.ar/biblioteca/INFOSUR-Nro5-2011-Beltran.pdf        [ Links ]

Bocco, G., Mendoza, M., & Masera, O. R. (2001). La dinámica del cambio del uso del suelo en Michoacán: Una propuesta metodológica para el estudio de los procesos de deforestación. Investigaciones Geográficas, 44, 18–36. Obtenido de http://www.oikos.unam.mx/Bioenergia/images/PDF/CC/CC-Bosques/la_dinamica_del_cambio_del_uso_del_suelo.pdf        [ Links ]

Brenner, N. (2001). The limits to scale? Methodological reflections on scalar structuration. Progress in Human Geography, 25, 591–610. doi: 10.1191/030913201682688959        [ Links ]

Chaves, E. E., & Rosero-Bixby, L. (2001). Valoración del riesgo de deforestación futura en Costa Rica. Uniciencia, 18, 29–38. Obtenido de http://ccp.ucr.ac.cr/bvp/pdf/bosques/chaves.pdf        [ Links ]

Consejo Nacional de Población (CONAPO). (2012). Índice de marginación por localidad 2010. México, D. F.: Autor.         [ Links ]

Cueto, V. R. (2006). Escalas en ecología: Su importancia para el estudio de la selección de hábitat en aves. Hornero, 21, 1–13. Obtenido de http://www.scielo.org.ar/pdf/hornero/v21n1/v21n1a01.pdf        [ Links ]

Deininger, K. W., & Minten, B. (1999). Poverty, policies, and deforestation: The case of Mexico. Economic Development and Cultural Change, 47, 313–344. Obtenido de http://www.jstor.org/stable/10.1086/452403        [ Links ]

Eastman, R. J. (2009). IDRISI Taiga Guide to GIS and image processing. Worcester, MA, USA: Clark Labs.         [ Links ]

Environmental Systems Research Institute (ESRI). (2011). Software ArcGIS, versión 10.0. Redlands, CA, USA: Autor.         [ Links ]

Evangelista, O. V., López, B. J., Caballero, N. J., & Martínez, M. A. (2010). Patrones espaciales de cambio de cobertura y uso del suelo en el área cafetalera de la Sierra Norte de Puebla. Investigaciones Geográficas, 72, 23–38. Obtenido de http://www.revistas.unam.mx/index.php/rig/article/view/19273        [ Links ]

Franco, M., S., Regil, G. H. H., González, E. C., & Nava, B. G. (2006). Cambio de uso del suelo en el Parque Nacional Nevado de Toluca, México, en el periodo 1972-2000. Investigaciones Geográficas, 61, 38–57. Obtenido de http://www.redalyc.org/pdf/569/56906105.pdf        [ Links ]

Frías-Armenta, M., López-Escobar, A. E., & Díaz-Méndez, S. G. (2003). Predictores de la conducta antisocial juvenil: Un modelo ecológico. Estudios de Psicología, 8, 15–24. doi: 10.1590/S1413-294X2003000100003        [ Links ]

García, C. L., Teich. I., & Balzarini, M. (2011). Una aproximación multivariada para identificar la escala de análisis en estudios de asociación entre comunidades de aves y la composición del paisaje. Revista de la Asociación Argentina de Ecología de Paisaje, 2, 26–34.         [ Links ]

Gomben, P., Lilieholm, R., González-Guillén, M. (2012). Impact of demographic trends on future development patterns and the loss of open space in the California Mojave Desert. Environmental Management, 49, 305–324. doi: 10.1007/s00267-011-9749-6        [ Links ]

Hunter, L. M., Gonzalez, M. J., Stevenson, M., Karish, K. S., Toth, R. E., Edwards, T. C., Lilieholm, R.J. & Cablk, M. (2003). Population and land use change in the California Mojave: Natural habitat implications of alternative futures. Population Research and Policy Review, 22, 373–397. doi: 10.1023/A:1027311225410        [ Links ]

Infante, G. S., & Zárate, L. G. P. (2003). Métodos estadísticos: Un enfoque interdisciplinario. México, D. F.: Editorial Trillas.         [ Links ]

Instituto Nacional de Ecología y Cambio Climático (INECC). (2011). Usos del índice de presión económica (riesgo) de la deforestación. Consultado 10-01-2014 en http://www.inecc.gob.mx/irdef-usos        [ Links ]

Instituto Nacional de Estadística y Geografía (INEGI). (1999). Conjunto de datos vectoriales (E14B13-E14B14-E14B15-E14B23-E14B24-E14B25-E14B33-E14B34-E14B35-F14D83-F14D84-F14D85) de las cartas topográficas escala 1:50000. Aguascalientes, Ags., México: Autor.         [ Links ]

Instituto Nacional de Estadística y Geografía (INEGI). (2000). Censo de población y vivienda 2000. Consultado 03- 10-2012 en http://www.inegi.org.mx/est/contenidos/proyectos/ccpv/cpv2000/default.aspx        [ Links ]

Instituto Nacional de Estadística y Geografía (INEGI). (2010). Conteo de la población y vivienda. Base de datos digital del periodo 1985-2010. Consultado 07-09-2013 en http://www.inegi.org.mx/        [ Links ]

Krannich, R. S., Carroll, M. S., Daniels, S. E., & Walker, G. B. (1994). Incorporating social assessment and public involvement processes into ecosystem-based resource management: Applications to the east side ecosystem management project. Walla, WA, USA: Interior Columbia Basin Ecosystem Management Project, USDA Forest Service.         [ Links ]

Kutner, M. H., Nachtsheim, Ch. J., Neter, J., & Li, W. (2005). Applied linear statiscal models. USA: McGraw-Hill.         [ Links ]

Lambin, E. F., Geist, H. J., & Lepers, E. (2003). Dynamics of land-use and land-cover change in tropical regions. Annual Review of Environment Resources, 28, 205–241. doi: 10.1146/annurev.energy.28.050302.105459        [ Links ]

Mahar, D., & Schneider, R. R. (1994). Incentives for tropical deforestation: Some examples from Latin America. In K. Brown, & D. W. Pearce (Eds.), The causes of tropical deforestation (pp. 56–78). London, UK: University College London Press.         [ Links ]

Montgomery, D. C. (2004). Diseño y análisis de experimentos. México, D. F.: Editorial Limusa.         [ Links ]

O`Neill, R. V. (1989). Perspectives in hierarchy and scale, in perspectives in ecological theory. Princeton, NJ, USA: Princeton University Press.         [ Links ]

PCI GEOMATICS. (2004). PCI Geomatics software. Canadá: Autor. Obtenido de http://www.pcigeomatics.com/        [ Links ]

Pineda, J. N. B., Bosque, S. J., Gómez, D. M., & Plata, R. W. (2009). Análisis de cambio del uso del suelo en el estado de México mediante sistemas de información geográfica y técnicas de regresión multivariantes. Una aproximación a los procesos de deforestación. Investigaciones Geográficas, 69, 33–52. Obtenido de http://www.revistas.unam.mx/index.php/rig/article/view/18003        [ Links ]

Pinedo, A., C., Pinedo, A., A., Quintana, M. R. M., & Martínez, S., M. (2007). Análisis de áreas deforestadas en la región centro-norte de la Sierra Madre Occidental, Chihuahua, México. Tecnociencia, 1, 36–43. Obtenido de http://tecnociencia.uach.mx/numeros/v1n1/data/analisis_de_areas_deforestadas.pdf        [ Links ]

Rojas-López, O., González-Guillén, M. J., Gómez-Guerrero, A., & Romo-Lozano, J. L. (2012). Renta de la tierra y pago de servicios ambientales en la Sierra Norte de Puebla. Revista Mexicana de Ciencias Forestales, 3(11), 41–56. Obtenido de http://www.scielo.org.mx/pdf/remcf/v3n11/v3n11a4.pdf        [ Links ]

Romero, H., & Fuentes, C. (2009). Análisis multiescalar de los cambios en los complejos dinámicos territoriales en la Cuenca del rio Itata. Chile: Universidad de Concepción.         [ Links ]

Rosete, V. F. A., Pérez, D. J. L., & Bocco, G. (2008). Cambio de uso de suelo y vegetación en la península de Baja California, México. Investigaciones Geográficas, 67, 39–58. Obtenido de http://www.scielo.org.mx/pdf/igeo/n67/n67a4.pdf        [ Links ]

Statistical Analysis System (SAS Institute Inc.). (2004) SAS/ STAT 9.1 User's Guide. Cary, NC, USA: Autor.         [ Links ]

Saab, V. (1999). Importance of spatial scale to habitat use by breeding birds in riparian forests: A hierarchical analysis. Ecological applications, 9, 135–151. doi: 10.1890/1051-0761(1999)009[0135:IOSSTH]2.0.CO;2        [ Links ]

Sarukhán, J., Carabias, J., Koleff, P., & Urquiza-Haas, T. (2012). Capital natural de México: Acciones estratégicas para su valoración, preservación y recuperación. México: Comisión Nacional para el Conocimiento y Uso de la Biodiversidad.         [ Links ]

Secretaría del Medio Ambiente y Recursos Naturales (SMRN). (2007). Diagnóstico socioeconómico y de manejo forestal. Unidad de manejo forestal, Zacatlán. México: Autor.         [ Links ]

Silva, A. L. C. (1995). Excursión a la regresión logística en ciencias de la salud. Madrid, España: Edición Díaz Santos, S. A.         [ Links ]

Skole, D. L., Chomentowski, H., Salas, W. A., & Nobre, A. D. (1994). Physical and human dimensions of deforestation in Amazonia. BioScience, 44(5), 314–322. Obtenido de http://www.jstor.org/stable/1312381        [ Links ]

Trucíos, C. R., Rivera, G. M., Delgado, R. G., Estrada, A. J., & Cerano, P. J. (2013). Análisis sobre cambio de uso de suelo en dos escalas de trabajo. Terra Latinoamericana, 31(4), 339–346. Obtenido de http://www.chapingo.mx/terra/?module=resumenes&volumen=31&numero=4        [ Links ]

U. S. Geological Survey (USGS). USGS global visualization viewer. Consultado 14-08-14 en http://glovis.usgs.gov/index.shtml        [ Links ]

Vitousek, P. M., Mooney, H. A., Lubchenco, J., & Melillo, J. M. (1997). Human domination of Earth's ecosystems. Science, 277, 494–499. doi: 10.1126/science.277.5325.494        [ Links ]