Resumo

OSORIO, Laura Paulina; MAS, Jean-François; GUERRA, Francisco  e  MAASS, Manuel. Analysis and modeling of deforestation processes: a case study in the Coyuquilla River Basin, Guerrero, Mexico. Invest. Geog [online]. 2015, n.88, pp.60-74. ISSN 2448-7279.  https://doi.org/10.14350/rig.43853.

We analyzed land use / land cover changes (LUCC) in the basin of the river Coyuquilla in the state of Guerrero, Mexico during the periods 1986-2000 and 2000-2011. Land use / land cover maps for 1986, 2000 and 2011 were elaborated using Landsat and SPOT images and previous cartography from the National Institute of Geography and Statistics (INEGI) and the National Forest Inventory. The analysis of the satellite images was carried out through the method of interdependent interpretation of multidate satellite data in order to assure the consistency of the classification. Maps' accuracy was assessed through the interpretation of a stratified random sample of verification sites. Census information from INEGI, which includes demographic and socioeconomic data, along with environmental data were gathered and integrated in a GIS database. The method of the weights of evidence was applied in order to identify the LUCC drivers and elaborate maps of change probability. These maps were evaluated using a receiver operating characteristic (ROC) analysis.

Overall map accuracy was 71%, 73% and 75% for 1986, 2000 and 2011 respectively. During the period 1986-2000, a loss of 16% of the forest cover that mainly affected the dry tropical forest with an annual deforestation rate of -3.7% was observed. During the period 2000-2011, a loss of 22% of forest cover was observed and oak forest was the cover with the highest deforestation rate (-5.54%). These results are consistent with previous studies in the region. The increase of induced pasture was the main cause of the loss of forest cover. The elevation, distance to roads, distance to human settlements and the slope were important factors to explain deforestation in the three forest types and the two periods analyzed. The assessment of probabilistic models through the computing of the area under the ROC curve index gives index values between 0.61 and 0.92. These results are well above a random model and show that the models were able to identify the most likely to LUCC areas for the next period (2000-2011) for the eight analyzed transitions. Therefore, the model presents a reasonable performance in identifying the forest areas more likely to be cleared. However, in order to understand the processes of changes and their causes, the statistical approach must be complemented by a study based on a social science approach.

Palavras-chave : Deforestation; prospective models; weights of evidence; geo-statistical analysis.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )