Use of spatial regression models in the analysis of burnings and deforestation occurrences in forest region, Amazon, Brazil

摘要

Through the PRODESDIGITAL Project, the National Institute for Space Research (INPE) has been mapping vegetal coverage in the Brazilian Legal Amazon using Landsat satellite images. INPE not only identifies deforested areas but also releases a daily map of burning areas. These fires often get out of control and end up accidentally invading areas of forest exploited by the lumber industry, agricultural plantations, and pastures. However, the deforestation and burning maps alone are insufficient for monitoring and control on a regional scale. The current study performs a dynamic analysis of the characteristics of burning occurrences in the state of Para, considering not only a possible spatial influence, but also the temporal dependence during the period 1999-2004. In the distribution analysis of burnings and deforestation in Para throughout this time, it can be observed that burnings mainly take place near the main highways in the state. Therefore, the currently deforested areas correspond to the mesoregions with good infrastructure access. These regions should be targets for farming technological investments. It can be shown that burning occurrences do not happen in a random way. Furthermore, the spatial diffusion process is faster for burnings than for deforestation. The spatio-temporal study showed that there was a change in the pattern records of burnings mainly in 2004, where the greatest quantity of records were located in the northeastern mesoregion of Para and great concentrations were also observed in southeastern and southwestern mesoregions.