Regresar

A multiscale analysis of LULC and NDVI variation in Nang Rong district, northeast Thailand

Abstract:

Landuse/landcover (LULC) dynamics in northeast Thailand are driven by a host of scale dependent relationships that are observed at the pixel to landscape levels, but may be influenced by decisions made at the household and/or community levels. To understand complex systems, it is important to examine the interplay between scale, pattern, and process so that landscape form and landscape function may be inter-related. Statistical relationships between plant biomass levels and selected social, biophysical, and geographical variables are assessed at nine different cell resolutions beginning at 30 m and extending to 1050 m. The basic intent is to examine the scale dependence of population and environment relationships in a northeast Thailand study site, a region that has experienced pronounced LULC change associated with deforestation and agricultural extensification and intensification to support lowland rice and upland cash crops. More subtle landscape changes are also occurring, those related to intra-annual crop phenologies and monsoonal rainfall patterns. For each variable and at successively coarser resolutions, cell values are calculated by hierarchical aggregation of the original 30 m grid. The beta-values of each variable are assessed and the R2-values of each multiple regression model are tracked over the nine spatial resolutions. Canonical correlation is used to link a group of environmental variables to a group of social variables across the same 9-scale steps to assess the scale dependence of their relationships. Pattern metrics and LULC proportions are also computed to ascertain the composition and spatial organization of the landscape for the same nine cell resolutions. Finally, multiple regression models are computed for two separate image dates within a single water year to explore intra-annual variation in normalized difference vegetation index (NDVI) and the level of explanation achieved through selected population and environment variables. In this scenario, spatial scale is held constant at 30 m, but distance decay coefficients, used to construct a population potential variable, are altered and model results tracked to assess NDVI sensitivities to population distribution parameters as well as wet and dry conditions and their seasonal differences on derived R2-values and beta-coefficients. LULC patterns in northeast Thailand are the products of scale dependent relationships that extend across thematic domains representing social, biophysical, and geographical factors. Results indicate the importance of population factors at finer scales and biophysical factors at coarser scales for explaining variation in plant biomass levels. © 2001 Elsevier Science B.V.