Integrating field sampling, geostatistics and remote sensing to map wetland vegetation in the Pantanal, Brazil

摘要

Development of efficient methodologies for mappingwetland vegetation is of key importance to wetland conservation. Here wepropose the integration of a number of statistical techniques, in particularcluster analysis, universal kriging and error propagation modelling, tointegrate observations from remote sensing and field sampling for mappingvegetation communities and estimating uncertainty. The approach results inseven vegetation communities with a known floral composition that can bemapped over large areas using remotely sensed data. The relationship betweenremotely sensed data and vegetation patterns, captured in four factorialaxes, were described using multiple linear regression models. There werethen used in a universal kriging procedure to reduce the mappinguncertainty. Cross-validation procedures and Monte Carlo simulations wereused to quantify the uncertainty in the resulting map. Cross-validationshowed that accuracy in classification varies according with the communitytype, as a result of sampling density and configuration. A map ofuncertainty derived from Monte Carlo simulations revealed significantspatial variation in classification, but this had little impact on theproportion and arrangement of the communities observed. These resultssuggested that mapping improvement could be achieved by increasing thenumber of field observations of those communities with a scattered and smallpatch size distribution; or by including a larger number of digital imagesas explanatory variables in the model. Comparison of the resulting plantcommunity map with a flood duration map, revealed that flooding duration isan important driver of vegetation zonation. This mapping approach is able tointegrate field point data and high-resolution remote-sensing images,providing a new basis to map wetland vegetation and allow its futureapplication in habitat management, conservation assessment and long-termecological monitoring in wetland landscapes.