Integration of vegetation inventory data and Landsat TM image for vegetation classification in the western Brazilian Amazon

This paper summarizes approaches for successional forest classification and develops a new approach based on the integration of vegetation inventory data and Landsat Thematic Mapper (TM) data in Rondonia, Brazil. Entropy (ENT) is calculated using tree height distribution from field vegetation inventory data and the adjusted entropy (adjENT) is then calculated by incorporating average stand height for each sample. The adjENT estimation model is developed using a linear regression analysis based on the integration of adjENT and measured TM reflectance. The successional stages are classified based on identified thresholds of adjENT values. The results shows that three successional stages (i.e., initial, intermediate, and advanced successional stages) can be classified with user's and producer's accuracies ranging from 77% to 79% and 61% to 89%, which are much improved comparing with the result from the maximum likelihood classifier. The strong correlation between adjENT representing forest stand structure and the TM spectral signatures may provide new insights for estimation of forest stand parameters, such as biomass, and for detection of forest degradation.