ESTIMATION OF TROPICAL FOREST AREA FROM COARSE SPATIAL-RESOLUTION DATA - A 2-STEP CORRECTION FUNCTION FOR PROPORTIONAL ERRORS DUE TO SPATIAL AGGREGATION

Retrieving area estimates from coarse spatial resolution land-cover maps is generally inaccurate due to the effect of spatial aggregation on class proportions. The objective of this study was to build a correction function with respect to proportional errors in order to estimate tropical forest areas from coarse resolution classifications. Previous studies have shown that the errors in the estimation of land-cover proportions depends on the spatial resolution of the map, the initial proportions of the landscape in the different land covers, and the spatial arrangement of the land covers at the initial resolution. The interaction between the proportional errors produced by spatial aggregation and the spatial pattern of cover types has been modeled in an operational way at a broad spatial scab. A correction function has been calibrated on the basis of tropical forest maps derived; from AVHRR 1 km data for the entire intertropical belt and a nonrandom sample of 13 Landsat TM scenes well distributed among the main tropical forest regions. The calibration model integrates a measure of the spatial fragmentation of the landscape in a two-step procedure. This calibration procedure improves by 25.5% the reliability of the retrieval of proportions of cover types from coarse resolution data by comparison to a simple correction function relating directly proportions at coarse and fine resolutions. The final coefficient of determination, of the model is 0.789.