1DETECTION OF BURNED FOREST AREAS IN CATALONIA USING A TEMPORAL SERIES OF LANDSAT MSS IMAGERY (PERIOD 1975-93)

The work that we present is a part of a general study aimed to characterize the fire regimes in Catalonia (NorthEast of Spain) and the effects of wildfires on regeneration dynamics of plant communities. For that purpose, a semi-automatic method was applied to de-tect burned forest areas in the 3 million ha region of Catalonia. The methodology employed more than a hundred of MSS images from Landsat satellites com-prising a period of time of 19 years. (1975-1993). They were geometrical and radiometrically corrected and the time series was registered. NDVI images were com-posed. Some masks were applied in order to avoid changes on plant cover dues to different causes. Sub-traction of consecutive NDVI images was employed to locate the forest areas affected by fire. This approach is based on the sudden decline that the plant commu-nities undergone when they burn. Linear regression models were used to fit the empirical changes observed for several test fires with NDVI differences between consecutive images. Detection thresholds were ob-tained from the models and applied to 31 windows of the images in order to minimize radiometric effects. Results were tested with administration inventories for the 1983-93 period. Omission errors are lower than 20 % for burned areas greater than 200 ha. Commis-sion errors are lower than 6 % for burned areas greater than 30 ha. A map series of fire history was created from the obtained burned forest areas. Fire recurrence is also depicted in order to locate hot spots. Some fire statistics are presented. Several reasons were argued in order to explain the errors of the implemented meth-odology and to understand them