Burned Area Mapping in a Mediterranean Environment Using Time-Series VEGETATION and Simulated PROBA-V Imagery by Employing an Object-Based Change Detection Approach

摘要

An object-based changed detection method was developed and assessed for accurately mapping burned areas by employing multi-temporal, PROBA-V and VEGETATION (VGT) NDVI data. Since the PROBA-V sensor is scheduled for launch in the near future, PROBA-V data were substituted with MODIS data. The method was assessed over the 2007 wildfires in the Peloponnese peninsula, Greece. Data captured over the study area were used to form a multi-temporal NDVI image which was then segmented into groups of spectrally similar adjacent pixels (objects). Consequently, the deviation between the multi-temporal NDVI signature of each object and that of an unchanged reference was measured by calculating the Mahalanobis Distance (MD) between them. Assuming that the deviation would be significantly greater for burned areas, objects with the highest MD values were considered burned. Burned area maps were produced based on an area derived MD threshold built on a preexisting knowledge of the percentage of the study area that was burned. PROBA-V data resulted to a kappa value of 0.81 and a burned area user's accuracy of 82.22% while the lower spatial resolution VGT data resulted to lower accuracy (0.73 and 75.54% respectively). Burned area maps were also produced based on MD threshold values derived from the distribution of MD values across the study area that did not require any preexisting knowledge and also resulted in equally accurate maps and PROBA-V out-performing VGT.