Department Geography and Environmental Studies, Stellenbosch University, Private Bag X1, Matieland, 7602, South Africa; Western Cape Nature Conservation Board, Scientific Services, P Bag x5014, Stellenbosch, 7599, South Africa; Department of Ecology and Evolutionary Biology, University of Connecticut, 75 N. Eagleville Road, Storrs, CT 06269-3043, United States; Council for Scientific and Industrial Research (CSIR), Meraka Institute, PO Box 395, Pretoria 0001, South Africa
De Klerk, H.M., Department Geography and Environmental Studies, Stellenbosch University, Private Bag X1, Matieland, 7602, South Africa, Western Cape Nature Conservation Board, Scientific Services, P Bag x5014, Stellenbosch, 7599, South Africa; Wilson, A.M., Department of Ecology and Evolutionary Biology, University of Connecticut, 75 N. Eagleville Road, Storrs, CT 06269-3043, United States; Steenkamp, K., Council for Scientific and Industrial Research (CSIR), Meraka Institute, PO Box 395, Pretoria 0001, South Africa
Fire is a critical ecological process in the fynbos of the south-western area of South Africa, as it is for all dwarf Mediterranean shrublands. We evaluated the potential of current publicly available MODIS burned area products to contribute to an accurate fire history of the fynbos. To this end, we compared the Meraka Institute's MODIS burned area product, based on the Giglio algorithm (termed the 'WAMIS' product) as well as the standard MODIS MCD45A1 burned area product, based on the Roy algorithm, with comprehensive manager-mapped fire boundary data. We used standard inventory accuracy assessment (number and size of individual burn scars) and confusion matrix techniques. Results showed promise for both burned area products, depending on the intended use. The MCD45A1 had low errors of commission (8.1-19.1%) and high consumer's accuracy (80.9-91.9%), but relatively common errors of omission, making it useful for studies that need to identify burned pixels with a high degree of certainty. However, the WAMIS product generally had low errors of omission (12.2-43.8%) and greater producer's accuracy (56.2-87.6%), making it a useful tool for supplementing manager-mapped fire records, especially for fynbos remnants occurring outside protected areas. © 2012 IAWF.