Relationships between vegetation indices and different burn and vegetation ratios. A multi-scale approach applied in a fire affected area

摘要

Vegetation indices have been widely used in remote sensing literature for burned land mapping and monitoring. In the present study we used satellite data (IKONOS, LANDSAT, ASTER, MODIS) of multiple spectral (visible, near, shortwave infrared) and spatial (1-500 meters) resolutions, acquired shortly after a very destructive fire occurred in the mountain of Parnitha in Attica, Greece the summer of 2007. The aim of our study is to examine and evaluate the performance of some vegetation indices for burned land mapping and also to characterize the relationships between vegetation indices and the percent of fire-scorched (burned) and non fire-scorched (vegetated) areas. The available satellite images were processed geometrically, radiometrically and atmospherically. The very high resolution IKONOS imagery was served as a base to estimate the percent of cover of burned areas, bare soil and vegetation by applying the maximum likelihood classification algorithm. The percent of cover for each type was then correlated to vegetation indices for all the satellite images, and regression models were fit to characterize those relationships. In total 57 versions of some classical vegetation indices were computed using LANDSAT, ASTER and MODIS data. Most of them were modified by replacing Red with SWIR channel, as the latter has been proved sensitive to burned area discrimination. IPVI and NDVI showed a better performance among the indices tested to estimate the percent of vegetation, while most of the modified versions of the indices showed highest performance to estimate the percent of burned areas.