![Sample our Earth Sciences journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5930/TOC-Subject-Sample-2021-Earth-Sciences.jpg"})
Abstract
The study undertaken investigates the use of multi-sensor characteristics for accurate assessment of burned areas at regular intervals for two districts of Punjab. Temporal datasets of (pre-burning and post-burning during the year 2006) LISS-III, LISS-IV, MODIS and AVHRR have been used in the quantitative estimation of burned areas. The multi-temporal image difference technique using three different indices (NDVI, NBR and GEMI3) was taken using the statistical threshold technique which has been calculated and verified by extensive ground-based measurements. The performance of the indices was evaluated using spectral separability analysis and GEMI3 was found as the best in spectral discrimination of burned and unburned surface. Moderate resolution LISS-III (23.5 m) data provides accurate estimation of burned area affected by fire, but frequent assessment is not possible due to low temporal resolution (24 days). On the other hand, MODIS (500 m) and AVHRR (1000 m), resulted in a high degree of bias in area estimation, but provide data at a high temporal rate. This bias in coarser resolution data is studied using the sub-pixel technique approach. The total burned area estimated from LISS-III data (GEMI3) is 98,697 ha and from MODIS and AVHRR data (sub-pixel) is 83,893 ha and 74,564 ha, respectively. The area estimates of LISS-III, MODIS and AVHRR were compared with LISS-IV data (aggregate to respective resolution) for the same dates in the year 2007 at selected locations. An overall regression fit was obtained for MODIS data at selected sites with a value of coefficient of determination (R 2) = 0.771. An overall linear regression fit of MODIS and AVHRR with the whole of the LISS-IV image was compared with sub-pixel estimates at MODIS (500 m), AVHRR (1000 m) and the results revealed a strong relationship with R 2 = 0.868 and R 2 = 0.817, respectively. The study suggested that for crop residue burning, a multi-sensor approach is effective for improving the accuracy of burned area estimation at coarser resolution.
Acknowledgements
This work has been conducted as a part of the IIRS-ITC joint educational /research programme study. The authors are thankful to Mr J.P.G. (Wan) Bakx, ITC for his valuable guidance and suggestions during the research work. The authors acknowledge their sincere gratitude to Director, NRSC, Rector, ITC for the encouragement and useful suggestions.
