Investigating the applicability of conventional vegetation indices for vegetation change detection in different environmental conditions

Vegetation indices have been developed to characterize and extract the Earth's vegetation cover from space using satellite images. For detection of vegetation changes, usually temporal images are independently analyzed or vegetation index differencing is implemented. A review on previous studies reveal that, in spite of developing several vegetation indices, to extract vegetation cover or vegetation changes usually NDVI and EVI are used. Therefore, the aim of this study is to compare and investigate the applicability of these indices in detection of vegetation changes in different climate and environmental conditions. For this purpose, several test sites in Malaysia, Iran and Italy with different environmental conditions including Tropical, Subtropical and Mediterranean were selected. Then, index differencing method using temporal Landsat-7 ETM+ and Landsat-8 OLI images belonging to the years of 2001 and 2014 were applied. In order to evaluating the accuracy of the output maps, confusion matrix was made to calculate overall accuracy and kappa index. Subsequently, commission and omission errors were calculated to assess nature of the errors in the results. Accuracy assessment analysis indicated that however the results of EVI in some of the test sites were acceptable, but in all of the test sites with variety of weather and environmental conditions, NDVI provided higher accuracy outcomes in detection of vegetation changes.