A comparison of vegetation indices, enhanced vegetation index 2 and normalized difference vegetation index to monitor vegetation cover: a case study of Kagombe Central Forest Reserve, Kibaale District

Abstract

Vegetation cover change is phenomenon that happens over time due to different factors. This change can occur as a positive occurrence through growth of vegetation or a negative change through deforestation, low vegetation health. Continued degradation leads to loss of vegetation cover. In this project, monitoring of the vegetation cover change is to be done using vegetation indices, Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index 2 (EVI-2). The objectives of the project were to monitor extents of vegetation change over the years using the vegetation indices under study i.e. EVI-2 and NDVI. The other was to obtain change or no-change areas of the forested areas with both indices. The last objective was to compare the values obtained from EVI-2 and those from NDVI when finding the image changes. The case study area is Kagombe central forest reserve in Kibaale district. To produce these change detection maps, image differencing was used by first producing the NDVI and EVI-2 maps to be used. The different indices maps undergo change detection of image differencing which produce change or no change maps. The results got from both indices are compared by with a reference sentinel 2 image. The classification of the NDVI and EVI-2 were compared to this image and the overall accuracy of NDVI WAS. The results are validated using high resolution images of Sentinel 2. The image was classified and the overall accuracy in NDVI was 82.465% and a kappa value of 0.7349. The accuracy for the EVI-2 was 85.96% and a Kappa value of 0.7837. This showed EVI-2 performed better than NDVI. The accuracy assessment was carried out on both of the change detection maps classified using the EVI 2 and NDVI values. It was compared to high resolution image and results were EVI 2 performed better with an overall accuracy of 92%, kappa coefficient of 0.92 whereas ndvi gave an overall accuracy of 90% and kappa coefficient of 0.90.