Seasonal effect on the accuracy of land use/land cover classification in the Bilate Sub-basin, Abaya-Chamo Basin, Rift Valley Lakes Basin of Ethiopia

Yimer, A. K.; Haile, Alemseged Tamiru; Hatiye, S. D.; Azeref, A. G. 2020. Seasonal effect on the accuracy of land use/land cover classification in the Bilate Sub-basin, Abaya-Chamo Basin, Rift Valley Lakes Basin of Ethiopia. Ethiopian Journal of Water Science and Technology, 3:23-50.

A correct and timely land use/land cover (LULC) classification provides indispensable information for the effective management of environmental and natural resources. However, earlier studies mapped the LULC map of Bilate Sub-basin using remote sensing images that were acquired for a single season. Hence, these studies did not consider the seasonal effects on the accuracy of LULC classification. Therefore, the objective of this study was to evaluate changes in classification accuracy for images acquired during wet and dry seasons in the Bilate Sub-basin. LULC of the study area was classified using the Landsat 8 satellite imageries. Based on field observations, we classified the LULC of the study area into 9 dominant classes. The classification for the two seasons resulted in a noticeable difference between the LULC composition of the study area because of seasonal differences in the classification accuracy. The overall accuracy of the LULC maps was 80%for the wet season and 90% for the dry season with Kappa coefficient values of 0.8 and 0.9 respectively. Therefore, the two seasons showed a significant difference in the overall accuracy of the classification. However, we discovered that when the classification accuracy was tested locally, that is for individual pixels, the results were not the same. In Bilate Sub-basin, several pixels (14.71%) were assigned to different LULC classes on the two seasons maps while 85.29% of the LULC classes remained unaltered in the two maps. According to the classification results, the season had a noticeable effect on the accuracy of LULC classification. This suggests that for LULC classification, multitemporal images should be used rather than a single remote sensing image.