Comparative evaluation of the accuracy of mapping irrigated areas using sentinel 1 images in the Bilate and Gumara watersheds, Ethiopia

Abstract

Precise estimation of irrigated areas is essential for effective water management, increased production, environmental conservation, and conflict resolution. Nonetheless, discrepancies frequently exist between estimated and actual irrigated areas. To address the data gaps in actual irrigation areas within Ethiopia, we utilized high-resolution remote sensing imagery. However, the accuracy of these images under varying climatic and landscape conditions was not fully substantiated. We conducted a comparative analysis between global irrigation map and local irrigated region maps within two distinct watersheds. Field data was gathered to both train and assess by employing random forest supervised classification algorithm. This algorithm was then applied to create accurate irrigation maps using high-resolution Sentinel-1 data for the Bilate and Gumara watersheds. During the irrigation seasons, maps of irrigated regions were produced using time-series imagery. Additionally, we employed maps indicating lands suitable for surface irrigation and applied post-processing techniques to refine the actual irrigated areas. The resulting accuracy was comparably high for both watersheds, with values of 88% and 87%. The kappa coefficients were 0.74 and 0.73, respectively, indicating a very good level of agreement. However, there were significant discrepancies between the global irrigation map and the local irrigated regions map in terms of spatial distribution and the extent of irrigation. This discrepancy necessitates further analysis of both products to decipher the underlying causes of their differences. We recommend for additional studies encompassing diverse watershed characteristics to improve irrigation area mapping via remote sensing. Our findings also validate the effectiveness of post-processing techniques in remote sensing applications.

Keywords

• Accuracy
• Bilate
• Gumara
• irrigation area
• random forest
• Sentinel