DETERMINING THE BEST ANNUAL RAINFALL INTERPOLATION FOR BILA WALANAE WATERSHED

Bila Walanae watershed is the largest watershed in South Sulawesi. Water resources management need to take sedimentation analysis into account. However, using USLE as sedimentation analysis requires annual precipitation data in a spatial form. On the other hand, there are many methods of spatial interpolation. The purpose of this study was to determine the best spatial interpolation method for annual rainfall Bila Walanae watershed. Monthly rainfall data from 41 observation points (14 inside watershed, 27 around outside) were used. This study compares four methods of spatial interpolations, such as Inverse Distance Weight (IDW), Thin Plate Spline (TPS), Ordinary Kriging (OK), and modified Regression Kriging (MRK). Additional data for MRK derived from TRMM rainfall estimates (2013-2014) and GPM (2015-2017). Interpolation parameters such as pixel size and variogram models determined in advance while number of neighbour (NN) is 14. Two IDW power (1 and 2) and five TPS regularization setting (1E-5, 1E-4 1E-3, 1E-2 and 1E-1) is used. Error is generated through the leave-one-out cross validation by calculating RMSE and MAE. Result shows that TPSE-1 generate small error with highest correlation (MAE=648; RMSE=799; r=0.45) followed by IDW (646;799;0.40), OK (666;810;0.43), and then MRK (1410;3176;-0.11). Disuse of outside points increases MAE about 9%. Result of TPSE-1 interpolation is then averaged over five years and compared with land cover. Relatively high annual rainfall pattern is seen in area with forest in southern part of the watershed and around Lake Tempe that has no forest. Projections RCP by 8.5 shows increasing precipitation pattern in those area. Therefore, further studies related to priority of those two regions are needed