Comparative Analysis of Different Spatial Interpolation Methods for Estimating Distribution of Monthly Rainfall (Case of Genale Dawa River Basin, Ethiopia)

Dereje, Fituma

Comparative Analysis of Different Spatial Interpolation Methods for Estimating Distribution of Monthly Rainfall (Case of Genale Dawa River Basin, Ethiopia)

Date

2021-02

Authors

Dereje, Fituma

Publisher

Addis Ababa University

Abstract

For most hydrological modeling and water resources development plan studies, the distribution of rainfall data is a very important input parameter. The spatial distribution of rainfall is usually accessible from the limited number of rain gauge stations especially in developing countries; this limitation increases the necessity of using a suitable interpolation method to obtain the spatial distribution of rainfall. Spatial interpolation methods are used to predict the values of the unknown place using known location points. In this study, the main objective to select a suitable interpolation method for estimating the monthly distribution of rainfall data for the Genale Dawa River Basin. Using the frequently used methods for comparison of estimating the Rainfall value in the kinds of literature done before and passing spatial interpolation method selection criteria, by considering the sample size available and spatial variability on monthly rainfall data in the study area. As result, four spatial interpolation methods were selected for this study namely: Inverse Distance Weighting (IDW), Spline with Tension (ST) Ordinary Kriging (OK), and Universal Kriging (UK) methods. The historical mean monthly Rainfall records from a total of 23 rain gauge stations selected for this study in the river basin which covers the area of 172,889km,2during 2000-2018 period used for the analysis were used to compare the interpolation methods. The performance evaluation of selected interpolation methods evaluated through a cross-validation test using the Geostatistical analyst tool. The spatial interpolation methods were compared with the (RMSE) Root Mean Square Error and Mean Absolute Error (MAE). The cross-validation result specifies that the Geostatistical spatial interpolation methods perform better than deterministic methods, particularly the OK interpolation method relatively has the lowest RMSE and MAE error metric at the River Basin level. Therefore, the OK is the suitable spatial interpolation method for estimating the mean monthly rainfall distribution for this study area. The UK, ST, and IDW methods are followed by in respectively. The Ordinary Kriging method similarly is the suitable method for estimating seasonal rainfall in the Belg and Kiremt seasons, whereas the UK is the suitable method for the Bega season. The suitable spatial interpolation method generating Map of areal mean rainfall distribution in each month, the monthly rainfall distribution decreased from northwest to southeast in the study area.