Determination of homogeneous regions for regional reference evapotranspiration estimation using the self-organizing map in western Taiwan

Although, the reference evapotranspiration (ET0) had been studied in many places, but when data at a given location are insufficient for a reliable estimation of the quantiles, a regional estimation must be performed to obtain ET0 in ungauged site. In this process of regional estimation, the sites must be assigned to homogeneous regions, because approximate homogeneity is required to ensure that ET0 is more accurate than at-site analysis. In this study, the self-organizing map (SOM) is applied to identify the homogeneous regions for reference ET0. First, several site characteristics are chosen as the relevant attributes in cluster analysis based on Penman-Monteith (PM) equation. Then, the SOM is compared with two traditional clustering methods, the K-means method and Ward's method, using the heterogeneity measure. Finally, the SOM is applied to actual meteorological data in western Taiwan to identify homogeneous regions for regional reference ET0 estimation. The heterogeneity measure indicates that the SOM can identify the homogeneous regions more accurately, compared to the other two clustering methods. In addition, the results show that the four regions are sufficiently homogeneous. The conclusion is that the SOM is more robust than the traditional clustering methods, and that it is recommended as an alternative to the identification of homogeneous regions for regional reference ET0 estimation.