Optimal Rain Gauge Network Design and Spatial Precipitation Mapping Based on Geostatistical Analysis from Co-located Elevation and Humidity Data

Measured rainfall data are important to many problems in hydrologic analysis and watershed management. The accurate estimation of the spatial distribution of rainfall requires a dense network of instruments, which entails large installation and operational costs. It is thus necessary to optimize the number of rainfall stations and estimate point precipitation at unrecorded locations from existing valued data. This paper serves 2 objectives: i) to establish a spatial representative rainfall stations from the entire existing network in the study area (i. e., rainfall-data optimization); and ii) to use of multivariate geostatistical algorithm for incorporating relatively cheaper elevation, humidity, and temperature data into the spatial prediction of rainfall at the study site. Yom river is the upstream control for the main Chao Phraya river basin. The technique was illustrated using annual and monthly rainfall observations measured at 326 rain gauge stations covering the entire basin and its vicinity. The precipitation prediction maps, generated by Thiessen polygon, inverse square distance, and ordinary Kriging, were used to determine the sensitivity of the rainfall data to the prediction results by constructing the covariance surface map. Optimal rain gauge network was designed based on the station redundancy and the homogeneity of the rainfall distribution. Ranking of sensitivity in terms of prediction error could define the priority of supplementary stations to satisfy the density of rainfall data in the basin. Digital elevation, humidity, and temperature models were incorporated into the spatial prediction of rainfall using multivariate geostatistical algorithms. The prediction performances of the geostatistical interpolation were cross validated with the straightforward linear regression of rainfall against elevation, humidity, and temperature. The results revealed that the multivariate geostatistical algorithm outperformed the linear regression, stressing the importance of accounting for spatially dependent rainfall observations in addition to the co-located elevation. The digital elevation data were highly correlated to monthly monsoon-induced precipitation in the study area. Humidity and temperature data exhibited a higher degree of correlation to the monthly precipitation data.