為了保護地下水資源，地下水位分佈的資訊極為重要。然而，濁水溪流域內地下水水位站的稀疏分佈卻限制了對於該區地下水位時空分佈的瞭解。本研究利用2006年至2015年間的地下水位月觀測資料，並結合普通克利金法（Ordinary Kriging, OK）和回歸克利金法（Regression Kriging, RK）以改善濁水溪流域地下水位的插值結果。為了瞭解降雨對地下水位的影響，RK所插值而得的地下水位將用於評估流域內地下水位與降雨間的時空交互作用，OK則僅使用於地下水位數據的插值。研究中共採用了31個地下水位站和12個雨量站觀測資料，高程和降雨量則利用RK合併為地下水位的附加變量。降雨數據由雨量站和CHIRPS觀測資料所組合而成，高程資料則由SRTM所提供。線性迴歸結果的相關係數（r）顯示，地下水位的觀測變動量中，有97％以上可以用地表高程數據來解釋，該結果表示高程數據可以作為雨量計數據的附加變量。然而由於多重線性問題，迴歸後的降雨資料不能與地下水位高程資料結合使用。相關係數、RMSE與NMSE的結果亦顯示，RK在時空上擁有比OK更強的預測能力，特別是地下水位極值的預測。空間上而言，雨季（5月和8月）地下水位升高，而地下水位擾動的最低值發生在乾季（3月和4月），多數出現於濁水溪流域的下游西部地區。此外，濕季地下水補給量與地下水水位的相關性亦相對高於乾季。濁水溪流域的地下水補給總量平均約1.40億立方公尺，為3.77億立方公尺，約為降雨量的37 ％ 上下。綜合上述，地下水資源的管理應集中在地下水補給率最高的濁水溪流域上游地區。;The sparse distribution of groundwater stations in Choushui River Basin limits spatial-temporal of groundwater level information in these region while this information was crucial needed to know for groundwater conservation purposes. This study reports on an effort to improve the interpolation of monthly groundwater level from groundwater stations using Ordinary Kriging (OK) and Regression Kriging (RK), spanning the period from 2006 to 2015. In order to know the effort precipitation to the groundwater level, the interpolation groundwater level of RK has used to assess spatial-temporal interactions between groundwater levels and recharge in Choushui River Basin. Therefore, a total of 31 groundwater stations and 12 rain gauges data have employed in this research. Basically, OK was done using groundwater level data only. Then, RK was tried to merge the elevation and precipitation as the additional variables for groundwater level. Precipitation data derived by combination rain gauge data and monthly rainfall of Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS). For elevation data, it was provided by Shuttle Radar Topographic Mission (SRTM). The correlation coefficient (r) of linear regression model proved that more than 97 % of the variability in groundwater levels observations can be explained by elevation data. It shows that elevation data can be included as an additional variables of rain gauges data. Conversely, precipitation data in regression model cannot be used in combination with elevation for groundwater levels due to multi-collinearity problem. The correlation coefficient (r), RMSE and NMSE reveals that RK has more robust prediction skill than OK in space and time, especially for prediction an extreme of groundwater level. Spatially, groundwater level elevated during wet months (May and August). The lowest level of groundwater level fluctuation was found to be from last of dry months (March & April), especially in the downstream west part of Choushui River Basin. Furthermore, groundwater recharge has derived and the correlation of groundwater recharge to groundwater level during the wet months was relatively higher than the dry months. Averagely, total amount of groundwater recharge at Choushui River Basin is about 1.40 billion m3 which represents 37 % of 3.77 billion m3 precipitation. As conclusion, the management of groundwater resource should be focused on the upstream area of the Choushui River Basin which has the highest groundwater recharge rate.
