氣溫及降水是決定生態系、植群之類型與空間分布之重要環境因子，如何利用分散的氣象測站資料推估出合理的氣候空間，以滿足各種環境生態模式應用上的需要，仍是迫切但尚待解決的全球問題。本研究以經過資料檢定程式的219個氣象站及877個雨量站資料做為空間推估分析所依據之默狀氣候資料，結合數值高程模型（Digital Elevation Model, DEM），以測站海拔、座標先行建立各月氣溫之迴歸模型，再套疊殘差之空間推估結果，可大幅提高氣溫推估之精確度，尤其在測站稀少的山區。降水空間推估方面，經交叉驗證（cross validation）之誤差均方根（Root Mean Square Error, RMSE）比較各種不同推估方法後，以先將資料對數轉換後再進行一般克利金（Ordinary Kriging,　OK）推估之結果最佳。本研究與奧勒岡州立大學PRISM模型（Parameter-elevation Regressions on Independent Slopes Model）、中央氣象局氣候圖集之結果相比對，發現除了山區一些極值差異外，其空間分布特徵是一致的，部分差異來源可能是選用測站和其資料品質的不同所造成。本文之結果可與其他生態和植被參數結合來繪制臺灣地區詳細之生熊氣候圖，最後本文也檢討空間推估在區域氣候研究尚待努力之方向。

　　　　　Temperature and precipitation are important environmental factors to determine the types and spatial distribution of ecosystem and vegetation. The technique of estimating reasonable spatial distribution from the separated meteorological station data for ecological model applications, are the critical but not solved completely issue in the world. We used the 219 meteorological stations and the 877 rainfall stations which had passed quality checking procedure as the base for spatial interpolation. Combining with Digital Elevation Model (DEM) database, the regression among the elevation, coordinates and monthly-average temperature at the meteorological stations were calculated first. Then they were overlay by the interpolated results of regression residual method to get the best spatial estimation of air temperature. The result improved substantially the spatial precision of air temperature, especially at the mountainous area where had sparse meteorological stations. After the comparison of different methods relayed to the cross validation of mot mean square error (RMSE), we found the logarithmic transform of precipitation data combined with Ordinary Kriging (OK) method provided the best result of precipitation spatial interpolation . The results in this study were compared to The Parameter-elevation Regressions in Independent Slopes Model (PRISM) of Oregon State University and the atlas of climate made by Central Weather Bureau. The spatial patterns among these sources were similar to each other, except the extreme values at mountain areas. It might be caused by the selection of meteorological stations and different data quality process. The results in this study could be combined with other ecosystem and vegetation parameters for making detailed eco-climate diagrams at Taiwan. We also gave discussion on the future work of spatial interpolation for regional climate research.