1961-1990臺灣的30年月均溫與月雨量空間推估-導入地形特徵

本研究選用臺灣地區1961年至1990年平均月雨量與月均溫資料，將測站資料空間推估成網格解析度300公尺的面量資料。推估方法參考克里斯多福戴利的獨立坡向指數高程迴歸模式(Parameter-elevation Relation on Independent Slop Model, PRISM)。該模式假設氣候因子與高度呈簡單線性關係，若推估地點與已知測站地形特徵相似，則在迴歸統計中測站會被給予較高的權重，反之亦然。此方法考量地點、距海遠近與坡向…等，可有效模擬不同地形特徵的氣候差異。本研究將雨量與溫度資料進行檢定，透過補遺基期較短的資料，使得推估成果能更佳反映1961至1990的平均氣候，最終採用240個溫度與1052個雨量測站進行空間推估。透過3次交互驗證法檢定均方根誤差，結果顯示PRISM比起克利金法或距離反比權重法有更好的推估成果。此結果，可供後續進行集水區資源管理、氣候變遷與植被分布推估應用。雖然目前已經有氣候地圖集以及其他推估成果，但是PRISM能提供更多空間上的細節。本研究於附件提供模式編碼供後續研究者使用。

關鍵字

獨立坡向指數高程迴歸模式；雨量；氣溫；空間推估；數值地形模型

並列摘要

The spatial climate data sets of 1961–1990 mean monthly precipitation and temperature were estimated for Taiwan areas. The result was converted to a map which the grid’s resolution were up to 300 meters. The spatial climate data was referred to the PRISM (Parameter-elevation Relationships on Independent Slopes Model) which was developed by Christopher Daly. This model was based on the climate-elevation simple linear regression and built terrain factor indices including distance, elevation, coastal proximity, topographic facets and vertical atmospheric. Stations entering linear regression are assigned weights based on terrain factor index. By using PRISM, it was effective to show the climate differences in the different terrain factors. This study deleted the qualitative data errors and adjusted the short-period-of-recorded data from 1961 to 1990. In the end, the spatial interpolation result used the data from the 240 temperature stations and 1052 precipitation stations By using 3-fold cross-validation to verify root-mean-square error (RMSE), PRISM was more accurate than Kriging and IDW. Therefore, the spatial climate data could be the reference for watershed resource management, climate change, mapping vegetation type and mapping ecological zones. Even though there were climate maps and other research, PRISM provides more spatial details. The model's source codes were attached in the end of the study.