臺灣地區月降雨及溫度1公里網格資料庫之建立（1960-2009）及其在近未來（2015-2039）的氣候推估應用

The Construction of Monthly Rainfall and Temperature Datasets with 1km Gridded Resolution over Taiwan Area (1960-2009) and Its Application to Climate Projection in the Near Future (2015-2039)

翁叔平(Shu-Ping Weng) ； 楊承道(Cheng-Tao Yang)

《大氣科學》 40卷4期 (2012/12) Pp. 349-369

摘要

蒐集散置於不同單位、時空分佈上不均勻的測站月降雨量暨月溫度資料，本研究利用測站的地理資訊，對缺失資料進行補遺，先產生時間上完整的測站資料後，再利用空間內插方法，獲得一組針對臺灣本島的高空間解析度（1公里）、長期的（1960/1～2009/12）網格化降雨和溫度資料庫。利用jackknife procedure，我們分析了源自補遺方法所導致的誤差大小。我們也檢驗了資料庫所呈現的降雨和溫度的季節循環和以恩索（El Nino-Southern Oscillation; ENSO）為目標的年際變異，並分別與較粗糙的APHRODITE （Asian Precipitation Highly-Resolved Observational Data Integration Toward Evaluation of Water Resources）網格化（0.25×0.25度）降雨資料以及CRU（Climate Research Unit）的網格化（0.5×0.5度）溫度資料，進行比較。使用日本MRI/JMA （Japan Meteorological Research Institute/Meteorological Agency）解析度20公里的區域氣候模式所模擬的歷史（1979～2003），以及推估的近未來（2015～2039）模式資料庫，本研究進一步應用BCSD （bias-correction and spatial disaggregation）統計降尺度方法到產製的高解析度資料庫，針對臺灣本島在近未來（2015～2039）的雨量及溫度（包含均温、最高溫與最低温）的可能變化進行推估。結果顯示：（1）近未來北台灣春季（4月）的降水有減少的趨勢；南部地區的夏季（7月）雨量則大增；（2）溫度在近未來的變化，則呈現普遍增溫的趨勢；但是，夏季（7月）溫度增加的幅度則小於其他季節。

關鍵字

資料補遺 ； 空間內插 ； 統計降尺度 ； 區域氣候模式 ； 氣候變遷推估

並列摘要

After collecting station rainfall and temperature data from various sources, which are spatiotemporally inhomogeneous in nature, this study first performs the missing data imputation by utilizing the station geographical information to create a complete time series at each station. Long-term (January 1960 - December 2009) monthly rainfall and temperature datasets with a 1km horizontal resolution are then constructed in terms of a conventional spatial interpolation technique. Adopting the Jackknife's take-one-out procedure, we also analyze and illustrate the data uncertainty in a gridded manner. Meanwhile, the depicted highly resolved seasonal climatology is analyzed and compared with the coarser APHRODITE (Asian Precipitation Highly-Resolved Observational Data Integration Toward Evaluation of Water Resources) rainfall (0.25° by 0.25°) and CRU (Climatic Research Unit) temperature (0.5° by 0.5°) datasets. Using the historical (1979-2003) and projected (2015-2039) products simulated by the Japanese MRI/JMA regional climate model (with a 20km resolution), this study is further applied with a BCSD (bias-correction and spatial disaggregation) statistical downscaling method to our high resolution datasets to project the possible rainfall and temperature (including the mean, maximum, and minimum values) changes over Taiwan in the near future of 2015-2039. The results show that (1) there exists a significant decreasing (increasing) trend in the springtime (summertime) rainfall over northern (southern) Taiwan, and (2) universally temperature-related change has an increasing trend across this island, though the increased amplitude is smaller in summer as compared to other seasons.

並列關鍵字

Data Imputation ； Spatial Interpolation ； Statistical Downscaling ； Regional Climate Model ； Climate Change Projection