地基雙偏極化微波輻射計為一接收由上而下大氣輻射的儀器，其所觀測的亮度溫度(Brightness
Temperature, Tb)資料可用來估算降雨。本研究使用的雙偏極化微波輻射計，具有18.7 GHz與36.5 GHz的垂直及水平極化頻道等二個頻率、四個頻道。第一及二頻道為位於18.7 GHz 的水氣吸收線區，第三及四頻道則位於36.5 GHz 的窗區。本研究目的為利用此輻射計估算降雨強度，其方法是使用雙偏極化微波輻射計Tb 與雨量計降水強度做定量相關性分析，利用兩者間之迴歸方程式求得。初步分析顯示，大氣中液態水含量隨著大氣垂直高度增加而呈指數下降；由連續24 小時觀測Tb和垂直極化與水平極化的差值分析顯示，降雨時，Tb 隨降雨量增加而升高，而極化差值(Polarization Difference, PD)則隨之減小。在誤差分析方面，觀測資料必須考量風速風向的影響，當風向為北北西且風速約達2 ms–1以上時，觀測資料因水滴沾附在感應器表面而產生異常變化。研究結果顯示，雨量計降雨強度與輻射計Tb之間的相關性，以累積60 分鐘的降雨量與Tb的相關性最佳。研究方法為利用18.7 GHz及36.5 GHz的Tb分別與實際降雨強度進行迴歸擬合，並使用獨立資料對兩迴歸方程式進行驗證。驗證結果顯示，兩迴歸方程式反演之降雨強度均方根誤差分別為2.07及2.08 mmhr –1。

Ground-based radiometer is an instrument which can measure the downwelling radiation, and we can estimate the rainfall rate by the brightness temperature (Tb) from its observation. The ground-based
radiometer used has four channels measuring horizontal and vertical polarizations at 18.7 GHz and 36.5 GHz, respectively. The first and second channels (18.7 GHz) are close to the absorption line of water vapor and the third and fourth are in window channel. The aim of this study is to estimate rainfall intensity from the correlation between Tb observed by ground-based radiometer and rainfall rate by rain gauge.
The preliminary analysis shows that the atmospheric water vapor and liquid water contents exponentially decay with height. The continuous monitoring of 24 hours for Tb and polarization differences shows that the
brightness temperatures increase and the polarization differences decrease with rainfall rates. In the aspect of error analysis, it is necessary to consider the influence of wind direction and speed. The wrong signals appear when the wind direction is north-northwest with wind speed greater than 2 ms–1 in favor of the wet of antenna by raindrops.
The research shows that the best correlation coefficient interval between Tb and rainfall rate is the 60-minute accumulated rainfall. This study fits two regressions between rainfall rates and Tbs of 18.7 GHz and 36.5 GHz, respectively, and then verifies the regressions by independent data. The root-mean-square errors between observed and estimated rainfall rates are 2.07 and 2.08 mmhr –1 for these two regressions, respectively.