本計畫之主要目的為發展一最佳陸上颱風即時定量降水預報模式，此模式預報內容包括小時降水量（或短時3/6/12/24小時累積降水量）及總降水量，以提供水庫現行運用要點防洪運轉時所需之具時效性及高準確性之預測降水量。本研究將以兩新近資料探勘之規則萃取技術分別建立兩個颱風即時降水預報模式，此兩技術分別為決策樹演算法與模糊決策樹演算法。 本計畫之工作流程包括八大項，包括颱風資料收集與分析、歷史事件選定與分類、線性迴歸雨量預報模式建立、決策樹雨量預報模式建立、簡易統計預報模式建立、模糊決策樹雨量預報模式建立、四種颱風降雨量預報模式比較以及最佳即時定量降水預報模式評估與應用等八項主要工作。其中資料收集將包括陸上雨量站雨量資料以及中央氣象局颱風期間各時段即時發佈之海上陸上颱風警報單內氣象資訊（即颱風中心氣壓、中心位置、暴風半徑、颱風預測速度與方向、中心最大風速、颱風預測位置等資料）。本計畫以淡水河流域石門水庫為研究範例，擬訂水庫上游集水區降水測站為降水預報代表站。 The major purpose of the project is to develop optimal real-time quantitative precipitation forecasting (QPF) model during typhoon periods. To provide the optimal real-time precipitations for operators on a reservoir flood-control system during typhoon, we propose the methodology to extract a set of optimal rules for forecasting precipitation. The results obtained from the optimal QPF model include the forecast of hourly rainfall (or 3-/6-/12-/14-hour accumulated rainfall) and total precipitation during typhoon periods. The two newly data mining technologies, namely decision-tree algorithm (ID3) and fuzzy decision-tree algorithm (FIDs) are employed in the extracting rules. The tasks to be carried out can be grouped into the following eight general categories: (1) collection of typhoon data, (2) selection and classification of the typhoon patterns, (3) building of the linear regression rainfall forecasting model, (4) building of the ID3 rainfall forecasting model, (5) building of the simple statistics rainfall forecasting model (such as climatology average method), (6) building of the FIDs rainfall forecasting model, (7) comparison of these four developed models, and (8) selection and application of the optimal real-time QPF model. The collected data involve the precipitations at rainfall station and the typhoon real-time warning document issued by Central Weather Bureau (the content including the data of pressure in the typhoon center, position of typhoon center, the radius of typhoon, the predicted moving speed and direction, the center maximum wind speed, and the predicted typhoon path). The developed methodology will be applied to establish the real-time optimal QPF model in rainfall stations of Shihmen reservoir watershed in the Tanshui river basin.