衛星降水資料（SPDs）可提供全球大規模空間覆蓋的降雨數據，且具有不同時間的解析度，因此預期可應用於高衝擊天氣（如颱風）和滑坡發生的相關性研究。本研究的第一個目標，即是客觀地評估 SPD 的性能。首先選定印尼的巴厘島省，因其自然災害脆弱性風險高而被選為測試評估研究區，也針對高衝擊天氣現象，選取有侵襲菲律賓呂宋島的颱風事件，進行在多重降水情景下對 SPD 進行全面性客觀的評估。第二個目標是應用適當的 SPD 來評估巴厘島滑坡發生的平均降雨強度和持續時間 (I-D) 以及累積強度和持續時間 (E-D) 閾值。第三個目標是應用降雨強度的優化調整來評估巴厘島滑坡發生的 I-D 和 E-D 閾值。評估 SPD 性能係採取一個客觀定量分析方式，亦即「連續統計測量」和「體積指數」，而另一方面，冪律技術則用於評析 I-D 和 E-D 閾值。分析結果顯示，IMERG 資料與雨量計觀測數據具有良好的一致性，且在侵襲菲律賓呂宋島的五個颱風事件中，降雨量的推估方面表現優於其他幾種，並且在不同風速下判釋強降水的能力也很高。 GSMaP 展示了在高海拔地區的強降水能力，而 IMERG 則有在低海拔強地區的較佳降水推估能力。 IMERG 數據集在每日、五天和季節尺度上表現出色，而 CHIRPS 在巴厘島省的月降雨量方面表現最佳。IMERG 數據集還描述了在低海拔地區的良好性能，而 GSMaP 在高海拔地區表現出更好的性能。在於應用 IMERG 資料方面，經由 I-D 和 E-D 閾值的結果，觸發巴厘島滑坡事件的主要降雨特徵是長期持續的高強度前期降雨。考察結果顯示20%的閾值最適合 IMERG 估計巴厘島省滑坡發生率。使用平均偏差強度 (MD) 和偏差因子 (BF) 調整降雨強度取決於I-D 和 E-D 閾值方面分別優於其他調整模型，概率水平分別為 5% 和 10%。綜合目前研究中的 I-D 和 E-D 閾值與過去研究的比較表明，使用E-D 閾值可以大幅降低 SPD 的不確定性，這表明未來可使用衛星降雨數據集，並且建立 E-D 閾值的高度可應用性。 ;Satellite Precipitation Datasets (SPDs) provide rainfall data on global spatial coverage and different temporal resolution have the potential to be applied in high-impact weather (typhoons) and landslide occurrence because the ground-based observation needs to maintain, the coverage observation is not widespread enough, and limited in the mountain areas. The first objective of this study is to evaluate the performance of SPDs objectively. In addition to the fair weather, SPDs under heavy precipitation events are investigated as well. Thus, Bali Province is chosen as the study area for its high risk of natural disaster vulnerability, while the typhoon events in the Philippines represent severe weather phenomena. The SPDs could expect to be evaluated comprehensively under various precipitation scenarios. The second objective is to apply the appropriate SPD in determining the mean rainfall intensities and duration (I-D) and cumulated intensities and duration (E-D) thresholds for landslide occurrences over Bali Province. The third objective is to apply the optimal adjustments of rainfall intensity in determining the I-D and E-D thresholds for landslide occurrences over Bali Province. Quantitative analysis used to assess the performance of SPDs are the continuous statistical measurement and volumetric indices. The power-law method was used to represent the I-D and E-D thresholds. The analysis results show IMERG dataset shows good agreement with rain gauge observations and performs significantly better in detecting rainfall during five typhoon events over the Philipines and also high capability to identify heavy precipitation in different wind velocities. The GSMaP demonstrated the highest ability to recognize heavy precipitation in high altitudes, while the greatest capability to identify heavy precipitation at low altitudes was demonstrated by IMERG. The IMERG dataset outperformed on daily, Penta-day, and seasonal scales, while CHIRPS achieved the best capability on monthly rainfall over Bali Province. The IMERG dataset also depicts good performance at low elevations, while GSMaP shows greater performance at high elevations. This study also demonstrated the result of the I-D and E-D threshold for landslides over Bali Province by using the IMERG early run dataset. The dominant rainfall characteristic triggering the landslide events over Bali Island is a long-term duration with high-intensity antecedent rainfall. The threshold of 20% is the most appropriate for IMERG in estimating the landslide occurrences over Bali Province. The adjustment of rainfall intensity using the mean deviation intensity (MD) and bias factor (BF) outperforms other adjustment models in determining the I-D and E-D thresholds at the probability levels of 5% and 10%, respectively. Comparison of the I-D and E-D thresholds in the current study with past studies exhibits that the E-D threshold can reduce the uncertainty in SPDs, this indicates a high possibility of using the satellite rainfall datasets to establish the E-D thresholds.
