臺灣位處地震活動頻繁地帶，提前地震預警的時效是相當重要且需求迫切。現行的強震即時警報系統僅能利用地震波傳遞速度的差異，在破壞性地震波到來的數十秒前發布即時警報，如此短暫的應變時間往往造成重大災損。自1988年以來許多研究利用衛星影像分析鄰近震央之斷層帶的地表溫度，已然發現地震前數天，地表溫度的變異明顯。然而以往的研究多半採用繞極衛星資料進行分析，因受限於觀測的時間週期，無法獲得斷層地帶連續性地表溫度或地熱排放訊號之變化。本研究應用自適應時空反射率融合模式（Spatial and Temporal Adaptive Reflectance Fusion Model, STARFM）於Landsat 8/9及Himawari-8資料，提供空間解析度30公尺的逐時高時空解析度地表溫度產品，與相同時空對應的Landsat 8/9資料比較後，獲得回歸線斜率0.994、決定係數（R平方值）為0.52的相關結果，顯示本研究產製的高時空解析度地表溫度產品，與Landsat 8/9衛星實際觀測資料極為吻合並可應用於後續分析。 本研究共選擇三起地震個案進行分析；個案一是2022年9月17日起，發生於臺東池上斷層的一系列地震，其中芮氏規模高於5的地震共有四次。透過分析9月1日至9月22日期間，池上斷層在無太陽輻射加熱的清晨時段，依舊存在較高的地表溫度，甚至有增溫之趨勢，配合地物類型大多屬於農田與樹林的地表特徵，增加了地熱向上傳輸的可能性。然而這類震前增溫的特徵並未出現在另外兩起位於花蓮的地震個案中，該兩起地震個案明顯呈現地震前地表溫度/地熱下降的趨勢。進一步分析震源機制（地殼錯動型式）後，發現該兩起位於花蓮的地震主要屬於逆斷層，在擠壓過程中阻礙地熱向上傳輸，導致衛星觀測地溫下降，待地震發生（地殼錯動）後，地熱才明顯逸散出地表，說明不同的斷層錯動方式可能影響地熱向上傳輸的效率及地表溫度。本研究對於後續斷層帶或地震活動預測應用，除繼續探討震源機制與地溫/地熱異常的互動外，將藉由衛星高時空解析融合影像在斷層帶區域熱點之分析，以及其位置、強度是否會因地震事件而變化等特徵，期能提供臺灣地熱資源開發之評估。 ;The current earthquake warning can only inform the masses a few minutes before the arrival of the earthquake wave. It is easy to cause serious damage in the case of less preparation time. Since 1988, the academic community has used satellite images to analyze the land surface temperature （LST） and point to the positive geothermal anomalies around the fault zone strong enough to be observed by satellites starting days in term of LST before the earthquake. However, these previous studies use polar-orbit satellites as a dataset which is not enough to see the detailed changes in geothermal anomalies. This study applied spatial and temporal adaptive reflectance fusion model （STARFM） to Landsat 8 and Himawari-8 images to synthesize Landsat-alike images with 30 m resolution hourly. The scatter plot shows a slope of 0.994 and an R-squared value of 0.52 between fused images and the corresponding Landsat images. The numerical results that are so close to the 1:1 correspondence and have correlations indicate that the fused images generated in this study are sufficient to be regarded as high temporal resolution Landsat 8/9 data for subsequent analysis. Three earthquake cases were selected for analysis in this study. The first case analyzed in this study is a series of large-scale earthquake events that occurred in Taitung Taiwan since 17th September, 2022. There are 4 earthquake events with a Richter magnitude scale greater than 5. By analyzing the LST extracted along the Chihshang fault zone in the period of 1st to 22nd September 2022, the result shows that the temperature in the southern section of the fault is relatively high and even the temperature rises in the early morning when is no solar heating. In addition, this study uses GIS stacked maps to confirm that the types of ground features in the southern section of the fault belong to farmland and trees, and there are almost no buildings. Such geomorphic features increase the possibility that the heat on the surface is transmitted upwards from the ground. However, this pre-earthquake warming trend did not appear in the other two earthquakes in Hualien, but there was a slight cooling trend. After further analysis, it was found that the two earthquake cases in Hualien had different focal mechanisms from Case One. Different fault dislocation methods may lead to different efficiencies in the upward transfer of geothermal heat. For example, the two earthquakes in Hualien belonged to reverse faults, and the heat in the ground is not easily transmitted during the compression of the strata. It was not until the earthquake occurred that the surface ruptured and the geothermal heat appeared. However, the characteristics of these events alone are not enough to predict seismic activity. In addition to the correlation between geothermal anomalies and focal mechanisms, follow-up research can also try to find out whether there are fixed hotspots around the fault, and whether their location and intensity will be changed by seismic events. Such hotspots may also assist in siting for geothermal resource development.
