| 摘要: | 近年來極端氣候事件、地緣政治危機與區域安全議題等逐年增加，造成全球各地大規模與區域性農糧危機上升，使糧食安全如同能源安全、金融安全等主要經濟安全核心受人注目，也是除了傳統政治、軍事與外交等安全議題外的「非傳統國家安全」重要一環，當國家面對大規模天然災害，導致農產歉收，若無法給予人民主觀感受的安全感，或客觀的糧食安全存在事實，人民將對政權與執政團隊，失去保障安全能力的信任感，進而可能造成國家內部政局動盪，獨裁、專斷政體甚至可能以國家利益為前提，藉由爭取區域安全平衡，透過傳統政、軍、外交與經貿等政策工具，形成對外擴權、爭戰的局面，影響區域穩定及和平。
隔鄰的中共政權，長期將保障糧食安全放在重要戰略地位，面對如2020與2021年的大規模洪災與農糧災損情形，是否會如實地展現農糧不足議題嚴重性，難以由其官方各項農產數據獲得真實情形，但從其事後大量從其他各國進口農糧產品，進而抬升國際農糧市場價格觀察，我國或東亞各國若缺乏對此類議題長期自主監控能量及態勢感知（Situation Awareness）能力，恐將影響日後各國糧食資源分配、進出口談判、農糧採購事宜等政策規劃，進而也會削弱我國對中國大陸政局提早應變能力，甚至危及我國家安全及區域政經穩定。
隨著衛星影像來源的多元化與資源的豐富度與日俱增，研究範圍的時空尺度隨之更為廣泛，當前衛星遙測影像對於大規模農糧災損資料獲取，已可快速提供跨時空方式的多時期、廣域面積數據（尤其在外人不易進入區域），整合歷年已公開的農糧調查統計資料，可獲得近即時的分析資訊，符合大面積監測的成本效益。但影像與空間資料的分析及計算量也相對暴增許多，針對此一科研發展趨勢及大數據分析挑戰，谷歌（Google）公司開發並公開名為「谷歌地球引擎（Google Earth Engine，GEE）」的雲端計算平台，提供全球尺度的地理空間分析服務，藉由Google巨大的線上計算能力，有效解決需要處理廣大面積及長時間監測的大規模地理數據及影像需求，本研究即利用此項網路資源，分析Sentinel-1雷達影像、Sentinel-2光學影像，以執行影像監督式分類、產製每月洪水覆蓋圖等流程，驗證將前述多元資料運用在災後區域糧食安全快速評估環節的可行性。;In recent years, extreme climate events, geopolitical crises, and regional security issues have increased year by year, resulting in the rise of large-scale and regional agricultural and food crises worldwide. Making food security, like energy security, financial security, and other major economic security core issues attract attention. It is also an essential part of "non-traditional national security" outside of the traditional political, military, and diplomatic security issues. When the country faces large-scale natural disasters and poor agricultural harvests, if the people cannot have a subjective sense of security or an objective fact of food security, the people will lose their trust in the government and the ruling team to ensure security. Under this situation, it may lead to political turmoil within the country. Dictatorship and authoritarian regimes may even take national interests as the premise. By striving for a regional security balance, and through traditional political, military, foreign, economic, and trade policy tools, they can form a situation of external power expansion and war, affecting regional stability and peace.
With the diversification of satellite image sources and the increasing abundance of resources, the temporal and geospatial scales of the research scope become wider. The current satellite imagery can quickly provide multi-period and wide-area data across time and geospatial for the acquisition of large-scale agricultural disaster damage data (especially in areas that are not easily or allowed accessible by outsiders). Coupled with the integration of agricultural survey statistics that have been published over the years, the near-real-time analytical information can be obtained, which is cost-effective for large-scale monitoring missions. However, the amount of analysis and calculation of imagery and geospatial data has increased dramatically.
In response to this scientific research development trend and the challenge of extensive data analysis, Google has developed and made public a cloud computing platform called "Google Earth Engine (GEE)" to provide global-scale geospatial analysis services. Google′s substantial online computing power can effectively solve the need for a large area and long-term monitoring. This research uses this internet resource to analyze Sentinel-1 SAR imagery and Sentinel-2 optical imagery. With this supervised classification and related processing procedures, we can produce the monthly flood coverage maps to verify the feasibility of applying them to the rapid food security assessment in post-disaster regions. |
