以中尺度成像光譜儀影像研究中國內蒙古自治區沙漠化時空變化歷程

Abstract

衛星影像可用於監測大規模地表變遷現象，不但同時提供了地表的生物、物理特性，也提供了低成本的研究資料，而且輕易跨越了政治與地形的限制。本研究利用2000-2006年Terra衛星中尺度成像光譜儀 (MODIS) 影像觀察中國內蒙古自治區之地表沙漠化分布及變遷，希望對沙漠治理有基礎幫助。在研究方法上將衛星影像透過去雲、運算等程序，並轉換為修正土壤植被指數 (MSAVI)，並以最大值合成法 (MVC) 產生歷年圖像，以供觀察地表綠度 (Greenness) 之空間分佈與時間變化；理論上在單位研究時間單元 (年) 內綠度最大的時候代表地面的植生量 (Biomass) 最大、最好的植生狀態，有效指示出了「乾燥狀況」最保守估計值。研究最後整合氣候與人為因子進行相關分析以了解現象發生原因。 由歷年的MSAVI觀察發現沙漠化並非一不可逆之地表植生退化過程，隨著時間其邊界在年間甚至會有大幅度的變動，大致上在研究期間內並未發現明顯的「退化」，但有發現很明顯的不穩定「波動」情形。因子相關分析方面，研究顯示對短年期的沙漠化現象而言，氣候因子的年累積降雨量是主要的影響地表植被生長的因子，不論是分佈或變動都呈現顯著的相關性與較高的相關值；人為因子，包括耕地面積、有效灌溉面積、羊畜養量及大牲畜畜牧量等，影響相對較有限，其相關係數無論在空間分佈上或時間變遷的分析都呈現較低或不顯著的數值結果，最後並利用空間單元的因子變遷性質濾定出最需要進行治沙計畫的地區，希望研究成果可供相關單位參考。

Desertification is an environmental problem of major importance in arid environments. Remote sensing has been an efficient technique and comprehensively applied in researching detailed analyses of spatial distribution and surface changing for the phenomenon of desertification. In this study, we examined the spatial and temporal ecological progress of IMAR in north China between Beijing and the Gobi desert. 2000~2006 biweekly/1 km resolution vegetation indices products, derived from MODIS instrument on Terra, were downloaded from Earth Observing System Data Gateway website. After cloud removal process, MVC method was adopted to produce annual MSAVI map, which conservatively estimate desertification condition. Furthermore, by using correlation analysis, this paper present studies on the relationship between climate, human activity and MSAVI in county-based. In the research, it shows that desertification is not an unconvertible retrogression of surface greenness, and the contour of isolated MSAVI values oscillated year by year. During the study duration, we did not detect significant vegetation degradation, but apparent vibration of biomass. Comparing with the 7 images we found the desert area shrunk and spread acutely. In regression analysis, it indicated that annual accumulated precipitation is the principal effective factor and showed higher positive correlation than every human activity factor in spatial distribution and temporal change. The results would provide authorities for arid region management and decision making.