以Landsat-TM及SPOT衛星影像監測高山地區土地利用變遷之研究

Using Landsat-TM and SPOT Images to Monitor the Landuse Changing on High Mountain Area

陳文福(Wen-Fu Chen)

《中華水土保持學報》 26卷3期 (1995/09) Pp. 183-196

https://doi.org/10.29417/JCSWC.199509_26(3).0004

摘要

爲確保集水區環境之安全，政府始終致力於山坡地超限利用的管理。惟限於人力、物力、法令不彰及蒐證之困難，一直難有成效。本研究乃著眼於大範圍之山坡地區土地利用濫測方法之研究，由遙測技術將衞星影像予以處理與分類並將分類之結果配合地理資訊系統之分析功能進行山坡地區土地利用變遷之偵測，並將不同年份土地利用分類之結果建成資料庫以供分析比較。其結果除了能掌握集水區土地利用變遷之趨勢外，亦能針對定點與定項之土地利用別加以探討與分析，以提供有用之資訊。本研究以合歡溪集水區與梨山地區爲範團，第1年因雲量過多與拍攝角過大造成幾何校正困難，故只買到美國地球資源探測衞星Landsat-TM之1984與1989兩年，雲量少於十分之一之第2、3、4三波段之兩張衞星影像加以分析。先以影像處理軟體ERDAS行影像處理及地表覆蓋分類，製成不同年份之土地利用圖，並以地理資訊系統軟體ARC/INFO將基本圖層數化且予以疊合與分析，以了解土地利用之變遷情況。第2年則只買到法國SPOT衞星1993年之一張影像進行與第一年類似之影像處理與分類。並加強山區現場之地質調査。所得結果與第一年者加以比較與分析，卽能了解不同年份與不同解析度之衞星影像間土地利用分類結果之差別。其結果顯示：（1）研究發現雖然高山地區由於天氣不穩定雲量常偏高，且SPOT衞星拍攝之角度超過24度則影像卽無法進行幾何校正，以致理想之衞星影像資料取得不易。但由一張SPOT衞星在1993年冬季拍攝的影像，與Landsat-TM在1984年8月及1989年6月間攝得之影像進行比較。經處理與分類後仍能獲得良好之土地利用變遷監測效果。（2）1989至1993四年間各土地利用別的改變面積最大者爲針葉林增加3.77平方公里，最小者爲濶葉林僅增加0.02平方公里，果園則減少了0.67平方公里，足見濫墾情形已漸趨緩和。但蔬菜面積卻大幅增加約1.449平方公里，值得注意。（3）由本研究之經驗建議：以地球定位系統測量訓練樣區並配合監督性分類法，將使具有高解析力之SPOT衞星影像分類結果更好。

關鍵字

集水區 ； 超限利用 ； 地理資訊系統 ； 地眞核對 ； 地球定位系統 ； 變遷監測 ； 訓練樣區 ； 監督性分類法

並列摘要

In order to improve the security of the watersheds in Taiwan, although our government has endeavored to forbit the overusing on hillslope lands, but the results have not been so satisfied because of some limits. This study focus on larger range of hillslope area and to find out the available methodology to moniter the landuse that on the hillslope. We used the different images of satellite to process and classfy and used the results to join the analysis function of GIS to erect the database and make change detection of landuse. With the results not only we can know the trends of landuse but we can also investigate and analyse the fixed position and fixed items of landuse, it can afford us relative and available information. The task of the first year of this study was to use the 2nd, 3rd and the fourth bands of the images, the cloud on those images must less than one-tenth, of Landsat-TM at 1984 and 1989 to process and then analyses with image processing software ERDAS. The results of classification was used to erect landuse map of various years and digitize the basic maps by ARC/INFO and then to overlay and analyse them. In the second year, we added the image of SPOT satellite of France at 1993 in the same study area and improved the ground truth checking. The results of classification was taken to compare with those of the first year and to analyse the difference of different resolution of images of different satellites. The results were as following: (1) Owing to the unstable weather of the high mountain area, although the cloud always much and the images could'nt be corrected when the taking angle of SPOT was over 24 degree and the ideal image was very few, but from the comparision of one image of SPOT was taken in winter with two images of Landsat-TM those were taken in summer, those were different in resolution, we could get good efficiency in land use change detection. (2) The difference of area of various landuse during 1989 to 1993 were 3.77 square meters increased in conifer areas, 0.02 square meters increased in hardwood, and 0.67 square meters decreased in orchard fields and we could see the situation of misuse of the watershed went less but we should take care that the area of vegetations fields was increased greatly. (3) The experience showed us that the results of the classfication of high resolution SPOT images will become better if we use GPS to survey the training areas to match the supervise classfication method.

並列關鍵字

Watershed ； Overusing ； GIS (Geographic Information System) ； Ground Truth Checking ； GPS (Grobal Positioning System) ； Change Detection ； Training Area ； Supervise Classification Method