Establishment of vegetation recovery model using satellite imagery

• Main Authors: Chih-Chuang Hsiao, 蕭志全
• Other Authors: 楊明德
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/00463732890329762610

碩士 === 國立中興大學 === 土木工程學系所 === 94 === Vegetation has been considered a great contribution to the prevention and mitigation of landslides. Being shacked by the Chi-Chi earthquake in 1999, the central Taiwan has a dramatic topographic change. Jioujiou Mountain is one case of the serious landslides induced by the Chi-Chi earthquake. Since then, the authorities have been spending a great amount of financial and labor costs in vegetation recovery so to avoid the further disasters possibly caused by mudslides and landslides. If we could predict the recovery pattern of vegetation, we could estimate what the natural power will do to the landslide, how to adopt appropriate recovery strategies, and when to expect an ideal vegetation recovery. This thesis applies satellite imagery and image processing techniques to estimate a vegetation recovery model for Jioujiou Mountain. With the analysis of 8 SPOT satellite images taken on 1999/04/01, 1999/09/27, 2000/04/18, 2001/01/02, 2001/11/10, 2002/10/17, 2003/02/07, and 2004/07/12 in Jioujiou Mountain, this research aims to establish the vegetation recovery model in Jioujiou Mountain since the earthquake. Shot before 921 earthquake, the image taken on 1999/04/01 is adopted as a reference image and compared with the other seven images. Through NDVI (Normalized Difference Vegetation Index), an overall vegetation condition can be estimated by remote sensing data. However, a proper radiometric correction is particularly important for an analysis of multi-date images. To decrease the radiometric error, two approaches, Linear Regression Normalization based on artificial structures and Histogram Matching in vegetation well-covered areas, are employed in this study in addition to the simple radiometric correction of NDVI itself. According to the mean and variation of DNDVI (difference of NDVI) from Linear Regression Normalization and Histogram Matching, the latter indeed makes better radiometric correction. In this research, the variation of NDVI generated from corrected SPOT images is used to represent the variation of vegetation in different phases. The ideal vegetation recovery model is obtained as follows:VRR=(e^(-0.00626*(t+808))-0.00636)/(-0.00626). This model predicts that the vegetation recovery rate in Jioujiou Mountain will reach 100 percent under a perfect condition after 660 weeks.