Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Chine → 2005 → Desertification Monitoring and Evaluation Based on Spectral Mixture Analysis

Northeast Forestry University (2005)

Desertification Monitoring and Evaluation Based on Spectral Mixture Analysis

李晓松; LiXiaoSong

Titre : Desertification Monitoring and Evaluation Based on Spectral Mixture Analysis

Auteur : 李晓松; LiXiaoSong

Grade : Master’s Theses 2005

Université : Northeast Forestry University

Résumé
As one of the most serious eco-environmental problem at present, desertification has got great concern throughout the world. China is one of the countries sufferring from desertification severely. The generation and development of desertification has brought critical impact to environment, natural resources, social economy and people’s daily life in desertification-prone region. Desertification monitoring and evaluation is a very important content in desertification context. Scientific and accurate evaluation of desertification can provide scientific basis for decision-making in combating desertification.Remote sensing, with the merit of large information capacity, huge observation scope, high accuracy and speed, has become the main means of regional desertification monitoring and evaluation. Although there have many researches on desertification monitoring and evaluation based on remote sensing, the mixed pixel problem of desertified land has not been paid attention to, which has influence to some extent upon the accuracy of desertification evaluation. Taking Landsat ETM+ images as data source, this article extracted desertification information of the Mu Us sandland at sub-pixel scale by means of spectral mixture analysis (SMA), and resolved mixed pixel problem effectively. Based on SMA results, we formulated a desertification evaluation index, Sand Vegetation Index (SVI) and conducted a desertification evaluation in the whole Mu Us sandland. The main results and conclusions are as following:1) The SMA results show a well correlation with NDVI, TC transformation and supervised classification results, indicating that the trend they reflect is very similar. And SMA results are higher quantitative and higher precision vegetation and soil information extraction, thus SMA is a very good method for extracting the desertification information.2) Unconstrained SMA results give a felicitous description to endmember abundance, though sometimes endmember abundance happens to be less than 0 or more than 1. Based on field investigation data, validations was made on the precision of psammophytic vegetation fraction. The results show that their linear coefficient reaches 0.982.3) Based on the vegetation information derived from SMA, SVI was formulated in order to conduct desertification evaluation. Compared with NDVI, SVI is more sensitive to different severity of desertified land, and is more suitable for desertification evaluation.4) According to the desertification evaluation, desertified land amount to 34840.059 km2 , about 93.5% of the Mu Us sandland, among which, land area with middle and serious desertification is 26627.448 km2 and accounts for 7155% of the Mu Us sandland. The results show that desertification situation is very serious in the Mu Us sandland.

Mots clés : desertification evaluation; spectral mixture analysis; mixed pixel; endmember;

Présentation (CNKI)

Page publiée le 10 septembre 2014, mise à jour le 19 septembre 2017