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

Accueil du site → Master → Chine → 2013 → Quantitative Method for Extracting of Desertification Information Based on TM Image

Lanzhou University (2013)

Quantitative Method for Extracting of Desertification Information Based on TM Image

马宗义; Ma Zong Yi

Titre : Quantitative Method for Extracting of Desertification Information Based on TM Image

Auteur : 马宗义; Ma Zong Yi

Grade : Master’s Theses 2013

Université : Lanzhou University

Résumé
Desertification is one of the most serious environmental problems in the world. It directly influences the living environment and social stability. Extracting desertification information accurately and quickly provides us important scientific basis for desertification control and rehabilitation. Thus it becomes one of the key issues of desertification research.On the basis of the comprehensive analyzing of present research and learning from the former research achievement, this thesis proposes an indicator system for desertification information extraction using TM image. First, according to the characteristics of the desert area, five indicators for the system were selected, which was guided by the index design principle. Then, the indicators were retrieved with appropriate methods. Finally, the desertification information was extracted quantitatively through the decision tree classification. The conclusions obtained are as follows:1. Choose the most suitable desertification concept.Different desertification concepts were compared according to the characteristics of quantitative extraction of desertification information, and the suitable one, which was put forward by Dong Yuxiang was chosen as the basis for the following study.2. Select appropriate indicators for desertification information extraction.Based on desertification conception, and combining with the characteristics of desert area, and under the guidance of the design principle, five indicators including surface albedo, vegetation coverage, MSAVI, soil moisture and texture character were selected and their relationships with desertification were analyzed. The above five indicators were conformed as indicators for desertification information extraction at the end.3. The suitability of using Landsat TM data to extract desertification information was demonstrated.At present, MODIS data and AVHRR data are used in plenty of research on desertification monitoring, but there comes its own shortcomings, which were discussed in this thesis. However, Landsat TM image has both higher spatial resolution and moderate time resolution, and has the following advantages:comprehensiveness, short cycle, convenient and speed, which can provide credible data source for dynamic change monitoring research on small and middle scale. Thus the suitability of using Landsat TM data to extract desertification information was demonstrated.4. Construct a desertification monitoring index, which could well distinguish desertification information from Gobi and saline-alkali soil land.By contrasting all essential factor’s Albedo-MSAVI feature space and the one without vegetation and water body, we found that vegetation and water body have a great influence on separation of Gobi, saline-alkali soil land and desertification land. In Albedo-MSAVI feature space without vegetation and water body, Gobi, saline-alkali soil land and desertification land showed well gathered characteristics. Therefore, by dividing the feature space, the Gobi, saline-alkali soil land and desertification land could be separated.5. Suitable methods on remote sensing retrieval of desertification indicators were selected.The retrieval methods of desertification indicators were discussed, and the suitable methods were selected. Then we retrieved them on Landsat TM image. And the distribution characteristics of the indicators in the study area were analyzed.6. The desertification information was extracted quantitatively through the decision tree classification.The desertification information was extracted quantitatively through the decision tree classification. The overall classification accuracy reached86.38%and Kappa Coefficient was0.8298. The field test shows that the indicators system can well extract different levels of desertification information in a convenient and efficient way.

Mots clés : Desertification; Information extraction; Indicator system; Index ofdesertification monitoring;

Présentation (CNKI)

Page publiée le 11 avril 2014, mise à jour le 11 avril 2018