A Study On Desertification Based On RS And GIS Methods In Dunhuang City

Desertification is one of the most serious environmental and social problems all over the world which causes great damage to the eco-environment and social economy. China is one of the countries that are severely threatened by desertification in the world. Remote sensing provides a fire-new measure for monitoring desertification. Due to the complexity of desertification, interactive interpretation is still the principal method even though it has several limitations, such as orientation error, low time-effectiveness, long cycle time, and huge workload etc. So it is very necessary and meaningful to establish an automatic classifying method for monitoring desertification.As one of the famous national historical city and the attractive tourist destiny in China, Dunhuang possesses a great number of cultural relics and historic sites which are noted all around the world and it has unique nature scenes. However, in recent decades, the rapid development of land desertification has threatened the local environment and people's livelihood. As for this, WEN Jiaobao, the premier of our country, pointed out that we must advance the desertification prevention in Dunhuang. Therefore, to know current states of desertification in Dunhuang has a very important practical significance for desertification prevention and protection of environment.Decision tree has great advantages in remote sensing image classification, which can organize multi-information efficaciously. In this study, vegetation fraction is the main index used in the classification system of desertification, and NDVI (normalized difference vegetation index) is the mostly used vegetation index in the estimation of vegetation fraction. Land surface temperature and texture features are introduced in classifying process to increase the precision of classification. Based on Landsat TM images, using decision tree, this study tries to establish an automatic classifying method for monitoring desertification, which has integrated the quantitative estimation result, namely vegetation fraction and land surface temperature, and the outcome of texture analysis. Finally, the development of desertification in Dunhuang in recent two decades and the distribution characteristics of desertification land are analyzed. The main conclusions are as follows:(1) The desertification in Dunhuang is very serious. According to the result of interactive interpretation, the area of desertification land in Dunhuang in 2004 reaches 7551km~2 , dominated by moderate and severe levels. Desertification land appears mainly in the south of Shule River and the spatial distribution pattern of desertification is related to the special types of land use for different reasons.(2)The desertification land in Dunhuang is expanding in recent two decades, and the degree of desertification changes. From 1987 to 2006, desertification land in Dunhuang is expanding and aggravating with the expanding speed of the latter decade is significantly faster than the former one. In 1987, it is dominated by the moderate desertification, in 1996 the severe desertification and in 2006 combination of moderate and severe desertification.(3)The major caused of desertification are location and climate background, hydrological, vegetation and humanity factors. The ways to control desertification are to enhance the power of curing the desertification according to the law, to develop sandy industry and manage it scientifically, to comply with natural and economic law to control desertification, to encourage people to take action in improving desertification policy and improve the public consciousness in policy, to intensify departmental cooperation and monitoring, to solve the water utilization problems reasonably and to strengthen dynamic monitoring of desertification.(4) This automatic classifying method for monitoring desertification, which has integrated the quantitative estimation result, namely vegetation fraction and land surface temperature, and the outcome of texture analysis, is feasible. The result of decision tree is high in comparability, but the precision is obviously lower than that of interactive interpretation.