Multisource Remote Sensing Data Applied To Extract Wind Erosion Information In Arid Area

Soil erosion has become the worldwide serious environmental problem,threatening the human sustainable development. Soil erosion concerns multi factors,multi spatial and temporal scales. Soil erosion monitoring is an important part of the application of RS and GIS. Remote sensing techniques have the ability to monitor the earth surface physical processes and detect the changes. The geography information system can play an important role in analysis and decision making. The remote sensing technology is widely used to detect and monitoring soil erosion due to rapidly and timely provide the information about properties, spatial distribution and extent of soil erosion.Multisource Information Fusion of RS image data is a type of enrichment of multi- detector information. Thus, Multisource RS image fusion, band together all the advantages of Multisource image data, can improve image analyses, understanding and identifying power. Multisource Information Fusion is one of the important method in modern image processing and analysis.This paper, based on RS technology, used 2008 TM image and 2008 SAR image as data source. Selecting 2008 TM image to extract study area's vegetation cover, soil texture, gradient and used these information to integrate with field practice data for comprehensive analysis, classifying the erosion degree of soil and confirming the classifying the scheme. Afterward, we used Maximum likelihood classification method to classify the TM and fusion of TM, SAR image respectively. We extracted soil erosion information and evaluated classification result we got before and after the fusion. Result shows that:(1) From result of soil erosion extraction we can see that only using TM image the overall classification accuracy is 88.7% . but confusion matrix shows that there still some confusion between different land types. Extremely eroded and strongly eroded which are in similar spectral characteristics has higher confusion degree.(2) Using fusion image classification image accuracy is increased to 95.04%, this is apparently higher than only using TM image classification accuracy is 88.72%, from confusion matrix we can see that though there are some confusions in different land types but its degree for more lower than just using the TM image. There are different backscatter coefficient for different ground roughness degree in RADAR image. As a result, blended division phenomenon is decreased especially between the extremely eroded and strongly eroded soil, and classification accuracy is increased.(3) Evaluation of the accuracy improvements in the detection of soil erosion features by fusing the two datasets. The fusion of TM and SAR data provided a unique combination that allowed more accurate identification of strongly eroded area, moderately eroded area and slightly eroded areas, as compared the results obtained by TM alone. Differences in the surface roughness determined variations in the amount of energy returning to the radar antenna, thus improving discrimination of classes exhibiting similar spectral behavior in the TM, and also it can be provide more information, higher spatial resolution and higher classification accuracy compared with TM image.