Study Of The Soil Salinization In Irrigated Area Based On HJ Satellite And GIS

Soil salinization is the most important and easily occured land degradation phenomenon in arid and semi arid area, which seriously affects the quality of ecological environment, and restricts the development of human society and economy. Soil secondary salinization in irrigated area has become the main factor limiting the development of ecology and economy, and a serious threat to the food production in China. Pingluo is a traditional agricultural irrigated area and major grain-producing county, whose soil salinization phenomenon is very serious. An urgent need is to have a comprehensive understanding of the extent and distribution of soil salinization in this area. In this paper, remote sensing and geographic information system technology are used to effectively monitor the soil salinization in the study area. The natural and artificial factors influencing the formation and development of soil salinization were analyzed in depth. On this basis, it realized the simulation and prediction of soil salinization in this area. The research conclusions are as follows:(1) The soil salinization in the study area had obvious accumulation phenomenon on surface, and there was a strong spatial variation in the soil surface layer. The redundancy analysis shows a strong correlation between K+、Na+、SO42-、Cl-and soil total salt content, as well as between HCO3-、Ca2+ and the pH value, while SO42- and Na+ have the strongest contribution to salinization. The soil salt content in each layer has a moderate intensity spatial correlation, the exponential model can be used to fit the semi variance function. There are some differences in the spatial distribution pattern of soil salinity in the surface layers and deep layers.(2) Linear spectral unmixing has been carried out for hyperspectral data of HJ satellite. The endmembers of water, salt, vegetation and dark material were extracted using pure pixel index and minimum noise. Compared and analyzed the methods of linear spectral unmixing with different constraint conditions. The linear spectral unmixing with full constraint condition shows the best result and a more explicit physical meaning. Based on this method, the application of the high spectral data of HJ satellite in classification and mapping of soil salinization was discussed.(3) Based on the agricultural planting patterns and phenology information in the study area, the land use/cover classification system was established by taking full advantage of HJ satellite CCD data. The NDVI time series which can reflect the change of vegetation information are constructed. The characteristic parameters of time dimension which can indicate the phonological information and show the variation between the classes were extracted. The decision tree based on expert knowledge was constructed by combining the spectral characteristic parameter, and the land use/Cover Classification of the study area was realized(4) Taking the actual control area of the samples as the research scale of soil salinization, natural factors and human factors affecting soil salinization were extracted by using the remote sensing data of HJ satellite、geographic data including DEM and the land use data. A canopy response salinity index with crop area as the weight was constructed as a comprehensive indicator in the quadrat scale, which can not only indirectly reflect the response for salinity of different crops, but also directly reflect the effects on soil salinization of different land use patterns. The BP neural network was used to establish the prediction model of EC with the index factors. The soil salinization in the study area was influenced by the natural factors and human factors. There were interactions and different scale effects between different factors, which could influence the prediction accuracy of salinization.