| Soil salinization is a serious threat to global food security and ecological environment.As one of the frontiers in land informatization,large-scale mapping of salinization information,quantitative prediction of soil salinity information and mastering the distribution pattern of soil salinization are of profound significance for soil salinization management.In this study,we selected the Xinjiang oasis region as the study area,regional soil salinity information was predicted based on the environmental similarity between sampled and unsampled sites with the help of MODIS data,DEM-derived variables and soil property data,based on the similarity soil mapping method.First,based on the theoretical principle of eco-geographical division,the study area was divided into different sub-areas,which reduces the complexity of land cover within the divisional unit.With typical samples,complete soil-environmental knowledge can be obtained.Then,based on the actual situation of the study area,we selected and inverted land surface parameters such as vegetation index,salinity index and topographic parameters,established a quantitative prediction model of soil salinity to try to predict the surface soil salinity content of different ecological geographic regions in the study area from 2010 to 2019.Based on the exploration of the temporal and spatial evolution of soil salinization,the driving factors of soil salinization are discussed.The key results have been summarized as follows:(1)Taking into account the ecological attributes of surface communities in Xinjiang,the environmental gradients affecting the distribution of communities,four types of 20 indicators,including temperature,moisture,topography and surface cover,were selected to establish the corresponding index system for ecological geographic divisions,and divided Xinjiang into six Eco-geographic regions.(2)Among the surface parameters involved in soil salt prediction,the salinity index(S1),salinity index(S2),topographic wetness index(TWI),brightness index(BI)and elevation(ELE)have high application potential in salinization prediction.The annual verification accuracy R~2 was basically above 0.6,with the best validation result in 2014(R~2=0.78)and the smallest root-mean-square error,which proved that the random forest method performs well in the prediction of soil salinity.The cross-regional accuracy verification error had certain fluctuations,but within an acceptable range,and the research method realizes the cross-regional soil salt prediction.(3)The spatial variation showed that the overall soil surface salinity in the study area was mostly below 8 g/kg,with non-salinized soil and lightly saline soil accounting for a relatively large proportion,moderately and heavily saline soils showing aggregated distribution,and very heavily saline soils were rarely seen.In addition,the degree of soil salinization in different subdivisions varies widely,and the soil salinity content is ranked from low to high as follows: Ili Valley Semi-Arid Oasis Region <Junggar Basin Arid Desert and Oasis Region < Tuha Basin Arid Desert and Oasis Region < Tianshan Mountain Arid Steppe Region < Tarim Basin Extreme Arid Desert,Gobi and Oasis Region.(4)The temporal variation showed that the soil salinity level was the highest in2013,and since then the salinity level has been decreasing year by year,with a significant decreasing trend in the salinity level in the arid desert and oasis areas of the Junggar Basin and a non-significant decreasing trend in other sub-regions.The driving analysis showed that the annual mean temperature and precipitation increased in 2013,while evaporation was in a decreasing stage,indicating that meteorological factors are not the main influencing factors of the salinization process in the region in recent years. |
PDF Full Text Request