中文摘要：

为研究中原黄泛区土壤盐分空间变异，以河南省封丘县为研究区，综合考虑引起土壤盐渍化的土壤盐分、地形、地下水位及矿化度、植被情况及其他影响因素，基于遥感影像和磁感式探测获得的土壤表观电导率等多源数据建立了区域土壤盐分综合评估模型，并对研究区分层土壤盐分空间变异进行评估。结果表明：对于0～60 cm土层利用多源数据进行模型构建中土壤表观电导率与光谱指数占主要比例，模型对于各层土壤盐分的评价精度0～60 cm土层优于≥60～120 cm土层。土壤盐分含量随着深度的增加而增大，变异系数在0.22～0.28之间，属中等变异强度。土壤盐分主要集中分布在研究区北部与东南部，尤其是东南角黄河沿线区域，且随着土壤剖面显示出从表现到深层逐渐增加的趋势。利用多源数据建立的分层土壤盐分综合评估模型对于区域土壤盐分解析具有较高精度。该研究为中原黄泛区土壤盐化消减与土壤质量提升提供了可靠新方法。

英文摘要：

Salinization and alkalinization are two of important land degradation processes in flood area of the Yellow River in central China. A synthesized model for assessment of regional soil salinity was established based on multi-source data including soil salinity, topographical variable, the groundwater level and mineralization degree, vegetation and other factors to the soil salinization. A total of 101 soil columns were sampled from the study area using grid sampling method, and then analyzed for soil electrical conductivity （ECe） and other soil properties. Auxiliary data used in this study to interpret variability of soil salinization were Landsat 5 TM data, apparent electrical conconductivity （ECa） measured using an electromagnetic induction instrument （EM38）, altitude derived from topographic map, the groundwater table and mineralization degree and soil pH. The spatial variability of soil salinity was assessed in Fengqiu County, Henan Province, China. Classification and regression tree was applied to obtain the relationships between ECe （0-120 cm） and the auxiliary data. The results showed that ECa accounted for the major proportion of model prediction from multi-source data in classification and regression tree model of total soil layer. Generally, ECaH （apparent soil electrical conductivity from EM38 horizontal mode） and spectral index （dvi： difference vegetation index, bi：soil index, int2： intentity, int1： intentity, ndvi： normalized difference vegetation index, si2： soil index and si1： soil index） were common variable for 0-60cm soil layer. For the 0-30 cm depth, plant index （ndvi and dvi）, soil index （si1, si2 and bi） and intentity （int1 and int2） had the highest influence on the model prediction followed by ECa. Plant index （dvi） accounted of more than 50% for 0-60 cm soil layer used in the model. Meanwhile, for≥60-120 cm, ECaV （apparent soil electrical conductivity from EM38 vertical mode） was the most important variable used in regression tree m