基于Logistic回归分析的土壤盐渍化易发性评价——以新疆南疆塔里木灌区为例

Evaluation of soil salinization susceptibility based on Logistic regression analysis:A case of Tarim irrigation area in southern Xinjiang

新疆南疆(简称南疆)的土地资源利用潜力巨大,是我国重要的后备耕地资源供给区,开展土壤盐渍化易发性评价,对于南疆增水增地选区与土地分区分类治理具有重要参考意义。以南疆塔里木灌区为研究对象,基于地下水溶解性固体总量(TDS)、地下水埋深、土壤质地、地貌类型、土地利用类型、土壤含盐量等实地调查数据,采用数据驱动方式的证据权进行土壤盐渍化单因素空间相关性分析,并通过Logistic回归模型进行多因素综合定量评价土壤盐渍化易发性。结果表明:(1)高的地下水TDS、浅的地下水埋深、细颗粒的土壤质地以及地貌类型中的冲(湖)积平原、土地利用类型中的草地与高土壤含盐量地区表现出较高的正空间相关性。(2)灌区冲积平原内的古河道,湖积平原内的古湖沼等槽型、封闭洼地是土壤盐渍化高易发区,面积约29 km2,占灌区总面积的1.3%;灌区农田周边的弃耕地为土壤盐渍化中易发区,面积约453 km2,占灌区总面积的20.5%;灌区周边广阔的沙漠及水域为土壤盐渍化低易发区,面积约1726 km2,占灌区总面积的78.2%。(3)综合模型评价置信度结果及野外实地调查现状表明,Logistic回归模型综合定量评价土壤盐渍化易发性的结果准确可靠,为土壤盐渍化易发性评价提供了新思路,也为南疆地区土壤盐渍化防治和后备耕地资源选区提供了科学依据。

The southern region of Xinjiang has great potential for land resource utilization and is a crucial reserve supplying cultivated land resources in China.Evaluating soil salinization susceptibility in this area is necessary,as it aids in making informed decisions regarding water and land selection and land zoning management in southern Xinjiang.Sources,including total dissolved solids(TDS)in groundwater,groundwater depth,soil texture,geomorphic type,land-use type,and soil salt content,were collected via field surveys.By employing data-driven evidence weight,the spatial correlations of a single factor contributing to soil salinization were analyzed,followed by a multifactor comprehensive quantitative evaluation of soil salinization using a Logistic regression model.The results showed the following:(1)Shallow groundwater depth,high groundwater TDS,fine-grained soil,alluvial(lacustrine)plain geomorphology,and grassland land-use types showed strong positive spatial correlation with high soil salt content.(2)High soil salinization-prone areas were identified,encompassing the ancient channel in the alluvial plain and ancient lake marsh in the lacustrine plain,covering~29 km2 and accounting for 1.3%of the total area.Additionally,abandoned farmland around the irrigated area was a prone area of soil salinization,covering~453 km2,accounting for 20.5%of the total area.In contrast,the vast desert and water areas around the irrigated area exhibited a low salinization-prone area of~1726 km2,accounting for 78.2%of the total irrigated area.(3)The comprehensive evaluation using the Logistic regression model was validated with field investigation,affirming its accuracy and reliability.This comprehensive quantitative assessment of soil salinization susceptibility provides a scientific basis for soil salinization prevention and control measures and for making informed decisions regarding the selection of reserved cultivated land resources in southern Xinjiang.