论文部分内容阅读
土壤盐渍化严重制约了农业可持续发展和生态安全,土壤盐渍化的精确评价分析,对土壤盐渍化的改善和治理具有重要的意义。本文以新疆焉耆盆地为研究对象,Landsat8 OLI遥感影像和实测采样数据相结合,提取地下水埋深(GD)、盐分指数(SI)、地表蒸散量(SET)和改进型温度植被干旱指数(MTVDI)建立了土壤盐渍化评价模型。结果表明:(1)结合野外实测土壤盐分数据,对BP神经网络模型进行训练。最终以最优的4-4-1结构的3层BP神经网模型对研究区土壤盐渍化进行了预测(R~2=0.864,RMSE=0.569)。相比传统多元线性回归模型(R~2=0.741,RMSE=0.767),神经网络模型对土壤盐渍化的预测精度更高;(2)土壤盐渍化分布与GD、SI、SET和MTVDI等存在较强的关联性,不同等级的土壤盐渍化是不同影响因素不同程度上组合而引起的结果,盐渍化土地主要分布在地下水位较低以及土地开垦之后没有利用的荒地区域;(3)整个研究区大部分区域受到不同程度的盐渍化影响,耕地退化为盐渍地导致该区域土壤盐渍化以及土壤次生盐渍化进一步加剧。
Soil salinization has severely restricted the sustainable development of agriculture and ecological security. The accurate evaluation and analysis of soil salinization is of great significance to the improvement and management of soil salinization. In this paper, we took the Yanqi Basin in Xinjiang as the research object, and combined the Landsat8 OLI remote sensing image with the measured sampling data to extract the groundwater depth (GD), salt index (SI), surface evapotranspiration (SET) and modified temperature vegetation drought index (MTVDI) Soil salinization evaluation model was established. The results showed that: (1) BP neural network model was trained according to field measured soil salinity data. Soil salinization in the study area was finally predicted by the optimal 4-4-1 BP neural network model (R ~ 2 = 0.864, RMSE = 0.569). Compared with traditional multivariate linear regression model (R ~ 2 = 0.741, RMSE = 0.767), neural network model has higher prediction accuracy for soil salinization; (2) Distribution of soil salinization and GD, SI, SET and MTVDI There is a strong correlation between different levels of soil salinization is a combination of different factors affecting the degree of results, salinized land mainly in the lower groundwater level and land reclamation after the unused wasteland area; (3 ) Most of the study area was affected by different degrees of salinization. The degradation of cultivated land to saline land resulted in the further increase of soil salinization and soil salinization.