基于Landsat OLI的绿洲灌区土壤盐度最优预测尺度分析

发布时间：2018-05-01 18:27

本文选题：土壤盐渍化 + Landsat　；参考：《中国农业科学》2017年15期

【摘要】：【目的】基于遥感提取的多尺度遥感指示因子和土壤实测电导率数据,借助统计分析,试图探寻适合干旱区典型绿洲灌区土壤盐度变异的最佳观测尺度和指征变量,为快速评估绿洲土壤盐渍化提供备选方案。【方法】以新疆渭干河-库车河绿洲为研究区,以野外采集的土壤盐度数据(采集0—10、10—20、20—40和40—60 cm土层土样,制备土壤饱和溶液并测试电导率(ms·cm-1))并将其作为预测对象(n=87),借助Landsat OLI遥感影像数据,利用栅格重采样(30—1 000 m)和领域滤波(原始分辨率为30、60、90、120、150、180、210 m,滤波尺度为3×3至31×31)两种方式,生成多个尺度若干种指示因子(主成分分析、缨帽变化、植被指数、湿度指数),共计获得1 078个(其中,栅格重采样生成352个,领域滤波生成726个)环境变量。在此基础之上,利用线性和非线性曲线模型分别拟合上述两种模式下土壤盐度和环境变量之间的相关性,进而找出最优环境因子和预测尺度。【结果】栅格重采样模式下能够较好响应各层土壤变异性的皆为非线性模式。其次,该模式下,拟合精度随着空间分辨率的降低而降低。此模式下最佳推理尺度为30 m,该尺度下最佳响应变量除了40—60 cm处为三波段差分指数(Three-band Maximal Gradient Difference,TGDVI)外,其余深度皆为扩展的归一化指数(Extended Normalized Difference Vegetation Index,ENDVI)。领域滤波模式下的最佳推理尺度为180 m(滤波尺度3×3),同时,各层最佳拟合变量皆为扩展的增强型植被指数(Extented Enhanced Vegetation Index,EEVI)。相比较栅格重采样模式,该模式下的拟合精度全面优于前者,各层依次提高14.60%、34.40%、32.10%和21.70%。【结论】基于领域滤波模式下,像元分辨率为180 m,窗口大小为3×3的ENDVI指数更适合预测本研究区土壤盐度的空间变异性。
[Abstract]:[objective] based on the multi-scale remote sensing indicator factors and soil electrical conductivity data extracted by remote sensing, and with the help of statistical analysis, this paper tries to find out the best observation scale and indication variable for soil salinity variation in typical oasis irrigation areas in arid areas. To provide an alternative scheme for rapid assessment of soil salinization in oasis. [methods] soil salinity data (0-10 ~ 10 ~ 20 ~ 20 ~ 40 and 40 ~ 60 cm) were collected from the oasis of Weigan River and Kuqa River in Xinjiang, and the soil salinity data were collected in the field (0 ~ 10 ~ 10 ~ 20 ~ 20 ~ 40 cm and 40 ~ 60 cm soil samples). The saturated solution of soil was prepared and its conductivity was measured and used as the prediction object. With the aid of Landsat OLI remote sensing image data, the raster resampling was used for 30-1 000 m) and domain filtering (original resolution was 30 0 090 1 201 180 210 m, the filter scales were 3 脳 3 to 31 脳 31 1), and the results were as follows: 3 脳 3 to 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31 脳 31. A number of indicator factors (principal component analysis, tassel change, vegetation index, humidity index) were generated for multiple scales. A total of 1,078 environmental variables were obtained (of which 352 were generated by grid resampling and 726 by domain filtering). On this basis, linear and nonlinear curve models are used to fit the correlation between soil salinity and environmental variables, respectively. Finally, the optimal environmental factors and the prediction scale are found. [results] in the case of grid resampling model, all the nonlinear models which can better respond to the soil variability in each layer are obtained. Secondly, the fitting accuracy decreases with the decrease of spatial resolution. In this model, the optimal inference scale is 30 m. The best response variable in this model is the three-band Maximal Gradient differential index three-band Maximal Gradient difference index (TGDVI) at 40-60 cm, and the other depth is extended Normalized Difference Vegetation index (ENDVI). The optimal inference scale is 180 m (filtering scale 3 脳 3 ~ 3), and the best fitting variables in each layer are extended augmented Enhanced Vegetation index (EEVI). Compared with the grid resampling model, the fitting accuracy of this model is better than that of the former, and each layer increases by 14.60%, 34.40% and 32.10%, and 21.70% respectively. [conclusion] in the domain filtering mode, The pixel resolution is 180m and the ENDVI exponent with a window size of 3 脳 3 is more suitable for predicting the spatial variability of soil salinity in the study area.
【作者单位】：新疆大学资源与环境科学学院/绿洲生态教育部重点实验室;
【基金】：国家自然科学基金(41661046) 中国博士后面上基金(2016M602909) 自治区科技支疆项目(201591101) 新疆大学博士启动基金(BS150248) 新疆维吾尔自治区重点实验室专项基金(2014KL005) 国家自然科学基金(新疆联合基金本地优秀青年培养专项(U1503302))
【分类号】：S156.41

【参考文献】