| 引用本文: | 吴家林,彭杰,白建铎,等.基于电磁感应数据的区域尺度土壤剖面电导率反演模型研究[J].灌溉排水学报,0,():-. |
| WU Jialin,pengjie,baijianduo,et al.基于电磁感应数据的区域尺度土壤剖面电导率反演模型研究[J].灌溉排水学报,0,():-. |
|
| 摘要: |
| 【目的】土壤盐渍化是新疆南部棉花高产的主要障碍因子，准确获取区域尺度土壤剖面盐分信息对棉田定额灌溉量计算、种植风险评价和种植结构调整具有重大意义。【方法】本文以南疆阿拉尔垦区为研究区，以田间尺度采集的30个不同盐渍化程度棉田的540个样点的0 ~ 0.375 m、0 ~ 0.750 m、0 ~ 1.000 m的土壤剖面电导率数据和对应的电磁感应数据为数据源，采用线性模型和非线性模型分别构建了田间尺度和区域尺度的土壤剖面电导率的电磁感应反演模型，并采用缩减建模样本量方法进一步检验了区域尺度模型的可靠性和稳定性。【结果】研究结果表明，多元线性回归(MLR)、偏最小二乘回归(PLSR)和主成分回归(PCR)建模方法的田间尺度模型R2在0.88 ~ 0.95，而对应的区域尺度模型R2在0.34 ~ 0.53。基于随机森林(RF)、神经网络(NN)和支持向量机(SVM)非线性建模方法构建的土壤剖面电导率的区域尺度电磁感应反演模型R2在0.60 ~ 0.85，其中RF模型的精度最高，0 ~ 0.375 m、0 ~ 0.750 m、0 ~ 1.000 m土壤剖面电导率的RF反演模型R2分别为0.80、0.85和0.84，相较于线性建模方法的区域尺度模型精度有明显的提高。RF区域尺度模型的样本数量由540个逐步缩减到240个，模型精度没有明显变化，表明采用区域尺度模型相较于田间尺度模型来讲，可大幅度降低土壤剖面样本采集数量，从而可显著提高采样效率和降低采样成本。【结论】研究结果为区域尺度的土壤剖面盐渍化调查提供了新的方法和思路。 |
| 关键词: 电磁感应;区域尺度;土壤电导率;反演模型 |
| DOI： |
| 分类号:S156.41 |
| 基金项目:兵团中青年创新领军人才项目(2020CB032),国家重点研发计划项目(2018YFE0107000)资助 |
|
| Research on Inversion Model of Soil Conductivity at Regional Scale Based on Electromagnetic Induction Data |
|
WU Jialin1,2, pengjie1,2, baijianduo1,2, wangjiawen1,2, jiwenjun3, wangnan4
|
|
1.College of Plant Science,Tarim university,Alar Xinjiang 843300;2.china;3.College of Land Science and Technology, China Agricultural University;4.College of Environment and Resources, Zhejiang University
|
| Abstract: |
| Abstract:【Objective】Soil salinization is the main obstacle to the high yield of cotton in southern Xinjiang. Accurately obtaining regional-scale soil profile salinity information is of great significance to the calculation of cotton field quota irrigation, planting risk assessment and planting structure adjustment.【Method】In this paper, the Alar Reclamation Area in southern Xinjiang is used as the research area. The soil profile conductivity data of 540 samples of 30 cotton fields with different salinization degrees collected at the field scale are 0 ~ 0.375 m, 0 ~ 0.750 m, and 0 ~ 1.000 m. With the corresponding electromagnetic induction data as the data source, the linear model and the nonlinear model were used to construct the electromagnetic induction inversion model of the soil profile conductivity at the field scale and the regional scale, and the method of reducing the modeling sample size was used to further verify the regional scale. The reliability and stability of the model.【Result】The research results show that the field scale model R2 of multiple linear regression (MLR), partial least square regression (PLSR) and principal component regression (PCR) modeling methods is 0.88 ~ 0.95, while the corresponding regional scale model R2 is 0.34 ~ 0.53. The regional-scale electromagnetic induction inversion model R2 of soil profile conductivity based on nonlinear modeling methods such as random forest (RF), neural network (NN) and support vector machine (SVM) is between 0.60 and 0.85, where the accuracy of the RF model The highest, 0 ~ 0.375 m, 0 ~ 0.750 m, 0 ~ 1.000 m soil profile conductivity of the RF inversion model R2 are 0.80, 0.85 and 0.84, respectively, compared with the linear modeling method, the accuracy of the regional scale model is significantly improved. The sample size of the RF regional scale model was gradually reduced from 540 to 240, and the accuracy of the model did not change significantly, indicating that compared with the field-scale model, the regional scale model can greatly reduce the number of soil profile sample collections, which can significantly improve Sampling efficiency and reducing sampling cost.【Conclusion】The research results provide new methods and ideas for the investigation of soil salinization at the regional scale. |
| Key words: Electromagnetic induction; Regional scale; Soil conductivity; Inversion model |
