| 基于EnKF和PF的沙壕渠灌域土壤含盐量监测模型研究 |
| |
| 引用本文： | 张智韬,陈策,贾江栋,殷皓原,姚一飞,黄小鱼.基于EnKF和PF的沙壕渠灌域土壤含盐量监测模型研究[J].农业机械学报,2023,54(6):361-372. |
| |
| 作者姓名： | 张智韬 陈策 贾江栋 殷皓原 姚一飞 黄小鱼 |
| |
| 作者单位： | 西北农林科技大学 |
| |
| 基金项目： | 国家重点研发计划项目(2017YFC0403302)和国家自然科学基金项目(51979232、51979234) |
| |
| 摘 要： | 为探究不同数据同化算法在时空尺度上监测土壤含盐量的可行性，以内蒙古河套灌区沙壕渠灌域为研究区域，采用高分一号卫星遥感图像作为数据源，通过EnKF算法和PF算法的同化观测算子和模型算子得到时空范围中的土壤含盐量变化情况。其中观测算子分为两步，首先通过PLS-VIP准则来筛选光谱指数作为自变量，再使用ELM模型建立基于不同时间不同深度的遥感监测土壤含盐量模型；模型算子为基于Hydrus-1D模型的数学模拟监测土壤含盐量模型。结果表明，基于ELM模型的土壤含盐量模型中，深度0～20 cm、20～40 cm和40～60 cm的平均IOA均在0.74以上，平均ME在0.14%以下，表明反演模型具有良好的精度；基于Hydrus-1D的数学模拟监测土壤含盐量模型中，3个深度平均IOA在0.79～0.89之间，平均ME在0.128%～0.137%之间，能够较好地反映土壤盐分在时间序列中的运移情况；EnKF算法3个深度IOA在0.820以上，ME在0.141%～0.157%之间，NMB在0.141～0.252之间，PF算法3个深度IOA在0.89以上，ME在0.090%～0.142%之间，NMB在0.0...
|
| 关 键 词： | 土壤含盐量 数据同化 集合卡尔曼滤波 粒子滤波 极限学习机 Hydrus-1D |
| 收稿时间： | 2022/10/27 0:00:00 |
Soil Salinity Monitoring Model of Shahaoqu Irrigation Area Based on EnKF and PF Algorithm |
| |
| ZHANG Zhitao,CHEN Ce,JIA Jiangdong,YIN Haoyuan,YAO Yifei,HUANG Xiaoyu.Soil Salinity Monitoring Model of Shahaoqu Irrigation Area Based on EnKF and PF Algorithm[J].Transactions of the Chinese Society of Agricultural Machinery,2023,54(6):361-372. |
| |
| Authors: | ZHANG Zhitao CHEN Ce JIA Jiangdong YIN Haoyuan YAO Yifei HUANG Xiaoyu |
| |
| Institution: | Northwest A&F University |
| |
| Abstract: | In order to explore the feasibility of different data assimilation algorithms in monitoring soil salinity on the spatio-temporal scale, the Shahaoqu Canal Irrigation Area in Hetao Irrigation District of Inner Mongolia was taken as the research area, and the Gaofen-1 satellite remote sensing image was used as the data source. The assimilation observation operator and model operator of EnKF algorithm and PF algorithm were used to obtain the changes of soil salinity in the spatio-temporal range. The observation operator was divided into two steps, firstly, the PLS-VIP criterion was used to filter the spectral index as the independent variable, and then the ELM model was used to establish the remote sensing monitoring soil salinity model based on different depths at different times; the model operator was a mathematical simulation monitoring soil salinity model based on the Hydrus-1D model. The results showed that in the ELM-based soil salinity model, the average IOA at the depths of 0~20cm, 20~40cm and 40~60cm were above 0.74, and the average ME was below 0.14%, indicating that the inversion model had good accuracy; in the Hydrus-1D based mathematical simulation monitoring soil salinity model, the average IOA at the three depths was between 0.79 and 0.89 and the average ME was between 0.128% and 0.137%, which could better reflect the transport of soil salts in the time series; the EnKF algorithm had IOA above 0.820 for three depths, ME between 0.141% and 0.157% and NMB between 0.141 and 0.252, and the PF algorithm had IOA above 0.89 for three depths and ME ranged from 0.090% to 0.142% and NMB ranged from 0.075 to 0.097, with better accuracy than the EnKF algorithm, which can well reflect the distribution of soil salinity in time and space. The assimilation scheme of Hydrus-1D model and ELM model based on EnKF and PF algorithms improved the accuracy of monitoring soil salinity, which can provide a basis for subsequent monitoring of soil salinity on a long time and large spatial and temporal scale, and can also provide a reference for the research of precision agriculture to control soil salinity. |
| |
| Keywords: | soil salinity data assimilation ensemble Kalman filtering particle filtering extreme learning machine Hydrus-1D |
|
| 点击此处可从《农业机械学报》浏览原始摘要信息 |
| 点击此处可从《农业机械学报》下载免费的PDF全文 |
|
