多源数据与土壤盐分运移模型同化及其不确定性研究

This project was proposed to reveal the mechanism of assimilation using multi-source monitoring data and soil salt-transport model, and to illustrate the uncertainty of the simulation of regional soil salt dynamic based on multi-source data assimilation. Taking the coastal salt-affected farming area as the experimental site which was characterized by active soil water-salt movement and intensive agricultural cultivation, this project adopted the idea of assimilating multi-spectral remote sensing (Landsat 8 OLI) and proximal sensing (Electromagnetic induction EM38-MK2) data to the model of soil water flow and salt transport. Approaches such as periodical soil sampling by manual and in-situ monitoring with sensors, field-scale and regional models, numerical simulation and experimental validation were employed in this project. The primary objectives were to investigate and simulate the law of soil water-salt movement under typical cropping systems in the coastal salt-affected farming area, to discuss the scale convergence mechanism of profile soil salinity inversion using multi-spectral remote sensing and proximal sensing data, to study the method assimilating multi-source data to model of soil water flow and salt transport, to specify the uncertainty of the simulation of regional soil salt dynamic based on multi-source data assimilation, and to analyze the error sources and sensitive parameters of the simulation of regional soil salt dynamic. The findings of this project could provide theoretical reference for utilizing the assimilation of multi-source data and numerical model to simulate regional soil salt evolution, and provide practical guidance for soil water-salt movement regulation and soil salinization degradation prevention in coastal salt-affected farming area.

为揭示多源监测数据与盐分运移模型的同化方法与机制，阐明基于多源数据同化的区域土壤盐分运移模拟的不确定性，本项研究针对水盐运动活跃、农业集约化程度高的滨海盐渍化农区，运用将多光谱遥感（Landsat 8 OLI）和近感（磁感大地电导率仪EM38-MK2）监测数据同化到水盐运移模型的思路，采用定期采样与典型样点原位监测、点尺度模型与空间模型、数值模拟与试验验证相结合的方法，旨在分析滨海盐渍农区典型种植制度下土壤水盐运动规律及其数值模拟，探讨基于遥感和近地传感器监测数据的区域土体盐分反演的尺度衔接机制，研究多源数据与水盐运移模型的土壤盐分数据单点同化方法，明确基于多源数据同化的区域土壤盐分运移模拟的不确定性，解析区域土壤盐分运移模拟的误差来源与敏感参数。本研究成果为利用多源数据和数值模型同化进行区域土壤盐分动态模拟提供理论基础，对滨海农区土壤水盐调控与盐渍退化阻控具有重要指导意义。

土壤水盐运移多尺度模拟的不确定性是制约盐渍化预报与精准治理利用的主要因素，深层次的原因是传统土壤水盐运移过程模型与大尺度的监测信息尺度衔接机制不明确。本项目针对水盐运动活跃、集约化程度高的滨海盐渍化农区，采用将近地传感监测数据同化到水盐运移模型的思路，研究了滨海盐渍农区典型种植制度下土壤水盐运动规律及其数值模拟，分析了基于近地传感器监测数据的区域土体盐分反演的尺度衔接机制，重点探讨了多源数据与水盐运移模型的土壤盐分数据单点同化方法，解析区域土壤盐分运移模拟的不确定性与误差来源。结果表明：滨海盐渍农区典型种植制度下土壤水盐运动受自然因素与人为因素的共同作用，尤其是灌溉量与灌溉水质，且根系吸水显著影响水盐运动过程，建立的包括水分运动、盐分运动、根系吸水和作物生长模块的HYDRUS模型可用于土壤盐分运动的模拟，并达到了率定和验证要求。揭示了基于近地传感反演的田间土壤盐分尺度拓展与衔接机制，构建了综合应用线性混合效应模型LME和限制性最大似然REML的土壤盐分剖面重构方法。研究发现集合卡尔曼滤波（EnKF）能有效地同化近地传感观测到HYDRUS数值模型中，较单一HYDRUS模型模拟结果，同化值有较大程度的改进，且基于序贯同化的EnKF对剖面土壤盐分的同化效果优于四维变分4-D Var法，土壤盐分数据EnKF同化有效深度为40-60 cm，随着深度的增加同化效果减弱。农田尺度下EnKF对剖面土壤盐分也表现出正同化效果，同化与未同化条件下的非饱和水力传导度K相对差值、土壤盐分Cs调整量变化范围逐渐减小，随着深度的增加同化效应亦减弱；EnKF数据同化过程中误差来源于多方面，除了模型参数，驱动参数、物理模型自身不确定性与误差外，状态变量数据集合大小、引入观测数据的频率与有效深度等因素均不同程度的影响数据同化的精度。本项目揭示了土壤水盐运移模型与近地传感等多源监测信息的数据同化原理及其不确定性来源，对土壤盐渍化精确预报、精准管理和水土资源高效利用具有重要科学意义。

内容获取失败，请点击重试