水质参数遥感反演光谱特征构建与敏感性分析

水质遥感监测是遥感的重要应用方向之一，作为传统水体采样化验的辅助手段，具有快速、大面积、无接触的技术优点。然而，目前内陆河湖水环境监测常用的遥感传感器大多是针对陆地观测或海洋水色观测而设计的，其性能指标的设计和设置并未考虑到内陆水体特性，限制了水质遥感定量监测的应用。针对这一问题，提出了一种基于变异系数和噪声占比指数的水质参数反演光谱特征构建方法，并通过光谱仿真模拟实验，研究光谱分辨率、信噪比以及辐射分辨率对典型水质参数反演模型的影响。选择上海市三项典型水质参数溶解氧（DO）、总磷（TP）和氨氮（NH₃-N）为研究对象，分别构建了水质参数反演光谱特征及相应遥感反演模型，然后开展光谱仿真模拟实验，定义了敏感度S和水质敏感微分指数CI，进行敏感性分析。最终从模型反演准确度和稳定性两个方面来评价仪器参数对水质参数遥感反演模型的影响。结果表明：光谱特征构建方法能有效确定水质参数反演特征波段。光谱分辨率对比值型的水质参数反演模型的影响较小，而信噪比和辐射分辨率对模型影响较大，随着信噪比和辐射分辨率的增加，水质参数模型精度和稳定性都有一定提升。综合仪器指标敏感性分析，可知当信噪比优于56 dB，辐射分辨率不低于9 bit，光谱分辨率适宜，能够较好地应用于内陆河湖水质遥感监测。该研究不仅可以为面向内陆河湖水质监测的传感器的研制提供参考和借鉴，还能为水资源监管部门进行水质遥感监测提供技术支持，有利于加快水环境智能化监测体系的构建。

Spectral Feature Construction and Sensitivity Analysis of Water Quality Parameters Remote Sensing Inversion

Water quality remote sensing monitoring is one of the important application directions of remote sensing. As an auxiliary mean of traditional water sampling and testing, remote sensing has the advantages of rapid, large-area and contactless. However, most remote sensing sensors commonly used in inland water monitoring are designed for land observation or ocean watercolor observation. The design and setting of sensor performance indicators do not consider the characteristics of inland water, limiting the application of water quality remote sensing monitoring. This study proposes a method for constructing the spectral characteristics of water quality parameter based on variation coefficient and noise ratio index, and study the influence of the spectral resolution, signal-to-noise ratio (SNR) and radiation resolution on typical water quality parameters inversion models through the spectral simulation experiments. Firstly, aimed at three main water quality parameters in Shanghai, we construct the spectral characteristics of dissolved oxygen(DO), total phosphorus(TP) and ammonia nitrogen (NH₃-N) respectively and establish remote sensing inversion models. Then, we carry out the spectral simulation experiment and calculate the sensitivity (S) and water quality sensitive differential index (CI) for sensitivity analysis. Finally, we evaluate the spectral resolution’s influence, SNR and radiation resolution on water quality parameters inversion models from two aspects of accuracy and stability. The results show that this method can effectively determine the bands of water quality parameters inversion models. Spectral resolution has little effect on the contrast-type inversion models, while SNR and radiation resolution greatly influence the models. With the increase of SNR and radiation resolution, the water quality inversion models’ accuracy and stability are improved to some extent. According to Comprehensive sensitivity analysis of sensors parameters, when the SNR is better than 56 dB, the radiation resolution is not less than 9 bit, and the spectral resolution is appropriate. It can be better applied to inland water quality remote sensing monitoring. This research can provide a reference for the development of sensors for inland water quality monitoring and provide technical support for water resources supervision departments to carry out remote sensing monitoring of water quality, which is conducive to accelerating the construction of an intelligent monitoring system for the water environment.