面向多类型土壤有机碳定量反演的天基高光谱探测参数研究

星载高光谱仪器的光谱通道以及光谱分辨率和信噪比等核心参数设置直接影响土壤有机碳定量反演精度。本研究开展了卫星载荷光谱分辨率、信噪比、光谱特征波段对不同土壤类型有机碳反演影响的研究,提出了基于大气传输模型、光谱分辨率分析模型、信噪比分析模型、特征波段的提取分析模型以及偏最小二乘回归反演模型的面向不同土壤类型有机碳监测的高光谱卫星“地面–大气–仪器–观测–反演”全链路仿真分析方法,实现了土壤类型、大气效应、仪器特性参数、反演方法的耦合影响分析。结果表明：①3种类型土壤有机碳反演的最佳光谱分辨率均在10~20 nm。②不同土壤类型对观测的信噪比需求不同。对于Phaeozem的有机碳监测,较另外两种土壤有更高的信噪比需求。③在不同特征波段提取分析方法下所需的最佳光谱分辨率和信噪比一致。不同类型土壤光谱数据提取出的特征波段不同,其中反演效果最佳的土壤类型为Chernozem,特征波段数为26个,R²=0.826 5,RMSE=3.438 9 g/kg。④反演模型与仪器特性参数无耦合关系,同一类型土壤不同反演算法的最佳光谱分辨率和信噪比需求一致。⑤Chernozem有机碳最佳反演参数需求为光谱分辨率15 nm,信噪比大于506.66,特征波段提取数为26个;Kastanozem有机碳最佳反演参数需求为光谱分辨率17 nm,信噪比大于331.42,特征波段提取数为22个;Phaeozem有机碳最佳反演参数需求为光谱分辨率15 nm,信噪比大于432.51,特征波段提取数为19个。

Study on Space-based Hyperspectral Detection Parameters for Quantitative Retrieval of Organic Carbon in Multiple Types of Soil

The spectral channel, spectral resolution, signal to noise ratio and other core parameters of space-borne hyperspectral instruments directly affect the accuracy of quantitative retrieval and prediction of soil organic carbon(SOC). In this study, the effects of satellite load spectral resolution, signal-to-noise ratio and spectral characteristic bands on the inversion of organic carbon in different soil types were studied, Based on atmospheric transmission model, spectral resolution analysis model, signal to noise ratio analysis model, a hyperspectral satellite ‘ground-atmosphere-instrument-observation-inversion’ full-link simulation analysis method for organic carbon monitoring of different soil types was proposed. And the coupling effect analysis of soil type, atmospheric effect, instrument characteristic parameters and retrieval methods was realized. The results showed that: 1) The best spectral resolution was in the range of 10–20 nm for soil organic carbon retrieval in different soil types. 2) Different soil types had different requirements for the observed signal-to-noise ratio, a higher signal to noise ratio requirement was needed for organic carbon monitoring of Phaeozem than the other two soil types. 3) The optimal spectral resolution and signal-to-noise ratio required under different feature band extraction and analysis methods were consistent. The characteristic bands extracted from the spectral data of different soil types were different, among which Chernozem had the best retrieval effect, with 26 characteristic bands, R²=0.826 5,RMSE=3.438 9 g/kg. 4) The retrieval model had no coupling relationship with the instrument characteristic parameters, and the best spectral resolution and signal-noise ratio requirements of different retrieval algorithms for the same soil type were consistent. 5) The best retrieval parameters for SOC contents of Chernozem, Kastanozem and Phaeozem were: spectral resolution15 nm, 17 nm and 15 nm, signal-noise ratio greater than 506.66, 331.42 and 432.51, and the number of feature bands extracted 26, 22 and 19, respectively.