高光谱成像
遥感
环境科学
大气辐射传输码
含水量
土壤水分
辐射传输
天蓬
反射率
植被(病理学)
土壤科学
光谱特征
地理
地质学
光学
物理
病理
考古
岩土工程
医学
作者
Chongya Jiang,Hongliang Fang
出处
期刊:International journal of applied earth observation and geoinformation
日期:2019-11-01
卷期号:83: 101932-101932
被引量:15
标识
DOI:10.1016/j.jag.2019.101932
摘要
Soil background reflectance is a critical component in canopy radiative transfer (RT) models. However, few efforts have been devoted to the development of soil reflectance models compared to other components in canopy RT models. In the spectral domain of soil reflectance, spectral vector models are more flexible than typical spectra models, but its characteristics and performance are poorly understood and validated. To improve the understanding of hyperspectral soil reflectance modeling, this study conducted a comprehensive diagnostic analysis on different spectral vectors derivation algorithms, the impact of training datasets on model performance, and the soil moisture effect in modeling. With improved understanding, a general spectral vectors (GSV) model was developed. The model employs three dry spectral vectors and one humid spectral vector derived from global dry and humid soil reflectance databases including 23,871 soil spectra (400–2500 nm), using a matrix decomposition algorithm. A comprehensive evaluation shows that separate modeling of dry and humid soils and the usage of global training data significantly improved the performance of spectral vectors model, while the choice of spectral vectors derivation algorithm has little influence on model performance. Overall, the GSV model accurately simulates global soil reflectance with an R2 of 0.99 and RMSE of 0.01, superior to the widely-used Price model. In particular, the performance of GSV was robust over various soil types and under different moisture conditions. Coupling with the GSV model substantially reduced errors of 3D and 1D canopy RT modeling. The proposed GSV model has great potentials for vegetation remote sensing studies.
科研通智能强力驱动
Strongly Powered by AbleSci AI