Scholarlink

Purpose <p>Soil moisture content (SMC) and soil organic matter (SOM) are vital for soil health, yet are often inverted separately, ignoring their interactions. This study addresses the need for a joint inversion approach.</p> Methods <p>To address this, this study proposes two core methods. First, a semi-empirical transfer model—the SOM-W Semi-Empirical Transfer Model—was developed based on spectral reflectance and SOM data from 76 soil samples in China. This model simultaneously incorporates parameters for both SOM content and SMC and was validated by establishing six moisture gradients. Second, a joint inversion method for SOM and SMC based on a Bayesian framework was introduced, which incorporates prior knowledge to constrain the model and improve inversion accuracy.</p> Results <p>The SOM-W model demonstrated high accuracy in fitting spectral reflectance within the 500–2400 nm range (<i>R</i><sup><i>2</i></sup><sub><i>p</i></sub> = 0.945, <i>RMSE</i><sub><i>p</i></sub> = 0.0094) and identified 20 key wavelengths characterizing spectral morphology, achieving precise full-wavelength curve fitting (<i>R</i><sup><i>2</i></sup><sub><i>p</i></sub> = 0.9976, <i>RMSE</i><sub><i>p</i></sub> = 0.0021). After further selection of 10 SOM-sensitive wavelengths, the prediction accuracy for SOM was significantly improved (<i>R</i><sup><i>2</i></sup><sub><i>p</i></sub> = 0.568, <i>RMSE</i><sub><i>p</i></sub> = 3.3858&#xa0;g/kg), effectively eliminating moisture interference. In the joint inversion, the Bayesian approach significantly enhanced accuracy compared to traditional gradient methods (TRF) and intelligent algorithms (PSO), with results of <i>R</i><sup><i>2</i></sup><sub><i>p_som</i></sub> = 0.767, <i>RMSE</i><sub><i>p_som</i></sub> = 3.8132&#xa0;g/kg, <i>R</i><sup><i>2</i></sup><sub><i>p</i></sub><i>_</i><sub><i>w</i></sub> = 0.967, <i>RMSE</i><sub><i>p_w</i></sub> = 0.5780%).</p> Conclusions <p>This approach not only effectively addresses the mutual interference between moisture and organic matter in soil spectra but also provides a novel technical pathway for soil spectral analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Semi-empirical model of soil organic matter and soil moisture content with bayesian joint inversion

  • 期刊论文

  • 发表日期：

    2026年3月24日

查看全文
  • Jiawei Xu,
  • Yuteng Liu,
  • Changxiang Yan,
  • Jing Yuan

摘要

Purpose

Soil moisture content (SMC) and soil organic matter (SOM) are vital for soil health, yet are often inverted separately, ignoring their interactions. This study addresses the need for a joint inversion approach.

Methods

To address this, this study proposes two core methods. First, a semi-empirical transfer model—the SOM-W Semi-Empirical Transfer Model—was developed based on spectral reflectance and SOM data from 76 soil samples in China. This model simultaneously incorporates parameters for both SOM content and SMC and was validated by establishing six moisture gradients. Second, a joint inversion method for SOM and SMC based on a Bayesian framework was introduced, which incorporates prior knowledge to constrain the model and improve inversion accuracy.

Results

The SOM-W model demonstrated high accuracy in fitting spectral reflectance within the 500–2400 nm range (R2p = 0.945, RMSEp = 0.0094) and identified 20 key wavelengths characterizing spectral morphology, achieving precise full-wavelength curve fitting (R2p = 0.9976, RMSEp = 0.0021). After further selection of 10 SOM-sensitive wavelengths, the prediction accuracy for SOM was significantly improved (R2p = 0.568, RMSEp = 3.3858 g/kg), effectively eliminating moisture interference. In the joint inversion, the Bayesian approach significantly enhanced accuracy compared to traditional gradient methods (TRF) and intelligent algorithms (PSO), with results of R2p_som = 0.767, RMSEp_som = 3.8132 g/kg, R2p_w = 0.967, RMSEp_w = 0.5780%).

Conclusions

This approach not only effectively addresses the mutual interference between moisture and organic matter in soil spectra but also provides a novel technical pathway for soil spectral analysis.