Loess Plateau Soil Organic Carbon Monitoring

Indirect Modeling with Optimal Visible Bands

Leveraging visible spectral response mechanisms, SOC is predicted indirectly via characteristic wavelength combinations of soil structure, aggregates, and enzyme activity, outperforming direct models.

Yang, Sha · Wang, Zhigang · Chenzhi, Zheng · Yuchao, Yang · Jiancheng, Zhang · Qiao, Xingxing · Rickard, William · Yu, Zhao · Meichen, Feng · Longmei, Gao · Yang, Wude · Wang, Chao

Computers and Electronics in Agriculture 2026

Specifications

Indirect model R_v²

{'zh': '0.74', 'en': '0.74'}

Indirect model RMSE

{'zh': '0.27', 'en': '0.27'}

Direct model R_v²

{'zh': '0.69', 'en': '0.69'}

Direct model RMSE

{'zh': '0.32', 'en': '0.32'}

Advantages

No field sampling or lab analysis

The prediction model relies solely on hyperspectral data in the visible range, allowing on-site scanning to obtain SOC content without field soil collection and laboratory chemical analysis.

Indirect model outperforms direct

The indirect model built from characteristic wavelength combinations of influencing factors (microbial activity, soil structure, and aggregates) achieves higher validation accuracy (R_v²=0.74) than the direct model (R_v²=0.69), with lower errors.

Visible bands suffice

The optimal characteristic wavelengths are concentrated in the visible range, eliminating the need for near-infrared or mid-infrared bands, thus enabling rapid monitoring with low-cost spectrometers.

Applications

• Soil Organic Carbon Monitoring：Replaces lab analysis with on-site scanning for rapid SOC quantification.
• Agricultural Soil Management：Supports carbon sequestration management and evaluates conservation tillage outcomes.
• Plant Factory Nutrient Monitoring：Visible-wavelength scheme can be integrated into rapid nutrient sensors.
• Hyperspectral Rapid Assessment：Enables high-accuracy inversion using only visible bands, cutting instrument costs.