AUTHOR=Zou Jinzhao , Wei Yanan , Zhang Yong , Liu Zheng , Gai Yuefeng , Chen Hongyan , Liu Peng , Song Qian TITLE=Remote sensing inversion of soil organic matter in cropland combining topographic factors with spectral parameters JOURNAL=Frontiers in Environmental Science VOLUME=12 YEAR=2024 URL=https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2024.1420557 DOI=10.3389/fenvs.2024.1420557 ISSN=2296-665X ABSTRACT=

Remote sensing has become an effective way for regional soil organic matter (SOM) quantitative analysis. Topographic factors affect SOM content and distribution, also influence the accuracy of SOM remote sensing inversion. In large region with complex topographic conditions, characteristic topographic factors of SOM in different topographic regions are unknown, and the effect of combining characteristic topographic factors with spectral parameters on improving SOM inversion accuracy remains to be further studied. Three typical topographic regions of Shandong Province in China, namely Western plain region (WPR), Central and southern mountain region (CSMR), Eastern hilly region (EHR), were selected. Topographic factors, namely Elevation, Slope, Aspect and Relief Amplitude, were introduced. Respectively, the characteristic topographic factors and spectral parameters of SOM in each region were identified. The SOM inversion models were built separately for each region by integrating spectral parameters with topographic factors. The results revealed that as for the characteristic topographic factors of SOM, none was in the WPR, E, RA, and S were in the CSMR, E and RA were in the EHR. In combination with characteristic topographic factors, the accuracy of SOM spectral inversion models improved, the calibration R2 increased by 0.075–0.102, the RMSE (Root mean square error) decreased by 0.162–0.171 g/kg, the validation R2 increased by 0.067–0.095, the RMSE decreased by 0.236–0.238 g/kg, and RPD (Relative prediction deviation) increased by 0.129–0.169. The most significant improvement was observed in the CSMR with the calibration R2 of 0.725, the validation R2 of 0.713 and the RPD of 1.852, followed by the EHR. This study not only contributes to the advancement of soil quantitative remote sensing theory but also offers more precise data support for the development of green, low-carbon, and precision agriculture.