Soil organic carbon (SOC) sequestration provides an opportunity to mitigate climate change impacts, since soils are the largest store of terrestrial carbon. Accurate estimates of SOC content across landscapes are therefore important to monitor and manage efficiently these SOC stocks. Mid-infrared (MIR) spectroscopy has been increasingly applied as a rapid, cost-effective, and accurate method for predictive soil analysis. This study assessed the performance of MIR spectroscopy for SOC prediction at a regional scale for remote landscapes in Iran. The potential for combining environmental covariates, including remotely sensed covariates and terrain attributes, with MIR variables to improve prediction was also tested. Soil samples were collected from 151 locations at two depths (0–5 and 5–15 cm) across a large study area (350 km²) and analysed for gravimetric SOC content. Partial least squares regression (PLSR) was used to model SOC from MIR spectra recorded on the samples and to obtain latent variables (LV) that were then used, either on their own or alongside environmental covariates, as input to a Cubist rule-based model. The Cubist model using the LV alone outperformed the PLSR model and produced a high prediction accuracy with an R² of 0.96, RPIQ of 5.61, and RMSE of 0.16% on the validation set. The inclusion of environmental covariates alongside LV did not improve the performance of the model compared with the model on LV alone (R² = 0.94, RPIQ = 4.81, RMSE = 0.19%). The high performance of the developed models indicates the high potential of MIR spectroscopy for SOC prediction in data-scarce areas.

由于土壤是陆地碳的最大储存地，因此固存土壤有机碳（SOC）提供了减轻气候变化影响的机会。因此，跨景观的SOC含量的准确估算对于有效监视和管理这些SOC库存非常重要。中红外（MIR）光谱已越来越多地用作预测性土壤分析的快速，经济高效且准确的方法。这项研究评估了MIR光谱在伊朗偏远地区区域范围内SOC预测中的性能。还测试了将环境协变量（包括遥感协变量和地形属性）与MIR变量组合以改善预测的潜力。在一个较大的研究区域（350公里）的两个深度（0-5和5-15厘米）的151个位置采集了土壤样品²）并分析了重量SOC含量。偏最小二乘回归（PLSR）用于根据样本上记录的MIR光谱对SOC建模，并获得潜在变量（LV），然后将其用作单独变量或与环境协变量一起用作基于Cubist规则的模型的输入。仅使用LV的Cubist模型优于PLSR模型，并在验证集上产生了较高的预测准确性，R ²为0.96，RPIQ为5.61，RMSE为0.16％。与单独使用LV的模型相比，与LV一起包含环境协变量并没有改善模型的性能（R ² = 0.94，RPIQ = 4.81，RMSE = 0.19％）。所开发模型的高性能表明MIR光谱在数据稀缺地区的SOC预测中具有很高的潜力。