Empirical relationships between environmental factors and soil organic carbon produce comparable prediction accuracy to machine learning

Abstract

Accurate representation of environmental controllers of soil organic carbon (SOC) stocks in Earth System Model (ESM) land models could reduce uncertainties in future carbon–climate feedback projections. Using empirical relationships between environmental factors and SOC stocks to evaluate land models can help modelers understand prediction biases beyond what can be achieved with the observed SOC stocks alone. In this study, we used 31 observed environmental factors, field SOC observations (n = 6,213) from the continental United States, and two machine learning approaches (random forest [RF] and generalized additive modeling [GAM]) to (a) select important environmental predictors of SOC stocks, (b) derive empirical relationships between environmental factors and SOC stocks, and (c) use the derived relationships to predict SOC stocks and compare the prediction accuracy of simpler model developed with the machine learning predictions. Out of the 31 environmental factors we investigated, 12 were identified as important predictors of SOC stocks by the RF approach. In contrast, the GAM approach identified six (of those 12) environmental factors as important controllers of SOC stocks: potential evapotranspiration, normalized difference vegetation index, soil drainage condition, precipitation, elevation, and net primary productivity. The GAM approach showed minimal SOC predictive importance of the remaining six environmental factors identified by the RF approach. Our derived empirical relations produced comparable prediction accuracy to the GAM and RF approach using only a subset of environmental factors. The empirical relationships we derived using the GAM approach can serve as important benchmarks to evaluate environmental control representations of SOC stocks in ESMs, which could reduce uncertainty in predicting future carbon–climate feedbacks.

Upscaling Soil Organic Carbon Measurements at the Continental Scale Using Multivariate Clustering Analysis and Machine Learning

“…Multiple environmental factors control SOC, including climate, plant productivity, edaphic properties, and topographic variables (McBratney et al., 2003). Moreover, understanding the spatial distribution of what variables drive SOC is crucial for predicting how ecosystems will respond to changing climate conditions (Doetterl et al., 2015; Gautam et al., 2022; Gonçalves et al., 2021; Mishra et al., 2022). Thus, it is critical to understand the key predictive drivers across different regions.…”

“…Mapping SOC stocks can help identify areas with high C sequestration potential (Rumpel et al, 2020;Smith et al, 2020;Vågen & Winowiecki, 2013) or regions more susceptible to climate change impacts (Ahmed et al, 2017). Significant efforts have been made to collect and upscale soil profile data for mapping SOC stocks at regional or global scale (Amundson, 2001;Batjes, 1996;Chaney et al, 2019;FAO & ITPS, 2020;Guevara et al, 2020;Hengl et al, 2014Hengl et al, , 2017Mishra et al, 2022;Ramcharan et al, 2018;Scharlemann et al, 2014;Stockmann et al, 2015;Tarnocai et al, 2009). Soil C mapping methods have been constantly evolving, leveraging refinement and innovation from various fields to enhance map accuracy.…”

Sentinel-1 Imagery Used for Estimation of Soil Organic Carbon by Dual-Polarization SAR Vegetation Indices

“…RF and SVM methods are already widely employed in soil attribute prediction and in digital soil mapping [17,48,49] and can be applied to both regression and classification problems. RF for regression works by building a collection of M regression trees that are random to each other.…”

“…Regression methods are sensitive to the correlation that exists between the predictors [53,54]. This is the case for linear regression, SVM, and RF [17,48,49,55]. Therefore, before splitting the data for training the models, a correlation filter was applied with a threshold equal to 0.80.…”

“…This methodology allows that in the subsampling process of the cross-validation, at each repetition, different data are in the k-folds and are used to adjust the hyperparameters of the model. Thus, it is possible to assess the degree of uncertainty of the models to the input data [48,54,57]. Since there are seven soil layers and three regression methods, this means that 2100 SOC prediction models were fitted.…”