Juhwan Lee

Large-scale assessment of soil organic matter (SOM) relies on spatial prediction to extend sparse observations to unsampled locations, but extensive soil databases do not necessarily ensure reliable predictions. Predictive uncertainty can arise from limitations in sampling, response-variable perturbation, and model specification, yet their relative contributions are rarely quantified. Here, we developed nationwide models of SOM in paddy soils across South Korea and used an ensemble random forest framework to decompose predictive uncertainty into sampling, response-variable perturbation, and model-specification components. We compiled 283,618 observations with soil, terrain, climate, and vegetation predictors. The model explained 50 % of the variation in independent SOM observations but underpredicted localized areas with high SOM. Model specification was the largest contributor to total predictive variance (41.9 %) and was the dominant uncertainty source at 55.8 % of locations, followed by sampling (29.3 %) and response-variable perturbation uncertainty (28.8 %). Exchangeable calcium and annual precipitation were the most influential predictors, although their importance rankings varied across model configurations. These results show that predictive performance alone does not reveal the sources of uncertainty in large-scale SOM assessment. Large datasets do not necessarily provide equal predictive information across the response distribution. Explicit decomposition of predictive uncertainty can identify whether improvements should prioritize additional sampling, improved response-variable measurements, or alternative model specifications, providing a basis for more targeted strategies to improve the reliability of large-scale SOM predictions.

Large-scale assessment of soil organic matter (SOM) relies on spatial prediction to extend sparse observations to unsampled locations, but extensive soil databases do not necessarily ensure reliable predictions. Predictive uncertainty can arise from limitations in sampling, response-variable perturbation, and model specification, yet their relative contributions are rarely quantified. Here, we developed nationwide models of SOM in paddy soils across South Korea and used an ensemble random forest framework to decompose predictive uncertainty into sampling, response-variable perturbation, and model-specification components. We compiled 283,618 observations with soil, terrain, climate, and vegetation predictors. The model explained 50 % of the variation in independent SOM observations but underpredicted localized areas with high SOM. Model specification was the largest contributor to total predictive variance (41.9 %) and was the dominant uncertainty source at 55.8 % of locations, followed by sampling (29.3 %) and response-variable perturbation uncertainty (28.8 %). Exchangeable calcium and annual precipitation were the most influential predictors, although their importance rankings varied across model configurations. These results show that predictive performance alone does not reveal the sources of uncertainty in large-scale SOM assessment. Large datasets do not necessarily provide equal predictive information across the response distribution. Explicit decomposition of predictive uncertainty can identify whether improvements should prioritize additional sampling, improved response-variable measurements, or alternative model specifications, providing a basis for more targeted strategies to improve the reliability of large-scale SOM predictions.

Large-scale assessment of soil organic matter (SOM) relies on spatial prediction to extend sparse observations to unsampled locations, but extensive soil databases do not necessarily ensure reliable predictions. Predictive uncertainty can arise from limitations in sampling, response-variable perturbation, and model specification, yet their relative contributions are rarely quantified. Here, we developed nationwide models of SOM in paddy soils across South Korea and used an ensemble random forest framework to decompose predictive uncertainty into sampling, response-variable perturbation, and model-specification components. We compiled 283,618 observations with soil, terrain, climate, and vegetation predictors. The model explained 50 % of the variation in independent SOM observations but underpredicted localized areas with high SOM. Model specification was the largest contributor to total predictive variance (41.9 %) and was the dominant uncertainty source at 55.8 % of locations, followed by sampling (29.3 %) and response-variable perturbation uncertainty (28.8 %). Exchangeable calcium and annual precipitation were the most influential predictors, although their importance rankings varied across model configurations. These results show that predictive performance alone does not reveal the sources of uncertainty in large-scale SOM assessment. Large datasets do not necessarily provide equal predictive information across the response distribution. Explicit decomposition of predictive uncertainty can identify whether improvements should prioritize additional sampling, improved response-variable measurements, or alternative model specifications, providing a basis for more targeted strategies to improve the reliability of large-scale SOM predictions.

Soil test data, derived from laboratory analyses of chemical properties such as organic matter (OM), provide a critical foundation for soil management and guiding fertilizer use. In Korea, the Rural Development Administration (RDA) has maintained a national soil test database (“HeukToram”) and associated soil attribute maps for commercial agricultural fields for several decades. However, the reliability of these data is often compromised by manual entry errors and reporting inconsistencies. This study evaluates the impact of error detection and correction on OM assessment at both sub-national and national scales. To address this, soil test records for paddy rice collected from the official dataset between 2022 and 2025 were analyzed and modeled ( n = 913,984). The dataset was processed through systematic screening for apparent errors. At the national level, the most frequent issue was repeated soil test values across consecutive sampling dates (16.37% of records), followed by values outside agronomically valid ranges (0.81%), records with all variables missing (0.06%), and those with identical values across all variables (0.02%). Depending on the type and extent of these errors, mean OM estimates deviated by up to 0.153 g kg ⁻¹ between the datasets, potentially affecting field-level fertilizer recommendations. The results indicate that even relatively minor errors in large-scale soil test datasets can lead to substantial misestimations in OM stocks and nutrient management requirements, the magnitude of which often remains unknown. Continuous quality control is therefore essential to maintain reliable soil assessments and to guide effective crop and soil management.

The Republic of Korea has extensive data (HeukToram) and maps related to soil attributes. Since the 1960s, the Rural Development Administration (RDA) has conducted nationwide soil testing and mapping for commercial fields. This Korean Soil Information System, which continues to expand through data collection by local Agricultural Technology Extension Centers (ATC), is the primary official resource for supporting growers with organic matter management and chemical fertilization. While the database provides broad national coverage and fulfills its initial purpose, its application is mostly limited to farm or field levels. Several challenges undermine the data's reliability and broader applicability, limiting its contribution to regional and global databases. This opinion paper addresses issues related to impractical data utilization, human errors in soil test results, spatial inaccuracies in soil maps, and the limitation of attributes and field-level soil profiling. It also proposes actions to improve the effectiveness and scope of these databases.

Machine learning (ML) has become more prevalent as a tool used for biogeochemical analysis in agricultural management. However, a common drawback of ML models is the lack of interpretability, as they are black boxes that provide little insight into agricultural management. To overcome this limitation, we compared three tree-based models (decision tree, random forest, and gradient boosting) to explain soil organic matter content through Shapley additive explanations (SHAP). Here, we used nationwide data on field crops, soil, terrain, and climate across South Korea (n = 9584). Using the SHAP method, we identified common primary controls of the models, for example, regions with precipitation levels above 1400 mm and exchangeable potassium levels exceeding 1 cmol⁺ kg⁻¹, which favor enhanced organic matter in the soil. Different models identified different impacts of macronutrients on the organic matter content in the soil. The SHAP method is practical for assessing whether different ML models yield consistent findings in addressing these inquiries. Increasing the explainability of these models means determining essential variables related to soil organic matter management and understanding their associations for specific instances.

Machine learning has become more prevalent as a tool used for biogeochemical analysis in agricultural management. However, a common drawback of machine learning models is the lack of interpretability, as they are black boxes that provide little insight into agricultural management. To overcome this limitation, we compared three tree-based models (decision tree, random forest, and gradient boosting) to explain soil organic matter content through Shapley additive explanations (SHAP). Here, we used nationwide data on field crops, soil, terrain, and climate across South Korea (n = 9,584). Using the SHAP method, we identified common primary controls of the models, for example, regions with precipitation levels above 1400 mm and exchangeable potassium levels exceeding 1 cmol+ kg-1, which favor enhanced organic matter in the soil. Different models identified different impacts of nutrients on the organic matter level in the soil. The SHAP method is practical for assessing whether different machine learning models yield consistent findings in addressing these inquiries. Increasing the explainability of these models means determining essential variables related to soil organic matter management and understanding their associations for specific instances.