Journal Paper Digests 2026 #5
- The influence of rewetting intensity on soil priming after drought
- Interactions between soil environmental factors and microbial communities consistently predict plant health
- Commentary: Structural equation models and causal claims in soil science and biogeochemistry – An equation-free “how to”
- Knowledge-Guided Machine Learning for Global Change Ecology Research
- Spatiotemporal Trade-Offs and Synergies Among Environmental Footprints of Grain Crop Production in China
- Combining physical models and machine learning for enhanced soil moisture estimation
- Potential for improving micronutrient supply and environmental sustainability by using underutilized crops in China
- Generating geochemical and mineralogy distributions of soil in the conterminous United States using Bayesian hierarchical spatial models
Generating geochemical and mineralogy distributions of soil in the conterminous United States using Bayesian hierarchical spatial models
Characterizing geochemical and mineralogical soil distributions across large spatial extents is essential for understanding mineral resources, ecosystem processes, and environmental risks. Rasters of soil geochemical distributions for the conterminous United States, however, are limited. We present a Bayesian modeling workflow and tool for generating predictive geochemical and mineralogy distribution maps for the conterminous United States using integrated nested Laplace approximation (INLA) with the stochastic partial differential equation approach. By modeling soil geostatistical data with environmental covariates (soil properties, topography, climate, and land cover), we generate predictive distributions of soil geochemistry that can be mapped or extracted for further analyses. As an example, we model the spatial distribution of trace elements in soil relevant to vertebrate health (cobalt, copper, iron, manganese, selenium, and zinc) and provide a workflow that can be used to generate and visualize predictive distributions of 39 other major and trace elements and 21 minerals of the soil survey, supporting a variety of ecological, environmental, and agricultural applications. Bayesian Modeling: Uses R-INLA to predict soil geochemistry across large spatial extents. Covariate Integration: Incorporates environmental variables to increase predictive accuracy. Raster Generation: Produces continuous geospatial layers of mineral and element distributions of the conterminous United States for a variety of applications.
Potential for improving micronutrient supply and environmental sustainability by using underutilized crops in China
Rice and wheat provide the bulk of calories in diets globally. However, foods made from these cereals are commonly in refined forms and are low in micronutrients and dietary fiber. Increasing the consumption of more nutrient-dense, underutilized cereals and beans (UCBs), such as millet, sorghum, mung bean, along with unrefined rice and wheat, could improve diet quality. Compared with rice and wheat, UCBs are generally cultivated using less intensive methods, resulting in a lower environmental impact, though their productivity is generally lower. This study explores how reallocating rice and wheat areas to UCBs, either alone or combined with greater use of unrefined rice and wheat, could potentially enhance micronutrient supply (iron, thiamin, riboflavin, calcium, zinc), while reducing water use and greenhouse gas emissions in China. A strategy combining area reallocation and greater use of unrefined rice and wheat increased micronutrient supply and dietary fiber by 12–82%, reduced environmental impact by 11–12%, and slightly increased energy supply (3%). These outcomes were achieved by reallocating 7.9 million hectares (Mha) of rice area (26% of the current total) and 1.7 Mha of wheat area to sorghum (+5.5 Mha), millet (+2.5 Mha), beans (+1.4 Mha), and oats (+0.2 Mha). As a result, the supply of UCBs and unrefined rice and wheat products increased, supporting healthier diets. Reallocating only 5% of the rice area would still yield improvements, especially for dietary fiber and iron ( 27%). These findings offer insights for rethinking the value of UCBs and supporting their integration into future food system strategies.
Combining physical models and machine learning for enhanced soil moisture estimation
Estimating soil moisture is crucial for agricultural management, water resource planning, and environmental monitoring. Traditional methods, whether based on physical models or machine learning, face limitations, with physical models suffering from reduced accuracy due to parameter sensitivity and environmental variability, and machine learning models struggling with interpretability and generalization. To address these challenges, this study introduces a novel hybrid approach that leverages the strengths of both physical modeling and machine learning to enhance soil moisture estimation. The hybrid model, ML-Phy-meteo, integrates the physical model (Hydrus-1D) results and meteorological data into a LightGBM framework, achieving optimal estimation accuracy across various soil depths. Quantitatively, ML-Phy-meteo exhibits superior performance across all depths, achieving an average root mean square error (RMSE) between 0.020 and 0.026 cm3/cm3, an average Nash–Sutcliffe Efficiency (NSE) ranging from 0.195 to 0.811, and an average Kling–Gupta Efficiency (KGE) from 0.623 to 0.860, thereby outperforming both standalone physical models and purely machine learning-based approaches. Notably, ML-Phy-meteo achieves high-precision predictions even in the absence of detailed soil texture and stratification data, with the machine learning component effectively compensating for the simplifications of the physical model. Among the machine learning methods used in the hybrid model, tree-based models (LightGBM and Random Forest) outperform deep learning models (LSTM) in terms of accuracy and robustness in handling noise and missing data, despite the latter’s smoother prediction profiles. These findings highlight the potential of hybrid models to overcome the inherent limitations of standalone physical or machine learning approaches, providing new ideas for future research and applications in soil moisture estimation.
Spatiotemporal Trade-Offs and Synergies Among Environmental Footprints of Grain Crop Production in China
Human agricultural activities have exacerbated multiple types of natural resource depletion and environmental impacts through complex interactions with land, water, carbon, and nutrient cycles, which can be measured as corresponding environmental footprints (EFs). However, the spatiotemporal trade-offs and synergies among multiple EFs in agricultural systems remain under-quantified, hindering effective mitigation strategies. Here, we propose an assessment framework of spatiotemporal trade-offs and synergies among multiple EFs of crop production, with a case study on blue water, green water, land, carbon, nitrogen, and phosphorus footprints for wheat, maize, rice, and soybean production across 31 Chinese provinces over 2000–2018. In total, 3630 pairwise EFs were analyzed. Results show that, although the EFs of unit mass crop production generally declined across provinces, national total EFs increased, with land, carbon, and phosphorus footprints rising by 16%, 17%, and 23%, respectively, during the study period. Synergistic interactions among EFs prevailed, comprising 50% positive and 32% negative synergies. The spatial distribution of trade-offs and synergies varies by crop and region. Land use intensity is the main factor limiting the positive EF synergies.
Knowledge-Guided Machine Learning for Global Change Ecology Research
Global change ecology demands predictive models that reconcile data-driven learning with mechanistic theory to address complex, interconnected ecosystem challenges. Traditional process-based approaches struggle with spatiotemporal parameterization, while purely data-driven machine learning approaches suffer from extrapolation, interpretability, and physical consistency. Knowledge-guided machine learning (KGML) bridges this divide by systematically integrating ecological principles (e.g., physical first principles, stoichiometry, process understanding, disturbance regimes) into how models are designed, trained, and adjusted to generalize across different ecosystems. The emerging KGML paradigm offers tremendous opportunities to advance the research of global change ecology. This review synthesizes KGML’s transformative potential, showcasing its capacity to enhance the prediction of carbon-water-nutrient cycles and other ecological processes and lay groundwork for ecological foundation models. Emerging applications in decision support and symbolic regression further illustrate its role in deriving actionable insights and novel theoretical hypotheses. Future directions emphasize adaptive integration of data and knowledge, uncertainty quantification, causal embedding in foundation models, and interdisciplinary collaboration to align KGML innovations with sustainability goals. By uniting ecological theory with AI advances, KGML offers a robust pathway to encompass ecosystem responses to global change, fostering scientific discovery and actionable solutions.
Commentary: Structural equation models and causal claims in soil science and biogeochemistry – An equation-free “how to”
Structural equation modeling (SEM) is a set of approaches that have seen exponential usage in the soil sciences as well as the related fields of agriculture and biogeochemistry. When correctly used and interpreted, SEM can be a powerful and flexible tool to test complex hypotheses on causality. However, the recent explosion of SEM usage in the soil sciences facilitated by user-friendly statistical programs has not been fully met by statistical expertise of users, reviewers and editors, ultimately leading to widespread contamination of the literature with inappropriate modeling and inflated or unfounded causal claims. The rise of such “SEM slop” poses a serious risk of an unreliable knowledge base and also undermines efforts and standards on what constitutes causality in the soil sciences. To address this, we diagnose major pitfalls in SEM, with an eye towards considerations specific to soil sciences, categorizable as three types: (1) Causal claims, including not satisfying causal criteria, lack of justified a priori models, not considering counterfactuals, and unqualified causal language; (2) Experimental design, including use in randomized complete block designs without complete pooling or multi-level models, inappropriate data type (e.g., ontological misalignment), and insufficient sample size; and, (3) Assessing the model, including incomplete or inappropriate model evaluation, non-qualified use of modification indices, and lack of robustness tests. There is a dual imperative for users as well as reviewers and editors to better implement and evaluate SEMs and claims of causality made with SEMs. To support this, we offer best practices and practical considerations on these three major pitfalls. These best practices will help SEM be appropriately employed as a powerful, nuanced statistical tool that benefits the soil science community.
Interactions between soil environmental factors and microbial communities consistently predict plant health
Intensive agricultural practices cause dysbiosis in soil nutrient levels and microbial communities, significantly affecting plant health and productivity. However, the mechanisms underlying the interactions between soil environmental factors and microbial communities, and their role in determining and predicting plant health, remain poorly understood. In this study, we collected soils planted with tomato in different health conditions, including healthy and bacterial wilt, Fusarium wilt, and nematode diseases, to identify key abiotic and biotic factors influencing plant health. Additionally, We fitted machine learning models using multidimensional data to classify plant health status. Our results revealed that diseased soils (bacterial wilt, Fusarium wilt, and nematode disease) exhibited significantly higher AP levels compared to healthy soils. Moreover, increased Amplicon Sequence Variants (ASVs) in diseased soils had lower network connectivity and were positively correlated with soil nutrient contents, pathogen abundance, and pathogen-supportive soil microbial functions, while being negatively correlated with plant defense-associated soil microbial functions. Both soil nutrient levels and the increased ASVs in diseased soil were stronger correlates of disease occurrence than other soil indicators. Optimal classification performance was observed when both soil environmental factors and microbial communities were considered, with AP emerging as the most influential indicator. In conclusion, excessive accumulation of AP was associated with disrupted microbial community structures, destabilized microbial networks, enhanced pathogen abundance, and impaired microbial functions, which collectively correlated with higher disease occurrence. These findings highlight the potential importance of optimizing soil nutrient management for supporting plant health.
The influence of rewetting intensity on soil priming after drought
Soil moisture is a key driver of soil organic matter (SOM) decomposition and the global carbon (C) cycle, and climate warming-induced extremes of rainfall and drought are intensifying the dynamics of soil C stocks. Although addition of labile C pools can trigger strong priming effects (PE) by stimulating decomposition of recalcitrant SOM, how varying rewetting intensity interacts with these exogenous inputs of C pools to affect PE remains unclear. In this study, we used an isotopic approach to examine the effects of rewetting intensity (40%, 60%, and 80% water holding capacity (WHC)) on PE during a 21-day incubation following intensive drought (20% WHC). Our results show that glucose addition induced a strong positive PE, with higher moisture (60% and 80% WHC) resulting in greater PEs. High moisture regulated microbial community composition and boosted microbial activity and turnover. These changes heightened microbial nitrogen demand, accelerated nitrogen mining, and intensified decomposition of stable C, leading to net soil-C loss. The correlation analysis shows that enhanced biosynthetic and degradative activity under higher moisture conditions facilitates the turnover of labile C, whereas reduced microbial diversity and metabolic intensity promote the stabilization of more persistent C forms. This study underscores the significant role of moisture in shaping PEs and soil C dynamics in the subtropical forest soils, offering insights into soil C sequestration in response to climate change.