Soils are the largest terrestrial carbon pool and support essential ecosystem functions. Soil microorganisms are central to these functions, driving processes like decomposition and carbon sequestration. Soil microbial biomass carbon (microbial carbon) is a key measure of microbial community size and function, acting as a dynamic carbon pool influencing climate feedbacks and linked to soil health. Understanding long-term microbial carbon dynamics is crucial for assessing the impacts of climate and land-cover change. While microbial communities are shaped by complex interactions of factors including temperature, moisture, land cover, pH, and elevation, limited data exists on long-term changes. This study aims to assess the spatial-temporal dynamics of global soil microbial carbon stocks between 1992 and 2013, identify key drivers of change, and pinpoint regions most sensitive to environmental alterations. This will help in developing informed management and conservation strategies to support the continuation of soil ecosystem functions. Climate change and land-cover change are the most significant and dynamic drivers influencing microbial carbon patterns, particularly due to anthropogenic activities. While increased land-use intensity generally leads to decreased microbial carbon, the interactions between climate change and land cover change on soil communities remain poorly understood. This study seeks to address this gap and provide a comprehensive assessment of the sensitivity of microbial carbon to these interacting factors at the global scale.

The literature extensively documents the importance of soil microorganisms in maintaining ecosystem functions. Studies have highlighted the role of microbial communities in nutrient cycling, carbon sequestration, and overall soil health. Several techniques exist for measuring microbial carbon, including fumigation, substrate-induced respiration, and PLFA analysis. Existing research shows that microbial carbon is influenced by a range of factors, including temperature, moisture, soil organic carbon content, pH, and land cover. However, there's a scarcity of long-term, spatially explicit data on microbial carbon dynamics, limiting our understanding of its response to global environmental change. Previous research has suggested a potential decline in microbial carbon due to climate change and land use intensification, but these studies often lack the global scope and long-term perspective needed to fully understand the phenomenon. This research builds on existing knowledge by leveraging a comprehensive global dataset and advanced modelling techniques to address the identified data gaps and provide a more holistic understanding of global soil microbial carbon dynamics.

This study utilized a global soil microbial biomass carbon dataset (updated from Xu et al., 2013), encompassing 762 independent entries from various ecosystems. The dataset was complemented by global environmental layers, including climate data (CHELSA), land cover (ESA CCI), soil properties (SoilGrids), elevation (WorldClim), and Normalized Difference Vegetation Index (NDVI) data from NOAA. Wetlands and bare areas were excluded from the analysis. The data was standardized, and a standardized land-cover type classification was created based on the ESA CCI data product, harmonizing classes across datasets. A random forest model was employed to predict soil microbial carbon from a set of environmental covariates. The model's performance was validated using a cross-validation approach, splitting the data into training and validation sets. To account for stochastic variability, 100 model runs were performed to assess the importance of environmental factors and the stability of predictions. The models examined the relationships between microbial carbon and environmental variables and generated global predictions of microbial carbon stocks annually between 1992 and 2013. Partial dependence plots were used to analyze the effects of individual predictor variables on microbial carbon. An environmental coverage analysis, using Mahalanobis distance and Area of Applicability methods, was implemented to identify regions where predictions could be made with high confidence. This analysis accounted for multi-dimensional outliers in environmental parameters. Rates of change in soil microbial carbon stocks were calculated for both global and regional scales, using quantile distributions for grouping. The significance of trends was evaluated using linear regression. To determine the relative contributions of climate change and land-cover change to microbial carbon dynamics, model predictions were generated with either climatic or land-cover variables held constant at their 1992 values. The Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) sub-regions were used for regional analysis. Microbial carbon stocks were estimated using predicted concentrations, soil bulk density, and the fraction of coarse fragments, following Hengl et al. (2014). Potential bias due to differing measurement methods for microbial carbon was assessed by comparing the results using only fumigation data and through the inclusion of measurement method as a predictor variable in the random forest model. The study rigorously addressed data limitations and potential biases, providing a robust assessment of global soil microbial carbon dynamics.

The study's key findings reveal a global decrease of 3.4 ± 3.0% in soil microbial carbon stocks between 1992 and 2013, equivalent to a loss of 149 Mt. This equates to a yearly loss of 7.09 Mt. Northern regions, characterized by high initial microbial carbon stocks, exhibited the most substantial declines, primarily driven by rising temperatures. While land-cover change showed a weaker overall impact globally, it exerted considerable influence at the regional level. Coniferous forests had the highest and most variable microbial carbon, followed by tropical forests, while croplands and shrublands showed the lowest values. The analysis of variable importance from random forest models consistently identified mean annual temperature as the most significant predictor, followed by soil organic carbon and soil pH. Partial prediction response curves showed mostly linear positive relationships for soil organic carbon and nitrogen content with microbial carbon, while non-linear relationships were observed for clay content, pH, and temperature. Spatial predictions revealed a strong latitudinal gradient in microbial carbon, with the highest stocks at high latitudes in the Northern Hemisphere and declining values towards the equator. Regional analysis using IPBES sub-regions highlighted significant microbial carbon losses in several regions, including North America, Eastern Europe, North Africa, South America, Southern Africa, and Central and Western Europe. The Caribbean was the only region exhibiting a significant increase. The analysis comparing model predictions with fixed climatic and land cover variables showed that climate change was the primary global driver of microbial carbon decline, while land-cover change exhibited context-dependent regional effects. Model validation yielded an RMSE of 65.0 mmol kg⁻¹, and a cross-validated R² of 0.40. Environmental coverage analysis demonstrated that the model's predictions could be applied with confidence to 50.2% of terrestrial surfaces, excluding glaciers. The study identified regions particularly vulnerable to microbial carbon loss, highlighting the risk of impaired soil ecosystem functions in these areas.

The findings confirm a significant decline in global soil microbial carbon stocks over the past two decades, with temperature increases as a major driver, particularly in high-latitude regions. These declines have substantial implications for soil health and ecosystem services. The significant regional variations in the relative importance of climate versus land-cover change underscore the context-dependent nature of these impacts. The study successfully utilized a large-scale dataset and advanced modelling techniques to analyze long-term dynamics, addressing a critical data gap in the field. The results emphasize the interconnectedness of climate change and land-use practices in influencing soil microbial communities. The identification of vulnerable regions provides crucial information for targeted conservation and management strategies. While the study focuses on microbial carbon as a key indicator of soil health, future research should integrate other measures of microbial community structure and function to further refine our understanding of these complex ecosystems. The model's limitations, such as the reliance on existing datasets and the use of yearly rather than seasonal resolution, highlight avenues for future research.

This study demonstrates a significant decline in global soil microbial carbon stocks driven primarily by rising temperatures, particularly in northern high-latitude regions. Regional variations highlight the complex interplay of climate change and land-cover changes. The findings emphasize the urgent need for strategies to mitigate climate change and implement sustainable land management practices to safeguard soil health and ecosystem services. Future work should focus on refining the model using higher resolution data and exploring other indicators of microbial community health to provide a more nuanced understanding of this essential aspect of global biogeochemical cycles. Further research should explore specific local management strategies that address the context-dependent effects of climate and land-cover change on soil microbial communities.

The study's reliance on existing datasets introduces limitations related to data availability and spatial heterogeneity. The limited temporal resolution (yearly) does not capture seasonal variations in microbial activity. The model's extrapolations to areas with limited data require caution, and uncertainties associated with soil bulk density estimations might affect microbial carbon stock calculations, particularly in high-latitude regions. The assumption of similar calibration among different microbial carbon measurement methods needs further scrutiny. Although these limitations were addressed by the methodology, caution is required when interpreting results, especially at fine spatial and temporal scales.