Emergent temperature sensitivity of soil organic carbon driven by mineral associations

Soil carbon–climate feedbacks are a major source of uncertainty in predicting the terrestrial biosphere's response to climate change. This uncertainty arises partly from poorly understood relationships between the temperature sensitivities of soil organic carbon (C) decomposition and stabilization processes – key parameters in soil biogeochemical models. While many studies have investigated temperature's effects on bulk soil C stocks and turnover rates at various scales, they often overlook the heterogeneity within soil organic matter. This is significant because different soil organic matter fractions (pools) – such as particulate or mineral-associated organic matter – can react differently to warming in both lab and field settings. Understanding and quantifying the temperature sensitivity of these pools globally is crucial for accurately predicting emergent feedbacks and soil organic C vulnerabilities to climate change. Warmer soil temperatures accelerate organic matter decomposition, often leading to bulk soil C stock losses. However, organic matter's decomposability can depend on its association with minerals. Mineral-associated organic matter comprises small, chemically or physically bound plant- or microbial-derived components, while particulate organic matter consists of larger, partially decomposed plant fragments. Mineral associations limit microbial access to organic matter, making mineral-associated C often older than particulate C. Soil biogeochemical models represent these pools explicitly or implicitly, impacting their dynamic behavior and long-term responses. We refer to mineral-associated C as 'protected' and particulate C as 'unprotected,' acknowledging the complexity of their bioavailability and persistence. Significant uncertainty remains in how protected and unprotected soil C responds to warming across broad scales, hindering model benchmarking efforts. Evaluating the transient dynamics of soil biogeochemical models under warming is challenging due to limited observations of soil C at long timescales. Laboratory and field studies show increased decomposition rates and C stock decreases with warming, but at longer timescales, microbial acclimation might limit warming's effect. Many experiments focus on bulk soil C, neglecting underlying fractions. An alternative approach uses spatial gradients to benchmark global models, though this is insufficient for constraining transient dynamics. This study assesses the climatological temperature sensitivity of bulk soil C stocks and the role of mineral-associated and particulate soil C pools in driving this emergent property in data and models.

Numerous studies have explored the impact of temperature on soil organic carbon (SOC) dynamics, focusing on bulk SOC stocks and turnover rates. However, these studies often overlook the inherent heterogeneity within SOC, neglecting the distinct responses of different SOC pools to temperature changes. Previous research has demonstrated that mineral-associated organic matter (MAOM) and particulate organic matter (POM) exhibit varying temperature sensitivities. MAOM, protected by its association with soil minerals, generally shows lower temperature sensitivity compared to POM, which is more readily available for microbial decomposition. This difference in temperature sensitivity has implications for SOC dynamics under changing climatic conditions. Several studies have attempted to quantify the temperature sensitivity of SOC using both laboratory incubations and field experiments. However, these studies often lack the spatial scale required to draw global conclusions or have not measured the relative responses of underlying soil C pools. The use of space-for-time substitution has been employed to infer long-term temperature sensitivities from spatial climate gradients, but this approach has limitations in capturing transient dynamics. Therefore, a comprehensive understanding of the temperature sensitivity of SOC requires a global-scale analysis that considers the heterogeneity of SOC pools and accounts for the limitations of using spatial gradients as proxies for temporal changes.

This study leverages a globally gridded data product of mineral-associated and particulate soil C derived from an observational synthesis of soil fractions to quantify the distribution of soil organic C between these two pools and their respective climatological temperature sensitivities at the global scale. The data product was analyzed to assess the relationship between soil C stocks (in protected and unprotected pools) and mean annual temperature (MAT), controlling for confounding factors like net primary productivity, mean annual precipitation, and clay and silt content using multiple linear regression. Climatological temperature sensitivity was calculated as the proportional decline in C stocks for a 10°C increase in MAT. The analysis was performed globally and separately for cool (<15°C) and warm (≥15°C) climate regions. For the model analysis, data were obtained from nine Earth System Models (ESMs) from CMIP6 and three offline land models. The slowest-cycling pool in most ESMs was considered analogous to mineral-associated C, allowing for a comparison of pool distributions and temperature sensitivities between data and models. The proportion of protected C was calculated for each model, and the multi-model mean was computed to compare with the observational data. Climatological temperature sensitivity was calculated for each model, following the same methodology as the data analysis. Comparisons were made between the observational data and model predictions for both the distribution of protected C and the climatological temperature sensitivity of each pool. The analyses focused on non-permafrost mineral soils with MAT > 0°C to avoid confounding factors associated with permafrost and water saturation. Details on data sources, processing, and statistical analyses are provided in the Methods section of the paper, along with details regarding the rationale for the choice of model pools and the challenges of comparing observations to models.

The study reveals that the climatological temperature sensitivity of unprotected (particulate) soil C is significantly higher than that of protected (mineral-associated) C globally. Specifically, unprotected C exhibits 28% greater temperature sensitivity than protected C, with this difference reaching 53% in cool climates (<15°C). In warm regions (≥15°C), the difference is smaller (15%). Both pools show weak temperature sensitivities in warm regions, suggesting that decomposition in cool regions is more temperature-limited. Globally, the proportion of protected C increases with increasing MAT, a trend not captured by most global models. The distribution of C among these pools significantly influences the emergent climatological temperature sensitivity of bulk soil C. A wide range of model predictions for the proportion of protected C (16% to 85%) exists across the analyzed Earth system models and offline land models. Around half of the ESMs underestimate the proportion of protected C compared to the observational data product, indicating an overestimation of C in faster-cycling pools. Several models adequately capture the proportion of protected C, while others considerably underestimate this proportion. At higher latitudes, the microbially explicit models show a lower proportion of protected C, consistent with observational data. However, some CMIP6 models show a larger proportion of protected C at higher latitudes, possibly due to the relative timescales of decomposition and vertical mixing processes. Models vary widely in their predictions of the climatological temperature sensitivity of bulk soil C. While some models capture the bulk temperature sensitivity, they often do not accurately represent the contributions of underlying pools. The study also finds that models tend to overestimate the climatological temperature sensitivity of protected C and underestimate the sensitivity of unprotected C in cool climates. This suggests potential biases in transient responses to warming, particularly in cool climates, with implications for soil C dynamics and age distributions under future scenarios. The discrepancy between model predictions and observed data highlights the need to consider the distribution and temperature sensitivities of underlying soil C pools to accurately predict soil C dynamics under climate change.

The findings highlight the critical need to improve the representation of soil organic carbon (SOC) pools and their temperature sensitivities in global climate models. The significant difference in temperature sensitivity between protected and unprotected SOC pools underscores the importance of considering pool-specific dynamics rather than relying solely on bulk SOC measures. The large variation in model predictions for the proportion of protected SOC reveals a key area for model improvement. The underestimation of protected SOC in many models suggests that these models may overestimate the responsiveness of SOC to climate change and CO2 fertilization, potentially leading to inflated estimates of future carbon sequestration. The underestimation of the temperature sensitivity of unprotected SOC in cool climates may also lead to an underestimation of SOC losses in these regions under warming. These findings directly address the research question by demonstrating the crucial role of underlying SOC pool distributions and their temperature sensitivities in determining the emergent temperature sensitivity of bulk SOC. This improved understanding has significant relevance to the field of climate modeling and carbon cycle prediction, allowing for more accurate assessments of climate-carbon feedback mechanisms. The observed discrepancies between models and data provide important benchmarks for evaluating and refining model formulations.

This study demonstrates that the distribution and climatological temperature sensitivities of underlying soil C pools are critical ecosystem properties that global models must accurately represent. While some models capture the overall temperature sensitivity of bulk soil C, they often fail to accurately reproduce the contributions of underlying protected and unprotected pools. This has implications for projections of soil C dynamics, ages, and ecosystem responsiveness to climate change. The underestimation of protected C in several models suggests an overestimation of the system's responsiveness to inputs, potentially leading to inflated estimates of carbon sequestration. Conversely, the underestimation of unprotected C sensitivity in cool climates may result in an underestimation of carbon losses in these regions. Future global models should explicitly report soil C pool distributions, allowing for more rigorous evaluations and refinements. Continued research on transient responses and detailed pool-specific data across diverse climate zones is needed for enhancing the accuracy and reliability of global carbon cycle–climate feedback projections.

The study's analysis is limited by the availability of globally consistent and high-resolution data on soil C pools, particularly in cool and sandy soils. While the study utilized the most comprehensive available data product, uncertainties remain regarding the accuracy of the data, potentially affecting the conclusions. The comparison of model pools with observational data relies on conceptual similarities and interpretations, acknowledging potential mismatches between operationally defined fractions and model states. Furthermore, the study focuses primarily on climatological temperature sensitivities derived from spatial gradients, which may not perfectly represent transient responses to warming. This limits the ability of the study to provide quantitative predictions of future soil C dynamics. However, the findings provide valuable benchmarks for evaluating and refining model parameterizations and reducing uncertainty in future projections of soil C dynamics under climate change.