Introduction

Earth System Models (ESMs) are indispensable tools for predicting the planetary response to climate change1. The accuracy and reliability of ESMs are crucial for informing climate projections that guide policy decisions. Soils store more carbon (C) than plants, the surface ocean or the atmosphere, and thus are critical for the functioning of the Earth system2. While ESMs are becoming increasingly complex, their predictions of soil organic C (SOC) stocks have improved only marginally in recent decades3,4.

Microbial communities process most of the C entering the soil, thereby shaping its fate5,6. Microbes metabolize multiple C sources, including detritus, root exudates, and microbial metabolites7. The energy needed to acquire C depends on whether the compounds can be taken up directly or require prior enzymatic degradation8. Additionally, microbial community composition and functioning are influenced by prevailing climatic conditions9,10,11. The general omission of microbial community structure and related processes in C cycle models has been suggested as one of the causes for their poor performance in predicting SOC stocks and their responses to climate change12,13.

Recognizing the impracticality of representing every conceivable microbial metabolic pathway, many models combine a spectrum of microbial processes into a single metric referred to as microbial C use efficiency (CUE)14,15. CUE, as a model parameter or as a system property emerging from multiple co-occurring processes, represents the fraction of C uptake allocated to the production of new microbial biomass16. Using this definition, CUE declines as more C is used for respiration to generate energy (for substrate uptake, cellular maintenance, enzyme production) or for exudation (extracellular enzymes, polysaccharides)17,18. This pragmatic approach streamlines the modeling of soil C cycling by incorporating the diverse fates of microbial C, including biomass production, respiration, and exudation, thereby providing a more comprehensive understanding of microbially-mediated C-pathways.

However, accurately integrating the spatial or temporal dynamics of microbial CUE into soil C models remains a significant challenge. Most of the current C cycle models either lack explicit representation of CUE or treat it as a constant value4, despite our understanding that CUE varies under different environmental conditions. For example, observations indicate significant variability in CUE at the global scale8, which may be partially attributed to inconsistencies among measurement techniques (Fig. 1a). Moreover, comparisons across ecosystems reveal that CUE is generally higher in grasslands than in croplands, with forests consistently showing the lowest CUE values, regardless of the measurement approaches used19,20 (Fig. 1c). CUEs derived from data assimilation21 are also lower than those from more direct measurement approaches (Fig. 1d).

Fig. 1: Variability of carbon use efficiency (CUE) at a global scale.
figure 1

a Observation-based CUE estimates at the global scale from C (13C and 14C) and 18O isotopic labeling, stoichiometric modeling and other methods. Data were collected from19,21,49,95,114. b CUE constants used in the MIcrobial-MIneral Carbon Stabilization model (MIMICS) for two litter types (diamonds). Metabolic litter comprises plant litter that decomposes easily, whereas structural litter is more resistant to decomposition131. c Observation-based estimates for different ecosystems using isotopic labeling114 or stoichiometric modeling19. d CUE values predicted using a microbial model assimilating information on SOC profiles21. Data assimilation integrates observed data into predictive models to refine model parameters and improve estimation accuracy.

Full size image

Several attempts have been made to reflect or incorporate CUE variations into models of litter22 or soil organic matter9,13 decomposition with the aim of assessing the implications for soil C cycling. For example, incorporating an empirically-derived negative relationship between microbial CUE and temperature into a microbial-explicit SOC model improved the simulation of contemporary soil C stocks23. Zhang et al.24 introduced the effects of substrate quality and soil fertility on microbial respiration, highlighting the joint control of litter quality and quantity on the steady-state SOC stocks. Wieder et al.25 enhanced the understanding of CUE variation by including two types of decomposers with differing substrate preferences and CUE (Fig. 1b). These examples suggest that more realistic representations of microbial C transformations have the scope for improving model predictions of soil C23,26. However, these predictions were poorly constrained by observational data, calling their reliability into question21,27,28.

In this Perspective, we synthesize our understanding of CUE regulatory factors and databases for constraining numerical models, with the aim of clarifying complexities, addressing controversies, and providing a holistic perspective on pathways to adequately reflect CUE variations in C cycle models and their consequences for simulated soil C stocks.

Data availability and challenges

Terminology and definitions of microbial CUE

The concept of microbial CUE, the fraction of C uptake that is used to produce microbial biomass16,17,18, is intuitively straightforward, but CUE definitions vary depending on the ecological processes involved, measurement methods, and scales of biological organization (e.g., population, community and ecosystem)14,17. Therefore, CUE can be regarded as an emergent parameter, encapsulating multiple processes within a single metric. It is useful in modeling as the number of processes that can be modeled is constrained by practical limitations (e.g., availability of data for calibration). Consequently, ecosystem models often simplify microbial process complexity, which in reality, escalates from the genomic to the ecosystem level (Fig. 2).

Fig. 2: Schematic representation of a cluster of models integrating observational constraints on CUE at population (CUEP), community (CUEC) and ecosystem (CUEE) scales.
figure 2

The genome-scale metabolic model predicts the movement of metabolites within a cell based on its genomic information. CUEP and CUEC can be validated by short-term incubation measurements, while CUEE requires long-term incubation measurements. Although the scales and processes governing CUE expand from individual cells to entire ecosystems, there is a practical limit to the extent they can be resolved in C cycle models.

Full size image

CUE is quantitatively expressed as the ratio of microbial growth (μ) to C uptake (U)16,29, that is, CUE = μ/U. This ratio encapsulates the efficiency with which microorganisms convert assimilated C into biomass. Microbial uptake involves C assimilation for growth (μ), respiration (R), and the secretion of extracellular enzymes and metabolites (EX). Geyer et al.14 introduced a nested conceptual framework for understanding CUE across different biological organization levels: population (CUEP), community (CUEC), and ecosystem (CUEE). This framework is useful for integrating C fluxes mediated by soil microbes into models at various ecological scales (Fig. 2).

CUEP reflects the species-specific functioning of microbial taxa (e.g., biosynthesis rate, exudate production) and thermodynamics of C substrate metabolism that limits the proportion of C uptake used for biosynthesis versus C lost from the cell (e.g., mineralized or exuded as metabolites). Typically measured in cultured populations, the CUEP formula adjusts for respiration (R) and exudation (EX) losses from the uptake, expressed as CUEP = \(\frac{U-R-{EX}}{U}\). CUEC incorporates additional environmental and community factors influencing microbial metabolism in natural communities consisting of multiple populations. It focuses on gross microbial production prior to the recursive substrate recycling of necromass and exudates, capturing the metabolic response of microbial communities to substrates over short durations (hours), and is similarly expressed as CUEC = \(\frac{U-R-{EX}}{U}\).

CUEE considers C retention as net microbial growth over longer time scales (days to months), taking into account the drivers of CUEP and CUEC as well as microbial biomass turnover. On these time scales, a significant proportion of microbial biomass is converted to necromass following microbial death (MD)30 such that CUEE = \(\frac{U-R-{EX}-{MD}}{U}\), encompassing all aspects of microbial C processing, including death and recycling processes.

Methods for measuring microbial CUE

Multiple approaches can be used to quantify CUE, such as isotopically labeling substrates31,32, stoichiometric modeling22,33 and others34. These methods rely on different assumptions and capture distinct microbial processes, which can explain the variability in CUE estimates across methods8,35,36 (Fig. 1a), including differences in the response of CUE to environmental changes37, and the relationship between CUE and SOC (Fig. 3a, b).

Fig. 3: Impact of different research methods on the SOC-CUE relationship and variability in incubation conditions across studies.
figure 3

Panels a and b illustrate the relationships between soil organic carbon (SOC) concentration and CUE based on a isotopic labeling methods (14C, 13C-labeled substrates, and 18O water) and b stoichiometric modeling. Panel c presents the incubation durations, while panel d shows the temperatures employed in studies using labeling and incubation methods. Data sources: a21, b19, and c, d20.

Full size image

The most common approach for measuring CUE is the tracking of isotopically labeled compounds (14C, 13C labeled substrate, or 18O water) introduced to the system. Carbon isotopes in microbial substrates enable the differentiation between C allocated to microbial biomass and that released through respiration. Although this labeling technique is widely used, its results can be influenced by the choice and combination of substrates31, as well as the incubation period14,38. A significant limitation of this approach is that measured CUE reflects only the efficiency of those microbes that use the introduced substrates, not the entire microbial community. Furthermore, the variation in incubation times and temperatures across different studies (Fig. 3c, d) presents a substantial obstacle to standardizing CUE measurements.

The method using 18O-labeled water is based on the incorporation of the 18O-atom into microbial DNA as a measure of growth as compared to catabolic C losses as CO232,39. This method has higher accuracy than the C labeling method as it is not substrate specific, does not perturb microbial metabolism like methods involving substrate addition, and exhibits comparatively less variability over time35. Nonetheless, this method faces limitations such as higher cost and demanding technical procedures. Concerns also arise regarding the method’s foundational assumptions, e.g., the presumption that water is the sole oxygen source for microbial DNA synthesis and the hypothesis that all microbial cells maintain a consistent DNA to biomass C ratio40. Furthermore, its applicability in dry soils is challenging41.

Stoichiometric modeling is a common method for indirectly estimating CUE, which is based on the assumption that microbes growing on plant detritus allocate C to produce enzymes and other necessary components to acquire nutrients in the appropriate elemental ratios at the whole-community scale29,33. This approach offers the advantage of requiring only a limited number of parameters, such as the activities of enzymes targeting C versus nitrogen (N) or phosphorus (P) acquisition and the C:N:P composition of the substrate and microbial biomass, which can be constrained by existing observations. However, it relies on highly simplified assumptions regarding elemental ratios and C allocation36. This approach inherently suggests lower CUE in soils with high SOC due to its focus on the metabolic costs of nutrient acquisition under conditions where nutrients are scarce relative to C. This outcome (Fig. 3b) starkly contrasts with the positive correlation between CUE and SOC observed using isotopic labeling techniques (Fig. 3a), which are commonly considered to provide a more realistic insight into the relationship between CUE and SOC. The isotope labeling method estimates microbial growth and CUE by tracking the incorporation of labeled atoms into biomass or DNA, reflecting intracellular biochemical transformations. In contrast, the stoichiometry model method estimates CUE by analyzing the activities of extracellular enzymes and the stoichiometric balance between organic matter and microbial biomass, focusing on extracellular metabolic processes42. Therefore, caution is advised when comparing results obtained from these two methods, even though they use the same term (CUE). We do not yet know the extent to which the stoichiometric and isotope methods are comparable. Until we understand which patterns can be accurately captured by the simpler stoichiometric method, we should rely on the more robust 18O method for measuring actual CUE and the 13C method for CUE associated with specific substrates.

In addition to the methods mentioned above, there are other less commonly used approaches, including the use of 18O in water vapor to minimize impact on soil moisture41, metabolic flux analysis17, and calorespirometry43. Each method offers unique advantages and faces specific limitations, grounded in their underlying assumptions and theoretical bases35,36,37. These limitations not only affect the accuracy of these methods but also introduce significant comparability issues. Consequently, there is an urgent need to improve current methodologies and integrate innovative techniques to more accurately assess soil microbial CUE.

Data gap

Given the methodological challenges in measuring CUE in situ, field assessments of microbial CUE are rare. The vast majority of existing CUE observations have been obtained from lab incubations. Yet, these CUE observations remain scarce at the global scale, a situation which is exacerbated by the lack of harmonization of observations from different measurement approaches. For some ecosystems, observations are few or even nonexistent, including ecosystems that play a critical role in the global C cycle, such as tropical rainforests, wetlands, and peatlands44,45.

Existing CUE measurements mostly come from studies of the litter and surface mineral soil16. Thus, our understanding of microbial CUE in subsurface soil remains limited, which is problematic as large amounts of C are stored in subsoils globally, and especially those of wetlands and peatlands. The few existing studies indicate that microbial CUE decreases with soil depth46,47 and that subsurface CUE may be less sensitive to warming31 but more sensitive to nutrient variations48.

Moreover, data on temporal variations in CUE are lacking. A commonly overlooked factor that may contribute significantly to CUE variability in soil ecosystems, regardless of methodology, is seasonality in CUE. Seasonal changes are associated with significant variations in substrate availability, temperature and moisture, all of which may have a substantial impact on the growth and respiration of soil microorganisms, thereby altering microbial CUE39. For example, CUE estimated using the 18O incorporation method ranged from 0.1 to 0.7 in soils from an agricultural field site and from 0.1 to 0.6 at a forest site within one year49. It has also been reported that soil microbial CUE exhibits significant fluctuations within a short period (daily) after rewetting50,51. This temporal dynamic in CUE values could contribute to the significant variability observed in CUE measurements.

Regulatory factors governing microbial CUE

The incorporation of soil microbial CUE dynamics into process-based models necessitates a comprehensive understanding of a range of regulatory factors influencing CUE (Fig. 4). CUE at a specific biological level is influenced by features of both the microbial community itself (biological controls) and its external environment (abiotic controls). These factors frequently interact, particularly at the community and ecosystem levels: abiotic controls can modify CUEC or CUEE by regulating biological controls, while biological controls may induce adaptation to abiotic factors, thereby influencing the impact of abiotic controls.

Fig. 4: Framework of biological and abiotic determinants of CUE in a carbon cycle context.
figure 4

The darker-colored area in the figure indicates biological controls; the lighter-colored area indicates abiotic effects. The arrows depict implicit relationships and the width of the arrows corresponds to the levels of scientific certainty: confident assertions are represented by thick lines, while less confident assertions are indicated by thinner lines. These confidence levels are based on the expertise of the authors.

Full size image

Biological controls

Microbial physiological state

Microbial CUE reflects the physiological state of microorganisms. Under natural conditions, only a small proportion (values vary from 1% to >20% in different studies52,53) of soil microbial cells are metabolically active, and soil respiration primarily originates from these metabolically active cells53. Nonetheless, a high fraction of microbial cells in the soil are in a potentially active state (10 to 60% of the total microbial biomass), meaning that they are ready to start using available substrates within a few hours after easily available substrate is added. The shifts in physiological states of these microbial cells, resulting from changes in temperature, moisture, or substrate availability, significantly impact CUE54. Consequently, CUEP or CUEC measurement methods relying on substrate addition may overestimate CUE14, and shifts in physiological state can lead to seasonal variations in CUE49.

Microbial community diversity and composition

Increased microbial diversity enriches the spectrum of metabolic functions within a community, potentially leading to greater microbial growth55 and CUEC by facilitating more efficient use of varied C sources10,56. The composition of microbial communities, notably the ratio of fungal to bacterial biomass (F:B), plays a critical role in determining CUEC57. Communities dominated by fungi can show higher CUEC, attributed to their higher biomass C to N) ratios (C:N) and their proficiency in decomposing complex organic materials58, or lower CUE due to the high costs associated with resource acquisition by decomposer fungi57. Therefore, this contrasting evidence from plant litter studies indicates that the relationship between F:B ratio and CUE is context-dependent57,59. Alternatively, an approach categorizing microorganisms into copiotrophs (r-strategists with low CUE) versus oligotrophs (K-strategists with high CUE) has been promising for estimating CUE60. For example, shifts from r-strategists to K-strategists explain increased CUEC along a successional gradient in the southeastern Tibetan Plateau61.

Changes in community composition may also enable microbial communities to alter their CUE in response to environmental changes or fluctuations62,63. For instance, long-term warming experiments indicate a decline in the temperature sensitivity of CUEC, suggesting that shifts in microbial composition can maintain CUEC despite changes in temperature and substrate quality31. Similarly, modeling studies suggest that changing microbial community composition can reduce the sensitivity of CUEC to substrate quality64 and soil moisture fluctuations65.

Biotic interactions

In the soil food web, biotic interactions such as mutualism, facilitation, competition, and predation can shape CUEC56. Interspecific microbial competition drives accelerated growth rates, accompanied by the release of secondary metabolites that can negatively affect CUEC66. Antagonistic interactions may trigger stress responses, further diminishing CUEC67. Conversely, facilitation enhances CUEC by broadening species-realized niches, alleviating environmental stress, and reducing extracellular enzyme production costs64. Biotic interactions at higher trophic levels, such as predation, can variably affect CUEC by altering microbial density and influencing the outcomes of interspecific competition68,69.

Abiotic controls

Temperature

Temperature significantly affects soil microbial CUE, with respiration often increasing more than growth in short-term incubations, resulting in a decrease in CUEP9,34,70. The impact on CUEC and CUEE is less clear63, likely due to varied responses among microbial taxa71,72 and interactive effects with other environmental factors38,39,46,73. Temperature shifts can lead to changes in community traits or select for taxa with distinct life strategies, known as trait modification and trait filtering, respectively74,75. However, limited research on how CUEP varies among different taxa in response to temperature impairs our ability to accurately predict changes in CUEC76,77,78.

The interplay between direct and indirect temperature effects on soil microbial CUEC and CUEE complicates our understanding of the impact of warming on CUE. Warming can intensify C-nutrient imbalances, potentially diminishing microbial CUE79, but it can also improve the efficiency of substrate utilization, thereby enhancing CUE32,72. Expected reductions in soil moisture due to increased evapotranspiration under warming conditions80 add another layer of complexity, with the combined impacts of temperature and moisture on microbial CUE remaining inadequately explored10,81. Some soil C models, including Millennial82 and MIMICS25 have begun to account for the temperature dependency of CUEC, indicating a growing recognition of the importance of including the dynamic response of microbial CUE to fluctuations in temperature.

Soil water availability

Increased soil moisture promotes microbial growth and CUE by improving substrate diffusivity and accessibility, and lowering investment in osmolyte synthesis, as long as conditions remain oxic8,10,83. Prolonged water stress reduces soil substrate accessibility and increases the need to synthesize osmolytes to survive during dry periods, leading to lower CUEC83, even though the taxa that remain active in dry conditions can maintain relatively high growth rates84. Furthermore, drought reduces plant C inputs to the soil83, thus potentially leaving microbes with fewer lower resources, resulting in lower CUE. The intricate interplay of drought-induced changes in microbial respiration and growth may leave CUE unchanged if the affected processes balance each other78. High levels of soil moisture may also reduce microbial CUE. As soil pores fill with water, air spaces and oxygen diffusivity decline, potentially leading to anaerobic conditions if saturation occurs. Under O2 limitation, soil microbes shift from aerobic to anaerobic respiration or fermentation, significantly reducing energy yield and leading to decreased microbial growth and CUE while having little impact on CO2 production rate due to upregulated biochemical rates83.

Microbial responses to rewetting of a dry soil also cause rapid changes in CUE, as shown in modeling studies50 and confirmed by empirical evidence51. Upon rewetting, respiration increases while growth lags behind, especially when the soil has been dry for a long period51. As a result, just after rewetting, CUE is low and then increases as growth recovers during the first days after rewetting. However, after this initial pulse of microbial activity, CUE peaks and decreases again as substrates released during rewetting are consumed51.

Nutrient availability

The availability of nutrients such as N and P significantly affects microbial growth and respiration according to the concept of stoichiometric homeostasis which assumes constrained biomass C:N:P ratios of microbial cells29,64. Consequently, CUE decreases with increasing substrate C-to-nutrient ratios and increases with nutrient amendment when organic substrates are nutrient-poor22,29. Several C cycle models, such as the one proposed by Manzoni et al.85 and its later implementation24, have integrated CUE dynamics as a function of stoichiometry. In contrast to the homeostasis concept, recent findings highlight the capability of microbes to store and use nutrients dynamically, contributing to a stable CUE across different environments by separating growth and respiration processes from immediate nutrient availability86. This resilience to nutrient stress suggests that future C modeling should incorporate microbial nutrient storage dynamics for enhanced predictive accuracy.

Soil pH

Soil pH influences microbial CUEC and CUEE by affecting the bacterial community composition and acting as a potential stressor87. It also impacts CUE by altering microbial community composition88, nutrient solubility83, and metal toxicity (e.g., aluminum87). Habitats with neutral pH generally have higher bacterial diversity and biomass compared to acidic or alkaline soils7. The response of community composition to a shift in soil pH from acidic to neutral corresponded with a significant increase in CUEC87,89. However, recent research indicates a complex interplay between soil pH, microbial community composition, and CUE dynamics, evidenced by both negative correlations90 and a U-shaped response curve, pinpointing a critical threshold at pH 6.491, although the calculations to document this are complex and may necessitate refinement.

Soil texture and structure

Microbial growth is intricately linked to substrate accessibility, which is influenced by soil environmental conditions like texture and soil structure. Approximately 40–70% of soil bacteria are associated with microaggregates and clay particles92. The structural complexity of the soil environment also plays a crucial role in shaping the community structure and function of soil microorganisms at the ecosystem level93. Heterogeneity of soil structure and composition creates diverse microhabitats that influence microbial interactions, diversity, distributions, and activity, as well as ecosystem processes like nutrient cycling and organic matter decomposition94. Still, limited information exists on the relationship between soil texture or structure and microbial CUE. A recent meta-analysis found a significant positive link between microbial CUEC or CUEE for glucose and soil clay content95, which was attributed to increased clay content enhancing substrate adsorption96, thereby limiting substrate availability to microbes97, and resulting in higher microbial CUEC or CUEE.

Substrate quality

Substrate quality, defined by the chemical characteristics of organic matter that influence its decomposability, such as the C:N ratio and molecular composition, significantly impacts soil microbial CUE98. A “high-quality” substrate typically has a lower C:N ratio, indicating a balanced N content relative to C, and a lower content of recalcitrant compounds, which generally leads to faster decomposition and higher CUE by providing C and nutrients that microbes require for growth and metabolism8. Compounds requiring multiple enzymatic steps for degradation can lead to reduced efficiency in building biomass. Polymeric substrates like lignin and cellulose need depolymerization before cellular uptake, whereas smaller substrates readily diffuse across membranes62. Takriti et al. (2018) found a positive association between soil CUEC and ratios of cellulase to phenol oxidase enzyme activity potential, which was considered to be indicative of soil organic matter (SOM) substrate quality46. Different substrates necessitate distinct metabolic pathways, resulting in different respiration rates per unit C assimilated8,99. Frey et al. (2013) observed lower microbial CUEC when soils were amended with oxalic acid or phenolic compounds compared to glucose, despite similar molecular sizes31.

Microbial CUE increases with the chemical energy per mole of C in the substrate, highlighting the importance of substrate chemistry for microbial CUE variability in soil8. This relationship is akin to the concept of energetic imbalance100, which parallels the idea of stoichiometric imbalance. The energy content of soil microbial biomass and substrate can be quantified by the degree of reduction (γ), which refers to the average number of electrons available per C atom for biochemical reactions, indicating the energy density of the substrate or biomass8. The degree of reduction of soil microbial biomass (γB) is typically around 4.2, while that of substrate (γS) usually varies between 1 (e.g., for oxalate) and 8 (methane)8. Most of the substrates used by soil microorganisms have a γS of 3 (e.g., various organic acids), 4 (e.g., glucose and other carbohydrates), and rarely 5 or higher (e.g., leucine, polyhydroxyalkanoates or lipids)8. When γS is lower than γB, the substrate’s energy content is insufficient to meet microbial demand, necessitating the oxidation of more substrate per unit of C assimilated, thereby reducing CUE101. These insights form the basis of the stoichiometric modeling for indirect CUE estimates.

SOC-CUE relationship

The relationship between CUE and SOC concentration at the ecosystem level can be positive, negative, or non-existent, depending on the interactions among multiple processes21,92,96,102,103,104. Higher CUE can lead to increased SOC through biosynthesis and accumulation of microbial by-products—facilitating SOC formation via the entombing effect16,102,105 — or conversely, trigger SOC decline through the priming effect by ramping up microbial biomass and enzyme activity9. While some studies suggest a negative correlation between CUE and SOC103,104,106, the majority of research supports a positive relationship21,74,107,108, indicating that higher CUE is often linked to increased SOC levels. In a recent study, Tao et al.21 employed observational data and data assimilation algorithms and found that, on a global scale, CUE is positively correlated with SOC concentration, arguing for CUE as the major determinant for SOC formation. However, subsequent arguments have raised methodological concerns which might have obscured the importance of microbial community dynamics27 and SOC stabilization processes109.

Indeed, the link between microbial CUE and SOC is contingent upon the stabilization of microbial necromass within soil aggregates or its association with minerals96,102,105. This stabilization process, pivotal for enhancing SOC, is significantly influenced by physico-chemical soil properties, which vary greatly and determine the potential for necromass protection110,111. Positive SOC-CUE relationships could be anticipated in soils with high physicochemical C stabilization potential and microbial communities that convert simple chemical substrates into necromass111. Conversely, when soil microbes face environmental stress, the relationship between CUE and SOC becomes less predictable. Particularly under conditions where nutrients are limited relative to carbon, the increased microbial respiration required to maintain stoichiometric balance leads to a decreased CUE29,33. Further reductions in CUE may be driven by environmental challenges such as low oxygen or pH88,106, as well as the physiological costs of microbial competition66. However, these stressors on microbial activity may differently affect SOC, potentially leading to either a negative or negligible correlation between CUE and SOC106. It’s worth noting that in organic-rich soils, such as peat, C stabilization relies more on the accumulation of undecomposed plant material than on necromass formation112, making the link between CUE and SOC less direct. Therefore, the CUE-SOC relationship in organic soils is expected to differ from mineral soils where C is mainly stabilized by mineral associations.

Additionally, it is important to recognize the distinct sensitivities of microbial CUE and SOC to environmental changes, as their responses are not synchronized. Microbial CUE can adjust rapidly, from days to months, in contrast to SOC, which may take years or even decades to respond to a measurable extent49,113. Data from two meta-analyses highlight this disparity, showing that although fertilization positively affects both CUEC and SOC37,114, the response ratios of CUEC were not significantly correlated with the response ratios of SOC, or even microbial biomass C content (Fig. 5a, c). Here, the “response ratio” is calculated as the ratio of the measured value in the treatment to the value in the control. Furthermore, the response ratios of microbial CUEC were not significantly related to treatment duration (within ten years of treatment) (Fig. 5b), whereas the response ratios of SOC increased significantly with experiment duration (Fig. 5d). Therefore, SOC gradually approaches a new equilibrium over several decades, whereas CUE achieves equilibrium almost immediately. This discrepancy underscores the importance of considering the state (SOC and microbial biomass) dynamics of an ecosystem when evaluating the interplay between microbial CUE and SOC dynamics.

Fig. 5: Contrasting responses of SOC and CUE to fertilization.
figure 5

Correlations between ln-transformed response ratios of microbial CUE and ln-transformed response ratios of (a) SOC and (c) microbial biomass C (MBC); and the correlation between experiment duration and ln-transformed response ratios of (b) CUE and (d) SOC. The response ratio is calculated as the ratio of the measured value in treatment to the value in the control. Data are from meta-analyses27,37,114. Both datasets include observations from all three methods of CUE measurement, i.e., C labeling, O labeling, and stoichiometry modeling as indicated by symbol colors in a–c.

Full size image

Using models and data across scales to clarify the microbial role in C cycling

Integrating genomic data with CUE and C models

With the rise of high throughput sequencing technology, the use of genomic datasets to help calibrate or validate C models has become both feasible and affordable. This capacity is especially valuable when predicting CUE115. As genomic data related to microbial traits becomes more readily available at both the population116 and community levels through metagenomics117, there is a growing need to effectively integrate this data into C cycle models. This integration requires models that can handle complex microbial interactions, from individual populations to entire communities (Fig. 2).

One way to integrate genomic data is by converting the genetic sequences of microbes into information on metabolic pathways (e.g., cellulose degradation, lignin degradation, nitrogen reduction, and fermentation) using genome-scale metabolic models (GEMs)118. GEMs take into account the microbe’s environment, such as substrate availability, and predict the transformation of metabolites within a cell based on its genomic information. This process allows for the calculation of CUE at the population level by analyzing substrate use and CO2 production118. For community-level CUE, GEMs can be combined into microbial community models that simulate interactions between different microbial taxa: The ‘computation of microbial ecosystems in time and space metabolic modeling platform’ (COMETS) extends GEMs to include dynamics of microbial growth and interactions, providing a tool for predicting CUEC under various environmental conditions115.

An alternative modeling approach at the community level is based on traits (e.g., quantity of cellulase produced, maximum rate of reaction (Vmax) of cellulose decay by cellulase, Vmax of cellulose-monomer uptake, and turnover rate), such as the DEMENT model, which uses data on microbial traits to simulate substrate use and CO2 production119. This model can predict both CUEP and CUEC under different environmental conditions and over time. However, translating genomic data into traits remains challenging120. Genomic datasets typically indicate the presence or absence of certain genes or pathways, but additional information, such as that from GEMs or experimental data, is necessary to accurately map these genes to functional traits in the models.

Validating genomic and trait-based models is crucial and can be achieved using community-level genomic datasets, which offer insights into microbial strategies that affect CUE, such as nutrient recycling and stress tolerance117,121. Combining these models with traditional CUE measurements and omics data allows for the creation of detailed maps of community-level CUE, offering new insights into C cycling dynamics and providing input information for C cycle models.

A major challenge in this field is the high computational demand of integrating omic data into complex models. One solution is the development of computational emulators that can simulate the dynamics of microbial models more efficiently, bridging the gap between detailed, small-scale models and broader applications in C cycle studies122. This approach promises to improve our understanding of microbial contributions to C cycling, leveraging the power of genomic data to inform and validate complex ESMs.

Harmonization of CUE measurements and aligning measured and modeled CUE

Harmonizing soil microbial CUE measurements across different methods, i.e., aligning results from different methodologies, poses a challenge due to the differences across measurement techniques. While adopting a universal protocol for CUE measurement—a single, standardized measurement method— would be ideal, it may not be feasible given the complexities of CUE. Therefore, a more practical approach involves providing a clear and comprehensive description of the methodologies used in different studies. This detailed reporting should include information on the physiological processes considered, such as maintenance, enzyme production, biomass generation, and mortality rates. This level of detail helps in understanding and comparing results across studies, as well as in selecting appropriate data for model calibration17.

In contemporary soil C models that explicitly incorporate microbial processes25,82, the CUE is close to empirically measured CUEC. To achieve a uniform approach to CUE measurement, microbial models that resolve key processes influencing CUE, such as uptake, respiration, exudation, and microbial death could be used17. Such models can generate CUE metrics that align with different measurement methodologies by incorporating a complete or partial set of these processes into their calculations. Furthermore, these models can be adapted to conduct numerical experiments with specific substrates or to incorporate isotopic tracers (e.g., 13C, 14C, 18O) to simulate outcomes from labeling experiments. This adaptability allows for the exploration of hypotheses regarding discrepancies in measurements under diverse conditions by modifying model boundary conditions. Additionally, microbial models serve as foundational tools for integrating microbial metabolism into broader global C models, potentially enhanced by machine learning emulators for improved scalability and applicability.

Constraining CUE using model-data fusion

Data assimilation encompasses a collection of techniques, including Bayesian inference, that refine biogeochemical models by integrating observational data. This process not only updates model parameters to reflect the most likely values based on available data but also quantifies their uncertainties, thus bridging the gap between empirical observations and theoretical models107. This approach is particularly valuable for parameters like microbial CUE, which are challenging to measure directly in the field due to technical limitations. An innovative application of data assimilation is demonstrated by Tao et al.21, who developed the PROcess-guided deep learning and DAta-driven (PRODA) approach123,124. This method integrates global-scale SOC data with a microbially explicit model to produce a global map of microbial CUE. PRODA employs traditional Bayesian data assimilation to estimate parameters at specific sites and then uses deep learning to extrapolate these site-specific parameter estimates to a global scale. The result is a set of parameters that optimally align with observed data, offering a detailed view of microbial CUE and SOC storage patterns worldwide, along with other soil C cycle dynamics such as decomposition rates, environmental impacts on soil respiration, and vertical C transport21.

Despite the potential of approaches like PRODA to harness large datasets for enhancing our understanding of the soil C cycle, their computational intensity—stemming from the extensive data sampling required by Bayesian inference—may limit their application in models with complex structures. The next wave of data assimilation techniques will likely integrate process-based models with deep learning algorithms more seamlessly121. Such advancements could offer quicker parameter optimization and facilitate comparisons across different models, paving the way for more accurate and comprehensive assessments of microbial CUE and C cycle dynamics on a global scale.

Long-term SOC records and ecosystem manipulation experiments

Ecosystem manipulation experiments and observations of natural gradients offer invaluable insights into how microbial communities and CUE adapt to global change factors. Especially insightful are field experiments (or studies leveraging natural gradients) that alter environmental factors such as soil temperature, precipitation patterns, or nutrient levels76,125 over long durations. These experiments provide critical data on the enduring effects of global change drivers on CUE, while simultaneously highlighting the limitations of current models and enhancing our comprehension of ecological processes. Integrating the results from these experiments with model simulations, supported by proven site modeling protocols and extra observational data, is crucial for steadily enhancing the accuracy and complexity of models126.

Incorporating radiocarbon (14C) data and long-term SOC records into models is also vital for refining CUE forecasts across longer (decadal to centennial) time scales. This temporal information is essential for capturing the dynamics of CUE over time, thereby improving the precision of models in depicting spatial and temporal fluctuations127.

Diagnosing CUE from existing models or simulation archives

In global C modeling, approaches to quantify the environmental impact on organic matter decomposition and stabilization differ significantly. An effective method for estimating microbial CUE at the ecosystem level as emerging from model simulations involves the calculation of the ratio between soil heterotrophic respiration (R) and gross decomposition (D) within these models. Gross decomposition refers to the sum of all C fluxes transferred between the modeled soil C pools that are mediated by microbial processes, excluding physically mediated transfers (e.g., sorption, aggregation, or leaching). This includes all C removed from organic matter pools, whether it is lost as CO2 or transferred to another pool (SI-Text 1). This ratio effectively quantifies microbial-mediated C losses from SOC pools, integrating both growth (anabolic processes) and respiration (catabolic processes). Under steady-state conditions, it is assumed that heterotrophic respiration aligns with microbial C uptake, resulting in the formula: CUE = 1 - R/D. The steady-state assumption implies that microbial communities and SOC stock are stable in time (i.e., in equilibrium with boundary conditions). This is an approximation of real systems where SOC varies due to anthropogenic and natural changes (e.g., Holocene climatic variations). This diagnosed CUE, emerging as a property inherent to the model, is not susceptible to the equifinality issues that can affect the underlying intrinsic model parameters (like CUEC), and it does not necessitate the incorporation of explicitly microbial models, offering a simplified yet insightful metric. These model-based CUE estimates, derived from long-term flux averages (e.g., 20 years), represent stable C stocks. In contrast, measurement-based estimates, taken over shorter periods, are more susceptible to significant CUE variations due to asynchronous fluctuations in components such as respiration and degradation, potentially introducing estimation inaccuracies. This timescale discrepancy likely accounts for the greater variability observed in measurement-based CUE compared to model-based CUE. We propose this “model-diagnosed CUE” as a novel metric, designed to estimate microbial CUE from model outputs without direct measurements of microbial uptake.

Analyzing diagnosed CUE and its relationship with SOC across various models, such as those evaluated in the Trends in the land carbon cycle (TRENDY) model intercomparison project2, facilitates the identification of differences attributable to unique model structures and assumptions. For example, warming-induced CO2 emissions should be higher in models with low diagnosed CUE compared to high CUE as the warming-induced stimulation of microbial activity will result in relatively more C being respired than cycled within the soil systems. This approach further allows the benchmarking and subsequent refinement of diagnosed CUE estimates using observed CUEE data.

For instance, we derived CUE estimates from simulations conducted with two different versions of the Organizing Carbon and Hydrology In Dynamic Ecosystems (ORCHIDEE) land surface model128, which differ in the SOC model deployed. The CENTURY SOC model (Fig. S1), which is widely used but does not resolve microbial processes, uses first-order decay, while the MIMICS model (Fig. S2) resolves microbial physiology, providing a more mechanistic understanding of microbial processes. The resulting global CUE maps (the average of simulation results over 20 consecutive years) revealed significant spatial variability (Fig. 6a, b). While the two maps showed a good correlation (Fig. 6c), the CUE values diagnosed from the MIMICS model were higher than those from the CENTURY model (Fig. 6d). These findings underscore the importance of incorporating observational data into model calibration efforts to enhance the accuracy and reliability of SOC predictions by realistically resolving CUE.

Fig. 6: Diagnosed CUE from two existing soil C models.
figure 6

CUE diagnosed from a nutrient-enabled version of the the Organizing Carbon and Hydrology In Dynamic Ecosystems land surface model (ORCHIDEE-CNP) deploying a soil module based on (a) the CENTURY model128, or (b) the MIMICS model with constant intrinsic CUEC132. c Correlation between diagnosed CUE values from the CENTURY-based model and the MIMICS-based model. d Distribution frequency of CUE for the two scenarios.

Full size image

In conclusion, the inherent structure of a model significantly shapes its outcomes, making the integration of empirical data with data-constrained models a fundamental step toward realistic predictions129,130. Precisely delineating the spatial and temporal dynamics of CUE in models that specifically address microbial activities is crucial for the reliability of their predictions of SOC status and dynamics. Moreover, future soil C models must navigate the intricate balance between the complex regulatory mechanisms of CUE, other processes governing SOC formation and stabilization, and the practicality of model use to promote more precise projections of CUE responses under diverse environmental scenarios. This Perspective underscores the importance of combining different data sources with sophisticated modeling techniques to refine global CUE predictions. By incorporating genomic data, standardizing measurement protocols, applying data assimilation practices and critically evaluating CUE within existing frameworks, our comprehension of the global dynamics of microbial CUE can be markedly improved. This Perspective provides a roadmap for establishing an effective modeling approach to accurately represent global soil microbial CUE and its interactions with other biological and abiotic processes that regulate SOC dynamics.