Rising global temperatures reduce soil microbial diversity over the long term

Significance

Global warming is posing a significant threat to soil ecosystems, including soil microbial diversity. This diversity is essential for sustaining ecosystem functions such as sequestration of soil organic carbon (SOC), a key component in mitigating climate change. Our global meta-analysis shows that warming significantly reduces bacterial and fungal diversity, with effects particularly strong under extended warming durations and in nutrient-poor soils. This decline in microbial diversity compromises soil ecosystem functioning, including the capacity to sequester SOC, potentially creating a feedback loop that further accelerates warming. These findings highlight the interconnected fates of microbial communities and SOC, underscoring the urgent need to integrate microbial biodiversity conservation into climate mitigation and ecosystem management strategies.

Abstract

Soil microbial diversity is crucial to sustaining ecosystem productivity and improving carbon sequestration. Global temperature continues to rise, but how climate warming affects microbial diversity and its capacity to sequester soil organic carbon (SOC) remains uncertain. Here, by conducting a global meta-analysis with 251 paired observations from 102 studies, we showed that, on average, warming reduced bacterial and fungal diversity (measured by richness and Shannon index) by 16.0 and 19.7%, respectively, and SOC by 18.1%. The negative responses of both soil bacterial and fungal diversity to warming became more pronounced with increasing warming magnitude, experimental duration, and decreasing soil nitrogen availability. Under the worst-case climate warming scenario (2010 to 2070, 3.4 increase in °C), soil bacterial diversity and fungal diversity are projected to reduce by 56% and 81%, respectively, over 60 y. Importantly, in addition to the direct impact of warming on SOC, warming-induced declines in microbial diversity also contributed to SOC losses. We highlight that prolonged warming could substantially reduce soil microbial diversity and decrease SOC sequestration, accelerating future warming and underscoring the urgent need for decisive actions to mitigate global climate change.

It is well known that the functioning and sustainability of ecosystems depend on their biodiversity (1). As the most diverse component of terrestrial organisms, soil microbes and their diversity play a key role in maintaining multiple ecosystem functions, including primary production, nutrient cycling, and soil carbon (C) storage (2, 3). The Earth is warming up (4), with far-reaching impacts on the structure and functioning of various biotic communities in ecosystems (5–7). However, it remains uncertain how climate warming affects soil microbial alpha diversity (hereafter “microbial diversity”), as previous empirical studies have reported positive (8, 9), negligible (10, 11), or negative (6, 12) effects of warming temperature on soil microbial diversity.

Warming is expected to have profound influences on soil microbial diversity because microbial species have widely differing temperature-dependent metabolic rates (13). On the one hand, rising temperatures may allow additional soil microbial taxa to thrive that were previously limited by cooler conditions (14). On the other hand, rising temperatures could lead to the local extinction of microbial species that cannot tolerate higher temperatures or are outcompeted by those better adapted to the warmer temperatures (6, 15). In addition to its direct thermal effect, warmer environments can enhance microbial diversity by increasing plant diversity and productivity (16), which provide more resources and niches for a greater number of microbial species to coexist (17). However, it is more likely that warming-associated environmental changes, such as reduced soil moisture and pH, create harsh conditions that selectively filter out species that are adapted to more humid or neutral to alkaline environments (6, 18, 19). This environmental filtering would favor drought- and acid-tolerant microbes while leading to the decline or loss of those unable to cope with these stressors, resulting in less diverse microbial communities (6, 18). In addition, warming may deplete soil C and nutrients, intensifying resource limitation (e.g., organic substrates) and further amplifying competition among the remaining taxa (20). Theoretically, bacterial and fungal communities might respond differently to warming due to differences in thermal sensitivity and energy requirements. Warming often leads to greater declines in fungal than bacterial diversity, as fungi are more thermally sensitive and metabolically constrained, whereas bacteria maintain higher diversity by shifting toward thermophilic taxa and adapting more readily to increased energy demands (21, 22).

Divergent empirical findings on the effects of experimental warming on soil microbial diversity could result from differences in warming magnitudes, durations, and background environmental conditions. Greater warming magnitude and prolonged duration can deplete key resources (7, 23), intensifying competition among microbial species, and thereby reducing their diversity. Moreover, in resource-poor environments [e.g., those low in soil C, nitrogen (N), and water], warming intensifies competition among microbial species for these critical resources due to accelerating microbial metabolism and leads to reduced diversity (20, 24–26), whereas in resource-rich environments, these negative filtering effects of warming on microbial diversity may be less pronounced (27). Meanwhile, experimental warming effects on soil microorganisms are more pronounced at colder temperatures but less evident in warmer conditions as microbial communities in warmer regions have already adapted to higher temperatures, making them less sensitive to additional warming (28, 29); however, this buffer has limits, as further warming may push already warm soils beyond the thermal maxima of many taxa, triggering sharp declines in diversity (30). Additionally, experimental warming effects on microbial diversity may differ with ecosystems (croplands, forests, and grasslands) due to dissimilarities in vegetation type and management practices (31, 32).

Warming and soil microbial diversity are both recognized as key drivers of soil organic C (SOC) cycling (3, 7, 33, 34), but little is known about how these factors jointly influence SOC dynamics. It is well established that warming could substantially reduce SOC directly by increasing microbial activity and respiration (7, 34); however, changes in soil microbial diversity under warming could further influence SOC dynamics, either exacerbating or mitigating C losses (11, 35). A reduction in microbial diversity could lead to decreased soil microbial biomass (36) and activity (37, 38), subsequently reducing the production of microbial necromass and residues (39, 40). Given that these microbial byproducts are crucial components of stabilized SOC pools (41), it is likely that a decline in microbial diversity would result in diminished SOC storage. Beyond the inputs from microbial necromass, more diverse microbial communities are expected to use resources more efficiently through complementary interactions between species, such as increasing microbial C use efficiency (3), the ratio of C retained in biomass for growth relative to the total C respired. This means that diverse microbial communities could store more C in their biomass, resulting in greater SOC accumulation during microbial turnover (42), rather than C loss through respiration. Understanding the links between soil microbial diversity and SOC under warming is crucial for predicting the fate of soil C in the Anthropocene.

Two recent global meta-analyses, comprising 36 and 27 studies, respectively, reported that experimental warming significantly decreased (43) or had no significant effect (18) on bacterial and fungal diversity, as measured by microbial richness and the Shannon index. Another meta-analysis of 29 studies reported that experimental warming significantly increased bacterial and fungal richness (29). These contrasting results likely reflect the limited number of studies included in previous analyses, which may not have captured the full range of experimental evidence across diverse ecosystems and may have been disproportionately influenced by a few extreme effect sizes. While these studies have made valuable contributions by synthesizing early experimental evidence and highlighting the potential effects of warming on belowground biodiversity, they did not examine how warming effects depend on key ecological moderators, especially background resource availability (e.g., soil N and water). Moreover, none assessed whether warming-induced changes in microbial diversity are linked to ecosystem functions such as soil C sequestration (SI Appendix, Table S1). Without this understanding, it remains difficult to predict where and when warming will most strongly alter microbial communities and how these changes may, in turn, influence ecosystem functioning.

To fill these knowledge gaps, we conducted a global meta-analysis of 251 paired observations from 102 studies that investigated the responses of microbial diversity (bacterial diversity and fungal diversity) and SOC to above- or below-ground warming across terrestrial ecosystems (SI Appendix, Fig. S1 and Table S2). We hypothesized that i) warming would decrease bacterial and fungal alpha diversity; ii) the negative effects of warming on microbial diversity would be more pronounced with higher magnitudes and longer durations of warming and in sites with low C, N (as indicated by high soil C:N ratios) and water availability; iii) warming-induced reduction in microbial diversity could indirectly decrease SOC sequestration through reduced soil microbial biomass and increasing respiration per unit of biomass (i.e., decreased microbial C use efficiency). The effects of warming were quantified using the log response ratio (lnRR) of warmed versus ambient values for microbial diversity and SOC (Materials and Methods). Given that both microbial richness (i.e., OTU, Chao1, and ACE) and Shannon index were commonly used to represent microbial diversity and show similar responses to warming treatment (SI Appendix, Fig. S2), we incorporated them simultaneously as measures of microbial diversity in our analysis, treating diversity index as a random effect. We used the microbial metabolic quotient (MMQ) as an inverse metric of soil microbial C use efficiency (44, 45).

Results and Discussion

Overall Warming Effects on Microbial Diversity.

Across all experiments, experimental warming of an average of 2.3 °C reduced bacterial diversity by 16.0% (95% CI, −18.9 to −13.1%; n = 143) and fungal diversity by 19.7% (CI, −23.7 to −15.6%; n = 108) (Fig. 1A). Consistent with our first hypothesis, we demonstrated the negative impacts of warming on soil microbial diversity, extending those negative responses of soil microbial diversity to warming derived from natural grasslands (6, 12) to a diverse range of ecosystem types. Our results are consistent with one recent comprehensive global meta-analysis, which included 36 studies (43). However, they contrast with two previous global meta-analyses based on 27 and 29 studies (18, 29), respectively (SI Appendix, Table S1). Our analysis, which includes 102 studies—about three times the number of studies used in previous analyses—offers a more robust and comprehensive examination of the effects of warming on soil microbial diversity. Collectively, our findings suggest that climate warming reduces microbial alpha diversity, potentially through mechanisms such as thermal filtering and competitive exclusion, as suggested by previous studies showing selection for thermophilic taxa under elevated temperatures (6, 15). However, our study did not directly assess taxonomic shifts or community composition changes, and future research integrating taxonomic and functional profiling is needed to better understand how warming reshapes microbial community structure and function.

Fig. 1.

Effects of warming on soil microbial diversity in relation to soil pH and moisture. (A) Logarithmic response ratios (lnRR) of bacterial diversity, fungal diversity, soil pH, and soil moisture. (B and C) Bivariate relationships between lnRR of microbial diversity with those of soil pH (n = 72 bacterial and 46 fungal diversity) and soil moisture (n = 72 bacterial and 52 fungal diversity). In panel A, values are bootstrapped mean and 95% CI. The number of observations is shown beside each attribute without parentheses with the number of studies in parentheses. In panels B and C, blue and red lines represent the bacterial- and fungal-specific biodiversity responses, respectively, with their 95% CI shaded. The significance (p) is calculated based on one-side F-test.

Furthermore, studies that simultaneously measured microbial diversity and at least one soil factor (soil pH or soil moisture) showed that experimental warming increased soil pH (n = 72) (Fig. 1A), making the soil less acidic, in contrast to our original expectation that warming would lower soil pH (6, 18). This pH increase likely reflects accelerated decomposition of soil organic matter under warming, which thins organic coatings on minerals and reduces sorption capacity, allowing organic acids to be leached away and driving a net rise in pH (46). Responses of both bacterial and fungal diversity to experimental warming were negatively associated with those of soil pH (Fig. 1B), indicating that warming-induced increases in soil pH could contribute to the decline in microbial diversity for both groups. In these soils, the warming-induced shift toward greater soil alkalinity (as indicated by a rise in soil pH above 7 in SI Appendix, Fig. S3) could reduce the availability of key nutrients such as phosphorus and iron, leading to a decline in microbial diversity (19). In addition, warming decreased soil moisture (n = 73) (Fig. 1A), and the responses of both bacterial and fungal diversity to experimental warming were positively related to those of soil moisture (Fig. 1C), suggesting that warming-induced soil drying can reduce microbial diversity by limiting water and other essential resources, as lower soil moisture hinders substrate diffusion (6, 47).

Key Determinants of Warming Effects on Soil Microbial Diversity.

As the magnitude (range: 0.3 °C to 6.8 °C) of experimental warming increases and its duration (range: 1 y to 12 y) extends (SI Appendix, Table S3), the negative responses of bacterial and fungal diversity to experimental warming become more pronounced (Fig. 2 A and B and SI Appendix, Table S4), suggesting that prolonged global warming could lead to significant and widespread losses in microbial diversity in the future. These increasingly negative impacts on soil microbial diversity with longer warming durations parallel the more pronounced declines observed in soil moisture (SI Appendix, Fig. S4D) and SOC dynamics (7) as well as the shift from positive to negative effects of warming on soil respiration (48) and plant productivity (49), highlighting the critical link between soil microbial diversity and ecosystem C cycling in a warming world. Importantly, the effects of warming on microbial diversity were highly dependent on background soil C:N ratios (Fig. 2C and SI Appendix, Table S5), changing from neutral to negative with increasing soil C:N ratios. This change is likely driven by the fact that excessive warming strengthens N limitation in N-poor soils (as indicated by high soil C:N ratios), increasing competition for N in the soil and ultimately reducing microbial diversity (20).

Fig. 2.

Effects of warming on soil microbial diversity in relation to warming magnitude, warming duration, and background soil C:N ratios. (A) In relation to warming magnitude; (B) in relation to warming duration; (C) in relation to soil C:N ratios. Slope estimates are partially dependent, derived from the most parsimonious model. Blue and red lines represent the bacterial- and fungal-specific biodiversity responses, respectively, with their 95% CI shaded. The significance (p) is calculated based on one-side F-test.

The effects of warming on soil bacterial and fungal diversity remained consistent regardless of background SOC content, mean annual temperature (MAT), and mean annual aridity index (AI) (SI Appendix, Fig. S5). In addition, despite significant differences in warming magnitude and warming duration between ecosystem types, the effects of warming did not differ with ecosystem type (croplands, forests, and grasslands) for bacterial diversity and fungal diversity (SI Appendix, Fig. S6 and Table S5). Similarly, the lnRR of richness and Shannon index showed consistent relationships with warming magnitude and duration (SI Appendix, Table S6), as well as with environmental factors (SI Appendix, Table S7). Our dataset includes 102 experiments conducted across Asia, Europe, North America, South America, and Africa, spanning both temperate and tropical regions (SI Appendix, Fig. S1), with the majority of studies originating from China and the United States. These globally distributed experiments consistently demonstrated negative effects of climate warming on soil microbial diversity. However, the effect of warming on soil fungal diversity marginally differed among warming approaches (i.e., heating cable, infrared heater, and open-top chamber), while bacterial diversity was not differentially affected by different warming approaches (SI Appendix, Fig. S7). Specifically, heating cables, which apply warming belowground, had a less pronounced negative effect on soil fungal diversity compared to aboveground warming (i.e., infrared heaters and open-top chambers). This less pronounced effect may stem from lower evapotranspiration and therefore reduced moisture loss in belowground warming, which results in less water competition than that experienced under aboveground warming (50). Given that the majority of our studies used aboveground heating (only 22 of 102 involved belowground warming; SI Appendix, Fig. S7), and lack simultaneous warming above- and belowground to better simulate climate change, our findings may be somewhat underestimated. We, therefore, emphasize the need for future studies to incorporate both above- and below-ground warming to more accurately capture the multifaceted impacts of global warming on soil ecosystems (34).

Global Mapping of Soil Microbial Diversity under Climate Warming.

By combining our collected soil microbial diversity responses with spatially explicit estimates of soil C:N ratios and soil surface temperature change, we reveal the global patterns in the vulnerability of soil bacterial and fungal diversity to climate warming. Under the worst-case climate warming scenario [Shared Socioeconomic Pathways 5-8.5 (SSP5-8.5), 2010-2070, 3.4 increase in °C, SI Appendix, Fig. S8], bacterial diversity is projected to decline by 56% (Fig. 3A), while fungal diversity is expected to decrease by 81% over 60 y (Fig. 3B). Regionally, our maps reveal that the most substantial reductions in microbial diversity under global warming are expected in Russia, Canada, and Australia (Fig. 3). These areas are characterized by higher soil C:N, particularly in high-latitude regions and certain Australian sites (51), which likely exacerbate the overall negative effects of warming on microbial diversity. The greater loss of soil microbial diversity in high-latitude regions mirrors the impacts on soil C stocks (7), suggesting that global warming has consistent effects on soil microbial diversity and C storage worldwide. These findings emphasize the need to prioritize these regions for climate change mitigation strategies to tackle the challenges posed by warming on ecosystem functioning and services. It is worth noting that extended duration (60 y) introduces uncertainties in estimating long-term effects, as soil microbial diversity could acclimate to warming (33), potentially compensating for additional impacts over time. Moreover, it is important to note that none of the studies included in our meta-analysis were conducted in the regions projected to experience the greatest diversity losses. As such, projections for these areas should be interpreted with particular caution due to the lack of direct empirical data.

Fig. 3.

Effects of the worst-case scenario warming on soil microbial diversity upscaled from 102 warming experiments. Changes in bacterial diversity (A) and fungal diversity (B) under the SSP5-8.5 scenario from 2010 to 2070. Global estimates for the warming effects on soil microbial diversity were upscaled using the relationships from their key drivers, i.e., warming magnitude, warming during, and background soil C:N ratios, synthesized through this meta-analysis. Values in the legend reflect the average warming-induced changes in soil microbial diversity (%) within each 1 km pixel.

Linking Microbial Diversity and SOC under Warming.

In studies that simultaneously reported changes in microbial diversity and at least one C variable [SOC, soil microbial biomass C (SMBC), or MMQ], experimental warming led to a significant decrease of 18.1% in SOC (CI, −21.3 to −14.9%; n = 97) and 15.9% in SMBC (CI, −21.3 to −10.4%; n = 66), and stimulated MMQ by 83.0% (CI, 66.5 to 99.4%; n = 34) (Fig. 4A). Moreover, the negative effects of experimental warming on SOC became progressively more pronounced with increasing warming magnitudes and experimental durations (SI Appendix, Fig. S9 A and B and Table S4).

Fig. 4.

The influence of soil microbial diversity on SOC sequestration under warming. (A) lnRRs of SOC, SMBC, and MMQ to warming treatment. (B–D) Bivariate relationships between lnRR of microbial diversity with those of SOC, SMBC, and MMQ. (E and F), structural equation model depicting influences of microbial diversity on SOC through SMBC (n = 91, P = 0.792) and MMQ (n = 47, P = 0.536). In panel A, the values are bootstrapped mean and 95% CI. The number of observations is shown beside each attribute without parentheses, along with the number of studies in parentheses. In panels B–D, the blue and red lines represent the bacterial- and fungal-specific biodiversity, respectively, with their 95% CI shaded. In panels E and F, accumulated warming was calculated by warming magnitude multiplied by warming duration. Solid and dashed lines represent positive and negative effects, respectively. The numbers adjacent to the arrows are the standardized coefficient (r) of the relationship. R²_marginal and R²_conditional represent the level of variance explained by all paths from the fixed effects, and the sum of fixed and random effects, respectively.

Simple bivariate plots revealed that the responses of bacterial diversity and fungal diversity to warming were positively correlated with those of SOC (Fig. 4B) and SMBC (Fig. 4C), while negatively associated with those of MMQ to warming (Fig. 4D). Our results from the structural equation model (SEM) revealed that increased warming magnitude and duration of warming (accumulated warming, calculated as warming magnitude multiplied by duration) was associated with lower SOC, both directly and indirectly. The indirect effects arose from the negative impact of warming on microbial diversity, which, in turn, reduced SOC via the positive relationships between microbial diversity, SMBC, and C use efficiency (as characterized by decreasing MMQ) (Fig. 4 E and F).

Our analyses highlight that warming would decrease SOC by reducing microbial diversity, improving the understanding of how warming affects biodiversity–ecosystem functioning relationships (52). Furthermore, our SEM demonstrated that decreased microbial diversity could drive SOC losses by decreasing SMBC and microbial C use efficiency. The positive impact of microbial diversity on SMBC may result from improved resource acquisition and use efficiency of soil microbes, promoting the biomass of microbial communities (2, 53). As a consequence, faster microbial growth and larger living microbial biomass in more diverse microbial communities could contribute to increased microbial necromass (36) and SOC accumulation (41). In addition to increased SOC inputs, larger living microbial biomass could also increase SOC outputs via microbial respiration (54); however, more diverse microbial communities are typically more efficient in C use, investing more C in biomass production and less C in respiration, thereby enhancing microbial C use and promoting SOC accumulation (3, 55, 56). Although the indirect negative effects of warming on SOC through microbial diversity (r = −0.03 and −0.11 for pathways via microbial biomass and C use efficiency, respectively) were smaller than the direct effects (r = −0.23 and −0.21 for two pathways, respectively) (Fig. 4 E and F), our results suggest that warming-induced declines in microbial diversity could play a significant role in driving SOC losses under warming conditions. However, the alternative structural equation showed that warming could also indirectly decrease soil microbial diversity via its direct negative effect on SOC (SI Appendix, Fig. S10). Therefore, we urge further studies to thoroughly investigate the relationship between SOC and soil microbial diversity, particularly in the context of climate change.

In summary, our study provides robust evidence of the ubiquitous negative effects of warming on soil microbial diversity across biomes and continents. In addition, we also identify the source of variation that contributes to the conflicting warming effects on soil microbial diversity from single-site studies. The different warming magnitudes, warming duration, and background soil N availability collectively shape the effects of warming on soil microbial diversity. Furthermore, the decrease in soil microbial diversity, to some extent, explains soil C losses under warming and supports the opinion that biodiversity in terrestrial ecosystems plays an important role in sustaining ecosystem functioning (2). Given that soil C is crucial for mitigating climate change and supporting ecosystem productivity (57, 58), we suggest that developing microbial diversity conservation strategies by slowing global warming, in addition to previously recommended strategies including reducing applications of chemical fertilization and pesticides and adding microbial probiotic or organic amendments into soils, could be effective solutions to maintain or even increase soil C (59). Our findings are of notable importance in demonstrating the long-term warming effect on soil microbial diversity and SOC globally and in guiding the growing efforts to use soil microbial diversity conservation for C sequestration. To avoid the most damaging effects on biodiversity and ecosystem functioning, controlling the warming rate is crucial for the future.

Materials and Methods

Data Collection.

We collected data on the effects of warming on microbial diversity by searching for peer-reviewed publications from Web of Science (Core Collection; http://www.webofknowledge.com), China National Knowledge Infrastructure (https://www.cnki.net), and three previous meta-analyses (18, 29, 43) with the search term: “warming or elevated temperature or rising temperature AND bacteria or fungi or microbial community AND soil.” We employed the following criteria to select relevant studies: i) they were purposely designed to test the responses of soil microbial diversity to warming, ii) they had at least one metric of soil microbial diversity, and iii) the magnitude and duration of warming treatments were reported. When a study included different warming magnitude, warming duration, or ecosystem types, they were considered as distinct observations. When multiple publications included the same data, we recorded it only once, treating each site in the selected publications as a distinct study. Finally, 251 paired observations from 102 studies were selected following PRISMA guidelines for meta-analysis (SI Appendix, Fig. S11). While we incorporated all studies used in previous meta-analyses, all data extraction and analyses were conducted independently in the present study. Each paired observation represents a comparison between warming and control treatments within the same site and study, ensuring consistency in soil type, vegetation, and background environmental conditions. Studies were classified as field experiments if warming was applied directly to natural or seminatural ecosystems in situ and as laboratory experiments if warming treatments were applied under controlled indoor conditions (e.g., soil incubation or greenhouse studies). In our dataset, 96% of experiments were conducted under field conditions (SI Appendix, Table S2). To ensure that experimental setting did not bias our results, we conducted a separate analysis using only field warming experiments. Field warming experiments showed very similar results to the full dataset (SI Appendix, Fig. S12).

For each study, we extracted the means, replication numbers, and SD of soil bacterial and fungal diversity in control and treatment groups, if reported. Microbial Shannon index and richness are the most commonly used metrics of soil microbial alpha diversity (60) and were therefore used as indicators of soil microbial diversity in this study. The OTU, Chao1, and ACE indices were used to quantify soil microbial richness in the original studies. In addition to soil microbial diversity metrics, SOC, SMBC, microbial respiration, soil pH, and soil moisture were also extracted from each study. MMQ was calculated as microbial respiration per unit of SMBC and was used to characterize microbial C use efficiency, since MMQ serves as a useful inverse metric of microbial C use efficiency (44, 45). In the original studies, SMBC was determined using the fumigation–extraction method. Our dataset included microbial respiration data derived from both field experiments (92% of all data) and laboratory measurements. For field measurements where only total soil respiration was reported, microbial respiration was estimated by multiplying soil respiration by 0.63 (61). The warming effects on MMQ were consistent across field and laboratory datasets (SI Appendix, Fig. S13A), and results based solely on field data closely matched those obtained from the full dataset (SI Appendix, Fig. S13 B–D).

We also extracted latitude, longitude, ecosystem types (cropland, forest, and grassland), MAT, warming magnitude, and warming duration from the original publications. The AI for each site was retrieved from the Consultative Group on International Agricultural Research–Consortium for Spatial Information Global AI dataset (62). Background SOC content, soil C:N ratios were used as proxies for background C and N availability (63). Given that all soil data focused on topsoil (0 to 20 cm) and sampled during the growing season (May to October), we did not consider the potential effects of soil depth and sampling season on microbial diversity.

Linear Mixed Modeling.

The lnRR was employed to quantify the effect size of warming treatments on selected variables, which was expressed as $lnRR=\mathit{ln}\left(\raisebox{1ex}{$x_w$}\!\left/ \!\raisebox{-1ex}{$x_c$}\right.\right)$, where $x_w$ and $x_c$ are the means of each observation in the warming treatment and control group, respectively. Although inverse variance was also widely used as weighting of lnRR in meta-analyses (64), sampling variances were not reported in 51 of the 102 studies in our dataset. Therefore, we used replication numbers (65) to weight the lnRR as follows: $\raisebox{1ex}{$\left(N_w\times N_C\right)$}\!\left/ \!\raisebox{-1ex}{$\left(N_w+N_C\right)$}\right.$, where $N_w$ and $N_c$ are the replications of each observation in warming treatment and control group, respectively. While alpha diversity metrics can be influenced by sampling effort and rarefaction approaches, our meta-analysis addresses this issue by calculating lnRR within each study, ensuring that warming and control treatments are compared under identical sampling conditions. By capturing relative changes within studies, lnRR reduces the impact of methodological variation on effect size estimates and enables valid comparisons across studies (66).

The soil bacterial and fungal diversity were considered as response variables and analyzed separately. We used the mixed-effect model to evaluate the overall warming effects on soil bacterial and fungal diversity and to examine the influence of warming magnitude (M, °C) and warming duration (D, years):

$$\mathit{ln}RR={\beta }_0+{\beta }_1·M+{\beta }_2·D+{\pi }_{\mathit{study}}+{\pi }_{\mathit{index}}+\epsilon ,$$ [1]

where β_i and ε are coefficients and sampling error, respectively; π_study was random effects accounting for the autocorrelation among observations within each “Study” while π_index was random effects accounting for the potential influence of variation in used soil microbial diversity indices. The analysis was conducted with the restricted maximum likelihood estimation with the lme4 package (67). All continuous variables were scaled, with ${\beta }_0$ representing the overall mean lnRR at the average values of M and D (68). We statistically compared the linear and logarithmic functions of M and D with study as the random effect, using Akaike information criterion (AIC). The analysis showed that the linear M and D resulted in lower, or similar AIC values (SI Appendix, Table S8). We then tested an alternative model that included the interaction between M and D as fixed effects. However, the model without interaction terms proved to be better, as indicated by lower AIC values (SI Appendix, Table S9). Thus, we used (Eq. 1) to model our subsequent analyses. As most of our models violated the assumption of normality based on Shapiro–Wilk’s test on model residuals, bootstrapped the fitted coefficients by 1,000 iterations was used (69). Further, there is no publication bias in our meta-analysis models (SI Appendix, Table S10) based on Egger’s test using metafor package (66). To test whether warming effects on microbial diversity changed geographically, we added the environmental variables (E, i.e., background SOC content, soil C:N ratios, ecosystems, MAT or AI) or warming method (WM) to Eq. 1:

$$\mathit{ln}RR={\beta }_0+{\beta }_1·M+{\beta }_2·D+{\beta }_3·E\left(or WM\right)+{\pi }_{\mathit{study}}+{\pi }_{\mathit{index}}+\epsilon .$$ [2]

Model Simulation.

To illustrate the global response of microbial diversity to warming, we derived the predictions based on the most parsimonious model (i.e., lnRR ~ magnitude + duration + soil C:N ratios). To project the worst-case changes in bacterial and fungal diversity under future scenarios, we used the SSP5-8.5 (70), as used in the Coupled Model Intercomparison Project 6. The SSP is considered as new version of emission trajectories or Representative Concentration Pathways (71), with SSP5-8.5 representing unmitigated emissions (70). We calculated the ensemble mean temperature of 5 climate models downloaded from CHELSA Version 2.1 at 1 km resolution (72) for the 2041 to 2070 period to account for the uncertainty among those model projections. The warming magnitude was determined using the ensemble means of SSP5-8.5 for 2041 to 2070 subtracted from the current mean temperature (°C) for 1981 to 2010 (72), respectively, representing a warming duration of 60 y (2010 to 2070). Assuming that the warming accumulatively increased across time, we divided the calculated warming magnitude by 2 to represent the warming magnitude between current and future scenarios (i.e., SSP5-8.5 for the 2041 to 2070 period). We obtained soil C:N ratios raster data at 1 km spatial resolution from SoilGrids (73). To match the spatial resolution (i.e., 1 km) of warming magnitude with that of the soil C:N ratio, we resampled soil C:N ratios to 1 km by calculating the mean of every 4,250 m pixel. While this dataset offers global coverage, its coarse spatial resolution and reliance on predictive modeling can lead to inaccuracies in heterogeneous landscapes (74). Therefore, while our global C:N–based projections highlight broad patterns, they should be interpreted with caution, especially for finer-scale or highly variable environments.

SEM.

To mechanistically understand the effects of experimental warming on soil microbial diversity and SOC, we first examined the lnRRs of SOC, SMBC, and MMQ in response to M and D with study as the random effect:

$$\mathit{ln}RR={\beta }_0+{\beta }_1·M+{\beta }_2·D+{\pi }_{\mathit{study}}+\epsilon .$$ [3]

Then, we examined the bivariate relationships between the responses of soil microbial diversity and the responses of SOC, SMBC, and MMQ to experimental warming. Subsequently, we employed the SEM to examine the effect of experimental warming magnitude and duration on the lnRRs of SOC, both directly and indirectly, via the lnRR of soil microbial diversity, SMBC, and MMQ based on the hypothesized causal pathways (SI Appendix, Fig. S14). We pooled bacterial diversity and fungal diversity together as microbial diversity since they had similar responses to experimental warming. As the combined warming magnitude and warming duration together indicate the warming effect for a given area, we employed accumulated warming in this analysis, which was calculated as the warming magnitude multiplied by warming duration (7). The SEM was performed using the R package piecewiseSEM with Study as the random factor (75).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Original data and R scripts can be found in ref. 76.