Experimental warming accelerates positive soil priming in a temperate grassland ecosystem

Abstract

Unravelling biosphere feedback mechanisms is crucial for predicting the impacts of global warming. Soil priming, an effect of fresh plant-derived carbon (C) on native soil organic carbon (SOC) decomposition, is a key feedback mechanism that could release large amounts of soil C into the atmosphere. However, the impacts of climate warming on soil priming remain elusive. Here, we show that experimental warming accelerates soil priming by 12.7% in a temperate grassland. Warming alters bacterial communities, with 38% of unique active phylotypes detected under warming. The functional genes essential for soil C decomposition are also stimulated, which could be linked to priming effects. We incorporate lab-derived information into an ecosystem model showing that model parameter uncertainty can be reduced by 32-37%. Model simulations from 2010 to 2016 indicate an increase in soil C decomposition under warming, with a 9.1% rise in priming-induced CO₂ emissions. If our findings can be generalized to other ecosystems over an extended period of time, soil priming could play an important role in terrestrial C cycle feedbacks and climate change.

Fig. 1. Study design and objectives.

The schematic of the study design illustrates soil sampling from warmed and control plots, followed by comprehensive analyses of soil and microbial properties and mechanisms, and subsequent field model optimization based on lab-derived data/model. The research objectives of this study are to ascertain: (i) the effects of experimental warming on soil priming; (ii) the microbial mechanisms underlying soil priming; and (iii) the potential for incorporating soil priming and associated microbial mechanisms into ecosystem models to enhance model performance and reduce uncertainty. To address these objectives, eight surface soil samples (0–15 cm depth) were collected in 2016 from warmed (targeted continuous heating at +3 °C above ambient temperature) and control plots (n = 4) at a long-term (7-year) experimental warming site in a tallgrass prairie ecosystem in the US Great Plains of Central Oklahoma (34°59’N, 97°31’W). Following geochemical measurements and Dissolved Organic Matter (DOM) analysis, the samples were incubated for one week with ¹³C-labeled wild oat (Avena fatua) straw powder to simulate plant litter decomposition, with additional treatments involving ¹²C-labeled straw and no straw serving as isotopic control and background, respectively. Active degraders in both warming and control samples were identified using qSIP analysis to further explore the microbial mechanisms underlying soil priming. Subsequently, the lab incubation datasets were integrated into a lab-scale Microbial-ENzyme Decomposition (MEND) model to simulate the 7-day incubation period. This lab-MEND model informed the prior parameter range for a separate field-scale MEND (field-MEND) model, which assimilated field warming experiments conducted from 2010 to 2016 to simulate soil C decomposition. Concurrently, the field-MEND model was compared with the Terrestrial ECOsystem (TECO) model to validate the effectiveness of incorporating microbial data into the MEND model for improving performance and reducing uncertainty. Soil, plant-straw, and bacterial symbols, as used in our previous study are adopted here.

Fig. 2. Warming enhanced the priming effect and restructured active bacterial communities.

a Response ratios of relative abundances of DOM between warming and control samples in 2016 (Warming vs. Control). Red symbols indicate significantly positive response ratios, while blue symbols indicate significantly negative response ratios. Grey symbols represent non-significant response ratios. Each symbol represents the average ± 95% CI of four biological replicates (n = 4) of warmed or control samples. Significance is denoted as follows: *p ≤ 0.05 and **p ≤ 0.01, as determined by using the one-sided Response Ratio test. No adjustments were made for multiple comparisons, and exact p-values are provided in the Source Data file. b The overall microbial respiration or priming effect during the 7-day incubation with ¹³C-labeled straw. The bars represent the average ± standard error of four biological replicates (n = 4) of warmed (red) or control (blue) samples. Significance is denoted as follows: **p ≤ 0.01 and ***p ≤ 0.001 determined by using one-sided permutation ANOVA. Exact p-values are provided in the Source Data file. c Abundance of active and total bacterial community after the 7-day incubation with plant litter. The bars represent the average ± standard error of four biological replicates (n = 4) of warmed (red) or control (blue) samples. Significance is denoted as follows: ***p ≤ 0.001, determined by using one-sided permutation ANOVA. Exact p-values are provided in the Source Data file. d The maximum-likelihood phylogenetic tree of active bacterial ASVs (decomposers) across all samples. The phyla colors are defined as follows: Firmicutes (taupe brown), Gammarproteobacteria (lavender purple), Betaproteobacteria (pastel pink), Alphaproteobacteria (light orange), Actinobacteria (dusty pink), Bacteroidetes (eggplant purple), Unclassified (light khaki), Thaumarchaeota (lime green). W: warmed samples; C: control samples; rrn: 16S rRNA gene. e PCoA analysis based on Bray-Curtis dissimilarity metric showing that taxonomic composition of active bacterial communities are different between warmed (red) and control (blue) samples. f Yearly means of relative abundance of active bacterial ASVs in in situ warmed samples during 2010–2016. The least-squares mean values were determined by the linear mixed-effects model. Each bar represents the mean ± standard error of 28 biological replicates (n = 28) of in situ warmed (red) or control (blue) samples over yearly repeated measures during 2010–2016. Significance is denoted as follows: ^#p ≤ 0.1; *p ≤ 0.05; **p ≤ 0.01; and ***p ≤ 0.001, determined by using two-sided ANOVA. No adjustments were made for multiple comparisons, and exact p-values are provided in the Source Data file. Source data are provided as a Source Data file.

Fig. 3. Warming stimulates the C-decomposing capacity and activity of active microbial communities.

a PCoA analysis showing that the functional gene composition of activebacterial communities is significantly different between warmed (red) and control (blue) samples. b The congruence between taxonomic (triangle) and functional gene compositions (circle) of active community assessed by a Procrustes analysis optimized through a PCoA plot (red: warming samples; blue: control samples). c The congruence between DOM (triangle) and functional gene composition (circle) assessed by a Procrustes analysis optimized through a PCoA plot (red: warming samples; blue: control samples). For (b, c), The overall fit of the Procrustes transformation is reported as the M²_1,2 value. Significance is assessed using a two-sided PROcrustean Randomization Test (PROTEST), wtih 999 permutations. d Response ratios of GeoChip signal intensities of C-decomposing genes between the warming and control samples. Red symbols represent significantly positive response ratios, while blue symbols represent significantly negative response ratios. Grey symbols represent non-significant response ratios. Each symbol represents the average ± 95% CI of four biological replicates (n = 4) of warmed or control samples. Significance is denoted as follows: *p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001 as determined by using the one-sided Response Ratio test. No adjustments were made for multiple comparisons, and exact p-values are provided in the Source Data file. Source data are provided as a Source Data file.

Fig. 4. Potential mechanisms of how warming enhances the priming effect.

a PLS model showing the relationships among soil temperature, soil properties, active bacterial community and priming effect. The active bacterial community composition (β-diversity) is represented by the PC1 to PC7 from the PCoA analysis based on Bray-Curtis dissimilarity metric. Directions for all arrows are from independent variable(s) to a dependent variable in the forward selected PLS models (p < 0.05 for both R²_Y and Q²_Y); only the most relevant variables (variable influence on projection > 1) are presented. Each number without parenthesis near the pathway is the PLS partial R² (Eq. (2)) and the significance is based on permutational test (1000 times) of PLS R²_Y. Each number in the parenthesis is the coefficient of determination (R²) between the two connected variables and the significance is based on Pearson correlation test or Mantel test (for β-diversity). The arrow width is proportional to the strength of the relationship determined by the PLS partial R². Significance is indicated by *0.01 <p ≤ 0.05; **0.001 <p ≤ 0.01; and ***p ≤ 0.001. b Linear regression between Mineral N (NH₄⁺ + NO₃^-) from the field and the primed C determined in laboratory (red: warming samples; blue: control samples). c Linear regression between plant biomass from the field and the primed C determined in laboratory (red: warming samples; blue: control samples). For (b and c), p values are calculated using a one-sided permutational test, constrained by treatment and block factors. Source data are provided as a Source Data file.

Fig. 5. Microbial- ENzyme Decomposition (MEND) model calibration and simulation.

a–c Comparison of lab-MEND simulated and observed cumulative microbial respiration in the 7-day qSIP incubation experiment: (a) CO₂ respired from the soil without straw addition, (b) CO₂ respired from the soil with straw addition; (c) ¹³CO₂ respired from the soil with straw addition; For (a–c), the error bar represents mean ± standard deviation (SD) (red: warming samples; blue: control samples). Observed CO₂: mean of 4 biological replicates (n = 4); Simulated CO₂: mean of 24 hourly simulation values (n = 24); MARE: mean absolute relative error (see Methods for details). d The field-MEND and TECO model parameter uncertainty assessed by the Coefficient of Variation (CV). The bars show the mean CV values of the 13 parameters for field-MEND and 10 parameters for TECO. The bars with different colors represent five model experiments and the above table columns of the same color list the information used in calibration process for each model. The + label indicates usage of the information when doing calibration. Blue: TECO model; Yellow: field MEND calibrated with R_h and microbial genes; Grey: field MEND calibrated with R_h, microbial genes and active fractions; Red: field MEND calibrated with R_h, microbial genes and Lab-MEND derived parameters; Light blue: field MEND calibrated with R_h, microbial genes, active fractions and Lab-MEND derived parameters. Lab-MEND derived parameters indicates replacing the parameter prior ranges with the parameter uncertainty ranges derived from lab-MEND calibration before calibration of field-MEND, which included five parameters (i.e., p_EP, V_g, α, K_D, and Y_g). Significance is indicated by *0.01 <p ≤ 0.05; **0.001 <p ≤ 0.01; and ***p ≤ 0.001 with two-sided Wilcoxon tests. e, Comparison of the field-MEND simulated and observed field heterotrophic respiration (R_h) under control (blue) and warming (red). R²: coefficient of determination. No adjustments were made for multiple comparisons, and exact p-values are provided in the Source Data file. Source data are provided as a Source Data file.