Environmental drivers of increased ecosystem respiration in a warming tundra

Tundra ecosystems store vast amounts of carbon and are warming faster than the global average. The balance between carbon uptake via photosynthesis and carbon release via ecosystem respiration (ER) will determine the fate of tundra carbon stocks. Previous studies show variable ER responses to experimental warming across sites (for example, increases of 8–52%), and tundra sites have been under-represented in meta-analyses. The persistence of ER responses over multi-decadal timescales and the environmental mechanisms underlying spatial and temporal variability remain uncertain. Indirect warming effects (for example, soil moisture, nutrient availability, permafrost thaw, biotic community changes) and context-dependencies (climate, soil properties, vegetation and microbial communities) could drive variability in ER responses. This study aims to quantify the magnitude and variability of warming impacts on growing-season ER across the tundra and to disentangle the roles of warming duration, indirect warming-induced changes in environmental conditions and context-specific environmental drivers.

Past meta-analyses and multisite studies have largely focused on the magnitude and climatological correlates of ER responses to warming, often reporting modest average increases and substantial heterogeneity. Reported tundra ER increases range widely (8–52%) across studies, and some analyses suggest short-lived increases that wane within a decade, while others indicate sustained responses. Indirect mechanisms involving soil moisture, nutrient cycles (notably nitrogen), vegetation structure and microbial composition have been hypothesized to mediate ER responses, but mechanistic evidence at biome scale has been limited. Moreover, tundra has been underrepresented in global syntheses of respiration responses, and ecological mechanisms linked to soil biogeochemistry (for example, total nitrogen, C:N ratio, pH) and rhizosphere priming have been proposed but not rigorously tested across broad tundra gradients.

Study design: The authors compiled 24,035 daytime growing-season ER measurements across 136 datasets (unique experiment × year combinations) from 56 open-top chamber (OTC) warming experiments at 28 arctic and alpine tundra sites on four continents, spanning 0–25 years of warming at the time of ER measurement. OTCs passively increased air temperature; experiments varied in OTC dimensions, deployment (year-round or seasonal), and ER measurement systems. Data inclusion required paired ER data from warmed and control plots during the growing season. ER data were standardized to g CO₂ m⁻² d⁻¹; outliers were removed (±3 s.d.). Growing-season mean ER per treatment per dataset was computed. Effect sizes: Primary effect size was Hedges’ g standardized mean difference (SMD) between warmed and control plots; a secondary effect size was the log ratio of means (ROM), later expressed as percent change. Meta-analysis: A multivariate meta-analysis (metafor rma.mv) estimated the pooled warming effect on ER, weighting datasets by inverse sampling variance and accounting for nesting (experiment/dataset) and repeated measures via an autoregressive (CAR) component. The same framework tested warming effects on microclimate, soil, vegetation and microbial variables. Temporal patterns: Warming duration (years since OTC installation) was analyzed using metaregression across four age classes (0–5, 5–10, 10–15, ≥15 years) and within classes to detect nonlinear temporal responses and persistence. Indirect warming effects: Single-factor metaregressions related ER SMDs to warming-induced changes (SMDs) in environmental variables measured at plot level, including microclimate (air, soil temperature, soil moisture), soil properties (SOM, TC, TN, C:N, bulk density, pH; mineral and/or organic layers; organic layer depth), vegetation (functional group cover, aboveground biomass, community height) and microbial metrics (bacterial/fungal biomass, fungal:bacterial ratio). Context-dependencies: Single-factor metaregressions related ER SMDs to ambient environmental context (controls): climate zone, permafrost probability, control-plot microclimate, soil properties (including mineral-layer TN and C:N), soil pH class, soil moisture class, soil carbon stock, organic layer depth, vegetation class and NPP. Model validation and bias checks included funnel/profile plots, residual diagnostics, and multiple sensitivity analyses to assess methodological biases, microclimate interactions, sampling design imbalance and robustness to data-matching decisions. Spatial upscaling: A two-factor metaregression using ROM as response and mineral-layer TN and C:N as predictors (ROM = 0.05 − 0.16 × TN + 0.01 × C:N; n = 39) was combined with gridded SoilGrids (ISRIC) mineral-layer TN and SOC data (derived C:N) at 1 km resolution to predict percent change in ER for a 1.4 °C warming across the arctic and circumarctic alpine region. Monte Carlo simulations (n = 100 per grid cell) propagated uncertainties from soil inputs and model parameters, and baseline ER fields (soil + plant respiration) were used to estimate absolute ER changes.

- Experimental OTC warming increased growing-season ER by a mean Hedges’ SMD of 0.57 (95% CI 0.44–0.70; n = 136), corresponding to a 30% increase (ROM 0.26; 95% CI 22–38%). - Warming treatment effects: mean air temperature increased by 1.4 °C (95% CI 0.9–2.0; n = 77), soil temperature by 0.4 °C (95% CI 0.2–0.7; n = 118), and soil moisture decreased by 1.6% (95% CI 0.8–2.4; n = 111). - Partitioned respiration (subset n = 9 experiments): both autotrophic (SMD 0.44 [0.08–0.80]) and heterotrophic respiration (SMD 0.92 [0.36–1.48]) increased significantly with warming; ROM-based percent increases were approximately 57% (autotrophic) and 55% (heterotrophic). - Persistence and temporal pattern: The mean positive ER response persisted across ≥25 years. A nonlinear temporal pattern was detected: the magnitude of the positive ER response decreased during years 5–9 (QM = 63, P < 0.001, n = 28) but increased during years 10–14 (QM = 5.6, P < 0.05, n = 15); no significant overall trend across all 25 years. - Indirect warming effects (mineral layer): Larger warming-induced increases in total nitrogen (TN) (QM = 5.4, P < 0.05, n = 42) and smaller reductions in pH (QM = 4.2, P < 0.05, n = 27) were associated with stronger ER increases. - Context-dependencies: Sites with lower ambient mineral-layer TN (nutrient-poor) or higher C:N ratios showed larger ER increases with warming (TN: QM = 6.3, P < 0.05, n = 43; C:N: QM = 4.7, P < 0.05, n = 39). Climate zone, permafrost probability, vegetation class and NPP were not significant predictors of response magnitude. - Spatial upscaling (for 1.4 °C warming): predicted ER increases of about 32% (s.d. 85%) for the arctic tundra (from 1.2 to 1.5 Pg C yr⁻¹; +0.37 Pg C yr⁻¹, s.d. 0.99) and 25% (s.d. 40%) across arctic + alpine tundra combined (from 3.4 to 4.3 Pg C yr⁻¹; +0.86 Pg C yr⁻¹, s.d. 1.36). Western/eastern Siberia and parts of the Canadian Arctic Archipelago were identified as particularly sensitive regions. Uncertainty was dominated by input soil data (~82% of total).

The synthesis demonstrates that tundra ER increases substantially and persistently with experimental warming, driven by both plant-related and microbial respiration. Variation in response magnitude is best explained by soil biogeochemistry of the mineral layer, specifically nitrogen availability (TN) and pH, and by context-dependent nutrient status (TN, C:N). These findings support mechanisms whereby warming stimulates nitrogen cycling and rhizosphere priming, particularly in N-limited systems, enhancing both decomposition and root respiration. Contrary to expectations, broad climatic context (for example, macroclimate, permafrost probability) and vegetation structural classes did not explain variability once soil biogeochemistry was considered, underscoring the mediating role of local soil conditions over macroclimatic or vegetation-type differences. The nonlinear temporal pattern (dip around 5–9 years followed by renewed strengthening at 10–14 years) suggests overlapping fast (kinetic) and slower (biogeochemical/community) processes that evolve with time under warming. The persistent positive ER response implies a potential shift of tundra carbon balance toward higher CO₂ emissions unless matched by compensatory increases in plant CO₂ uptake, highlighting the importance of integrating mineral-soil nitrogen processes and plant–soil linkages into Earth system models.

This study provides a biome-wide, mechanistic synthesis showing that experimental warming increases tundra growing-season ecosystem respiration by about 30% on average and that this enhancement persists over at least 25 years. The magnitude of ER increases is strongly governed by local mineral-layer soil conditions—especially total nitrogen, C:N ratio and pH—and by warming-induced changes in these properties, implicating coupled plant–microbe processes and rhizosphere priming. Spatial upscaling indicates substantial potential increases in ER across arctic and alpine tundra under modest warming, with sensitive regions in Siberia and the Canadian Arctic Archipelago. These insights can inform and benchmark land surface and carbon–climate models by emphasizing soil biogeochemical controls at depth. Future work should expand ER partitioning across sites and years, include non-growing-season fluxes, reduce spatial sampling gaps, and improve gridded soil datasets to lower uncertainty in upscaled predictions.

- Partitioning data for autotrophic and heterotrophic respiration were available for only 9 datasets, limiting precision on component contributions. - Environmental driver data coverage was uneven across datasets, necessitating single-factor metaregressions and reducing power for multifactor inference. - Funnel plot asymmetry suggests potential small-study or heterogeneity effects; however, inclusion of unpublished data reduces classical publication bias concerns. - Spatial upscaling uncertainty is dominated by gridded soil input data quality and representativeness. - Analyses focus on growing-season ER; lack of year-round (especially winter) ER may underestimate annual responses. - Gross primary productivity data were not included; plant responses were approximated via NPP/biomass proxies.