Soil moisture-atmosphere coupling accelerates global warming

Anthropogenic greenhouse gas (GHG) emissions since the Industrial Revolution have warmed Earth’s climate, but warming rates vary regionally. Prior studies show that land-atmosphere coupling—especially through soil moisture, snowpack, and vegetation—modulates energy and mass exchanges with the atmosphere, shaping regional climate responses and extremes. Soil moisture–atmosphere coupling (SA) has been linked to stronger and more frequent high-temperature extremes and heat waves. However, a coherent, global assessment of how SA affects overall warming rates and heat extremes, and how this influence evolves under different future emission scenarios, has remained unclear. This study addresses that gap by quantifying SA’s contribution to global and regional warming and extreme heat occurrence and intensity across modern, mid-century, and late-century periods under contrasting emissions pathways.

The paper builds on extensive literature showing that land-surface conditions modulate atmospheric processes and climate variability (e.g., Seneviratne et al. 2006; Koster et al. 2004; Betts et al. 1996). SA feedbacks have been implicated in amplifying heat extremes in Europe and globally, with precipitation deficits and soil moisture anomalies enhancing hot-day frequency and intensity (e.g., Mueller & Seneviratne 2012; Fischer et al. 2007; Vogel et al. 2017). Vegetation and CO2 physiological responses can further modulate extremes (Lemordant & Gentine 2019; Skinner et al. 2018). Prior work also highlights uncertainties in coupling strength in models and reanalyses. This study advances the literature by isolating SA’s contribution to warming trends and extreme heat under CMIP6 experimental designs and by decomposing radiative forcing terms that connect soil moisture changes to surface air temperature.

• Models and experiments: Six CMIP6 models (CESM2, CMCC-ESM2, EC-Earth3, IPSL-CM6A-LR, MIROC6, MPI-ESM1-2-LR) were used. Experiments include LS3MIP’s LFMIP-pdLC (soil moisture fixed to 1980–2014 climatology), ScenarioMIP SSP1-2.6 (low emissions) and SSP5-8.5 (high emissions), and the historical experiment. Extra-tropical summer is defined as JJA in the Northern Hemisphere and DJF in the Southern Hemisphere. Multi-model means were used to reduce inter-model uncertainty; latent heat flux is from five models (MIROC6 lacks this field).
• Isolation of SA effect: SA impacts were isolated as the difference between fully coupled runs (historical/SSP) and the fixed-soil-moisture LFMIP-pdLC run under the same external forcings (sea surface temperature, sea ice, CO2). Time horizons: modern (1995–2014), mid-term (2040–2059), long-term (2080–2099).
• Time series processing: Lanczos low-pass filtering (11 weights; cutoff 1/11) was applied to remove interannual variability for 1980–2099 in select analyses.
• Extreme high-temperature metrics: At each grid point, the 90th percentile of daily surface air temperature (tas) was computed. Thresholds were defined from LFMIP-pdLC for each time period. The SA-induced changes in (a) probability of exceeding the threshold and (b) intensity (exceedance magnitude in °C) were calculated as differences between coupled and fixed-soil experiments. Probability distribution functions of regional mean tas for Europe (40–60°N, 20–50°E) and central North America (28–55°N, 88–110°W) were derived from daily data per model (e.g., shown for MPI-ESM1-2-LR), avoiding multi-model averaging for PDFs.
• Decomposition of SA-driven tas changes: Using a surface energy balance framework, SA-induced changes in surface air temperature were decomposed into radiative forcing terms linked to surface albedo, evapotranspiration (latent heat), shortwave transmissivity (cloud effects), air emissivity (longwave), and aerodynamic resistance, plus a residual circulation term. Four models with required radiation diagnostics (CESM2, CMCC-ESM2, EC-Earth3, IPSL-CM6A-LR) were used for decomposition.
• Contribution to warming trend: Under SSP5-8.5, trends (2015–2099) were computed for total warming (fully coupled), SA-only (difference: coupled minus fixed-soil), and GHG-only (fixed-soil) over global land excluding Antarctica and specified regions (Northern mid-latitudes, Southern subtropics, Europe, North America).

• Amplification of warming: SA amplifies warming over most land areas in all scenarios. Under SSP5-8.5, SA-driven warming accelerates with time due to the ‘warmer climate – drier soil’ feedback, exceeding 0.5 °C over extratropical land by late century.
• Extreme heat probability: SA increases the likelihood of extreme high temperatures by about 10% globally (excluding Antarctica) and by more than 30% over large parts of North America and Europe under high emissions by century’s end. SA contributes less than 10% under modern conditions but approaches 20% in mid-high latitude NH and subtropical SH by late century.
• Extreme heat intensity: SA raises the intensity of extreme high temperatures by more than 1.5 °C globally and up to ~8.0 °C over North America and Europe under SSP5-8.5 by late century.
• Regional PDFs: Over North America and Europe, SA causes rightward shifts and flattening of tas PDFs, increasing the probability of extreme high temperatures (NA: +52.5%; EUR: +30.8%) under high emissions by late century; without SA, extreme heat probabilities would drop by roughly one-third (NA) and one-quarter (EUR).
• Physical mechanisms: GHG-driven soil drying reduces evapotranspiration, decreases cloud cover, increases surface-received shortwave radiation, and shifts surface energy partitioning from latent to sensible heat, producing non-linear warming—strongest in Northern mid-latitudes (EUR, NA) and Southern subtropics. Some arid regions (e.g., Sahara, Arabian Peninsula) show slight SA-related cooling where local evapotranspiration changes are negligible and non-local circulation effects dominate.
• Accelerating feedback: Under SSP5-8.5, progressively drying soils correlate more strongly with evapotranspiration reductions, enhancing sensitivity and accelerating SA-driven warming. Sensitivity increases are evident in surface and root-zone soil moisture.
• Radiative forcing decomposition: SA-induced warming is primarily driven by positive forcing from reduced evapotranspiration and increased shortwave transmissivity. Over EUR and NA in the long-term future (SSP5-8.5), combined positive radiation (sum of terms) reaches 33.5 ± 14.1 and 32.8 ± 16.7 W m⁻², increasing from 15.4 ± 9.9 and 22.1 ± 9.2 W m⁻² in the modern period—equivalent to increases of 117.5 ± 276.5% (EUR) and 48.4 ± 53.3% (NA). Under SSP1-2.6, combined radiation also rises but is less than half of that in SSP5-8.5 (23.1 ± 5.9 and 25.3 ± 12.6 W m⁻²).
• Mitigation effects: Under strong mitigation (SSP1-2.6), SA-induced warming and its influence on extreme heat weaken notably across regions and in all models examined.

The findings demonstrate that soil moisture–atmosphere coupling significantly amplifies and accelerates GHG-driven warming by drying soils, reducing evapotranspiration, enhancing surface shortwave input, and increasing sensible heat flux. This process intensifies both the frequency and severity of extreme high-temperature events, especially in North America and Europe, and grows stronger over time under high emissions. The decomposition analysis attributes most SA-induced warming to reduced evapotranspiration and increased shortwave transmissivity, explaining the strong regional signals. These results directly address the research question by quantifying SA’s contribution to global and regional warming and extremes and showing its temporal acceleration under SSP5-8.5. The study underscores the importance of early mitigation: strong emissions reductions can substantially weaken SA-driven warming and limit extreme heat escalation. It also highlights that both local SA feedbacks and potential non-local circulation effects shape regional outcomes, motivating integrated assessments of these mechanisms in future work.

This study establishes that soil moisture–atmosphere coupling is a robust amplifier of global warming and a key driver of increasing frequency and intensity of extreme heat, with an acceleration under high-emission scenarios due to the feedback between warming and soil drying. By isolating SA’s contribution using CMIP6 experiments and decomposing its radiative drivers, the work clarifies mechanisms and quantifies regional impacts. Under high emissions, SA-driven warming surpasses 0.5 °C over extratropical land by late century, and extreme heat risks expand markedly in North America and Europe. Strong mitigation (SSP1-2.6) meaningfully weakens these effects. Future research should jointly assess local and non-local SA influences on temperature and circulation, improve parameterizations of land-atmosphere coupling, and explore ecosystem-based strategies to maintain soil moisture and reduce SA amplification of warming.

• Model and parameterization constraints: Only six LS3MIP CMIP6 models were available; some lack key diagnostics (e.g., MIROC6 latent heat flux), and uncertainties in SA parameterizations can affect projections.
• Scenario uncertainty: Future radiative forcing pathways are uncertain, leading to scenario-dependent outcomes.
• Regional variability and processes: Non-local circulation effects and monsoon-related precipitation uncertainties (e.g., Indian summer monsoon) can produce diverse regional responses and projection spread.
• Diagnostics availability: Radiative decomposition used four models due to missing radiation inputs in two models, potentially limiting generality of term-by-term attributions.
• Fixed-soil experiment idealization: LFMIP-pdLC fixes soil moisture climatology, which is an idealized setup and may not capture all feedback complexities.