Widespread shift from ecosystem energy to water limitation with climate change

The study addresses whether and how climate change is shifting terrestrial ecosystems from energy limitation toward water limitation, a key question for food and water security, CO₂ uptake, and evaporative cooling. Ecosystem function depends on soil moisture and energy availability, but the net effect of climate change and rising atmospheric CO₂ on moisture limitation is uncertain. Water and carbon cycles are coupled via vegetation: photosynthesis assimilates CO₂ while stomata regulate transpiration, linking surface energy balance and evaporative cooling. Regional ecosystem responses hinge on whether conditions are energy- or water-limited. Despite observed vegetation greening largely driven by CO₂ fertilization, signals of changing water availability and their impacts on transpiration and surface energy fluxes remain debated. The authors propose to jointly assess energy and water limitations using a functional ecosystem metric to reconcile disparate trends and better understand future shifts in ecosystem regimes.

Prior work shows mixed evidence on trends in soil moisture, evapotranspiration, and aridity from observations, reanalyses, and models, leading to inconsistent conclusions about drying/wetting and dryland expansion. Some studies focus on water supply (soil moisture), others on atmospheric water demand and meteorological indices (precipitation with potential evaporation), which can yield different outcomes. In contrast, increases in energy availability and associated warming are robust, influencing photosynthetic enzymes (e.g., RuBisCO), vapor pressure deficit, stomatal conductance, and atmospheric evaporative demand. Vegetation greening and CO₂ fertilization have altered LAI and water use efficiency, confounding interpretation of water limitation trends. Large-scale climate variability (e.g., ENSO) modulates interannual ecosystem water limitation. These mixed findings highlight the need to simultaneously consider energy and water constraints on ecosystem function.

The authors develop and apply an ecosystem limitation index (ELI) to distinguish water- versus energy-limited regimes using CMIP6 simulations. ELI is defined as the difference between Kendall rank correlations of monthly anomalies: ELI = cor(SM', ET') − cor(SW', ET'), where SM is soil moisture, ET is terrestrial evaporation, and SW is incoming shortwave radiation; primes denote detrended, de-seasonalized monthly anomalies. Positive ELI indicates water-limited conditions; negative ELI indicates energy-limited. As an alternative energy proxy, air temperature anomalies replace SW to test robustness. They also compute ELI using plant transpiration anomalies (ELL) instead of total ET, and partial correlations cor(SM', ET'|SW') and cor(SW', ET'|SM') to reduce confounding. Data: Monthly outputs from 11 CMIP6 models at 2×2° resolution, using historical (1980–2015) and SSP5-8.5 (2015–2100) simulations. The full 1980–2100 period is split into 12 decades. Within each decade and grid cell, linear trends are removed and seasonal cycles computed; anomalies are months with T > 10°C (“warm months”). Grid cells with fewer than 30 data points per decade are excluded. The “warm land area” comprises grid cells with complete ELI time series from 1980–2100 in at least four models. Trend estimation: Theil–Sen slopes quantify trends in ELI and related variables; significance is assessed with Mann–Kendall (Kendall’s tau). Spatial maps show multimodel means and significance stippling. Attribution: Decadal ELI variability (12 data points) is modeled via multivariate linear regression using all combinations of candidate predictors (incoming SW, SM, ET, LAI, aridity index, and variants), selected by AIC with multimodel inference. Models within ΔAIC < 4 and adjusted R² > 0.5 are retained. The most important predictor per grid cell is identified by relative explained variance (relaimpo; averaged across models weighted by Akaike weights). Sensitivity tests include alternative R² thresholds and adding land-use proxies (crop and tree fractions). Additional diagnostics include LAI evolution, ET partitioning (T/ET), and comparison to reference ET datasets. Robustness to using air temperature as the energy variable is evaluated.

- Global ELI increases steadily from 1980 to 2100 across CMIP6 models, indicating a shift from energy limitation toward water limitation. This trend persists when using air temperature as the energy proxy and when expressed versus global warming since 1980; all models show increasing ELI beyond 1.5°C warming. - Spatially, ELI increases over about 73% of the warm land area; positive trends are significant (P < 0.05) over 36% versus 5% for negative trends. Strongest increases occur in North America (NAM), South America (SAM), Central Europe (CEU), Northern Eurasia (NEU), and East Asia (EAS), with pronounced signals over forested regions. - The fraction of warm land that is water-limited expands from 70% (1980) to 77% (2100), an increase of ~7 million km². Transitional zones migrate poleward, especially in the Northern Hemisphere (NAM, NEU). - Seasonally, the duration of water-limited conditions lengthens by up to 6 months in 43% of the warm land area, with only ~3% showing decreases. Water-limited periods extend into months previously energy-limited or cold, and the peak intensity of water limitation increases. - Drivers: Increasing incoming shortwave radiation reduces energy limitation globally, while regional decreases in soil moisture exacerbate water limitation (notably in parts of South America). Global soil moisture trends are modest and uncertain across models; ET increases until ~2030 then diverges across models. LAI shows consistent increases with greater multimodel spread, raising the T/ET fraction and vegetation influence on surface fluxes. - Attribution: Incoming shortwave radiation is the single most important predictor of ELI trends, dominating about 20% of the warm land area globally and 27–57% within regions of interest where models perform well. However, multiple variables (SW, SM, ET, LAI, aridity index) are required to explain global patterns. - Alternative formulations: Using transpiration anomalies (ELL) yields similar but slightly weaker trends; using partial correlations slightly weakens but does not alter the increasing ELI trend. Normalized trends show ELI changing more notably than its individual components, highlighting combined energy–water effects and ecosystem feedbacks.

By jointly analyzing energy and water constraints on ecosystem function, the study reconciles conflicting signals from individual hydrometeorological variables and indices. The consistent rise in incoming shortwave radiation and warming alleviate energy limitation broadly, while regional soil moisture decreases and increased atmospheric demand shift ecosystems toward water limitation. The ELI captures functional ecosystem responses (via ET anomalies) to hydrological and radiative forcing, revealing widespread regime shifts in both space and time beyond what single-variable trends imply. Regional analyses show coherent increases in water limitation across forested mid- to high-latitude regions and parts of subtropical regions, with tropical contrasts (e.g., SAM vs. Central Africa) linked to differing soil moisture/precipitation trends. Lengthening and intensification of water-limited seasons suggest growing exposure to drought stress, with implications for CO₂ exchanges, evaporative cooling, and land–atmosphere feedbacks. Dominance of shortwave radiation in attribution underscores the central role of energy trends, but multiple variables contribute regionally, reflecting local climate, vegetation, and soil characteristics.

The study introduces and applies the ecosystem limitation index (ELI) to demonstrate a widespread, robust shift from energy-limited to water-limited ecosystem function over 1980–2100 across CMIP6 models. This shift is driven by reduced energy limitation due to increasing incoming shortwave radiation and, regionally, by soil moisture declines, with LAI increases and changing ET partitioning enhancing vegetation’s role. The water-limited fraction of warm land expands (70% to 77%), and water-limited seasons lengthen and intensify. Incoming shortwave radiation emerges as the most important predictor of ELI trends, though a suite of variables is needed to explain global patterns. These findings imply heightened risks to food and water security, land degradation, biodiversity loss, weakened terrestrial CO₂ sequestration, and more frequent/intense extremes. Future work should reduce model uncertainties by improving representations of CO₂ fertilization, water use efficiency, dynamic vegetation, and rooting depth; incorporate nutrient limitations; and expand observational constraints to better validate ecosystem regime shifts.

Results rely on CMIP6 model outputs with inherent uncertainties in process representations (e.g., CO₂ fertilization effects on LAI and water use efficiency, dynamic vegetation), potential oversensitivity to CO₂, and missing processes such as adaptive deep rooting. Models have difficulty capturing evaporative regime changes, contributing to multimodel spread in ELI means and trends, especially in key regions. Nutrient limitations (N, P) are not included due to sparse observations and uncertainty, yet may increasingly constrain transpiration. While these factors likely affect the magnitude of ELI trends, the sign and spatial patterns are considered robust across models. Observational validation is limited, though comparisons to ET datasets and a conceptual soil moisture model support ELI’s ability to reflect water-limited conditions.