Vapor pressure deficit, or VPD, is a deceptively simple quantity with outsized consequences. It combines air temperature and relative humidity into a single measure of atmospheric demand for water. As the climate warms, the atmosphere’s capacity to hold water vapor rises faster than actual humidity in many regions, so VPD climbs. That climbing deficit tugs harder on every leaf and every patch of bare soil, pulling moisture out through stomata and soil pores. Previous research had established VPD as a key source of uncertainty in projections of future evapotranspiration, but most studies had zoomed in on the physiology of individual leaves or forest canopies. What remained missing was a quantitative, planet-wide picture of how VPD and evapotranspiration interact across deserts, grasslands, croplands, and rainforests alike—a gap the new study set out to close.

To build that picture, the team assembled an unusually rich data stack covering 1981 to 2020. Evapotranspiration and its two major components—transpiration from plants and bare-soil evaporation—came from the Global Land Evaporation Amsterdam Model, version 4.2a, a widely used product built on the Penman–Monteith framework. Vapor pressure deficit was calculated from two-meter air temperature and dew point temperature in the ERA5-Land reanalysis, ensuring consistency with the forcing used in the evaporation model. Land-cover classifications came from NASA’s MODIS MCD12C1 product, leaf area index from the GIMMS dataset, and soil moisture from GLEAM itself. Crucially, the researchers cross-validated their headline results against two fully independent datasets: a VPD product derived from the Climatic Research Unit’s CRU TS v4.09 archive and the P-LSHv2 evapotranspiration dataset with explicit soil moisture constraints.

The methodological core of the study is a clever adaptation of the space-for-time approach. Rather than tracking how a single location responds to VPD through time—a signal easily contaminated by long-term trends—the team used a moving-window strategy on a common 0.1-degree grid. For each target pixel, they compared it with neighboring pixels within a five-by-five window, retaining only those that shared the same dominant land-cover type, differed by less than 10 percent in the fractional cover of that type, and sat within 100 meters of the same elevation. The Theil–Sen slope estimator, a nonparametric technique resistant to outliers, then translated the differences in evapotranspiration and VPD among these carefully matched pixels into an apparent local sensitivity, expressed in millimeters of water per hectopascal of vapor pressure deficit.

The resulting map is striking. About 60.7 percent of global land area shows positive apparent sensitivity of evapotranspiration to VPD, with a global mean of 293.3 plus or minus 62.3 millimeters per hectopascal. Positive sensitivity dominates warm, humid regions, where abundant soil water allows ecosystems to meet the atmosphere’s escalating demand. Negative or weak sensitivity clusters in arid and cold regions, where water supply, not atmospheric demand, is the binding constraint. The land-use breakdown adds another layer of nuance: densely vegetated areas show the strongest sensitivity at 405.3 millimeters per hectopascal, followed by grasslands at 342.3 and barren land at 200.4, while croplands—shaped by irrigation and management—respond most weakly at 78.1.

When the researchers decomposed evapotranspiration into its components, a fundamental asymmetry emerged. After statistically controlling for root-zone soil moisture and precipitation using partial correlation analysis on detrended annual series, VPD remained positively correlated with plant transpiration across 65.8 percent of valid land pixels, with the strongest and most significant positive relationships concentrated in North America, Europe, and northern Asia. Bare-soil evaporation told a different story: its correlation with VPD was weaker overall and turned negative across 59.1 percent of pixels, particularly in water-limited landscapes such as Australia, southern Africa, and Central Asia. In other words, drier air reliably squeezes more water through plants wherever roots can still find it, but over parched soil it often accelerates nothing—because there is simply no water left to evaporate.

Perhaps the most consequential finding concerns thresholds. Using piecewise linear regression checked against generalized additive models, the team identified pronounced aridity-dependent transition points in the VPD–evapotranspiration relationship. In arid regions, the dominant transition occurs only at a VPD of 1.90 kilopascals; in semi-arid regions, 1.46 kilopascals. But in semi-humid and humid regions the thresholds collapse to 0.49 and 0.47 kilopascals respectively. The interpretation is subtle but important: wetter ecosystems reach their response transition under far lower atmospheric dryness, meaning that as VPD continues its relentless rise, humid and semi-humid regions may hit their turning points earlier than anyone expected, while dry regions face chronic suppression once water limitation takes over. These thresholds are statistical transition points in the observed relationship, the authors caution, not fixed ecological limits, and their exact values may shift with vegetation type, soil water storage, and the data products used.

To untangle the web of interacting drivers, the researchers deployed structural equation modeling separately for each aridity zone, treating temperature, precipitation, and solar radiation as exogenous climatic drivers, root-zone soil moisture as an intermediate water-supply variable, leaf area index as a vegetation-state variable, and evapotranspiration as the final response. The results revealed that VPD was negatively related to root-zone soil moisture in every aridity zone, with standardized coefficients ranging from −0.25 in arid regions to −0.66 in semi-humid ones. The direct pathway from VPD to evapotranspiration was strongly negative in arid and semi-arid regions—standardized coefficients of −0.79 and −0.55—weak and non-significant in semi-humid areas, and moderately negative in humid ones. Summing direct and indirect effects, the total effect of VPD on evapotranspiration was negative in all four zones, strongest in arid regions at −0.96. The paradox is resolved once the full pathway structure is visible: higher VPD does not merely demand more water, it also depletes soil moisture, stresses vegetation, and ultimately throttles the flux it drives.

The implications ripple outward through climate science and water management. Global VPD rose at a rate of 0.002 kilopascals per year between 1981 and 2020, with positive trends across 92.9 percent of land area, yet evapotranspiration increased only slightly and unevenly, declining across nearly a third of land pixels—concentrated in Africa, central South America, southwestern North America, and eastern Australia. Regions with the fastest VPD increases often saw falling evapotranspiration, a mismatch that climate models struggle to reproduce and that recent work suggests can lead to significant overestimation of climate-driven evapotranspiration increases when land–atmosphere feedbacks are ignored. Because VPD also suppresses photosynthesis while promoting water loss, its asymmetric effect has already been implicated in the leveling off of global vegetation water-use efficiency since 2000, offsetting some of the benefits of rising carbon dioxide.

The authors are candid about the uncertainties that remain. Evapotranspiration products can disagree on annual estimates by nearly 50 percent, and the partitioning of total evaporation into transpiration, interception, and soil evaporation is still hotly contested. Soil moisture in reanalysis products carries biases, especially in drylands, which could shift how much of the VPD effect is attributed to direct stress versus the indirect soil-moisture pathway. The structural equation models capture directional statistical associations, not fully reciprocal feedbacks between evaporation, soil moisture, and atmospheric humidity. Even so, the study delivers something the field has lacked: quantitative constraints on where and how rising atmospheric dryness will push terrestrial water fluxes, and where it will instead deepen drought. For forecasters, the message is that a single global sensitivity value is meaningless—adaptation planning, from early-warning systems in humid basins to hard-nosed water management in drylands, must be tailored to the thresholds of each climate zone.

Subject of Research: The influence of vapor pressure deficit changes on global terrestrial evapotranspiration

Article Title: The influence of vapor pressure deficit changes on global terrestrial evapotranspiration

Article References: Miao, Y., Zhu, G., Wang, Y., Huang, E., Wang, Q., Gun, Y., Zheng, Z., Yang, J., Li, W., & Liu, Z. (2026). The influence of vapor pressure deficit changes on global terrestrial evapotranspiration. Hydrology and Earth System Sciences, 30(19), 6249-6264. https://doi.org/10.5194/hess-30-6249-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-6249-2026

Keywords: vapor pressure deficit, evapotranspiration, hydrology, climate change, remote sensing, GLEAM, ERA5-Land, soil moisture, transpiration, aridity thresholds, land-atmosphere feedback, water cycle