Human-induced warming accelerates local evapotranspiration and precipitation recycling over the Tibetan Plateau

The Tibetan Plateau (TP), known as Asia’s water tower, feeds twelve major river systems and supports nearly two billion people. It is ecologically fragile yet highly sensitive to climate change, having warmed at more than 0.3 K per decade over the past 60 years, far above the global average. Consequent environmental shifts include glacier retreat, lake expansion, altered streamflow, permafrost degradation, and greening. Observations show a dipole in precipitation: drying along parts of the Himalayas and wetting over the inner TP. Prior work emphasized weakening Indian monsoon and aerosol forcing, but quantitative attribution of large-scale circulation changes versus local moisture recycling remains incomplete. This study asks how human-induced greenhouse gas (GHG) warming alters the balance between external moisture advection (monsoon and westerlies) and local precipitation recycling over the TP, and how these processes explain past and future hydroclimatic changes. Using a 3D Lagrangian moisture-tracking framework, water budget decomposition, and CMIP6 single-forcing and scenario simulations, the authors investigate the mechanisms behind observed changes and assess future sensitivity of the TP hydrological cycle to local land–atmosphere interactions.

Previous studies linked TP precipitation changes to a weakening Indian monsoon and modifications of the Asian westerly jet, often tied to anthropogenic aerosol forcing. However, quantitative evidence connecting circulation changes to upwind moisture sources and TP precipitation remained limited, and mechanisms for inner TP wetting under weakened westerlies were unclear. From a moisture recycling perspective, TP warming and associated land changes could intensify local evapotranspiration and recycling. A comprehensive review of precipitation recycling ratios over the TP shows wide ranges depending on method and scale: isotopic and budget analyses often report higher recycling (summer ~30–80%, annual ~53–63%), while 2D Eulerian methods commonly yield lower values (~10–40%) due to neglect of cascading recycling and vertical structure. Reported values across 16 studies span ~10–80% (summer) and ~10–63.2% (annual), reflecting methodological and domain-definition differences. This study positions its 3D Lagrangian estimates between these groups and addresses methodological limitations in prior work by capturing cascading effects and three-dimensional transport.

Datasets and period: Reanalysis and observations include CERA-20C (primary driver for tracking), ERA5, HARv2, TP Reanalysis (TPR), GPCP, GPCC, CRU precipitation, GLEAM evapotranspiration, and HadCRUT5 surface temperature. Multi-product ensembles were constructed for evapotranspiration and precipitation over 1983–2019 to benchmark trends and evaluate CERA-20C biases. Spatial domain: TP outlined by 3000 m elevation; basin and watershed boundaries from HydroSHEDS. 3D Lagrangian tracking: FLEXPART v10.4 was run globally in domain-filling mode (5 million parcels; 91 model levels; parcel motion every 15 min, outputs every 3 h) driven by CERA-20C model-level fields for 1971–2010. Moisture source attribution used the WaterSip diagnostic, which accounts for cascading precipitation–evaporation cycles and discounts early uptakes by en-route losses. A spatially and seasonally varying RH threshold scheme was developed to minimize biases between Lagrangian-estimated precipitation and surface precipitation. Tracking terminated when >99% of source was traced or after 20 days (optimal-tracking mode). The approach yields precipitation footprints (backward moisture-source fields) and recycling ratios. Circulation regime classification: A compact monsoon–westerly index (MWI = u300 · u850) was devised to identify overturning (monsoon) vs barotropic (westerly/easterly) regimes. Regimes relevant to TP precipitation were mapped seasonally to attribute contributions from the Indian monsoon, East Asian monsoon, and various westerly regimes to TP precipitation at grid scale. Moisture budget decomposition: Using the vertically integrated water balance, interdecadal precipitation change (δP) between 1991–2010 and 1971–1990 was decomposed into contributions from evapotranspiration (δE), thermodynamic humidity effect (δTH), dynamics (δDY), transient eddies (δTE), nonlinear (δNL), storage (δST), and a residual. Vertical integration used 10 hPa to surface; near-surface humidity terms unavailable in reanalyses were absorbed into the residual. CMIP6 detection/attribution and projections: DAMIP single-forcing experiments (historical-all, hist-GHG, hist-aer, hist-nat) and ScenarioMIP (SSP2-4.5, SSP3-7.0, SSP5-8.5) were assessed. From 12 DAMIP-participating models, the best five (CNRM-CM6-1, MRI-ESM2-0, ACCESS-ESM1-5, ACCESS-CM2, GFDL-ESM4) were selected based on skill in reproducing the observed inner wetting–Himalayan drying pattern. Trends analyzed over 1971–2014 (historical) and 2015–2100 (future), including precipitation, temperature, 500-hPa specific humidity, 400-hPa dry static stability, and 500-hPa divergence/winds. Statistical analysis: Mann–Kendall tests assessed trend significance; t-tests assessed MWI significance. Uncertainties reported as interannual standard deviations, interquartile ranges from bootstrapping (budget terms), or inter-model spread (25th–75th percentiles).

- Moisture source attribution and regimes: - Up to 94.5 ± 2.0% of TP precipitation is traceable via Lagrangian tracking. - Annual precipitation recycling ratio is 46.5 ± 2.2%; summer (JJA) 52.2 ± 2.7%; non-summer 41.0 ± 2.5%. - Terrestrial sources account for 85.2 ± 1.7% of TP precipitation; direct oceanic contribution ~9.3 ± 1.2%. - In summer, Indian monsoon supplies 24.6 ± 2.7% and mid-latitude westerlies 7.0 ± 0.6% of TP precipitation; in non-summer, westerlies contribute 29.0 ± 2.0% and Indian monsoon 11.6 ± 1.4%. - Large areas (62.6–74.3% of TP) are TP-dominated in source footprints, underscoring cascading moisture recycling and terrestrial self-dependence. - Observed hydroclimate trends (recent decades): - Near-surface air temperature warming over TP at 0.33 ± 0.07 K dec⁻¹, exceeding global 0.20 ± 0.02 K dec⁻¹. - Evapotranspiration increased 5.8 ± 2.2 mm yr⁻¹ dec⁻¹; precipitation increased 6.6 ± 6.6 mm yr⁻¹ dec⁻¹ (≈1.3 ± 1.3% dec⁻¹) in ensemble products. - E–P ratio shows a robust upward trend of 0.8 ± 0.7% dec⁻¹, indicating growing potential for precipitation recycling. - Lagrangian recycling ratio increased 0.77 ± 0.54% dec⁻¹ (≈5.7 ± 4.0 mm yr⁻¹ dec⁻¹), confirming intensifying internal water cycling and stronger influence of local ET on precipitation. - Fractional contributions decline for external conveyors: Indian monsoon −0.50 ± 0.44% dec⁻¹ (≈−3.7 ± 3.3 mm yr⁻¹ dec⁻¹); westerlies −0.45 ± 0.29% dec⁻¹ (≈−3.3 ± 2.1 mm yr⁻¹ dec⁻¹). - Mechanisms from moisture budgets and circulation: - Inner TP wetting is primarily driven by increased evapotranspiration (δE) and enhanced moisture convergence (humidity-driven δTH plus dynamic δDY), outweighing drying by transient eddies (δTE). - Himalayan drying corresponds to increasing moisture divergence that offsets rising local ET. - Trends show weakening westerlies (easterly tendency at 500 hPa) and slowing Indian summer monsoon (northeasterly tendency over the Indian Ocean), consistent with reduced external moisture transport. - Soil moisture trends support positive land–atmosphere feedbacks sustaining inner wetting and, conversely, feedbacks that exacerbate Himalayan drying. - Role of forcings and future projections (CMIP6 MME5): - Anthropogenic aerosol forcing explains much of the recent inner wetting–Himalayan drying contrast via cooling; as aerosols diminish and GHGs dominate, overall TP wetting emerges. - Dual GHG effect: warming boosts lower-tropospheric humidity and ET/recycling, while top-heavy heating increases 400-hPa dry static stability, fostering anticyclonic tendencies over the TP and weakening southerly monsoons. - Future ET rises faster than precipitation: under SSP5-8.5, ET trend ≈15 ± 0.4 mm yr⁻¹ dec⁻¹; relative growth of ET (2.9 ± 0.07% dec⁻¹) exceeds that of precipitation (2.6 ± 0.08% dec⁻¹). E–P ratio increases across scenarios, especially in non-summer. - Sensitivity to warming (SSP3-7.0 example): TP ET scales at 5.1 ± 0.07% K⁻¹ vs precipitation at 4.1 ± 0.17% K⁻¹, both outpacing global land rates. - Despite a projected decrease in 500-hPa southerly winds over southern TP (−10 ± 1.1% dec⁻¹), meridional moisture transport increases (2.7 ± 0.2% dec⁻¹) due to higher specific humidity. - Strong correlation between ET and recycling ratio (r = 0.72, p < 0.01) indicates increasing dependence of TP precipitation on local ET.

The study disentangles the relative roles of local precipitation recycling and external moisture advection in TP hydroclimatic change. Observational analyses and 3D Lagrangian tracking reveal that warming has intensified evapotranspiration and precipitation recycling while contributions from the Indian monsoon and westerlies have declined. Moisture budget decomposition shows inner TP wetting arises mainly from increased ET and enhanced convergence from within the plateau, whereas Himalayan drying is driven by increasing divergence that counteracts local ET. Detection–attribution and scenario analyses highlight a dual GHG effect: warming increases humidity and ET, strengthening recycling and overall wetting, but top-heavy heating stabilizes the mid–upper troposphere, favoring anticyclonic conditions and weakening southerly monsoon inflow across the Himalayas. Consequently, the TP hydrological cycle becomes increasingly self-constrained—more reliant on terrestrial sources and local land–atmosphere feedbacks and less on external advection. This has important implications for water resource reliability, ecological resilience, and climate prediction across High Mountain Asia and downstream basins.

This work provides robust quantitative evidence—via 40-year 3D Lagrangian tracking, moisture budget decomposition, and CMIP6 single-forcing and scenario experiments—that recent TP precipitation changes are driven by intensified precipitation recycling and enhanced moisture convergence that offset diminishing monsoon and westerly transports. The TP’s water cycle is becoming more self-constrained and sensitive to local land–atmosphere interactions under human-induced warming. Projections indicate ET will continue to outpace precipitation growth, increasing the E–P ratio and dependence on recycling, while GHG-induced atmospheric stabilization weakens southerly monsoons. These insights inform water-tower management and risk mitigation for downstream societies. Future research should expand DAMIP participation to reduce model spread, improve land–atmosphere coupling and representation of complex topography and convection, enhance observational networks over the TP, and conduct intercomparisons across isotopic, Eulerian, and Lagrangian methods to reconcile precipitation recycling estimates and uncertainties.

- Model selection and spread: Only a subset of CMIP6 models (MME5) reproduced the observed inner wetting–Himalayan drying; appreciable inter-model spread remains in projected water budgets. - Forcing experiments coverage: DAMIP Tier-1 participation is limited, constraining cross-model robustness. - Methodological constraints: 3D Lagrangian trajectories are subject to uncertainties from convection, turbulence, rain evaporation, and numerical diffusion; water budget residuals are sizable over complex topography. - Data limitations: Sparse in situ observations over the TP necessitate reliance on reanalyses and satellite/model products; near-surface humidity gaps affect budget closure. - Scale and boundary dependence: Precipitation recycling ratios depend on region definition and shape; literature uses inconsistent TP boundaries, complicating comparisons across studies; isotopic and Eulerian approaches have their own sampling and structural biases.