Warming, increase in precipitation, and irrigation enhance greening in High Mountain Asia

Understanding changes in vegetation greenness is essential for predicting, mitigating, and adapting to climate change. Satellite observations of leaf area index (LAI) show global greening, primarily attributed to CO2 fertilization, but also influenced locally by land management and precipitation trends. Greening affects hydrologic connectivity, water and energy fluxes, and atmospheric dynamics, making it critical to identify its drivers. High Mountain Asia (HMA), encompassing Asian mountain ranges around the Tibetan Plateau, is hydrologically complex, with strong cryosphere–biosphere–hydrosphere interactions, heterogeneous topography and land cover, and large populations reliant on its water resources. HMA is experiencing rapid warming and changes in precipitation, while India and China exhibit some of the highest greening rates globally due to climate and land-use changes. In HMA, greening is largely moisture-induced. This study aims to identify the principal climatic and anthropogenic drivers of greening across HMA by jointly analyzing vegetation, soil moisture, snow cover, terrestrial water storage, precipitation, and temperature from 2003 to 2020 using a multivariate, remote-sensing-based framework.

Global greening has been widely documented, with ~79% attributed to CO2 fertilization and nitrogen deposition, though land use and climate trends also contribute. In South and East Asia, especially China and India, strong greening trends have been linked to climate variability, land-use change, and afforestation programs. Prior work indicates moisture-induced greening in South Asia and that changes in temperature and precipitation shape vegetation dynamics on the Tibetan Plateau and Central Asia. HMA’s precipitation estimates are uncertain due to sparse ground observations and complex terrain. GRACE-based analyses have advanced understanding of terrestrial water storage (TWS) but are limited in spatial resolution. The study builds on these insights by integrating multiple satellite and reanalysis products and applying partial information decomposition to disentangle unique and redundant contributions of water and climate variables to greening.

The study performs a multivariate remote sensing analysis at annual resolution (2003–2020), examining LAI, soil moisture, snow cover fraction, TWS, precipitation, and air temperature over HMA. Data were analyzed at basin scale and at 0.5° grid (coarsest dataset resolution, consistent with GRACE mascons); all other datasets were upscaled to 0.5°.

• Vegetation: MODIS MCD15A2H v6 LAI (500 m, 8-day) used to quantify greening trends.
• Snow: MODIS MOD10CM monthly snow cover fraction (0.05°) used to assess cryospheric changes.
• Soil moisture: ESA CCI v05.2 daily combined product (active and passive microwave).
• Terrestrial water storage: GRACE CSR RL06 mascons (300–500 km) to capture changes in TWS.
• Meteorology: Precipitation from multiple gridded datasets (ERA5, IMERG, CHIRPS, APHRODITE, HAR, PRINCETON); air temperature from ERA5. Statistical analyses included Mann–Kendall trend testing for monotonic trends in LAI, precipitation, temperature, soil moisture, snow cover, and TWS at 95% confidence. To attribute drivers, Partial Information Decomposition (PID) decomposed mutual information into unique, redundant, and synergistic components. First, contributions of soil moisture (SM) and snow cover (SC) to TWS and LAI were quantified. Then, precipitation (P) and temperature (T) contributions to SM and SC were analyzed. Greening was attributed as follows: (i) precipitation-driven where SM uniquely explains LAI and P uniquely explains SM; (ii) irrigation-driven where SM uniquely explains LAI, SM increases are not explained by P/T, and the area is irrigated; (iii) warming-driven where SC uniquely (or redundantly with SM) explains LAI due to decreases in snow increasing SM and lengthening the growing season. Synergistic information was negligible and omitted. Potential lags were ignored due to annual aggregation.

Greening is heterogeneous across HMA, strongest at low to mid-elevations (<4000 m). By land cover: evergreen and mixed forests (~13% of HMA) show LAI increases of ~0.011 m2 m−2 yr−1; croplands (~18%) ~0.01 m2 m−2 yr−1; grasslands (~16%) ~0.0036 m2 m−2 yr−1 on average.

• Irrigation-driven greening: Predominant in the Ganges–Brahmaputra and Indus basins. Croplands exhibit the highest LAI increases (up to 0.04 m2 m−2 yr−1 in Ganges–Brahmaputra; 0.03 m2 m−2 yr−1 in Indus). TWS shows strong declines (up to ~10 cm yr−1 in Ganges–Brahmaputra and ~4 cm yr−1 in Indus), consistent with groundwater depletion. Soil moisture has increased (~2% yr−1) due to irrigation, not precipitation or temperature, and uniquely explains LAI changes in these irrigated regions.
• Warming-driven greening: Widespread across eight of eleven basins (e.g., Tibetan Plateau, Hwang Ho, Ili, Amu Darya, Syr Darya, Tarim). Air temperature increases (up to ~0.6 °C on average in the Tibetan Plateau and nearby ranges) coincide with decreases in snow cover fraction (> −0.4% yr−1) and increases in soil moisture (up to ~1% yr−1), resulting in more available water and longer growing seasons. In higher-elevation Yangtze areas (>1500 m), despite increased precipitation, greening is mainly controlled by snow cover decreases; reported local air temperature increases are ~0.2 °C yr−1. In several basins, TWS decreases (~0.1–1 cm yr−1) accompany snow loss.
• Precipitation-driven greening: Observed in mid/low-elevation, forested areas in the southeast HMA, including non-irrigated parts of the Indus and monsoon-dominated basins (Irrawaddy, Si, Song Hong). Here, SM uniquely explains LAI/TWS and is primarily driven by P. Non-irrigated Indus shows increasing precipitation (0.03–0.08 mm day−1 yr−1), with modest TWS increase (<0.1 cm yr−1). Monsoon basins show increased precipitation (~1 mm day−1 yr−1) but small TWS decreases likely linked to anthropogenic withdrawals; notable TWS declines occurred 2003–2006 in Si and Song Hong and 2012–2020 in Irrawaddy despite rising P and SM.
• Spatial attribution: Fig. 2 delineates three primary drivers—precipitation, warming (via snow decrease), and irrigation—whose influence depends on elevation, land cover, and land use.

Irrigation-driven greening covers more than half of the Ganges–Brahmaputra and about 22% of the Indus basin, producing the largest LAI increases in HMA. Irrigation-induced soil moisture changes can significantly alter land–atmosphere coupling and potentially local climate, with observed slight air temperature decreases (<0.03 °C yr−1) in irrigation- and precipitation-controlled areas, likely due to enhanced evapotranspirative cooling associated with increased vegetation. Warming-driven greening areas generally see increased precipitation, amplifying greening. While afforestation has contributed to greening in parts of China, the analysis indicates climate as the dominant driver across forested basins in the region. Feedbacks between greening and precipitation likely operate on longer timescales than those examined here; over the study period, precipitation variations primarily drove forest greening where precipitation was identified as the driver. The results emphasize the need to account explicitly for anthropogenic water use (irrigation and pumping) and cryospheric changes in understanding and modeling HMA’s coupled water–energy–carbon dynamics.

This study provides a holistic attribution of greening across High Mountain Asia, identifying three dominant drivers—intense irrigation, decreases in snow cover associated with warming, and increases in precipitation—whose relative influence varies with elevation, land cover, and land use. Irrigation is the primary driver of the strongest greening in the Ganges–Brahmaputra and Indus, with concurrent soil moisture increases and marked TWS depletion. Warming reduces snow cover and lengthens growing seasons, enhancing greening across many high-elevation basins, while precipitation increases drive greening in monsoon-influenced, forested low- to mid-elevation regions. These findings underscore the importance of incorporating irrigation dynamics and cryosphere–hydrosphere–biosphere interactions into Earth system models and projections. Future work should improve observational constraints on precipitation in complex terrain, resolve human water use impacts, examine lagged and seasonal processes, and assess feedbacks between greening and regional hydroclimate at longer timescales.

Key limitations include: (1) Uncertainty in precipitation datasets over HMA due to sparse ground observations and complex topography; multiple products were used to mitigate but cannot eliminate this uncertainty. (2) GRACE TWS’s coarse spatial resolution (300–500 km) constrains attribution at fine scales and necessitated upscaling all datasets to 0.5°. (3) Annual aggregation ignores potential seasonal and lagged relationships among precipitation, soil moisture, snow cover, and LAI. (4) PID-based attribution relies on statistical information measures and assumptions (e.g., low synergy, attributing SM increases to irrigation where not explained by P/T), which may not capture all confounding processes. (5) Potential local land-use changes (e.g., afforestation) are not explicitly modeled and may co-occur with climatic drivers.