The western transition region in the central Himalaya constitutes the study region (WCH: 29° 38′–32° 13′ N and 77° 13′–80° 50′ E) (Figure 1). During summer, glaciers in the region are known to be fed by the Indian summer monsoon (ISM). In contrast, winter snow accumulation is influenced by a precipitation regime driven by mid-latitude winter westerlies. Long-term meteorological records of higher elevation are not readily available. Analyses of gridded Climate Research Unit (CRU) records indicate that annual regional precipitation is ∼800 mm, of which the warm-wet summer months (April-September) receive about 80%. The mean annual temperature varies around 5.5°C, with a minimum (−9.4°C) in January and a maximum (∼18.5°C) in July (Harris et al., 2014, Supplementary Figure S1). Long-term hydro-climatological records derived from regional tree-ring studies show a trend of declining precipitation from both the ISM and winter westerlies (Sano et al., 2012, 2017; Singh et al., 2019). However, winter precipitation has increased in recent decades (Treydte et al., 2006; Yadav et al., 2017). In contrast, tree-ring-derived regional temperature climatology shows an increasing trend irrespective of the season, with a prominent increase in the winter temperatures (Shah et al., 2019; Gaire et al., 2020; Panthi et al., 2021).

The tree-ring cellulose δ¹⁸O records from three sites distributed across the studied region (Figure 1), comprising the dominant tree species (Table 1), were used to reconstruct the climatology of regional atmospheric humidity content (AMC). The first site (Manali: Sano et al., 2017) is located at the northwestern periphery of the ISM incursions. The second site (Uttarakashi: Singh et al., 2019) encompassed δ¹⁸O records from three species belonging to two different plant functional types (PFTs) that differ in their phenological cycle (Table 1): one PFT included an evergreen conifer and the other was a broadleaf deciduous species (Aesculus indica). The growth period of A. indica (April to September) coincides with the warm-wet phase in the central Himalaya. The δ¹⁸O records from the third site (Jageshwar: Xu et al., 2018) covered the longest period (∼ 400 years) and similar to site one, were composed of a dominant conifer species (Table 1). Therefore, based on previous isotopic studies (Sano et al., 2017; Xu et al., 2018; Singh et al., 2019) and existing high inter-annual correlation among δ¹⁸O records from WCH (Table 2), a ∼400 years-long regional tree-ring δ¹⁸O chronology was produced utilizing mean values of these records.

These studies followed standard sampling (two cores per tree) and dendrochronological procedures such as mounting, surface smoothing, tree-ring widths measurement, cross dating and quality control as per the standard methodology. Further, isolation of cellulose from whole wood and oxygen isotope analyses were carried out by the respective work within standard analytical precision and presented in the common δ-notation against VSMOW. For a detailed dendro-isotopic procedure one may refer to Sano et al. (2017), Singh et al. (2019) and Xu et al. (2018).

In-situ inter-annual meteorological records are mainly available for lower elevations (< 2000 m asl). Therefore, long-term temperature and precipitation datasets for the study region were obtained from CRU (TS.3.22, 0.5° x 0.5°, monthly, 1901–2015; Harris et al., 2014). The monthly reanalysis AMC dataset (0.5° x 0.625°) was obtained from MERRA-2 (i.e., total precipitable water vapour, http://giovanni.sci.gsfc.nasa.gov) from 1982–2015 (n = 34 years). Monthly soil moisture data, from 1982 to 2015, were obtained from the NOAA Climate Prediction Center (0.5° x 0.5°) (https://psl.noaa.gov/data/gridded/data.cpcsoil.html)

MODIS: Regional trends in leaf area index (LAI: MOD15A2H) and evapotranspiration (ET: MOD16A2) since 2000 were deduced using 500 m resolution, 8-days composite MODIS products. Products were clipped using the shapefile for the WCH and pixel values were exported using ERDAS imagine to ASCII format. A MATLAB code was prepared to calculate the average LAI value of each image. LAI and ET computations were performed following the processing manual (www.reverb.echo.nasa.gov) (Supplementary Table S1). Regional sensitivity of ET to LAI (∆ET/∆LAI) was derived following Zeng et al. (2016).

LANDSAT: To detect changes in valley-scale vegetation area and greening trends (> 2,500 m asl), we utilized 30 m resolution L1T Landsat data since 1972 with different onboard sensors (MSS, TM, ETM + , OLI) (a total of 766 images, Supplementary Table S1). Changes in the seasonal vegetation area of eight evenly distributed glacial valleys in WCH (Supplementary Table S2) were derived during summer (April–September: when both evergreen and deciduous canopies are green) and winter (October–March: evergreen species only) from the NDVI (Normalized Difference Vegetation Index: > 0.3). The images were downloaded (https://glovis.usgs.gov), processed and analysed to detect multi-year seasonal vegetation trends. The watershed boundary of each valley (four valleys each in Garhwal Himalaya and Kumaun Himalaya: Supplementary Table S2) was delineated by utilizing mosaicked and re-projected tiles of 30 m Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model v2 (http://gdex.-cr.usgs.gov/gdex/). Available L1T datasets with less than 10% cloud cover were selected and referenced to the Geographic Coordinate system UTM Zone 44N. In order to bring time-series data to a common scale, digital number values from each image were converted to radiance using gain and offset values specified in the image metadata. Selected Landsat image were transformed into NDV time series using the following equation: N D V I =   N I R −   R e d N I R + R e d where “NIR” and “Red” are wavelengths in near-infrared and red bands, respectively. Five-year seasonal image composites, using maximum value composite procedure, were used, except for the period from 1972–1980, where a nine-year composite was generated due to limited availability of cloud-free images. This compositing process strongly reduces the effects of clouds, snow, seasonal differences caused by solar angle differences and noise. Reducing the data volume also minimizes errors in phenological changes.

CERES: Multi-year radiative forcing changes at the scale (Synoptic 1° × 1°; monthly; Product: EBAF-TOA 4.0) of the eight WCH glacial valleys (Supplementary Table S2) were observed with Cloud and Earth Radiation Energy System (CERES), one of the most important global satellite detectors to monitor radiation and energy budget. Net radiative forcing and its components (shortwave and longwave), at the surface (SFA), and at the top of the atmosphere (TOA) were downloaded from the official website (https://ceres.larc.nasa.gov/) at a monthly scale from March 2000 to December 2018. Longwave, shortwave, and net fluxes under clear and all-sky conditions were used to compute radiative forcing at the TOA and SFA following the equations (1 to 9) given in Bao et al. (2019) and details of the computation of radiation components can be found in the supplementary table (Supplementary Table S3).

The present study utilized the tree-ring cellulose δ¹³C-derived proxy of annual glacier mass balance reconstruction over 273 years of four benchmark glaciers in the region (Singh et al., 2021a). This reconstruction is based on previous and relevant proxy-based glacio-hydrological studies, based on the assumption that climate indirectly connects the mass balance of a glacier with tree-ring C isotope ratios (Borgoankar et al., 2009; Shekhar et al., 2017; Zhang et al., 2019; Singh et al., 2021a). These four benchmark glaciers are in-situ monitored glaciers that are evenly distributed across the study region: Dokriani glacier (in DOK valley), Chorabari glacier (in MAN valley), Tipra Bamak (in CHA valley) and Dunagiri glacier (in NAN valley) (Figure 1; Supplementary Table S2).

To understand the relationship between atmospheric humidity (AMC) and hydroclimatic signal in evergreen conifers and broadleaf deciduous δ¹⁸O data, simple Pearson correlations were applied with a response function approach. To investigate the correlations of δ¹⁸O data with the monthly AMC averages in both cases, correlations were analysed at confidence intervals of 95% and 99%. In the case of evergreen conifers, response function analysis was performed from August of the previous growth year through to September of the current year. In the case of broadleaf deciduous species, with a growth cycle from April to September, the response function was performed for this period. To further corroborate the response functions, we plotted 3-month moving correlation coefficients between conifer δ¹⁸O and mean AMC from 1982/1983 to 2014/2015 (Table 3).

Based on the above significant correlations, we established linear regression models for AMC reconstructions: 1) for previous year August to September (pAugest—September) based on δ¹⁸O of regional conifers, and 2) for April to September based on δ¹⁸O of the broadleaf deciduous species A. indica. Reconstructions were performed using the LM module in R (ggplot2 package, Wickham, 2016). After applying statistics to check the strength of our linear regression model (specifically, the F-test, residual standard error, root mean square error, Durbin–Watson test), we applied Akaike’s and Bayesian information criteria for model evaluation and selection (Table 3). Therefore, a linear regression model was employed for the reconstruction of annual (pAugest—September) AMC over the past 395 years (1621–2015) and the corresponding empirical equation is: A M C a n n u a l = 27.392 − 0.487 ×   δ 18 O r e g i o n a l   c o n i f e r s

Here, δ¹⁸O_{regional conifers} is the mean chronology of regional evergreen conifers and AMC is annual atmospheric moisture content (Kg m⁻²). Detailed calibration–verification statistics are presented in Table 3, which indicate the strength and reliability of our reconstruction. Validation tests including the number of sign agreements between reconstructed series and observed AMC records, and cross-correlation between reconstruction and measurements were significant (p < 0.001).

Furthermore, a linear regression model was also employed for the reconstruction of summer season (April–September) AMC over the past 196 years (1820–2015) by utilizing δ¹⁸O of a broadleaf deciduous species (A. indica). The corresponding empirical equation is: A M C s u m m e r = 25.885   −   0.388   ×     δ 18 O d e c i d u o u s

Here, δ¹⁸O_deciduous is chronology of A. indica and summer AMC is atmospheric moisture content (Kg m⁻²) during April to September. Model and calibration–verification statistics indicate reliability and strength of this reconstruction model (Table 3). Validation tests including the number of sign agreements between the reconstructed series and observed AMC records, and cross-correlation between reconstruction and measurements are significant (p < 0.001). Finally, October to March (winter) AMC was estimated from the above reconstructions.

The leave-one-out cross-validation method (LOOCV; Michaelsen, 1987) was used for the entire calibration period (1982/1983–2014/2015) and verifying the reconstruction (Table 3). This method is most suitable when the length of observed records is short (Shekhar et al., 2017; Zhang et al., 2019; Singh et al., 2021a); each observation is successively withdrawn, and a model is estimated based on the remaining observations and a prediction is made for the omitted observations. The LOOCV analysis was performed using package “caret” (Kuhn, 2015). Statistics such as sign test, product mean test (PMT), reduction of error (RE), and correlation coefficients were calculated to evaluate the similarity between observed and estimated values (Table 3). The PMT and RE statistics provide a rigorous test of the association between actual and estimated series. Positive values indicate the predictive capability of the model. A positive RE is evidence of a valid regression model (Fritts, 1976). In addition, other statistics, viz., root mean square error, coefficient of efficiency, mean absolute error, and Durbin–Watson test were carried out to evaluate the linear regression model (Table 3). The reconstructions were standardized using Z-scores and smoothed with 11-years or 21-years fast Fourier transform to highlight the common climate signals. To understand regional scale-climatic features associated with present seasonal atmospheric moisture reconstruction, we performed cross-field spatial correlation with gridded seasonal precipitation (CRU TS 4.05) using KNMI Climate Explorer (http://climexp.knmi.ni; van Oldenborgh and Burgers, 2005).

Tree-ring oxygen isotope ratio (δ¹⁸O) is a precise recorder of regional hydroclimatic condition (Baker et al., 2016). High temporal coherency between site-chronologies expands the spatial scale and increases the reliability of dendroclimatic reconstructions (Kahmen et al., 2011; Xu et al., 2017; Zeng et al., 2017; Xu et al., 2018). Moreover, the climatic sensitivity of tree ring δ¹⁸O is further enhanced when high correlations among site-chronologies allow the combination of tree species of diverse functional types (e.g., Xu et al., 2018; Singh et al., 2019). Cellulose δ¹⁸O chronologies have been extensively utilized to study the spatial climate heterogeneity over the HKT, covering century to millennium-scale variability in regional hydroclimate, e.g., changes in precipitation, atmospheric moisture, cloud cover, and vapor pressure (Treydte et al., 2006; Sano et al., 2012; Sano et al., 2013; Hochreuther et al., 2016; Grießinger et al., 2017; Sano et al., 2017; Xu et al., 2018; Huang et al., 2019; Singh et al., 2019; Managave et al., 2020).

As tree roots take up soil water without fractionation between the heavier and the lighter water molecules, therefore, a major part of the isotopic signature in tree rings should reflect the variation in precipitation or atmospheric humidity δ¹⁸O (Baker et al., 2016; Lehmann et al., 2018; Lehmann et al., 2020; Managave et al., 2020). The mechanisms associated with physical fractionation of stable oxygen isotopes over lower latitudes indicate that tree-ring δ¹⁸O values are mainly regulated by δ¹⁸O of atmospheric humidity and are dominated by Rayleigh distillation via amount effect and or local moisture recycling (Baker et al., 2016). However, depending upon the atmospheric humidity condition, basin-intrinsic ecophysiological processes (e.g., isotopic composition of soil water, leaf-water enrichment, and oxygen isotope exchange reactions of photosynthates with water) may have a considerable effect on tree-ring δ¹⁸Ovalues (Baker et al., 2016). Nevertheless, in the Himalayan environment with higher atmospheric humidity, tree-ring δ¹⁸O records have been widely utilized to reconstruct century to millennium-scale variability in regional precipitation and hydroclimatology (Roden et al., 2005; Lehmann et al., 2018).

Over the central Himalaya, utilizing levels of correlations (0.11–0.94) among δ¹⁸O chronologies of five sites, Xu et al. (2018) showed a decline in the strength of the ISM since the last 180 years. Moreover, δ¹⁸O chronologies of different functional types (conifers and deciduous) from a WCH site (Uttarakashi) confirmed declining ISM precipitation in the region (Singh et al., 2019). These studies indicate that irrespective of site and tree species, interannual changes in ISM-derived atmospheric moisture are closely related to the El Niño–Southern Oscillation, but correlation strength varied synchronously across the central Himalaya (Sano et al., 2012; Sano et al., 2013; Sano et al., 2017; Xu et al., 2018; Singh et al., 2019).

However, our analyses of all six δ¹⁸O records across the central Himalaya (Table 1) reveal a remarkable climatic distinction between the eastern (ECH) and western (WCH) part of the central Himalaya (Supplementary Figure S2). Studies even suggest increasing climatic heterogeneity over the last 6 decades (Singh et al., 2021a). Towards the eastern side, δ¹⁸O records indicate normal ISM conditions during the 20th century (Bhutan: Sano et al., 2013; Hochreuther et al., 2016). In contrast, δ¹⁸O chronologies from the central (Ganesh: Xu et al., 2018) and western regions (WCH) (Manali: Sano et al., 2017; Uttarakashi: Singh et al., 2019; Jageshwar: Xu et al., 2018; and Humla: Sano et al., 2012) unanimously show declining monsoon precipitation, as reflected in enriched tree-ring δ¹⁸O values (Figure 2). More specifically, mean values of δ¹⁸O chronologies progressively decline eastwards with a substantial difference (8‰–10 ‰) between the Bhutan site (19.24 ‰) and WCH chronologies (29.31 ‰) (except Humla) (Table 1, Supplementary Figure S2). This difference likely indicates a greater influence of the Bay of Bengal branch of the ISM on the eastern Himalaya (Perry et al., 2020; Singh et al., 2021a). A correlation matrix of six central Himalayan site chronologies found remarkably high correlations (0.55–0.75) among the WCH chronologies (Manali: Sano et al., 2017; Uttarakashi: Singh et al., 2019 and Jageshwar: Xu et al., 2018). However, inclusion of the Humla δ¹⁸O chronology (Sano et al., 2012) substantially lowers the correlation (0.23–0.5) (Table 2). Therefore, this study utilizes only the coherent WCH chronologies to reconstruct the regional changes in AMC. The WCH δ¹⁸O record spanning the last four hundred years indicates a prominent shift towards a drier phase after the 1960 s (Figure 2). This recent drier phase in regional hydroclimate has generally been attributed to anthropogenic emission-induced warming and resultant reduced land–ocean thermal contrasts (Xu et al., 2018).

The Pearson correlations with a response function approach showed the relationship between atmospheric humidity (AMC) and hydroclimatic signal in evergreen conifers and broadleaf deciduous δ¹⁸O data. The response function analysis for evergreen conifers from August of the previous growth year to September of the current year (pAugest-September) revealed a significant negative correlation (r = −0.671; p < 0.001). However, during peak winter (December and January) the correlation was positive (Figure 3A). Previously, a positive correlation with winter atmospheric humidity was observed in the study region. The correlations for these winter months were negative with the precipitation (Sano et al., 2012; Sano et al., 2017; Xu et al., 2018; Singh et al., 2019). A time lag induced by winter snowfall and frozen conditions could plausibly explain this consistent behaviour across the study region. Moreover, to corroborate the response function, we plotted 3-month moving correlation coefficients between evergreen conifer δ¹⁸O and mean AMC from 1982/1983 to 2014/2015. This enhanced the correlations considerably, even in the winter months (Figure 3B).

In the case of broadleaf deciduous tree species, having an annual growth cycle between April and September, the response function analysis showed a significant correlation (r = −0.63; p < 0.001) (dsl). Three-month moving correlation coefficients between deciduous δ¹⁸O and mean AMC improved the relationship substantially. This is in parity and has been shown with 3-month moving averages of growing season LAI dynamics, the correlation between ET and LAI, and the sensitivity of ET to LAI (∆ET/∆LAI: mm day⁻¹ (m² m⁻²) (Figure 3D). This indicates that the above biophysical parameters including their sensitivity remain tightly coupled to active vegetation growth from April to September. In contrast, for the non-growing winter season months, the sensitivity of ET to LAI, in particular, remained low and decoupled.

The correlation (r = −0.671, p < 0.001, n = 34) between WCH conifer δ¹⁸O chronologies and observed AMC was strong enough to establish a significant calibration model. A linear regression model was thus employed for the reconstruction of annual AMC (pAugest–September) over the past 395 years (1621–2015). The equation accounts for 45% of the AMC variance during the observation period (1982–2015) (Table 3; Figure 5B).

Smoothing of resultant annual AMC reconstruction with an 11-years moving average and break-point analyses indicates decreasing atmospheric moisture since the mid-19th century and a prominent decline in recent decades (since the 1960 s) (Figure 4A). Our results on seasonal disaggregation of annual AMC indicate that the decline in moisture influx in the region could be ascribed to both major atmospheric circulations, i.e., the ISM and the westerly circulation (Figure 5). Paleoclimatic evidence such as tree-ring δ¹⁸O chronologies (Sano et al., 2012; Sano et al., 2013; Sano et al., 2017; Xu et al., 2018; Singh et al., 2019), speleothems (Kotlia et al., 2012; Liang et al., 2015) and ice-core records from the central Himalaya (Kaspari et al., 2008) show that regional hydroclimate shifted towards a drier phase from the mid-19th century onwards. Studies indicate a reorganisation of hemispheric atmospheric circulation following the Little Ice Age. Moreover, a recent decline in AMC (since the 1960 s) could be attributed to increasing temperatures and climate change (Xu et al., 2018).

An approximately three hundred year-long glacier mass balance (GMB) variability record comprising four benchmark glaciers of WCH has previously been reconstructed based on cellulose δ¹³C records of two regionally dominant evergreen conifer tree species (Abies pindrow and Picea smithiana) (Singh et al., 2021a) (Figure 4B). We show the multi-decadal coherence between GMB and AMC. Over the common record of 273 years, the 51-years moving correlations between atmospheric moisture and ice mass display a shift in the relationship since the mid-19th century, which became even more pronounced after the 1960 s (Figure 4C).

To find the cause of this shift in the AMC—GMB relationship, we disaggregated annual AMC on a seasonal scale (summer and winter accumulation seasons). Approximately 200 years (1820–2015) of AMC record of the summer accumulation season (April–September) was derived based on a dominant deciduous tree species (A. indica) that completes its annual phenological cycle between April and September and remains physiologically dormant during winter (Figures 3C,D) (Singh et al., 2021a). As discussed above, a strong correlation (r = −0.626, p < 0.001, n = 34) was observed with summer season AMC (1982–2015) and the corresponding equation accounted for 39% of the AMC variance during the observation period. Further, model calibration–verification and validation tests statistics (Table 3) indicate the strength of the reconstruction. Thus, the winter season (October–March) AMC was derived from the annual and summer season AMC.

A decadal-scale change in the vegetation is expected to modulate the regional radiative energy balance (McPherson, 2007). Numerous observations from the orogen have indicated a widespread greening and vegetation expansion including in the subnival vegetation (Shen et al., 2015; Silva et al., 2016; Anderson et al., 2020; Parida et al., 2020; Sigdel et al., 2020). In this context, we focussed at the glacial valley scale vegetation trend in WCH through MODIS (250 m; 2000–2015) and 5 decades (1970–2016) of Landsat imageries (30 m) (Supplementary Table S2). Analyses indicate a differential greening behaviour in the plant functional types. The deciduous vegetation coverage has expanded rapidly (37.2 km² yr⁻¹) compared to conifers (13.7 km² yr⁻¹) (Supplementary Figure S4). Results of Landsat imageries corroborated a rapid increase in deciduous vegetation with respect to evergreen conifers at the valley-scale (Supplementary Figure S5). On average, deciduous vegetation coverage increased by 24% (15%–35%) relative to the pre-1994 level across the WCH glacial valleys. Current mean deciduous coverage area in the valleys is 36%, varying between 22% and 49% in different valleys. Furthermore, both MODIS- and Landsat-derived seasonal green area coverage estimates indicate that approximately 40% (30%–50%) of the glacial valley area rapidly turns-up green with deciduous canopy development (April—May) (Supplementary Figures S4, S5).

Decadal-scale environmental changes resulting from warming, augmented moisture and nutrient flow from snow-glacier melting, enhanced pre-monsoon rainfall, and disturbances such as forest fires might have favoured the expansion of deciduous species. Thus, given widespread greening and thermophilization (a preferential expansion of broadleaf deciduous vegetation over evergreen conifers), we contend comprehensive changes and a gross-scale alteration in surface–atmosphere interactions having direct implication on regional radiative balance. In addition, favoured expansion of deciduous vegetation on the valley slopes may even have repercussions on seasonal radiative balance, as deciduous species remain leafless throughout the winter and rapidly turn green with the advent of the spring season.

The above factors and associated changes in albedo, canopy conductance, aerodynamic properties and atmospheric moisture flux are expected to perturb the regional radiative balance. Therefore, the CERES radiation dataset (2000–2018) covering glacial valleys in the WCH region was utilized to access shortwave (SRF), longwave (LRF) and net radiative forcing (NRF). Multi-year monthly average NRF at both the surface and at the top of the atmosphere (TOA) showed unimodal annual behaviour (Figure 7). NRF at TOA showed a maximum cooling effect (−45 to −55 W m⁻²) during the monsoon months and remained below −25 W m⁻² during the rest of the months. Interestingly, surface NRF from October to May (except monsoon months) indicates a heating effect or a feeble cooling (Figure 7).

Further, we performed seasonal-scale analyses. During the post-monsoon to winter season, NRF at TOA indicated a cooling effect (−5 and −20 W m⁻²). In contrast, surface NRF indicated net heating (1.0–10 W m⁻²). Similarly, during spring or pre-monsoon season, NRF at TOA remained between −10 and −30 W m⁻², while surface NRF indicated a weak cooling (0.6−14 W m⁻²). However, with the advent of the monsoon, both TOA and surface NRF resumed a cooling effect (Figure 7). Annual SRF (at both the surface and TOA) showed unimodal behaviour with a net cooling effect showing peak cooling (−70 to −100 W m⁻²) during the summer monsoon months. Interestingly, the magnitude (−15 and −50 W m⁻²) of TOA and surface SRF remarkably deviated during winter to pre-monsoon seasons with a mean difference of 5–10 W m⁻² (Figure 7). Annually, LRF at both the surface (38 W m⁻²) and TOA (24 W m⁻²) had a heating effect. Across the seasons, the magnitude of surface LRF varied slightly (monsoon: 30 W m⁻²; winter to pre-monsoon: 45 W m⁻²). In contrast, LRF at TOA showed a unimodal behaviour with maximum heating (45–50 W m⁻²) during the monsoon season and comparatively less heating (10–20 W m⁻²) during winter to pre-monsoon seasons. These results, particularly radiative alterations from winter to pre-monsoon, suggest it is highly possible that enhanced winter season humidity coupled with winter warming (Shah et al., 2019; Gaire et al., 2020; Panthi et al., 2021) makes the environmental conditions favourable for vegetation growth in the following growing season (Huang et al., 2022).

These multi-year results indicate that the ongoing changes in vegetation albedo and in ecohydrology deserve future analyses to fully understand the impact of the vegetation changes on local and regional climate. In particular, it would be worth exploring the possibility that the energy increasingly released as latent heat by the expanding vegetation is subsequently advected and released at a higher elevation during condensation, favouring snowmelt (Harder et al., 2017). Additionally, the effects of the increased cloudiness associated with increased evapotranspiration (Duveiller et al., 2018) have to be evaluated in terms of the energy budget.