1 Introduction

Western Himalayas (WH) are one of the most ecologically fragile mountainous regions that have greater concern of climate change vulnerability due to frequent high intensity rain occurrences in the recent decades (e.g., Beniston et al. 1997; Beniston 2003; Teegavarapu 2012). Annual precipitation in this region comes largely during the winter and spring months and is associated with eastward propagating western disturbances (WDs) which are synoptic-scale weather systems commonly originating from the regions over the Mediterranean, Caspian and Black Seas (e.g., Pisharoty and Desai 1956; Mooley 1957; Rao and Srinivasan 1969; Dhar et al. 1984; Dimri 2006; Dimri et al. 2015; Madhura et al. 2015; Cannon et al. 2015), although the genesis locations of WDs can exhibit significant variations (Hunt et al. 2018). The WH also receives considerable rainfall during the summer monsoon months (June through September) associated with synoptic-scale tropical disturbances embedded in the monsoon circulation and their interactions with the equatorward advancing extratropical upper level circulation (e.g., Rao 1976; Bookhagen and Burbank 2010; Saeed et al. 2011; Kulkarni et al. 2013; Vellore et al. 2016; Priya et al. 2016). Past studies have documented the circulation features and evolutionary characteristics of WDs using observational data and model simulations (e.g., Singh et al. 1981; Roy and Bhowmik 2005; Puranik and Karekar 2009; Dimri et al. 2015; Chand and Singh 2015; Madhura et al. 2015; Cannon et al. 2015, 2016; Hunt et al. 2018). Precipitation during the months from December through April (DJFMA) contributes to nearly half of the annual precipitation over the Karakoram and Hindu Kush ranges of the WH (Kapnick et al. 2014; Cannon et al. 2015; Hunt et al. 2018) and is important for the growth of wheat, barley, mustard, gram, lentil, and sesame crops, and also for accumulation of snow which is a primary water reserve for the subsequent dry periods (Lang and Barros 2004; Rees and Collins 2006; Immerzeel et al. 2010; Bolch et al. 2012; Hasson et al. 2013).

Available station observations over the WH region indicate coherent long-term warming trends in wintertime surface air temperatures (e.g., Shrestha et al. 1999; Kothawale and Rupa Kumar 2005; Bhutiyani et al. 2007). On the other hand, there is less spatial coherence of precipitation trends over the WH region (Archer and Fowler 2004; Dimri and Dash 2012). Recent observational studies further suggest that there is an increasing trend in the number of wet days over the WH during the past few decades (Klein Tank et al. 2006; Choi et al. 2009). Cannon et al. (2015) investigated variations in WD activity and their relationship with heavy precipitation in the Karakoram (KH) and Central Himalayas (CH) over the period 1979–2010. They found indications of increasing trend in synoptic scale activity associated with WDs influencing local heavy precipitation over the KH, whereas the CH region experienced a decreasing trend of local heavy precipitation during this period.

Variations in the subtropical westerly jet and the activity of WDs can be modulated by large-scale climate drivers like Madden Julian Oscillation, El Nino/Southern Oscillation, Arctic Oscillation, North Atlantic Oscillation and Pacific Decadal Oscillation (eg., Barlow et al. 2005; Yadav et al. 2009; Kar and Rana 2014; Ahmad et al. 2014; Cannon et al. 2015, 2017; Krishnan and Sugi 2003). It is not yet clear whether the increasing trend of synoptic scale WD activity and associated heavy precipitation over the KH, as noted by Cannon et al. (2015), is linked to global climate change. A recent study by Madhura et al. (2015) reported a significant rise in the observed winter-to-early spring extreme precipitationFootnote 1 events over the WH region in the past few decades, which appears to have connections to amplitude variations of WD activity. Their inferences broadly indicated a possible linkage between the increased WD variability and enhanced baroclinicity of the background upper level westerly winds arising from the elevation dependent climate warming signal over the Tibetan Plateau.

A related scientific problem pertaining to the WH region, the so-called “Karakoram anomaly”, has additionally come to the focus of investigations in recent times (e.g., Hewitt 2005; Gardelle et al. 2012; Kapnick et al. 2014; Kääb et al. 2015; Forsythe et al. 2017). The aforementioned studies suggest that regions in the Karakoram Himalayas have experienced an exceptionally stable glacier behavior in recent decades, as opposed to glacier shrinkage observed in many other places. Climate model simulations indicate that changes in non-monsoonal wintertime frozen precipitation over the Karakoram Himalayas appear to shield this region from significant glacier thickness losses under warming climate (Kapnick et al. 2014; Kääb et al. 2015). In a recent study, Norris et al. (2018) have examined the contrasting climatic trends between the Central and Karakoram Himalayas based on 36 years (1979–2015) of dynamically downscaled simulations using the Weather Research and Forecasting (WRF) model. Their results indicate that the dominance of an anti-cyclonic trend over the Central Himalaya, both during winter and summer, has led to significant warming and reduced snowfall in recent years, whereas the Karakoram region in the western part of High Asia has been influenced by a cyclonic trend with no warming. While reliable assessment of changes in mass balance of the Himalayan glaciers remains a more challenging problem because of the complex regional atmospheric circulation dynamics and feedbacks (Forsythe et al. 2017), it is rather feasible to investigate the influence of climate change on the synoptic scale WD activity and precipitation over the Western Himalayas (Ridley et al. 2013). With the above perspectives in view, the present study is aimed at understanding the linkage between changes in WD activity and non-monsoonal (DJFMA) precipitation over the WH in a changing climate. In climate simulations, certain degree of realism is indispensable for better confidence on the outcomes. In this study, a verification of simulated variables is performed with the available observations/circulation products to envisage that degree of realism required for interpreting the plausible dynamical linkages between changes in WD activity and precipitation over the WH. The analysis performed in this study uses observed and reanalysis products and high-resolution climate model simulations, which are described in the following section.

2 Datasets and model setup

Datasets used in the present study include the gridded (0.5° × 0.5°) observed precipitation products from the Asian Precipitation Highly Resolved Observational Data Integration Towards Evaluation of water resources (APHRODITE) for the period 1951–2007 (Yatagai et al. 2012); daily precipitation totals derived from the Tropical Rainfall Measuring Mission (TRMM) 3B42 V7 based on TRMM Multi-Satellite Precipitation Analysis (TMPA) for the period (1998–2016) at a horizontal resolution of 0.25° longitude × 0.25° latitude (Ref: https://pmm.nasa.gov/data-access/downloads/trmm). We also use the observed gridded (0.5° × 0.5°) monthly surface temperature dataset (http://badc.nerc.ac.uk/view/badc.nerc.ac.uk) from Climate Research Unit (CRU) during 1951–2005 and reanalysis circulation products from the European Centre for Medium-Range Weather Forecasts (ECMWF, ERA-40) for the period 1958–2002 and ERA-Interim Reanalysis for the period 1979–2015 (Dee et al. 2011). A brief description of the model is given in the following.

2.1 Model description and numerical experiments

The Laboratoire Meteorologie Dynamique Zoom General Circulation Model (LMDZ4 GCM) used in this study is the atmospheric component of the Institut Pierre Simon Laplace (IPSL) coupled model (Hourdin et al. 2006; Dufresne et al. 2013). The latitude–longitude grid configuration of the LMDZ4 model is stretchable over the entire globe, so that it can be used for global climate studies as well as for regional climate investigations adapted with finer horizontal grid resolution. The Z in LMDZ stands for zoom, which allows the feasibility to use the model as a variable resolution GCM with much finer grid resolutions over regions that appear to have potential impacts of climate change, for example, the regional climate aspects over the South Asian monsoon region (e.g., Ramarao et al. 2015; Krishnan et al. 2016). The telescopically zoomed grid configuration (grid size ~ 35 km) employed in LMDZ4 demonstrated realistic simulations of the Indian summer monsoon (Sabin et al. 2013). The horizontal grid configuration consists of 360 × 180 grid points in longitude and latitude directions, respectively and the model has 19 layers in the vertical. The grid configuration and the model topography for the Himalayan region are shown in Fig. 1. The moist convective processes are parameterized following the Emanuel (1993) scheme. For more details about representation of physical processes in the LMDZ4 GCM, the reader is referred to Hourdin et al. (2006), Dufresne et al. (2013) and Sabin et al. (2013).

Fig. 1
figure 1

a The LMDZ4 model horizontal grid configuration and model topography (shaded; units in meters) over South, West and Southeast Asia and adjoining regions. Also shown is the domain of the Western Himalayas (WH, 70°E–80°E and 32°N–39°N), as referenced in the text. The boxes GP1 (58°E–62°E, 32°N–36°N) and GP2 (75°E–78°E, 28°N–30°N), which are chosen following Cannon et al. (2015), correspond to centers of daily anomalies of geopotential height at 200 hPa influencing the lag-0 precipitation over the Western and Central Himalayas, respectively. b Topographic relief map for the box region shown in a. The Hindu Kush and Karakoram and Central Himalayas (CH) are indicated in the figure. Cross-wire marks the location of the Indian state of Jammu and Kashmir

Full size image

This study uses two long simulations of the LMDZ4 model, which have been conducted for the period 1886–2005. The first one includes both natural forcing (i.e., varying solar irradiance, volcanic effects) and anthropogenic forcing due to increasing concentration of greenhouse gases (GHGs), aerosol emission, and changes in land use and land cover and is referred to as the “historical run”—HIST. The second experiment includes natural forcing only and is referred to as “historical-natural”—HISTNAT. In the present study, we have analyzed the HIST and HISTNAT simulations for the period 1901–2005. In addition to these two simulations, a future projection experiment is also performed for the period 2006–2095 following the Representative Concentration Pathway 4.5 (RCP4.5) scenario with a radiative forcing of + 4.5 W m−2 at the end of twenty-first century (see Dufresne et al. 2013 for more details). In the RCP4.5 scenario, CO2 concentration stabilizes at 543 ppmv in 2150, while global emissions of aerosols and their precursors are projected to decrease through the twenty-first century (Westervelt et al. 2015) and land use changes incorporate reforestation during the twenty-first century (Davies-Barnard et al. 2014). The model is driven by prescribed sea surface temperature (SST) lower-boundary conditions derived from the IPSL-CM5 coupled-model, employed for climate change simulations documented in the 4th Intergovernmental Panel on Climate Change assessment report (IPCC 2007). A bias correction is applied for the SST and sea-ice fields by superposing the SST/sea-ice anomalies obtained from the IPSL-CM5A-LR (LR stands for low resolution; latitude–longitude grid sizes 1.875° × 3.75°) simulations on the observed climatological mean SST/sea-ice (see Krishnan et al. 2016 for details). Precipitation events over the WH associated with the activity of WDs, embedded in the upper-level subtropical westerlies, are prevalent during the winter and early spring months (eg., Cannon et al. 2016; Hunt et al. 2018). Following this reasoning, observations and LMDZ4 model simulations (HIST, HISTNAT and RCP4.5) valid for the DJFMA months are considered for the analysis in this study.

3 Analysis of observed and simulated climatic variations over WH

Here we present an analysis of the time-mean features and synoptic-scale variability associated with WDs, using the observed and reanalysis datasets, as well as the LMDZ4 HIST simulations. In particular, we focus on evaluating the fidelity of the model simulation (HIST) in representing the observed aspects of time-mean circulation, precipitation distribution over the WH and characteristics of synoptic-scale WDs during the DJFMA months, as will be discussed in the following subsections. Since the observed and reanalysis datasets are mostly available during the post-1950s, the spatial analysis of the HIST and HISTNAT simulations is presented for the period 1951–2005. The aim of this analysis is to basically comprehend the realism of the HIST simulation to be able to capture the dynamical linkage between the synoptic-scale activity of WDs and heavy precipitation occurrences over WH, which in turn is essential for reliable assessments of future climate change projections in the region.

3.1 Mean temperature, precipitation and circulation during DJFMA

A comparison of the observed and simulated (HIST) mean DJFMA climate for the period 1951–2005 is shown in Fig. 2. The overall spatial distribution of mean surface temperatures in the observations (CRU) and HIST simulation show values < − 5 °C over the high-elevations of WH and warmer temperatures (> + 10 °C) over the Indian landmass south of the Himalayan range (Fig. 2a, b). While the simulated temperatures exhibit significant cold bias along the high-elevated southern slopes of the Himalayas, the bias is rather smaller over of the plains of the Indian subcontinent. The observed (APHRODITE) and simulated (HIST) spatial patterns of mean precipitation indicate a northwest–southeast oriented band of precipitation (Fig. 2c, d) along the entire stretch of the Himalayan topography. The use of high-resolution (grid size < 35 km) is particularly important for capturing precipitation features in regions of narrow orography and steep topographic gradients (Sabin et al. 2013). Note that the precipitation amounts in the WH are relatively higher in the HIST simulation as compared to the APHRODITE dataset. It is also noteworthy to mention that the simulated wet bias is in part likely due to inadequate gauge observations over the WH mountainous region and their representation in gridded precipitation datasets (e.g., Kim et al. 2015), although a similar wet bias has also been noted over this region from earlier modeling investigations (see Dimri et al. 2013). The HIST simulated annual cycle of climatological mean precipitation over the WH region shows dominant contributions from the DJFMA months, with an overestimation as high as 1 mm day−1 compared to the APRHODITE dataset (Fig. 2e). The corresponding HIST simulated surface air temperature averaged over the WH region shows a cold bias relative to observations (CRU, Fig. 2e). Furthermore, it is interesting to note that a large fraction (about 80% or more) of the simulated total precipitation during the DJFMA months comes from frozen hydrometeors (Fig. 2f), while it is mostly of liquid type during the summer monsoon JJAS months over the WH region.

Fig. 2
figure 2

Spatial maps of DJFMA climatological mean. Surface air temperature (°C) (a, b) and precipitation (mm day−1) (c, d) from observations (CRU and APHRODITE) and model simulation (HIST). Also included are the climatological annual cycles of e observed and simulated surface air temperature and precipitation averaged over the WH and f percentage fractions of snowfall and rainfall with respect to the total precipitation over the WH from HIST

Full size image

The activity of WDs during winter and early spring significantly depend on the position and intensity of the mean sub-tropical westerly jet (Cannon et al. 2015; Hunt et al. 2018). Figure 3a, b shows the mean horizontal winds at 200 hPa for the DJFMA season from the ERA-interim reanalysis and HIST simulation respectively. It can be seen that the mean latitudinal position of the axis of the subtropical westerly jet, which is located to the north of 25°N in the longitude band between 10°E and 100°E, is generally captured in the simulation, while the magnitude of winds averaged over the jet core (60°E–90°E, 25°N–35°N) is overestimated in the HIST simulation by about 1.8 ms−1. An intensification of the upper-level subtropical/mid-latitude westerly flow to the east of 120°E and northward shift of the jet core over eastern China, Korea and Japan is also seen in the simulation.

Fig. 3
figure 3

Spatial map of DJFMA climatological mean horizontal winds (ms−1) at 200 hPa. a ERA-Interim, b HIST. Shading denotes wind speed (ms−1)

Full size image

3.2 Characteristics of synoptic-scale activity associated with WDs

Spectral characteristics of upper-level subtropical westerly jet provide useful information to infer synoptic-scale variations associated with the WDs. We applied power spectrum analysis on daily time-series of 200 hPa zonal winds averaged over the region 40°E–80°E, 25°N–35°N for the ERA-Interim and HIST simulation (Fig. 4a, b). Significant synoptic variance with time-scales in the range of 4–10 days are evident in both the ERA-Interim and HIST datasets, although the peaks within the synoptic spectrum show subtle differences between the two datasets (Fig. 4). It can be noticed that the synoptic scale variations, both in ERA-Interim and HIST, clearly exceed the red-noise background variance. This indicates that the HIST simulation captures broadly the observed synoptic-scale variations of the upper-level subtropical westerly jet. Additionally, one can notice in Fig. 4 significant low-frequency variations with time-scales greater than 10 days. Dimri (2013) emphasized the importance of intra-seasonal variations (time-scale ~ 16 days) of winter precipitation over the WH. Cannon et al. (2015) have suggested the role of nonlinear interactions among different modes of variability in modulating the WD activity over the Himalayan region.

Fig. 4
figure 4

Spectral variance (m2 s−2) of daily zonal winds at 200 hPa averaged over the region (40°E–80°E, 25°N–35°N) of the subtropical westerly jet. a ERA-Interim, b HIST. The blue and green lines correspond to the lower and upper confidence bounds of the spectral variance. The red line is the red noise spectrum

Full size image

Cannon et al. (2015) examined the relationship between WDs and heavy precipitation in the Karakoram Western (WH) and Central Himalayas (CH) using lag correlations between heavy precipitation and 200 hPa upper-level geopotential height (H200). They noted that the H200 anomaly center associated with the lag-0 of heavy precipitation for the WH is around the region (58°E–62°E, 32°N–36°N) located ~ 15° to the west of the Karakoram; whereas the center of negative H200 anomalies associated with the lag-0 of heavy precipitation for the CH is around the region (75°E–78°E, 28°N–30°N). For further details, the reader is referred to Cannon et al. (2015), in particular Figs. 4, 5, 6, 7 and 8 of their paper. Using a similar approach, we investigated the relationship between WDs and heavy precipitation over the WH and CH from the HIST simulation. Figure 5 shows the time-series of H200 anomalies averaged over the boxes GP1 (58°E–62°E, 32°N–36°N) and GP2 (75°E–78°E, 28°N–30°N) based on the ERA-Interim dataset and HIST simulation. The GP1 and GP2 boxes basically represent centers of action (H200 anomaly) influencing lag-0 precipitation over the Western and Central Himalayas respectively. It is noted that the standard deviations of the H200 anomaly fluctuations are relatively higher for GP1 [114 (155) gpm in ERA-Interim (HIST) respectively] relative to GP2 [101 (125) gpm in ERA-Interim (HIST) respectively]. It is also noted that the H200 fluctuations in the HIST experiment have higher amplitudes compared to the ERA-Interim. Spectral variance of the time-series of H200 anomalies averaged over GP1 and GP2 are shown for the ERA-Interim and HIST simulation in Fig. 6. The broad spectral range of significant synoptic variance for GP1 and GP2 is between 3.3 and 10 days and exceeds the red-noise background variance. While it is realized that there are some differences in the variance within the synoptic spectrum between the two datasets, the important point is that the HIST simulation broadly reproduces the synoptic-scale variations around the centers of action (viz., GP1 and GP2).

Fig. 5
figure 5

Daily time-series of H200 anomalies (gpm) averaged over GP1 and GP2 respectively. a, b ERA-Interim, c, d HIST

Full size image
Fig. 6
figure 6

Spectral variance of daily H200 [(gpm)2] anomaly time-series averaged over GP1 and GP2 respectively. a, b ERA-Interim, c, d HIST. The blue and green lines correspond to the lower and upper confidence bounds of the spectral variance. The red line is the red noise spectrum

Full size image
Fig. 7
figure 7

Composites of lag-0 precipitation anomalies (mm day−1) over the WH and CH, associated with synoptic-scale WDs, constructed by taking the difference between the 15th percentile and 85th percentile of the 4–15 days band-pass filtered time-series of GP1 (for WH) and GP2 (for CH), respectively. This approach is similar to Cannon et al. (2015). a–c Anomaly composites [WH, CH and difference (WH − CH)] based on the APHRODITE precipitation dataset. d–f Same as a–c except for HIST

Full size image
Fig. 8
figure 8

a Spatial map of trend in daily precipitation (DJFMA) exceeding the 90th percentile for the period 1951–2007 based on the APHRODITE dataset. The units are mm (57 year)−1. Dotted areas are trends exceeding the 95% confidence level based on the Mann–Kendall test. b Time-series showing interannual variations of daily precipitation (DJFMA) exceeding the 90th percentile over the Karakoram WH (70°E–80°E, 32°N–39°N) from the APRHODITE (1951–2007) and TRMM (1998–2015) datasets

Full size image

We applied a 4–15 days band-pass filter on the GP1 and GP2 indices to retrieve the synoptic scale variability in the H200 anomalies. The 4–15 days filtered time-series of GP1 and GP2 are shown in the Auxiliary Figure A1. To understand the connection between extreme precipitation and the synoptic variability of WDs, we constructed composites of precipitation anomalies based on the difference (15th percentile minus 85th percentile) of the synoptic time-series GP1 and GP2. The spatial maps of precipitation anomalies associated with the GP1 index for the APHRODITE and HIST datasets are shown in Fig. 7a, b, respectively. The corresponding maps for the GP2 index are shown in Fig. 7c, d and the difference maps (GP1 minus GP2) are shown in Fig. 7e, f. It can be noticed that the composite anomaly pattern, based on the percentile difference (15th minus 85th) of the GP1 index, shows precipitation enhancements over the WH. For the anomaly pattern based on the GP2 index, precipitation enhancements are seen over the Central Himalayas, with slight westward extensions into the Jammu-Kashmir region. Also note that the HIST simulation reproduces the precipitation patterns associated with the GP1 and GP2 indices (Fig. 7c, d). The (WH minus CH) difference maps of precipitation anomalies indicate accompaniment of excess precipitation over the Hindu Kush-Karakoram region with deficit precipitation over the Central Himalayas, which is captured well in the HIST simulation (Fig. 7e, f) and consistent with Cannon et al. (2015). Having seen the promising capability of the HIST simulation in bringing out the observed characteristics of WDs and their linkage to DJFMA precipitation variability in the Himalayan region, the long-term variability of WDs and regional precipitation will be explored in the next section.

3.3 Long-term variability in extreme precipitation and WD synoptic activity

Figure 8a shows the spatial pattern of trend in daily DJFMA precipitation extremes (i.e., daily precipitation amount exceeding the 90th percentile) during 1951–2007 based on the APHRODITE data. A notable feature in Fig. 8a is the significant increasing trend [~ 6 mm (57 year)−1] of extreme precipitation over the WH region. One can also see a negative trend of precipitation extremes over the CH. This opposite polarity of extreme precipitation trends is consistent with observed trends in synoptic activity over the two regions as reported by Cannon et al. (2015). The time-series of interannual variation of daily precipitation extremes (ie., precipitation exceeding the 90th percentile) over the region (70E–80E; 32N–39N) in the WH based on APHRODITE (Fig. 8b) shows a significant increasing trend during the period (1951–2007). Increasing trend in the number of wet days has been observed at some stations in the WH region by earlier studies (eg., Klein Tank et al. 2006; Choi et al. 2009). In taking note of the trends in daily precipitation extremes (Fig. 8), we also realize that changes in observational quality, sparseness and intermittence of the number of stations used in gridded datasets can affect long-term trends. Keeping this in mind, we additionally examined the interannual variations of daily precipitation extremes over the WH region using the TRMM dataset for the period (1998–2016), which also displays a significant rising trend (Fig. 8b). It must be mentioned that the APHRODITE and TRMM datasets of daily precipitation show a high correlation (~ 0.6) over the WH region for the period 1998–2007 (figure not shown), which is consistent with Cannon et al. (2015).

Further corroborations to precipitation enhancement over the WH region and increasing trends of snow cover over the Indus basin during 2000–2011 has been reported by Singh et al. (2014) based on analysis of the Moderate-resolution Imaging Spectroradiometer (MODIS) product. On the other hand, it has been noted that the Ganga and Brahmaputra basins indicate a slight decline of snow cover during the same period (Singh et al. 2014). Increasing trends in the areal extent of snow-cover over the Indus and Chenab river basins of the WH are also evidenced in the high-resolution observations from the Indian Remote Sensing (IRS) Resourcesat-1 and 2 AWiFS data for the period 2004–2014 (SAC 2016). Additionally, tree-ring based hydroclimate reconstructions extending back to several centuries in the past, both from the high-mountain Karakoram areas in northern Pakistan (Treydte et al. 2006) as well as the north Indian region of Jammu and Kashmir (Yadav et al. 2017), also substantiate that the most wettest periods over the WH have occurred during the recent few decades. We also examined year-to-year variations in the synoptic scale activity based on the daily H200 anomalies corresponding to the GP1 index. We computed standard-deviations of the daily time-series of the 4–15 days band-pass filtered GP1 index for each of the DJFMA seasons using the ERA-40 (1958–2002) and ERA-Interim (1979–2015) datasets (Fig. 9). One can notice that the interannual variations of the time-series in Fig. 9 are accompanied by a significant positive trend in the standard-deviation of the GP1 index, which also lends support pointing to an increasing tendency in the amplitude variations of synoptic scale WD activity.

Fig. 9
figure 9

Time-series of standard-deviation of daily filtered (4–15 days band-pass) GP1 index (gpm) computed for every DJFMA season using H200 anomalies from ERA-40 (1958–2002) and ERA-Interim (1979–2015) datasets. Note the increasing trend of amplitude variations associated with the synoptic-scale WD activity

Full size image

4 Climate change and non-monsoonal precipitation response over the WH

The preceding discussions focused on the increasing trend of daily precipitation extremes over the WH and enhanced amplitude variations of synoptic-scale WD activity during the recent few decades, although it is not yet clear if these changes are linked to global climate change. Motivated by the ability of the HIST simulation in representing the broad synoptic-scale characteristics of WDs, here we examine long-term variations in synoptic-scale WD activity and associated precipitation extremes over the WH under climate change.

4.1 Comparison of time-mean surface response between HIST and HISNAT

Figure 10 shows difference maps of the DJFMA time-mean (1951–2005) surface temperature, precipitation, snowfall and rainfall between the HIST and HISTNAT experiments. The simulated surface temperature response to climate change is evident from the significant warming (> 1 °C) seen over the Indian landmass, Indo-China, and the Tibetan Plateau. Note that the magnitude of simulated warming is < 0.6 °C over the Hindu Kush and Karakoram Himalayas (HKH) (Fig. 10a). The HIST minus HISTNAT difference map of precipitation shows prominent increase in precipitation over the Karakoram and Western Himalayas and decrease over the Central Himalayas and the Indo-Gangetic plains (Fig. 10b). It is interesting to note that much of the precipitation enhancement over the Karakoram WH is largely due to increased snowfall (frozen precipitation, see Fig. 10c), with relatively modest changes in liquid precipitation over the region (Fig. 10d). While taking note of the statistically significant increase in the simulated snowfall over the Karakoram WH (Fig. 10c), we also recognize that the systematic cold bias of the model can affect the simulated snowfall. Results from a recent climate modeling study by Kapnick et al. (2014) suggest that non-monsoonal winter precipitation response under climate warming can protect the Karakoram glaciers from snowmelt. Norris et al. (2015) analyzed the influence of synoptic-scale flow patterns associated with two extreme winter snowfall events in January 1999 and March 2006 over the WH and CH, respectively using the WRF model simulations. They noted that the extreme snowfall over WH in January 1999 was driven by terrain-locking of long-wave trough over West Asia, which resulted in persistent snowfall at various locations of the WH. On the other hand, the March 2006 extreme snowfall event over CH was more widespread spatially and was accompanied by an amplification of trough further south and east along the Himalayan ridge, thereby pointing to the strong internal dynamics of the regional extreme snowfall.

Fig. 10
figure 10

Difference maps of the DJFMA time-mean (1951–2005) fields between HIST and HISTNAT. a Surface temperature (°C). b Total precipitation (mm day−1). c Snowfall (fraction of total precipitation). d Rainfall (% of total precipitation). Hatched areas represent differences exceeding the 95% confidence level based on a two-tailed student’s t test

Full size image

4.2 Daily precipitation extremes and WD activity

Figure 11a shows the spatial pattern of trend in the daily DJFMA precipitation extremes (ie., precipitation exceeding the 90th percentile) during 1951–2005 based on the HIST simulation. A clear increasing trend of precipitation extremes can be noted in the simulation over the WH, whereas areas over eastern Nepal and CH show negative trends. Figure 11b shows the time-series of interannual variation of daily precipitation extremes over the WH (70E–80E; 32N–39N) based on the HIST and HISTNAT simulations during (1951–2005); and also the future projection for the period (2006–2095) following the RCP4.5 scenario. It can be seen that the HIST simulation of daily precipitation extremes over the WH exhibits a clear positive trend, whereas the HISTNAT simulation for the same period shows no detectable long-term trend. The future projection also generally indicates an upward trend with smaller increase of precipitation extremes over the WH until the mid-twenty-first century and a rapid rise thereafter in association with enhanced global warming under the RCP4.5 scenario. From the foregoing analysis, it can be inferred that the increasing trend of precipitation extremes in the WH is largely attributable to human-induced climate change. One can also note that the time-series in Fig. 11b exhibits multi-decadal variations of precipitation extremes, in addition to long-term trends. This point will be discussed further in the next section.

Fig. 11
figure 11

Spatial map of trend in daily precipitation (DJFMA) exceeding the 90th percentile. a HIST experiment for the period 1951–2005 [units: mm (55 year)−1]. Dotted areas represent trends exceeding the 95% confidence level based on the Mann–Kendall test. b Time-series showing interannual variations of daily precipitation (DJFMA) exceeding the 90th percentile over the WH (70°–80°E, 32°N–39°N) from the HIST (brown), HISTNAT (blue) and RCP4.5 (red) experiments. A 5-year moving average has been applied on the time-series

Full size image

Additionally, we performed an analysis of interannual variations of synoptic scale WD activity from the model experiments, which allows us to interpret the increasing trend of simulated precipitation extremes over the WH region. Figure 12 shows the time-series of the year-to-year standard-deviations of GP1 index for the HIST and HISTNAT experiments and the RCP4.5 future projection. A clear rising trend in the standard-deviation of the GP1 index can be noted both in HIST and RCP4.5, whereas the trend is insignificant in the case of HISTNAT. The magnitudes of the trends during the period 1951–2005 are found to be [+ 7.5 m (55 year)−1] in HIST and [− 1.75 m (55 year)−1] in HISTNAT, respectively. The corresponding trend for the RCP4.5 projection during 2006–2095 is found to be [+ 10 m (90 year)−1]. These results essentially point to the importance of climate change in enhancing the amplitude variations of the synoptic-scale WD activity. Furthermore, it is interesting to note that the long-term trends are accompanied by a superposition of multi-decadal variations of the GP1 index. The underlying mechanisms for these multi-decadal variations are not clear. Cannon et al. (2015) noted that the activity of WDs can undergo multi-annual variability due to complex teleconnections and non-linear interactions among various modes of climate variability. From Figs. 11 and 12, it can be deduced that the rising trend of precipitation extremes over the WH in HIST and RCP4.5 emerges as a consequence of enhanced amplitude variations of the synoptic-scale WD activity. In the following section, we shall discuss a plausible mechanism through which climate change can modulate the synoptic scale WD activity.

Fig. 12
figure 12

Time-series of standard-deviation of daily filtered (4–15 days band-pass) GP1 index (gpm) computed for every DJFMA season using H200 anomalies from the HIST (orange), HISTNAT (blue) and RCP4.5 (red) experiments. A 5-year moving average has been applied on the time-series

Full size image

4.3 Background upper-level circulation response under changing climate

The difference map of simulated mid-tropospheric circulation and temperature between HIST and HISTNAT is shown in Fig. 13. It is interesting to note that the difference map exhibits a well-defined large-scale stationary-wave pattern over the region poleward of 25°N, and a less wavy response over the low-latitudes (south of 20°N). Note that the mid-latitude pattern comprises of an anticyclonic anomaly centered around (110°E, 45°N) and a deep southward intruding trough on the western side of the Tibetan Plateau, with trough axis approximately located at 55°E. This circulation configuration directs strong southerly flow to the WH from the Arabian Sea. Notice also that the mid-tropospheric temperature difference (HIST minus HISTNAT) shows strong warming ~ 1.5 °C centered around the anticyclonic anomaly, as compared to the weaker temperature anomaly (< 0.5 °C) associated with the trough. The zonal mid-tropospheric temperature gradient between the anticyclonic and cyclonic anomalies, seen over the eastern and western flanks of the Tibetan Plateau, is over 2 °C which suggests that the large-scale mid-latitude stationary-wave pattern is consistent with thermal-wind equilibrium. A warm core anticyclonic anomaly is also evident over the Mediterranean sector and a cyclonic anomaly located to the east of Japan, which are indicative of the circumglobal feature of the mid-latitude stationary-wave pattern (Fig. 13).

Fig. 13
figure 13

Difference map of time-mean (1951–2005) simulated mid-tropospheric circulation (ms−1) and temperature (°C) between HIST and HISTNAT

Full size image

Atmospheric stationary waves are generated through zonal asymmetries in orography, land–ocean contrasts and SST gradients. Nigam and DeWeaver (2003) explored the NH winter-time stationary wave response to the combined and individual influences of heating, mountains and transient fluxes of heat and momentum. They noted that the wintertime circulation response at 300 hPa to heating largely influences the upper-level jets over the Asia-Pacific and Atlantic sectors, while the response to mountains showed anomalous height excess over the Himalayan–Tibetan complex and height deficits (negative anomalies) downstream over the region of Japan (see their Fig. 13c). The mid-latitude stationary pattern in Fig. 13 bears resemblance with the orographically forced wintertime stationary wave patterns discussed by Nigam and DeWeaver (2003). Furthermore, it is seen that the mid-tropospheric stationary-wave response in the future projection (RCP4.5, figure not shown) is rather similar in terms of the spatial pattern of response shown in Fig. 13, but for larger magnitudes of mid-tropospheric warming in the future. Our understanding suggests that the changes in the background mean circulation shown in Fig. 13, in response to climate change, can potentially favor enhanced amplitude variations of WD activity. In particular, the quasi-stationary southward intruding trough-like feature to the west of the Tibetan Plateau appears to promote baroclinicity of the mean upper-level westerlies to support the development of large-amplitude synoptic-scale WDs. The WDs can be further anchored as topographic Rossby waves on the windward side of the WH (see Hara et al. 2004), so as to direct moist southerly flow from the Arabian Sea to enhance orographic precipitation (Pisharoty and Desai 1956; Cannon et al. 2015; Madhura et al. 2015). Terrain-locking of troughs over West Asia is also known to be an important mechanism for producing extreme snowfall over WH (Norris et al. 2015, 2018). The composite anomaly pattern of winds and temperature at 500 hPa, based on the percentile difference (15th minus 85th) of the GP1 index, shows an intensified cyclonic anomaly over West Asia with strong southerly flow directed towards the WH (Fig. 14a, b). On the other hand, the downstream ridge-like feature around 80°E tends to inhibit the Central Himalayan precipitation. It is noted that the HIST simulated synoptic-scale trough-ridge features generally concur with the ERA-Interim anomalies, albeit a slight tilt in the simulated structure. Furthermore, increase of precipitable water along the southerly flow into the WH is also noted both in the simulation and ERA-Interim reanalysis (Fig. 14c, d). While an overestimation of precipitable water is noted in the HIST simulation over the east coast of China and West Pacific, the simulated precipitable water anomalies extending from the Arabian Sea towards the WH compare well with the ERA-Interim reanalysis.

Fig. 14
figure 14

Composites of lag-0 anomalies of circulation and moisture, associated with synoptic-scale WDs, constructed by taking the difference between the 15th and 85th percentiles of the 4–15 days band-pass filtered time-series of GP1 from ERA-Interim (top row) and HIST simulation (bottom row). a, b Wind (ms−1) and temperature (°C) anomalies at 500 hPa. c, d Precipitable water (kg m−2)

Full size image

5 Elevation dependent warming of the Tibetan Plateau and links to WD activity

One of the noteworthy features of global warming is the elevation dependency of the climate warming signal characterized by pronounced warming at higher elevation mountain environment relative to lower elevations (eg., Beniston and Rebetez 1996; Diaz and Bradley 1997; Pepin et al. 2015). The Tibetan Plateau, which stretches over a vast area of ~ 2.5 × 106 km2 with an average elevation of nearly 4500 m, is one of most sensitive regions to global climate change (Liu and Chen 2000; Liu et al. 2009). Several high elevation stations (> 4000 m a.s.l) over the Tibetan Plateau experienced rapid warming in annual mean temperatures during 1961–2012 by about 0.36 °C decade−1, while low elevation sites (< 500 m a.s.l) have observed a slower warming rate ~ 0.05 °C decade−1 (see Liu and Yan 2017 and references therein). In particular, stronger warming of about 0.47 °C decade−1 has been observed at high elevations during the boreal winter (Liu and Yan 2017). While the elevation-dependent warming (EDW) of the Tibetan Plateau involves multiple interactive feedbacks, including snow-albedo interactions, water vapor-cloud-radiation interactions, surface fluxes, aerosol-forcing (Pepin et al. 2015), a comprehensive and integrated understanding of the mechanisms of EDW over the complex terrain of the Tibetan Plateau remains a major challenge and warrants further research.

In the context of EDW, we examined the temperature response to climate change at higher elevations of the Tibetan Plateau from the model simulations. Figure 15a shows the difference map (HIST minus HISTNAT) of surface temperature for elevations > 3 km a.s.l. We notice a strong warming (> 1.14 °C) signal over the large area of the eastern Tibetan Plateau. On the other hand, the simulated warming is relatively less (~ 0.56 °C) in the western side, which was earlier noted in Fig. 10a. The east–west gradient of surface warming response across the Tibetan Plateau is also prominently seen in the future RCP4.5 projection (Fig. 15b). To comprehend the connection between the strong warming of the eastern Tibetan Plateau and the associated mid-tropospheric warming (Fig. 13), we examined trends in the thickness of atmospheric layers using the hypsometric equation (Holton 2004). Figure 16 shows the thickness trends for different layers from the HIST and HISTNAT simulations. The trends were calculated for the period 1951–2005 using DJFMA thickness values averaged over the eastern part of the Tibetan Plateau (90°E–120°E, 30°N–38°N) shown by red bars. For the HIST experiment, the thickness layers show significant positive trends from the surface up to the upper troposphere, with maximum trend around 400 hPa. We also calculated the trends of thickness layers over the region to the west of the Tibetan Plateau (40°E–70°E, 30°N–38°N), which are shown by blue bars in Fig. 16. The thickness layers in HIST show relatively smaller positive trends over the western side as compared to those on the eastern side. The HISTNAT experiment generally does not reveal significant thickness trends both over the eastern and western flanks of the Tibetan Plateau. The analysis in Fig. 16 suggests that the climate warming response over the Tibetan Plateau need not be restricted to the surface alone, but can also extend vertically into the middle and upper levels.

Fig. 15
figure 15

Difference map of time-mean simulated surface temperature (°C) for mountainous elevations > 3 km. a [HIST–HISTNAT]. b [RCP4.5–HISTNAT]

Full size image
Fig. 16
figure 16

Linear trends of different thickness layers from the HIST and HISTNAT simulations for the period 1951–2005. The units are m (55 year)−1. Red bars indicate trends calculated using the DJFMA layer thickness values averaged over the eastern Tibetan Plateau (90°E–120°E, 30°N–38°N) from the HIST experiment. Blue bars correspond to trends of thickness values averaged over the region to the west of the Tibetan Plateau (40°E–70°E, 30°N–38°N) from the HIST experiment. The corresponding trends from the HISTNAT experiment are shown by blue and red circles respectively

Full size image

Madhura et al. (2015) indicated that the EDW over the Tibetan Plateau can induce significant mid-tropospheric warming gradients which can enhance the baroclinicity of the background westerlies. While this possibility is supported in the HIST experiment, it is interesting to note that the mid-tropospheric thickness layers over the eastern side of the Tibetan Plateau tend to increase at a significantly faster rate as compared to their western counterpart (Fig. 16), thus giving rise to a positive zonal gradient in the geopotential height field and a thermal balance of the wind field (Fig. 13). It is suggested that this backdrop of large-scale stationary-wave pattern, with increased baroclinicity and westerly trough located to the west of the Tibetan Plateau, can be conducive for synoptic-scale WD passages to undergo large amplitude modulations. The linkage between extreme precipitation over the WH and large-amplitude synoptic-scale troughs associated with WDs over West Asia, indeed is well captured in the high-resolution HIST simulation (see Figs. 7a, d, 14).

6 Summary and discussions

Winter and early spring precipitation over the Western Himalayas (WH) is vital for sustaining the regional water resources, agriculture and the mountainous ecosystems. Major part of this regional precipitation is produced by eastward propagating synoptic-scale western disturbances (WD) originating from the Mediterranean. Precipitation observations over the last three decades indicate a rising trend in daily precipitation extremes over the WH and having associations with synoptic-scale WD activity (Cannon et al. 2015; Madhura et al. 2015). While recognizing that sparseness and intermittence of the number of stations used in gridded precipitation datasets can affect long-term trends in precipitation extremes over the WH region, it is noteworthy to mention that atmospheric reanalysis datasets provide corroborative support pointing to an increasing tendency in the amplitude variations of the synoptic scale WD activity. Whether the aforementioned WH regional changes during the recent decades are attributable to climate change, however, has remained hitherto unclear. This scientific issue has been examined in the present study using three long-term climate simulations from a global variable-grid model (LMDZ4) with finer horizontal grid resolution (~ 35 km grid) telescopic zooming over the South Asia. The first two experiments (HIST and HISTNAT) cover the twentieth century period 1900–2005. The HIST experiment includes both natural and anthropogenic forcing, whereas the HISNAT includes only natural forcing. The third experiment is performed in continuation with HIST into the twenty-first century period 2006–2095, following the Representative Concentration Pathway 4.5 (RCP4.5) scenario.

Realism of winter precipitation simulations over the WH warrants finer representation of orography in climate models, which has been a major concern in understanding the Himalayan regional precipitation response to climate change using coarse resolution models (grid size > 200 km) based on the Coupled Model Intercomparison Project (Palazzi et al. 2013). In this regard, the use of high-resolution (grid size < 35 km) LMDZ4 simulation has been particularly advantageous for the present study to better comprehend, not only the regional aspects of precipitation variability over the Himalayas but also the linkages between the synoptic-scale WDs and precipitation extremes over the Western and Central parts of the Himalayas. It is noteworthy to mention here that the high-resolution LMDZ4 HIST simulation clearly demonstrated the ability in reproducing the observed features of the DJFMA time-mean circulation, the position, intensity and spectral variance of the upper-level westerly jet stream and broad synoptic-scale characteristics of WD activity. Prompted by these promising features from the HIST simulation, we further examined the impacts of climate change on the synoptic-scale WD activity and variability of precipitation extremes over the WH region. The findings of this study indicate that the rising trend of surface temperature and precipitation extremes over the WH during the 20th are attributable largely to anthropogenic forcing, and these trends cannot be explained by natural forcing alone. Furthermore, these trends are projected to rise more significantly during the twenty-first century.

A detailed analysis of the HIST simulation suggests that the impact of climate change on the surface temperature response at high elevations (> 3 km a.s.l) is associated with stronger warming over the large areal expanse in the eastern part of the Tibetan Plateau as compared to the western part. This differential thermal response sets up a positive zonal temperature gradient between the eastern and western parts of the Tibetan Plateau. Long-term station observations (1961–2012) over the Tibetan Plateau also provide corroborative evidence for the strong increasing trend of surface temperatures over the eastern side of the Tibetan Plateau (Liu and Yan 2017). It is also interesting to note that the warming gradient across the Tibetan Plateau appears to have an impact on the background circulation. A comparison of the DJFMA time-mean circulation between the HIST and HISTNAT experiments, shows warming of the middle and upper-tropospheric temperature at a faster rate over the eastern side of the Tibetan Plateau as compared to the western side, in response to climate change. A large-scale thermally balanced mid-latitude stationary-wave pattern, characterized by anticyclonic anomaly over the eastern side of the Tibetan Plateau and a trough over the western side together with a downstream ridge-like feature over Central Himalayas, emerges as a consequence of this differential warming. The aforesaid trough over the WH, as manifested in the simulated response of the background circulation to climate change, is conducive for enhancing the amplitude of synoptic-scale WD passages. The HIST simulation shows that anchoring of large-amplitude troughs on the windward side of the WH tends to favor orographic precipitation enhancements in association with strong southerly flow of moisture from the Arabian Sea, whereas the downstream ridge-like feature inhibits the Central Himalayan precipitation. Mean surface temperatures are often below freezing point, as low as − 10 °C, over high-elevated locations (eg., Leh, Elevation 3.53 km a.s.l) in the Western Himalayas during the winter and early spring months. The maximum increase of surface temperature in this region during the last century is about 1.6 °C (Bhutiyani et al. 2007). Given the low average surface temperatures over the high elevations of the Karakoram and Western Himalayas, the dynamical controls of enhanced amplitude variations of WDs appear to have outweighed the local surface warming effects on winter snowfall during the last few decades. While the central focus of this study is the impact of climate change on the synoptic-scale WD activity, further investigations are warranted to unravel the multi-scale associations of the subtropical westerly jet with various climate drivers (Cannon et al. 2015). Likewise, the physical mechanism of EDW over the Tibetan Plateau and its link to large-scale atmospheric circulation patterns needs further exploration (Pepin et al. 2015).