This article is licensed under a 
1. Introduction
High Mountain Asia (HMA) serves as a major water source for large rivers in Asia (Immerzeel and Bierkens 2012). HMA consists of the Tibetan Plateau (TP), surrounded by the mountain ranges of Tien Shan, Pamir, Hindu Kush, and the Karakoram in the west, the Himalayas in the south and southeast, and Qilian Shan in the east. Over 1.4 billion people in various countries, including Afghanistan, Bangladesh, Bhutan, China, India, Kazakhstan, Kyrgyzstan, Mongolia, Myanmar, Nepal, Pakistan, and Tajikistan, depend on water originating from HMA. The topography and atmospheric circulation patterns mainly define the region’s climate variability (Maussion et al. 2014; Schiemann et al. 2009; Webster et al. 1998). The region has strong longitudinal (west–east), latitudinal (north–south), and vertical climate gradients (Yao et al. 2012). The climate of HMA is driven by the interaction of the Indian summer monsoon, dominant in the southeastern parts, while winter westerly winds dominate the climate in the western region. The steep north–south orography results in strong temperature gradients that drive the moisture-laden winds from the Indian Ocean to the landmass, and ultimately the moisture is released under the orographic influence.
Consequently, the southern and eastern parts of HMA receive nearly 80% of the yearly precipitation from June to September during the monsoon season, which occurs as rainfall at low altitudes and snow at high altitudes (Bookhagen and Burbank 2010). Conversely, the westerly winds in the west, originating from the Mediterranean, contribute around 50% of annual precipitation during winters, mainly falling as snow (Bao and You 2019; Rees and Collins 2006). The interaction of the monsoon and westerlies, where the monsoon has a dominant role in the summer months and westerlies in the winter months, influences the climate of the interior TP (Frauenfeld et al. 2005; Yang et al. 2014; You et al. 2015b).
The climate of HMA has witnessed many changes in recent decades. There are spatially consistent and statistically significant warming trends over the different regions of HMA (Krishnan et al. 2019; Liu and Chen 2000). Studies based on ground-based observations have consistently reported warming trends over the TP in the past (Kosaka and Xie 2013; Liu and Chen 2000; Yan and Liu 2014). Tien Shan, Central Asia, and the Hindu Kush Himalayas (HKH) region have observed similar warming trends (Aizen et al. 1997; Hu et al. 2014; Ren et al. 2017). In contrast to temperature, precipitation shows a more considerable interannual variability, and inconsistency in trends for different regions in HMA (Fowler and Archer 2006; Palazzi et al. 2013; Ren et al. 2017; Shrestha et al. 2000; You et al. 2015b; Zhan et al. 2017). The climate variability in recent decades has resulted in changes in the cryosphere (glaciers, snow cover, and permafrost) and hydrology (water availability, seasonality, and hydrological extremes like floods and droughts), which in turn affect society (Bolch et al. 2012; Immerzeel et al. 2010; Jin et al. 2020; Kääb et al. 2012; Kang et al. 2010; Shean et al. 2020; Wijngaard et al. 2017; Yang et al. 2010; Yao et al. 2012).
Past studies used monthly-scale station data to derive the historical trends for different regions in HMA (Cao et al. 2013, 2017; Duan and Xiao 2015; Guo and Wang 2012; Khattak et al. 2011; Liu and Chen 2000; Shrestha et al. 1999, 2000; Xu et al. 2018; Yan and Liu 2014; Yang et al. 2014). Attempts were made with remote sensing techniques to calculate trends (Qin et al. 2009; Salama et al. 2012; Zhong et al. 2011). Some recent studies used the general circulation model data downscaled with fine-resolution regional climate models to calculate the long-term trends (Amato et al. 2019; Zhang et al. 2017). Moreover, some studies combined in situ and reanalysis data to understand the spatial pattern of historical climate change (An et al. 2017; Krishnan et al. 2019; Madhura et al. 2015). However, these studies are either scattered around the basin, national and regional levels. Further, these scattered studies use different data, coarser spatiotemporal resolution, and approaches. Variability in approaches, data, and methods makes it even more challenging to align and compare the changes around different regions in HMA. Studies that use consistent, observed, and remotely sensed data integrated with numerical models at a higher spatial resolution over the entire HMA region are required to resolve the climate variability in the region.
Existing hydrometeorological stations, mostly located in valleys lower than 4000 m, are sparsely distributed in the region (Pepin et al. 2015; Qin et al. 2009). The complex topography and harsh conditions in the mountains impose difficulties in managing the ground stations. Therefore, climate signals are biased toward these station observations at lower elevations (An et al. 2017; Palazzi et al. 2013). Remotely sensed satellite measurements from geostationary thermal infrared and polar-orbiting passive microwave sensors are useful for deriving precipitation measurements based on cloud-top brightness temperature and spectral scattering due to large ice particles, respectively. However, the uncertainty is high due to sensor signals’ limitations in penetrating the clouds and correctly estimating the precipitation falling as snow at high altitudes (Immerzeel et al. 2015). Nevertheless, remotely sensed products or products merged with gauge observations, in recent decades, have proven to be a cost-effective and reliable tool to understand precipitation patterns and trends at various spatial and temporal scales (Gehne et al. 2016). Among other remotely sensed products (or merged products), the Tropical Rainfall Measuring Mission (TRMM), the Climate Hazard group Infrared Precipitation (CHIRPS), the Multi-Source Weighted-Ensemble Precipitation (MSWEP), the Climate Prediction Center morphing technique (CMORPH) and the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network (PERSIANN) and are most commonly used in the HMA region for a wide range of applications (Ashouri et al. 2015; Beck et al. 2017; Funk et al. 2015; Huffman et al. 2007; Joyce et al. 2004; Yatagai et al. 2012). The direct use of such products to derive climatological and hydrological trends often requires validation and correction based on in situ observations (Gebregiorgis and Hossain 2015; Gehne et al. 2016). The correction is often hampered by the mismatch in resolution and insufficient geo-statistical interpolation accuracy of situ data caused due to sparse gauge distribution. Moreover, these global remotely sensed (or merged with gauge observation) products have higher uncertainty in assessing the correct precipitation amount in an environment with complex mountainous topography such as HMA (Cheema and Bastiaanssen 2012; Mei et al. 2014). A common scientific consensus, based on a plethora of studies, could be made that all these remotely sensed (or products merged with gauge observation) products have large and variable biases relative to the gauge data in HMA (Andermann et al. 2011; Cheema and Bastiaanssen 2012; Guo et al. 2015; Tong et al. 2014; You et al. 2015a).
A gridded reanalysis product—which is a result of data assimilation from multiple sources: airborne balloons, scatterometer radiosonde, dropsonde, aircraft measurements, satellites, and ground-based radar-gauge composite—provides an alternative to the sparse and inconsistent point-scale observations to find spatial patterns of change (Alexander et al. 2006; Li et al. 2022). Even though biases between reanalysis and in situ observations are present, the reanalysis products have shown good reliability in resolving the climatological mean, anomalies, and normalized trends (Donat et al. 2014; Simmons et al. 2010). Given the high variability in the climate of HMA, this paper aims to assess trends in annual and seasonal air temperature and precipitation and a range of climate change indices for the high-altitude regions. We use the state-of-the-art ERA5 high-resolution reanalysis data to derive the trends (Hersbach et al. 2020). These trends help us to detect similarities and contrasts in recent climate change over the entire HMA using one consistent dataset.
2. Study area
The HMA, consisting of the TP and its surrounding high mountain ranges along with its 18 downstream river basins, is considered for this study (Fig. 1a). Within this region, the Hindu Kush Himalayan (HKH) range along with the TP covers an area of over 5 million km2 with an average elevation of ∼4000 m above mean sea level (MSL) (Yao et al. 2012). The areal extent considered in this study is 57°–113°E and 22°–47°N. Given the large extent of the study area, the overall climate is variable. For example, in the west, the Helmand, Amu Darya, and Syr Darya River basins have a dry continental climate characterized by cold winters and hot summers (Chen et al. 2011), while in the east, the Yangtze basin has a subtropical climate with maximum rain between April and October (Gu et al. 2018). The northwestern basins, Balkash, Junngar, and Alaguy are influenced mainly by midlatitude westerlies and cold inflows from the polar region. The TP, generally below 0°C and temperature decreasing from east to west, experiences cold winters and wet summers, with maximum precipitation during July and August (Frauenfeld et al. 2005). Climatic differences enhance the spatial variation within each basin’s higher and lower altitudes (Krishnan et al. 2019; You et al. 2017). A significant part of the southeastern basin’s precipitation is from the southwestern Indian monsoon between June and September. The winter monsoon brings rain to the northwestern part of the HKH. At high altitudes, precipitation mainly falls as snow, whereas at lower altitudes it mainly falls as rain. The Central Asian basins receive annual precipitation of ∼211 mm, ranging from less than 50 mm in the desert areas to higher than 2000 mm on the windward slopes (Deng and Chen 2017).
(a) The river basins analyzed in this study (black boundaries). Gray lines represent the upstream region of each major river basin. The background represents the elevation of the region. The arrows represent the major atmospheric circulation system, red for monsoon and blue for westerlies, in HMA [the arrows are adapted from Yang et al. (2014)]. Also shown is the spatial distribution of mean annual (b) precipitation (mm) and (c) temperature (°C) during 1979–2018 across HMA.
Citation: Journal of Applied Meteorology and Climatology 62, 2; 10.1175/JAMC-D-21-0045.1
3. Data and methods
We use historical climate data from the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 dataset covering the years 1979–2018 (40 years) (Hersbach et al. 2020). The ERA5 is an improved (atmosphere, ozone, land, and ocean wave component) and high-resolution successor of the ERA-Interim (Dee et al. 2011). The ERA5 uses observations from over 200 satellite instruments or conventional data types, including ground-based radar–gauge observations, PILOT, radiosonde, dropsonde, buoys, and aircraft measurements. The ERA5 data are available at an hourly time scale and 31 km × 31 km spatial resolution for 137 vertical pressure levels. Surface or single-level data are also available, containing two-dimensional parameters such as precipitation, 2 m temperature, top-of-atmosphere radiation, and vertical integrals over the entire atmosphere.
The utility of ERA5 in complex terrain such as HMA, where the in situ meteorological observations are sparsely and unevenly distributed, has been investigated in several studies (see Table 1). ERA5 has been validated in several regions of the HMA, for instance, the Pamir region of Tajikistan (Zandler et al. 2019), the Indus basin (Dahri et al. 2021b), mainland China (Jiang et al. 2020; Lei et al. 2022; Tang et al. 2020), Tibetan Plateau (Hu and Yuan 2021; Ji and Yuan 2020), and HMA (Sun et al. 2021). The study by Lei et al. (2022), on the basis of 666 ground stations over China, showed that annual total precipitation was correctly simulated when compared with the daily total in ERA5. Authors suggested ERA5 performs well for simulating intensity indices such as annual total wet days precipitation, max 5-day precipitation, very wet days, max 1-day precipitation amount and extremely wet days. Frequency indices such as extreme heavy precipitation days were consistently well simulated across China except northwestern China in ERA5. ERA5 better simulated the consecutive dry days than the consecutive wet days. Similarly, Sun et al. (2021) found that ERA5 precipitation generally captures the seasonal variations of ground observations (monsoon and westerlies patterns) and the broad spatial distributions of precipitation in both magnitude and trends in HMA. Interestingly, ERA5 well represents the precipitation seasonality of the Tarim basin, which is not reflected in GPM and outputs from the regional climate model. Moreover, ERA5 data have been used to understand the precipitation and large-scale atmospheric systems (Lai et al. 2021; Wang et al. 2020; Yu et al. 2021; Zhu et al. 2020), snow cover and snow depth (Lei et al. 2023), glacier and snowmelt simulation (Bhattacharya et al. 2021; Kraaijenbrink et al. 2021), and hydrological and water balance simulations (Dahri et al. 2021a; Khanal et al. 2021; Sun et al. 2021).
Summary of the findings related to ERA5 data reported in several studies.
We further assess the performance of six different gauge-based gridded precipitation products (CHIRPS, TRMM, PERSIANN, GLDAS, CFSV2, and GPM) and compared them with ERA5 over the HMA (Fig. A1 in the appendix). The average annual precipitation patterns show more or less similar large-scale spatial patterns. ERA5 seems to be wetter in the southern and eastern parts relative to the other regions of HMA. The differences are also noticeable in region-aggregated climatological analysis. CFSV2 seems to be on the drier side for the monsoon period where as the GLDAS is drier for the dry seasons. In general, ERA5 overestimates the precipitation in comparison with the six products. Satellite-derived products are of insufficient quality to capture the magnitude of mountain precipitation (Immerzeel et al. 2015). Authors found that the amount of precipitation required to sustain the observed mass balances of the large glacier systems in the Indus basin is far (by a factor of 5–10) beyond what is observed at valley stations or estimated by gridded precipitation products.
There is no perfect product in HMA as the ground-based observations at upper altitudes of HMA are lacking (Sun et al. 2021). Precipitation products are not equally good enough for all the basins in HMA. Even though ERA5 overestimates the precipitation in comparison with the other products (satellite, gridded, reanalysis, and model simulation), the aforementioned studies show that the ERA5 has a wide range of utility. ERA5 captures the seasonal variations of ground observations and the broad spatial distributions of precipitation in both magnitude and trends when it comes to data-scarce upstream regions of HMA. Thus, we conclude that ERA5 is good enough to use to understand the climatic trends in HMA.
The daily aggregated surface level precipitation sum and mean temperature are used in this study to derive historical climate indicators as described in Table 2. We analyze each climate indicator on an annual scale and seasonal scale: (i) winter [December, January, and February (DJF)], (ii) summer or premonsoon [March, April, and May (MAM)], (iii) monsoon or rainy [June, July, August, and September (JJAS)], and (iv) postmonsoon or autumn [October and November (ON)].
List of climate indices used in this study.
The climate indicators analyzed for this study are consecutive dry days (CDD), heatwave duration index (HWDI), highest 5-day precipitation amount (RX5), heavy precipitation days (R10), wet days precipitation (R95P), and compound indices (COMP95). The first five indices used are defined and described in Zhang et al. (2011). We define COMP95 as the number of days when both precipitation and temperature are greater than the 95th quantile values of their distributions. The COMP95 is used as a proxy for floods because it indicates the “Warm-Wet” quadrant of the precipitation and temperature distribution. A similar approach is used in Khanal et al. (2021) and Lutz et al. (2016). For convenience, we use CDD and HWDI as the proxy for droughts and their related hazards, and R10, RX5, R95P, and COMP95 as a proxy for flood-related hazards. The relationships between the climate indices and floods/droughts are often complex. Changes in the trend of these indicators may not directly lead to floods or droughts. However, they would certainly lead to changes in the antecedent conditions and state of the precursors (saturation content of the soil, groundwater state, rain on snow, etc.) related to floods and droughts (Merz et al. 2014; Nied et al. 2014). For instance, the increasing trend in R10 and RX5 would mean higher episodes of precipitation events that may impact the soil moisture conditions. Any medium to high precipitation episode on saturated soil most likely results in floods (Khanal et al. 2019; Merz et al. 2014; Nied et al. 2014). We also calculate the standardized precipitation anomaly (SPA) and standardized temperature anomaly (STA) on a regional scale to understand the regional patterns of floods and droughts.