1 Introduction

In High Mountain Asia (HMA), cool-season precipitation and the resulting spring and summer glacial melt provides water resources for hundreds of millions of people, but also presents risks for many extreme weather conditions (Kääb et al. 2012; Hewitt 2005). Recent work has shown that atmospheric rivers (ARs), long conduits of strong moisture transport, are significant contributors to winter and spring precipitation in HMA (Nash et al. 2021). ARs occur in a variety of locations across the globe and are associated with extreme precipitation, flooding, lightning, landslides and anomalous snow accumulation (Cannon et al. 2018; Nash and Carvalho 2020; Oakley et al. 2018; Zhu and Newell 1994, among others). In HMA, ARs contribute to extreme precipitation and are associated with flood events in the Nepal and Bay of Bengal areas (Thapa et al. 2018; Yang et al. 2017). Nash et al. (2021) found and characterized three distinct types of ARs producing above-average precipitation in northwestern, western, and eastern HMA. Moreover, they determined that there are typically between 9 and 11 HMA ARs per month in the winter and spring, contributing between 40 and 60% of total seasonal precipitation. However, on some occasions, a single strong AR event can provide up to a quarter of that precipitation, with precipitation totals exceeding 100 mm day\(^{-1}\) increasing rainfall-related risks, such as landslides and flooding.

Many studies have investigated long-term climate trends over HMA. In western HMA, Norris et al. (2019) identified positive trends of cloud ice and liquid cloud, indicating the higher frequency of extratropical cyclones in recent years. Nash et al. (2021) demonstrated that of the three types of HMA ARs, Northwestern and Western HMA ARs are primarily associated with extratropical cyclones, where the warm, moist air from the AR is advected in the area ahead of the cold front. Given this information, it is likely there have been changes in the frequency or intensity of HMA ARs, although this has yet to be quantified. Furthermore, Wang et al. (2014) observed upward trends in the height of the 0 \(^{\circ }\)C isotherm (hereafter, the freezing level) during summer in HMA. Changes in winter freezing levels have yet to be quantified in HMA, but increases in the freezing level are likely to result in decreased frozen precipitation, particularly during ARs. Previous studies have observed the increase of the freezing level during an AR, as extratropical cyclones associated with an AR are typically warmer than those without (Lundquist et al. 2008; Neiman et al. 2008, 2011). Above-average freezing levels during ARs can increase the likelihood of precipitation-related hazards because the fraction of rain to snow at higher elevations results in increased runoff and snow melt (Guan et al. 2016).

Espinoza et al. (2018) demonstrated that under the RCP 8.5 warming scenario, the frequency of HMA ARs is expected to increase by 6–8% while the intensity of integrated water vapor transport (IVT) is expected to remain the same between 2073 and 2096. Kirschbaum et al. (2020) showed that increases in extreme precipitation in HMA has the potential to increase landslides by 10–70% more in the years 2061–2100. Increases in ARs and their intensity could potentially increase precipitation and precipitation-related hazards; therefore, it is important to understand recent changes in AR properties to determine their influence on local warming and precipitation trends.

This study highlights the importance of long-term trends in the freezing level associated with HMA ARs by contrasting two events that both resulted in extreme precipitation across western HMA. These two events featured greatly differing freezing level heights and thus outcomes regarding precipitation-related hazards. Advanced Weather Research and Forecasting (ARW-WRF, hereafter WRF) simulations at 6.7 km resolution are used to differentiate between the mesoscale characteristics of these two events. The finer spatial resolution of this model largely overcomes the typical limitations of scarce observational data and coarse reanalysis resolution (> 27 km) amidst the complex topography of HMA. Focus is placed on mesoscale characteristics that are important to extreme precipitation, such as water vapor flux, the orientation of the AR relative to topography, the height of the freezing level, and the orographic mechanisms related to precipitation in the foothills of HMA.

The organization of this paper is as follows: Sects. 2 and 3 describes the data used for this analysis and outlines WRF model set up. Section 4.1 describes thermodynamic trends during HMA AR events using 36 years of dynamically downscaled reanalyses over HMA. We evaluate changes in the freezing level and moisture, focusing on areas where HMA ARs typically result in above-average precipitation during the winter. Sections 4.2 and 4.3 outlines the selection of two extreme AR events that had similar overall characteristics but had different freezing levels and precipitation amounts. Section 4.4 compares the synoptic patterns of both events. Using the WRF model, Sect. 4.5 examines the mesoscale meteorology of two ARs associated with extreme precipitation, emphasizing the differences between an AR event with an above- and below-average freezing level. We summarize our results in Sect. 5.

2 Data

2.1 AR detection: tARget v3

To detect ARs, we use the Tracking Atmospheric Rivers Globally as Elongated Targets (tARget) algorithm version 3 which was applied to global, 6-hourly ERA-Interim data from 1979 to 2015 (Guan and Waliser 2019). This AR detection algorithm is useful for the HMA region as it detects ARs via relative IVT intensity thresholds, which is particularly useful during the winter in HMA, as there is, on average, little to no moisture (Nash et al. 2021). Nash et al. (2021) identified three main types of ARs that reach HMA in winter and spring months using tARget v3. We use the resulting classification of HMA AR types in this study to focus on Northwestern and Western HMA AR Types that resulted in extreme precipitation.

2.2 WRF setup

This study uses 36 years of Climate Forecast System Reanalysis (CFSR) (Saha et al. 11a). Results are similar for 34.09\(^{\circ }\)N and 74.02\(^{\circ }\)E, except the moisture flux for the February 2010 AR extended almost all the way to 400 hPa, peaking at 0.8 m s\(^{-1}\) between 750 and 600 hPa (Fig. 11b). Possible explanations for the deeper profiles of water vapor flux during the 2010 event include a stronger AR, a longer-duration AR (possibly allowing more time for moist parcels to rise), and a warmer air mass requiring more moisture to reach saturation. However, future work is needed to more fully quantify the relationships between AR / IVT intensity, duration, temperature, and the vertical profile of water vapor flux at inland locations.

Fig. 11
figure 11

a Climatological vertical profile of horizontal water vapor flux (m s\(^{-1}\)) based on WRF at 34.87\(^{\circ }\)N, 72.66\(^{\circ }\)E for all days when AR conditions are met during the months of December, January, or February between 1979 and 2015 at this location (blue line and box-and-whisker plots show the distribution of the 284 events), and vertical profile of horizontal water vapor flux (m s\(^{-1}\)) based on WRF at the same location on 5 January 1989 12:00 UTC (red solid line) and 8 February 2010 06:00 UTC (red dashed line). The box extends from lower to upper quartiles of the data, with a black line at the mean. The whiskers show the range of the data from the 5th percentile to the 95th percentile, and outliers are shown as points past the end of the whiskers. b Same as (a) but for 34.09\(^{\circ }\)N and 74.02\(^{\circ }\)E. The locations of both points are identified by the black triangles in Figs. 1b, 6e–h, 9e,f, and 10e,f

Full size image

5 Conclusions

This study shows that between 1979 and 2015, southerly IVT has significantly increased in western India and Pakistan during Western HMA ARs, indicating that in recent decades, there has been an increase in the intensity of Western HMA ARs. Additionally, the height of the freezing level has significantly increased across southern Asia during HMA ARs. One consequence of these findings is that there is significantly less frozen precipitation during HMA ARs with an above-average freezing level compared to those with a below-average freezing level. Should future trends continue as currently observed, western HMA will see an increase in the intensity of ARs with an above-average freezing level. With more liquid precipitation during these events, there is a higher likelihood of risk for associated natural hazards such as landslides and floods.

To further highlight the importance of the freezing level on resulting precipitation in western HMA, this study focused on two impactful western HMA ARs: one that occurred during below-average freezing level conditions and one that occurred during above-average freezing level conditions. Both ARs transitioned from Northwestern to Western HMA ARs, were quasi-stationary over this area, featured greater than the 85th percentile of IVT for Western HMA ARs, and resulted in greater than the 85th percentile of precipitation for these storm types, largely due to a long duration of orographically lifted moisture within the AR plume. We used dynamically downscaled CFSR at 6.7 km spatial resolution to compare their mesoscale characteristics to determine the influence of the freezing level on orographic precipitation.

The below-average freezing level AR occurred in January 1989, lasted for just under 4 days, and resulted in about 175 mm of precipitation across western HMA. The above-average freezing level AR occurred in February 2010, lasted for about 5 days, resulted in about 200–450 mm of precipitation, and was related to six separate landslide events in western HMA. Although freezing levels were only 50–600 m higher during the 2010 AR, this event resulted in 10–70% less frozen precipitation than the 1989 AR (Fig. 6). This is an extreme difference between two disparate events, but even in aggregate from 1979 to 2015, there was a 10–40% reduction in frozen precipitation during above-average freezing level ARs (Fig. 4).

This study illustrates the importance of mesoscale conditions in modulating the interaction of ARs, topography, freezing level, and precipitation-triggered landslides. During the 2010 AR, a deep moist layer was orographically lifted directly perpendicular to the topography near the foothills of HMA, resulting in a combination of rain and snow of about 150 mm day\(^{-1}\). This triggered multiple landslides across western HMA near and upstream of where the freezing level intersected with the topography, in the transition zone from rain to snow. Future studies seeking to improve the predictive skill of these destructive events will therefore need to consider both the synoptic and mesoscale environments in which they occur.

While freezing level likely plays a large role in determining the likelihood of landslides, other factors are also important. Naturally, storm intensity and total precipitation (liquid or frozen) plays a role. Moderate, long-duration precipitation interspersed with short-duration high intensity precipitation increases the likelihood of precipitation-triggered shallow landslides (Cordeira et al. 2019; Kirschbaum et al. 2020; Oakley et al. 2018). Other factors that may need to be considered are antecedent soil moisture conditions, and the possibility of rain-on-snow events, which have been shown to increase the risk for floods and landslides when they occur (e.g., Guan et al. 2016).

In summary, this work conclusively shows that from 1979–2015 across HMA, the freezing level has increased (1–4%), the intensity of Western HMA ARs has increased (2–16% increase in IVT), and that when the freezing level is above-average, there is significantly less frozen precipitation. Furthermore, the examples of below- and above-average freezing level ARs presented here demonstrate the importance of mesoscale processes in orographic precipitation and highlight the varying outcomes that can result across HMA from relatively small differences in freezing level height.