Dynamic Rain and Snowfall Patterns. Average annual precipitation derived from remote sensing ranges from 325 mm at elevations > 4500 m a.s.l. to 622 mm at elevations < 2225 m a.s.l. (Table 1). Temperature estimates derived from MODIS Land Surface Temperature & Emissivity (LST&E) (MOD11A1) were used to separate liquid from solid precipitation. Snow accounts for 76% of total precipitation at high elevations compared to 42% at elevations < 2225 m. During the 20-yr period from 2001 to 2020, there was a slight but insignificant overall increase in total precipitation in the Pamir region.
Table 1
Average annual precipitation components and temperatures throughout the Pamir in different elevation bands for the period from 2001 to 2020.
| < 2225 m a.s.l. | 2225–4500 m a.s.l. | > 4500 m a.s.l. |
Total average annual precipitation (mm) | 622 | 475 | 325.5 |
Average annual rainfall (mm) | 360 | 178 | 78.5 |
Range in annual rainfall (mm) | 218–457 | 113–239 | 35.7–118 |
Average annual snowfall (mm) | 262 | 297 | 247 |
Range in annual snowfall (mm) | 144–356 | 194–355 | 173–304 |
% precipitation as snow | 42% | 63% | 76% |
Average annual temperature (°C) | 15.9 | − 0.9 | − 9.6 |
Range in annual temperature (°C) | 14.4 to 17.5 | − 2.5 to 0.7 | -11.5 to − 8.0 |
Focusing on snow water at a scale of 10 x 10 km, some significant patterns emerge in the different elevation zones during this 20-yr period (Fig. 1a). For the combined areas above 4500 m, annual snow accumulation did not significantly change with time, but increases occurred during spring to mid-summer (0.29 mm y− 1) and decreases in snowfall occurred in early fall to early winter (largest decrease in December; − 0.88 mm y− 1). Mid-elevations (2225–4500 m) experienced overall weak (insignificant) positive trends in snowfall with a negative trend occurring in late summer to early fall. At lower elevations, a positive trend in annual snowfall was observed, particularly evident in late spring (1.3 mm y− 1; Fig. 1b). Snowfall increases were primarily concentrated in broad valleys of the Vakhsh River upstream of the Nurek reservoir. Many mid to high elevation areas in Vakhsh basin also had significant snowfall increases (Fig. 1a). In contrast, some mid-elevation areas of the central Panj basin experienced small increases in snow while other areas had minor decreases.
Spatial and temporal patterns of annual rainfall differed at various elevations (Fig. 1c). Above 2225 m, no overall temporal trends in rainfall were detected, although total precipitation increased in spring. At mid-elevations (2225–4500 m), snow water inputs declined in late summer; however, annual trends were slightly positive, but insignificant. Snow water inputs increased in early summer above 4500 m but decreased in early winter; no annual trends were observed. Isolated areas of rainfall decline occurred in higher parts of eastern Vakhsh basin, whereas throughout most of the mid to high elevation Panj basin, many areas experienced increased rainfall (Fig. 1c). The southern-most portion of the Panji basin in Afghanistan had isolated areas of rainfall decline. The largest increases in annual rainfall occurred throughout western portions of the Vakhsh and Panj basins (elevations < 2225 m). Overall, annual rainfall below 2225 m increased significantly during the 20-yr period at a rate of ≈ 0.5 mm y− 1 (Fig. 1d), with the greatest increases in spring and early fall.
Snow Persistence and Snowline Elevation Trends. Trends in snow persistence (SP; fraction of a year that snow is present on the ground) can help inform climate change. Mean SP in the Vakhsh and Panj basins varied between 0.45 ± 0.28 and 0.54 ± 0.3 from 2000 to 2020 (Fig. 2a). Interannual SP variability was rather high and no significant temporal trends occurred. The lowest SP values, 0.45 and 0.47, occurred during 2007 and 2008, respectively (Fig. 2b) and were mainly at mid and high elevations (i.e., > 2225 m; Fig. 2a). SP values in the mid and high elevations during the 20-yr period averaged 0.54 ± 0.2 and 0.79 ± 0.15, respectively. In contrast, low elevations (< 2225 m) had a mean SP value of 0.13 ± 0.13 with high interannual variability. The highest SP values were observed during 2009 below 2225 m with values ranging from 0.60 ± 0.19 at mid-elevation to 0.85 ± 0.12 at high elevations.
Another insight into climate change in the cryosphere can be derived from the evolution of long-term snowline elevations. Snow precipitation and coverage varied widely from one winter to another; however, snow elevations may evolve over decades due to climate change. Results of remote sensing analysis (see Methods) throughout the entire Vakhsh and Panj basins showed no significant trends of monthly snowline elevation. However, when examining only the high elevation range (4500–7495 m), small, but significant, increases in snowline elevations of 2.57 m yr− 1 (p-value: 0.03) and 1.23 m yr− 1 (p-value: 0.04) appeared in November and February, respectively, since 2001 (Fig. 2c). This means that the snow coverage tended to decrease since 2000. This increasing elevation of snowline in early winter is consistent with the significant decreasing trend of snowfall in December (− 0.88 mm y− 1) since 2000. Early winter is receiving less snow than in the past leading to a higher snowline after that time of the year.
Temperature trends. Temperature trends are important to understand how observed patterns of rain and snow relate to runoff. There are major temperature gradients across the Vakhsh and Panj basins with the highest temperatures in the western region grading sharply to lower temperatures at high elevations to the east (Table 1). Slightly increasing air temperatures during the 20-yr period occurred in parts of the high elevation glaciated areas in the Vakhsh basin (Fig. 3). Climate warming ‘hot spots’ are seen in the mid-high to high elevation central and southeast portions of the Pamir (Fig. 3). However, it is notable that these areas have alpine climates with low precipitation, thus any effects of climate warming on runoff produced from these areas may be minimal and considerably delayed. Interestingly, below 2225 m, temperature has been relatively stable during the 20-yr period, including terrain proximate to the Vakhsh River and its major tributaries (Fig. 3). These areas with little temperature increase correspond to some of the most productive agricultural lands in Tajikistan.
Other Dynamic Water Sources: Glacier Melt and Permafrost Thaw. Based on the World Glacial Inventory (https://nsidc.org/data/g011w230), about 5.5% of the total area of Tajikistan is covered by glaciers making it one of the most glaciated nations worldwide. The ‘supply index’ partly associated with glacier volume and water yield in the Amu Darya basin is higher than most major river basins26. Because more than half of the land area of Tajikistan is above 3000 m, 44% of the country has been identified as potential permafrost area27. Most of this permafrost is concentrated in the high-elevation Badakhshan region of the southern and eastern Pamir in Tajikistan and northern Afghanistan.
Approximately 13,000 glaciers reside in the Pamir (ice volume of 1300 km3)28, and about 8% of total discharge in the Amu Darya emanates from these12. Investigations of glacier area and mass change in the Pamir conducted in the past decade have shown different results based on geography, methods used, and time periods. To normalize these diverse findings, we employed a systematic literature review that revealed glacial mass loss is not necessarily directly related to glacier area. Elevational changes in Pamir glaciers estimated by various methods are typically less than or within the range of error estimates (Table 2). Studies in the adjacent Western Kunlun indicate minor, but significant, glacial elevation increases21,23. Other studies in the circum-Pamir show small gains of glacier mass during the first decade of this century followed by small losses; however, these values are mostly within error estimates24, 29. Studies at specific sites revealed opposite patterns or very little change in glacial mass during this century20,22,25. A recent study in the central Pamir concluded that glacier mass budgets were approximately balanced or had slight mass deficits from mid-1970’s to mid-2010’s30. Our summary of studies throughout the wider Pamir region confirms these findings (Table 2).
Table 2
Recent studies using various remote sensing techniques estimating glacial mass change (using elevation change as an indicator) within and proximate to the Pamir.
Specific Location | Glacier area | Elevation (m a.s.l.) | Glacial elevation change per year | Methodology | Source |
Eastern Pamir, W. China 38-39°N, 73-75°E | 2362.5 km2 | 3000 to > 7600 | -0.06 ± 0.16 m (2000–2009) 0.06 ± 0.04 m (2000–2016) | Geodetic using: SRTM, NASA HMA & ALOS-PRISM DEMs | Lv et al. 2020 |
Eastern Pamir, NW China 38-39°N, 74°40’ – 75°40’E | 1018 to 999 km2 | 3000 to 7719 | -0.15 ± 0.12 m (≈ 1971 – ≈2013) | ASTER, Cartosat-1, Landsat, SRTM DEMs, topographic maps & glacier inventories | Zhang et al. 2016 |
Eastern Pamir, Xinjiang, China 38°17’N; 75°07’E | 274 km2 | up to 7546 | -0.01 ± 0.30 m (1973–2013) | Geodetic using: Hexagon KH-9, ALOS-PRISM, Pléiades, Landsat & SRTM DEM | Holzer et al. 2015 |
Central Pamir | 2120 km2 | -- | -0.03 ± 0.24 m (1975–1999) | Randolph Glacier Inventory, KH-9 stereo images, SRTM DEM | Zhou et al. 2019 |
Western Pamir | 3178 km2 | 2800 to 7090 | 0.14 ± 0.13 m (1999–2011) | SRTM DEMs & SPOT5 stereo images | Gardelle et al. 2013 |
Pamir | 1000 ± 300 elev. Samples | -- | -0.46 ± 0.28 m (2003–2008) | ICESat altimetry data, SRTM DEM | Wang et al. 2017 |
Pamir | 4441 km2 | 3600 to 5750 | -0.05 ± 0.08 m (2000–2016) | ASTER DEMs | Brun et al. 2017 |
Pamir, Panj River basin | > 2000 km2 | 800 to > 7000 | -0.52 m (2002–2013) also considers snow | Empirical, lumped hydrological model estimates (J2000g) | Pohl et al. 2017 |
Pamir | 6500 km2 | -- | -0.48 ± 0.14 m (2003–2008) | ICESat altimetry data | Kääb et al. 2015 |
Pamir Alay | 809 km2 | 3300 to 4600 | -0.04 ± 0.07 m (2000–2016) | ASTER DEMs | Brun et al. 2017 |
West Kunlun, Xinjiang, China 35-36°N, 82-83°E | 1687 to 1695 km2 | 5000 to 7000 | 0.26 ± 0.07 m (2000–2016) | ASTER DEMs | Brun et al. 2017 |
West Kunlun, NW China | 12,500 km2 | -- | 0.05 ± 0.07 m (2003–2008) | ICESat altimetry data | Kääb et al. 2015 |
Given the scattered distribution of glacier studies in the Pamir, we examined precipitation and temperature parameters within six glaciated regions: Fann mountains, Pamir-Alay range, Fedchenko region, central Pamir, lower Panj basin, and Wakhan corridor (Fig. 4). Remote sensing analysis of solid precipitation and temperature during the last 20 year revealed some significant short-term trends for these climate parameters in most sub-regions (Fig. 4). All sub-regions experienced a significant increase in December temperature, while only the Fann, Fedchenko, and Pamir-Alay sectors had significant increases in annual temperature during the 20-yr period. Although warming dominated in December, temperatures were much lower than − 20°C in all regions. Thus, in the foreseeable future, these increases should not affect glacier melting. Furthermore, from early winter to early spring, a few short-term (4-week) temperature increases were observed in the lower Panj, central Pamir, Pamir-Alay, and Wakhan (Fig. 4). For other periods, temperature trends were highly variable with time. Snowfall exhibited no significant annual trends in all glaciated areas. Significant short-term decreases in snowfall occurred in the central Pamir (December), Pamir-Alay (late summer), and Wakhan corridor (March). In contrast, Fedchenko and lower Panj received increasing snowfall from 23 April to 20 May over the 20-yr period (Fig. 4). These late spring increases in snowfall will likely offset the winter temperature increases and help maintain glacier ice mass.
While glacial melt is usually assumed to contribute directly to streamflow3,14,31−33, runoff generated from melting glaciers in high elevation plateaus may be partly or even completely disconnected from major streams and rivers11,34. Glacial melt can recharge shallow or deep groundwater reservoirs, which eventually support baseflow in streams32,35, but in high glaciated plateaus such contributions may be small. These complex hydrogeomorphic effects are not captured in any of the hydrological models currently applied in this region. To elucidate the connectivity of glacial melt to streams, we overlaid maps of glaciers in this area on a digital elevation model (Alos Palsar 12.5 m) containing stream networks to provide better estimates of melt contributions (Fig. 5). Results showed that about 75% of Pamir glaciers were closely connected to first-order or larger channels in the Vakhsh-Panj basin. This analysis somewhat overestimates connectivity because some first-order streams are not connected to major river systems. Nevertheless, this points to invalid assumptions often made that all glacial melt is delivered to rivers and that all changes in melt will affect river discharge14,34.
Wide areas of permafrost exist in the Pamir-Alay (≈ 50,000 km2) and the Tien Shan (≈ 5500 km2), with continuous permafrost beginning at elevations of 4000 m and 3500 m, respectively36. Discontinuous and sporadic permafrost typically begins at elevations of about 300 to 400 m and 600 to 800 m lower than continuous permafrost, respectively. Based on the TTOP model37, we estimate that 23,951 km2 of continuous permafrost terrain exists in the Panj and Vakhsh basins (14% of the combined Panj-Vakhsh basin), with the majority located in eastern Tajikistan and the Wakhan corridor of Afghanistan at elevations > 3577 m (Fig. 6). Permafrost in most areas underlies the ‘active layer’, which is about < 1 to > 2 m thick and seasonally thaws and refreezes in response to thermal changes37.
Given that no studies have quantified streamflow contributions from summer thaw of the active layer, we estimated this contribution for current conditions in the outlined areas (i.e., Fig. 6) of the Pamir. Active layer thickness (ALT) is inversely related to elevation with current estimates ranging from 1 m (> 4700 m) to 3.5 m (< 4400 m)38. Streamflow contributions from permafrost thaw during July through August were calculated as subsurface flux from streambanks by adapting research findings from field studies in a similar environment39 and scaling them to different stream orders in the Vakhsh and Panj basins (see Methods). Stream and river segments were segregated into small (first- and second-order), medium-size (third-order), and large (fourth-order and above) reaches comprising channel lengths of 18,920.7, 2347.0, and 578.3 km, respectively, within the combined basins. Estimated total permafrost contribution to streamflow during the 2-month melt season was 638 x 106 m3, about 1.5% of the average annual river discharge for the combined Vakhsh-Panj basins. More than two-thirds of this contribution emerged in first and second-order streams, which were more numerous and where the entire streambank height (0.5 m) was assumed to consist of an active layer. Permafrost contributions from third- and higher-order channels were 0.33% and 0.16%, respectively, of average annual basin flows.