Dryland belt of Northern Eurasia: contemporary environmental changes and their consequences

The surface air temperature across the DLB region had been stable from the beginning of the 20th century until the mid-1960s, but has increased rapidly in the past five decades (figure 1(b)). These changes are mostly illustrated by changes during the winter and spring seasons with mean rates of change of 1.8 °C (100yr)⁻1 and 1.6 °C (100yr)⁻¹, respectively. In winter, the temperature variability is higher with the 1.8 °C (100yr)⁻¹ linear trend explaining 23% of the variation, whereas a slightly smaller warming trend explains 37% of the variation in spring temperatures. For autumn and summer seasons, the warming signals have been much smaller with rates of 0.7 °C (100yr)⁻¹ in autumn and 0.4 °C (100yr)⁻¹ in summer, with the changes concentrated solely in the past three decades. This warming shifts the temperature seasonal cycle, particularly the earlier dates of snowmelt (Arctic Climate Impact Assessment (ACIA) 2005, Bulygina et al 2013, Tomaszewska and Henebry 2018), earlier spring onset in the biosphere, and freshet in river discharge. In summer, the warming causes glacial retreat in Central Asia, the Caucasus, and Southern Siberia (Khromova et al 2014) and exacerbates water deficit for the DLB landscapes. In the past three decades, warming in the DLB has spread from Hungary in the West (Mátyás et al 2018) to Northeast China in the East (Zhao et al 2013) exaggerating dry weather conditions. These and many other negative consequences of the DLB warming could be avoided, were precipitation to increase. Indeed, oceanic warming must result in more evaporation, and the additional atmospheric water vapor could mitigate the water deficits in lands of the continental interior through increased precipitation. However, this water vapor must first be transported to the interior by atmospheric circulation.

The DLB receives an abundance of heat through insolation, although the drylands do not have sufficient holding capacity to retain heat in soil, biosphere, or hydrological objects. Consequently, day/night temperature differences are very high in comparison to maritime climates. Soil moisture that could mitigate strong diurnal temperature swings is limited in drylands and is often concentrated below the rooting zone. Dryland areas also have limited heat capacity and finite water storage in glaciers, interior lakes, and permafrost accumulated during the past pluvial epochs and/or during cold seasons. The remaining water that the region receives comes from atmospheric precipitation, ground water, and inland lakes. However, precipitation is not necessarily transferred inland via the atmosphere. The role of atmospheric circulation is critical here. There appear two competing major components of the global warming process that are, among others, responsible for water vapor transport into the interior drylands of Northern Eurasia.

Disproportional warming at high latitudes and in the Arctic (Arctic Climate Impact Assessment (ACIA) 2005, Blunden and Arndt 2017) decreases the equator-to-pole temperature differential with the latitudinal temperature gradient instrumental for the westerlies' circulation in the extratropics. In particular, in the Northern Hemisphere westerlies move the water vapor from the North Atlantic Ocean into the Eurasian interior. The weaker the westerlies, the less water vapor penetrates into the northern part of the continent exposed to the Atlantic and Arctic air transfer (i.e. most of Northern Eurasia). Another feature associated with weaker westerlies is the meandering of their flow and more frequent formation of atmospheric blocking (Mokhov et al 2013, Lupo et al 2014). These changes in atmospheric circulation lead, in the cold season, to a larger variability when unusually cold and warm weather conditions occur over the entire Northern Eurasia (Schubert et al 2014). In the warm season, it results in prolonged periods of days with and/or without precipitation (Groisman and Soja 2009, Zolina et al 2013, Lupo et al 2014).

In summer, the land warms more strongly than the ocean, which is projected to strengthen the monsoon circulation with global warming (IPCC 2013). The southern edge of the DLB is mostly blocked from the warm Indian Ocean, but its eastern edge is exposed to the Pacific Ocean and warm humid air from the Pacific penetrates northwestward. Thus, with the stronger monsoons, a more humid climate would occur on the eastern edge of the DLB and will be expected to continue in the future with continued global warming (Collins et al 2013). However, the observations in Northeast China (Guo et al 2013) and analysis of the eastern edge of the DLB do not support these expectations (figure 2).

Zoom In Zoom Out Reset image size

Figure 2 and table 1 show major statistics of cyclone characteristics for the four regions of the Asian part of the DLB. In figure 2, these regions are identified with four figure inserts that visualize the time series of the annual count of individual cyclones that crossed the regions. Tilinina et al (2013) showed that the total cyclone count is dependent of the spatial resolution of the post-USSR period tracks. Here we use the ERA-Interim reanalysis (Dee et al 2011) that has a spatial resolution of 1.5° × 1.5°. The climatology of the cyclone counts shown in table 1 depends on this resolution. The Asian DLB includes many mountain and plateau systems. Reanalyses have shown that these areas are very sensitive to the observational network used for their generation. In particular, when these networks have been changing, there can be spurious inhomogeneity in the data (Arsenault and Brissette 2014). Therefore, we used only the post-Soviet Union period (prior to this period the networks were much denser in Central Asia and in the eastern half of the DLB, but these data were less available to the developers of the reanalysis products).

The one central and two western regions of the Asian DLB in figure 2 receive most of their atmospheric water vapor from the West (Shver 1976, Kuznetsova 1983), but the easternmost DLB region—Northeastern China and Western Mongolia—resides in a monsoon climate benefiting from major water transport from the Pacific Ocean. Over the past three decades, we have not seen systematic changes in cyclone counts in this region (figure 2). In contrast, the numbers of cyclones in Central Asia, Northwest China, and Mongolia increased. We cautiously conclude that these increases indicate improvement in the water budgets of these drylands.

Water vapor transport into these regions during recent decades can provide an estimate of integral precipitation changes related to atmospheric circulation. Table 2 reports the results of processing of the ERA-Interim reanalysis data for the post-USSR period. It demonstrates that most of water vapor integrated over the atmospheric column comes from the West, while the southward water vapor transfer is substantially less. Moreover, the dramatic reduction of the eastward water vapor transport into Central Asia (∼15% during the last three decades) is not only the result of weakening of the westerlies but may be related to the simultaneous retreat of the Aral Sea (the 60°E longitude crosses its remnants). The Aral Sea has been shrinking since the 1960s with the largest decrease in the past two decades (Zavialov 2007, Gaybullaev et al 2012). In addition, large-scale climate oscillations can affect precipitation patterns in Central Asia (de Beurs et al 2018). Whatever the cause of the water vapor transport decline, it contributed to drier weather conditions in Central Asia (see section 2.5).

Atmospheric precipitation, its amount (total), form (frozen, mixed, or liquid), intensity (for rainfall), intraseasonal distribution, and the systematic changes (trends) remain the most variable characteristics of the DLB climate. Past studies have documented an earlier onset of spring across the entire Northern Hemisphere (Schwartz et al 2006), increases in daily rainfall intensity during the heavy rain events that may coincide with prolonged no-rain periods (Zolina et al 2010, Groisman et al 2013, Zolina et al 2013), and uncertainties in quantifying the changes in precipitation. The uncertainty emerges in this part of the globe both from the relatively sparse observational networks and from time-dependent systematic biases in precipitation records (Groisman and Legates 1995). Groisman et al (2014) quantified these biases for Russia, while Ding et al (2007) did the same for China. They showed how each improvement in rain gauge instrumentation, wind shielding of gauges, and observing routines resulted in increases of 'observed' precipitation while the actual 'ground-true' precipitation was quite different and had different decadal trends. As a result, the latest Second National Climate Change Assessment for the Russian Federation contains two estimates of the past precipitation changes over the nation rather than just one (Second National Climate Change Assessment for the Russian Federation 2014). The first set of estimates is based on high-quality observations from the national meteorological network and the second set of estimates is based on the same observations but corrected for the time-dependent biases that provide the real ground-true precipitation values. The first set of estimates reports nationwide increase of annual precipitation; whereas, the second set reports moderate multi-directional changes. For the Russian part of the DLB (steppes in the south of West Siberia and the Trans-Baikal region), the assessment reports a long-term decrease in precipitation. Using bias-corrected precipitation time series, Akhmadiyeva and Groisman (2008) reported only a 4% increase in annual precipitation over Kazakhstan during the 1990–2006 period compared to previous three decades (1960–1989). Furthermore, significant discrepancies between a global precipitation reanalysis and a fuller representation of regional precipitation stations was documented for Kazakhstan (Wright et al 2009).

Difficulties with the reporting of precipitation trends in other parts of the former USSR as well as in Mongolia are the same and, therefore, alternative approaches based on remote sensing land surface products were used (see Lioubimtseva and Henebry 2009, de Beurs et al 2015). In Kazakhstan in the 1990s, precipitation deficit growth was reported using normalized difference vegetation index (NDVI) analyses. Farther southward, Wang and Zhou (2005) and Ding and Chan (2005) reported an increase in precipitation in the northwest of Xinjiang province in China. Ding et al (2007) showed the importance of bias corrections for precipitation reports in China and addressed the possible time dependence of biases in these reports. The biases can be caused by changes in observations, but they can be also introduced by changes in other 'natural' factors that affect precipitation measurements. In particular, systematic continent-wide reduction in the near-surface wind speeds (see Ding et al 2007, Bulygina et al 2013) may 'increase' the observed cold season precipitation while the ground-true precipitation is unchanged or even decreased. Figure 3(a) shows that during most of the year (autumn, winter, and spring), there were no significant changes in the total precipitation over the DLB (a region-wide increase by 16 mm per 8 months).

Zoom In Zoom Out Reset image size

Liquid precipitation observations are less prone to biases than those for frozen precipitation; thus, changes in liquid precipitation can be presented in absolute numbers (in mm instead of percent). In figure 3(b), we show the latest changes in summer rainfall totals over the DLB during the past 13 years compared to the previous 25 years estimated from the MERRA-2 reanalysis output. These changes are mainly negative illustrating the Northern Eurasian heat waves and droughts described by Schubert et al (2014). These waves over the western part of the DLB (Hungary, the Ukraine, and European Russia) manifested themselves by several severe droughts (including the extreme drought of 2010). Over the eastern part of the DLB (the Baikal Lake Basin, Trans-Baikal area, and Northeastern China) the waves were responsible for increasing water deficit and severe forest fires centered in the Trans-Baikal area (Loboda and Chen 2017).

All components of the cryosphere in the DLB have been changing in recent decades, with most of these changes showing stable decreases (AMAP Arctic Monitoring and Assessment Programme 2011, 2017):

• Duration of seasonal spring snow cover has been steadily decreasing, making the onset of spring (e.g. the start of vegetation growth and spring freshet on rivers) earlier by several days (Bulygina et al 2011, Second National Climate Change Assessment for the Russian Federation 2014).
• Period with stable snow cover over Europe has become shorter and/or disappears causing river ice break-up, and increases in winter runoff (see section 2.5).
• Permafrost at the edge of the permafrost zone, especially in the areas of discontinuous permafrost, has begun to thaw (Romanovsky et al 2017).
• The depth of seasonal frozen ground thawing has increased from the Arctic to the mountains of Central Asia affecting infrastructure, such as roads and buildings (Shiklomanov et al 2017).
• The area and volume of land ice (i.e. glaciers) has been decreasing across the entire Northern Eurasia including the montane areas of the DLB (Shahgedanova et al 2010, Khromova et al 2014, Shahgedanova et al 2014, Kotlyakov et al 2015, Syromyatina et al 2015).

Ice storage on and below the ground is an important source of the water supply for the DLB. For centuries (or millennia) water was transported to the remote mountains of the Eurasian continent and, instead of entering the global water cycle via runoff, was fixed and stored there in frozen form. This storage served as a cushion for the water balance of the DLB regions located at the foothills of high mountains. Currently, this storage is being gradually depleted.

Ongoing global warming is more prominent at higher elevations than in lowlands (Barry 2008). The temperature increases, particularly during the warm season, escalating glacial melt and the thawing of subsurface ice. Although the abundant summer streamflow promises a strong future water supply in the drylands, the storage of water in glaciers is limited. As illustrated in figure 4, three rivers in neighboring valleys have distinctively different tendencies of their discharge depending upon the glaciation level at their higher elevations. Snow-free high dry plains and disappearing lakes in Northwest China (e.g. in Qinghai province) and West Kazakhstan are good examples.

Zoom In Zoom Out Reset image size

Runoff data for the Russian Federation as well as for most of the Central Asian countries show an increase in river discharge. However, it is not clear how persistent these increases will be in the future once several factors that are currently favorable for these increases change (supplement 2).

We have a complete set of socio-economic and meteorological data for the Central Asian nations, and our analysis of climate change impacts on different aspects of water availability here was made using a combination of economic census data with hydrological modeling (supplement 3). The dynamics of water use for domestic, industrial, and livestock needs have been simulated using the University of New Hampshire water balance model (WBM, supplement 3) using country-based statistical socio-economic information along with spatially distributed population density. The water scarcity index (WSI; Damkjaer and Taylor 2017) and water availability indexes (WAI; Schyns et al 2015) for the Central Asian countries were estimated with WBM simulations using only locally generated water resources (figure 5, left panels) and total available water resources (including inflow) (figure 5, right panels).

Zoom In Zoom Out Reset image size

The WSI is the ratio of annual freshwater abstractions (i.e. system losses due to water consumption, ET, and deep drainage) to the annual water availability. This information was aggregated over administrative units and adjusted using census data. Water stress is classified from high water stress (red in figure 5) to water abundance (blue in figure 5). The WSI neglects temporal and spatial variations as well as water quality data. There was a general and substantial decline in available water between the 1990s and 2014 (figure 5). The water security situation across the region is more stressful despite significant political and socio-economic transformations during the 1990s, which led to decreased water use and increases in river runoff in some montane areas, as mentioned above. The current trend toward increasing water stress results primarily from changing climatic conditions coupled with fast population growth (https://www.populationpyramid.net/central-asia/2017/).

One of the biggest episodes of land cover change over the steppes of the DLB was the so-called 'Virgin Lands' development during 1954–1964, when the area of arable land in Kazakhstan was expanded from 7–8 to 21–23 million hectares. These changes were also spread across the steppe zone of Western Siberia and Southern Russia but were centered over Northern Kazakhstan (Jackson 1962). This massive conversion of grassland to cropland resulted in statistically significant increases of monthly surface air temperatures by 0.3° to 0.5 °C during the spring, summer, and autumn seasons and significant changes in ET (Yesserkepova 1988). Even within the interdecadal variability of the near-surface temperature and humidity in this part of the world, the impacts of large-scale virgin land development are evident in figure 6.

Zoom In Zoom Out Reset image size

The latest land cover types and their changes within the Asian part of the DLB were examined using the MODIS land cover product (MCD12Q1V6; https://earthdata.nasa.gov/) through re-grouping the International Geosphere-Biosphere Program (IGBP) classification cover types between 2001–2016. This product provides global land cover types at yearly intervals (2001 to date) at 500 m spatial resolution and is a combined product derived from Terra and Aqua reflectance data. Version 6 of the MCD12Q1V6 data product was derived from different classification schemes, of which we used the IGBP classification scheme validated to stage 2 maturity level (Friedl and Sulla–Menashe 2015). Because land cover changes within a short time period (i.e. <3 years) are likely minimal and difficult to detect with MODIS imagery, we compared the difference between 2001–2014, 2002–2015, and 2003–2016 to ensure that the three differences are consistent and, hence, representative of the actual land cover changes during the 15-year study period.

We used the tabulate area function in ArcGIS to categorize area by cover type and their relative proportions by the 10 administrative units in the Asian part of the DLB (8,779,162 km²). We created a triad of image difference datasets, namely 2001–2014, 2002–2015, and 2003–2016, to compare change across six years, thereby making the findings robust (three years in the early period of observations and three in the last three years). We then labeled and visualized the pixels with consistent change (red), some change (blue, when one of the three triads shows a different change) and no change (light gray) with 5.58%, 0.18%, and 94.24% of the area respectively. The pixels with 'some change' are labeled with blue dots. They are nearly invisible in figure 7(b). The small number of such dots indicates that the post-classification land cover changes due to uncertainties in the land cover labels for any one year and do not seriously impact our conclusions about the regional land cover change.

Zoom In Zoom Out Reset image size

The DLB is dominated by arid barrens and temperate grasslands that accounted for 56.85% and 31.36% of the total land area (total = 88.21%), respectively, during 2001–2016 (figure 7(a)). The arid barrens are the largest on the Earth's surface and stretch from the Gobi Desert in Southwest Mongolia to the Western portion of Central Asia (Mildrexler et al 2006, Chen et al 2014). The grasslands included those on the Mongolia Plateau, which are experiencing higher warming trends than the global average due to a combination of high latitude, high elevation, and a continental climate (John et al 2018, Tian et al 2018). The croplands (4.80%), shrublands (1.48%), and forests (1.14%) are the next three major cover types in the region. However, the distributions of these cover types varied substantially by country. East Asian countries have a similar proportion of grasslands (45.86%) to those in Central Asia (45.50%). Kazakhstan has the highest grassland cover (85.2%) among all administrative units (figure 7(a)). Overall croplands in Eastern and Central Asian countries occupy 4.13% and 7.60%, respectively. Forests were found in Northeast Inner Mongolia, and on northern aspect slopes in the Tian Shan mountains of Central Asia.

Land use was also very extensive and intensive across the DLB, with 5.58% of the land experiencing consistent cover change from 2001–2016 (figure 7(b)). Overall, it appeared that countries in Central Asia experienced more cover changes than those in East Asia. Several visual hotspots included land cover changes that are apparent around the Caspian Sea in Turkmenistan, Uzbekistan, and Southwest Kazakhstan (I) and around the edges of the East Asian drylands of Western Xinjiang–Inner Mongolia–Mongolia (III and V). Other major hotspots of land cover change were found in the Yellow River Delta or Hetao region (IV), forested regions of Northeast Inner Mongolia (VII), Northern Mongolia (VI), and Northwestern Kazakhstan (II). These hotspots have been widely reported as being the results of major policy shifts by the individual countries and, to a lesser degree, the changing temperature and ET (i.e. water loss) (Jung et al 2011, Liu et al 2013, Chen et al 2014, Groisman et al 2017, Qi et al 2017, John et al 2018). For example, the large increases in forest cover in the Yellow River Delta and Northeast Inner Mongolia are likely due to afforestation/reforestation program of China's Green for Grain Program (Liu et al 2014), whereas hotspots in Western Xinjiang are due to the rapid expansion of agricultural lands. Meanwhile, hotspots in Central Asia may have been the direct results of elevated ET loss, salinization of soils, or new agricultural enterprises.

Biosphere modeling employed for this projection is described in supplement 3. Here we present and discuss the biosphere modeling output. In the baseline climate, drylands were projected to cover about 50% of the Siberian window that was allocated equally between steppe and semi-desert/desert. Simulations indicated that Siberian vegetation would be altered by the 2080s over the study area. Almost no change in the forest distribution was projected in the 20 GCM ensemble under both the RCP 2.6 and RCP 8.5 scenarios, while the extent of the drylands changed minimally. A 6% increase in drylands was projected under the RCP 8.5 scenario for the 20 GCM ensemble (table 3, figure 8(c)). According to the ensemble of the five driest GCMs under RCP 8.5, drylands would cover 63% of the study area with a 10% predominance of steppe. The semi-desert/desert area would increase only slightly compared to its current extent. Forest coverage would remain similar to the current conditions (i.e. ± 1.5%–2%), because the forest would shift into the current tundra distribution and the tundra would nearly vanish. Only in a dry climate would its coverage decrease by about 12.5%.

Zoom In Zoom Out Reset image size

The ecological consequences for forestry and agriculture in the rapidly changing environment of Southern Siberia will require adaptive management to adjust agricultural and mixed agro-forestry practices to the newly emerging forest steppe and steppe habitats of the late 21st century.

Increased tree mortality in a drier climate along the southern Taiga border would lead to accumulation of woody debris, which in turn, paired with increased 'fire' weather, could result in destructive fires and shift forests northward to wetter habitats. Grasslands that follow in the wake of the forest are better adapted to frequent fire events because of their shorter life cycle, a stronger adaptability to less precipitation and droughts, and their capability to recover due to the allocation of perennating plant parts below ground (Tchebakova et al 2009). Generally, the projected warmer and drier climates will promote drylands to extend over southern Siberia (Tchebakova et al 2010, Tchebakova et al 2011a) and the steppe area would increase by 50%. South of 60°N, tundra is observed only in the highlands on mountains and in the RCP 8.5 scenario it vanishes completely.

To minimize the negative consequences and to enhance the benefit of climate change in Siberian forests and drylands, adaptive measures need to be implemented, including forest restoration on failed forest lands and planting different crops suitable for the new climates. Due to recent and predicted climatic changes, concerns about food security increase. However, in the cold climates of Siberia, agriculture would benefit from climatic warming, only when and where appropriate infrastructure development and population growth permit.

Spatial distribution of livestock density on temperate Eurasian steppe in 2006 was estimated based on province/prefecture-level inventory data of livestock. The carbon ingested by livestock (C_graze) was calculated using the pasture system simulator in a boreal ecosystem productivity simulator (BEPS; Chen et al 2017). The distribution and C_graze are tightly controlled by sub-regional socio-economic conditions (Chen et al 2014). Spatial heterogeneity exists across and within different administrative divisions (figure 9). In the Mongol steppe, major consumption by livestock was concentrated in China, particularly in Inner Mongolia. The most extensively grazed grasslands were located in traditional natural pastures with relatively high productivity, such as the XilinGol Prairie in mid-Inner Mongolia, and in some areas of desert steppe in Southwestern Inner Mongolia. Livestock density and C_graze in those areas exceeded 800 sheep units per km2 and 30 gC m⁻² yr, respectively. In contrast, the livestock density and C_graze in Mongolia were much lower (John et al 2016). This contrast can be seen in the bordering areas of Eastern Mongolia (Dornod province), where livestock density is lower than 40 sheep units per km2 and Northeastern Inner Mongolia (Hulunbuir City), where density is higher than 200 sheep units per km².

Zoom In Zoom Out Reset image size

In the Kazakh steppe, C_graze was highly concentrated in the South, especially in the desert countries of Turkmenistan and Uzbekistan. These two countries contain huge numbers of livestock within limited pasture ecosystems. Livestock density and C_graze are generally higher than 500 sheep units per km2 and 12 gC m⁻² yr for the two countries (figure 9), while livestock density is generally lower than 40 sheep units per km² over the steppe of Kazakhstan.

Concern over increasing demands for animal products and preservation of grassland resources has led to the launch of regional conservation programs in Inner Mongolia since the late 1990s, including the Grain to Green Program (Liu et al 2008) and the Grazing Withdrawal Program (Chen et al 2017). As a result, improvements to livestock habitat and ecosystem service functions have been observed in recent years (Li et al 2012, Mu et al 2013). In contrast, the excessive stocking rates in the southern Kazakh steppe have not received enough attention from the scientific community or from governments (Blench and Sommer 1999, Mirzabaev et al 2016). Field investigations, large-scale assessments, and specific rehabilitation programs are urgently needed in this area.

Prior to 1990, the Central Asian economies were not economically independent. Most agricultural and livestock production was focused on meeting demands dictated by the central planners of the USSR (e.g. production of cotton and meat). As a result, the reduction of livestock numbers during the 1990s was mainly induced by institutional changes, loss of access to markets, and agricultural reforms following the dissolution of the USSR (de Beurs and Henebry 2004, Ojima and Chuluun 2008, Wright et al 2012, de Beurs et al 2015, John et al 2016). Clearly, the situation in the 1980s or 1990s cannot be considered as the baseline for future projections because of the very different governance and socio-economic systems.

Future projections show that, with the rapid population growth and dietary transitions associated with economic development, the demand for animal products will continue to increase. For example, global meat demand is projected to double by 2050 (Flammini et al 2017). This new demand level will lead to continuously increasing production of meat and milk from East and Central Asia (Bruinsma 2003).

This situation would require future expansion of the scale of pasture systems and/or enhanced feed efficiency, such as through concentrated animal feeding operations. In order to meet the increasing demand, pasture land would need to be expanded by ∼8%, and feed efficiency enhanced by ∼30%, compared to the 1990s level in East and Central Asia (Wirsenius et al 2010). However, both arable land and water resources limit further feed production expansion in many areas of the DLB (Qi et al 2012, Qi et al 2017).

Meanwhile, large-scale unregulated grazing patterns have caused large reductions in biodiversity and productivity, and led to degradation and desertification, even in historically productive and stable pasture systems (Yusupov 2003, John et al 2009, Miao et al 2009, Chen et al 2014, John et al 2018). Therefore, addressing the question of how to enhance feed efficiency in a sustainable way is urgently needed to meet the challenges from both human activities and climatic changes to the pasture systems of the DLB.