Future changes in precipitation over Central Asia based on CMIP6 projections

Multimodel simulations and projections from CMIP6 archive are used in this study (table 1; Eyring et al 2016). Based on monthly data availability, including precipitation, evaporation, near surface air temperature, specific humidity, horizontal wind and vertical velocity, the historical simulations and projections under four scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5) from 15 models are analyzed. The scenarios are the combination of Shared Socioeconomic Pathways (SSPs; O'Neill et al 2017) and forcing levels of the Representative Concentration Pathways (RCP). Only the first realization (r1i1p1) from each model is used. The outputs are interpolated onto the same resolution of 2.5° × 2.5° by the bilinear interpolation method to avoid the interpolation from low-resolution to high-resolution.

2.2.1. Moisture budget

A moisture budget analysis is used to investigate possible mechanisms responsible for the changes in precipitation (Seager et al 2010, Chou and Lan 2012), which can be expressed as

$$\begin{equation}P = - {\partial _t}\left\langle q \right\rangle - \left\langle {{\nabla _h} \cdot {V_h}q} \right\rangle - \left\langle {{\partial _p}\omega q} \right\rangle + E + \delta \end{equation} \tag{ 1 }$$

where P, E and q are precipitation, evaporation and specific humidity, respectively. ${V_h}$ and $\omega$ are horizontal wind and vertical velocity, respectively. < > denotes a vertical integration from surface to tropopause. Thus, $- \left\langle {{\nabla _h} \cdot {V_h}{\text{q}}} \right\rangle$ and $- \left\langle {{\partial _p}{{\omega \rm{q}}}} \right\rangle$ represent the horizontal moisture advection and vertical moisture advection, respectively. ${{\delta }}$ is the residual term, which is attributed to sub-seasonal transient eddies.

Neglecting the small changes in the tendency term, the changes in precipitation are balanced by those in evaporation and moisture advection, which is written as

$$\begin{equation}P' = - \left\langle {{V_h}{\nabla _h}q}\, \right\rangle^{\prime} - \left\langle {\omega {\partial _p}q} \right\rangle ' + E' + \delta ',\end{equation} \tag{ 2 }$$

where ( )' denotes the monthly anomalies relative to the climatology. The horizontal moisture advection $- \left\langle {{V_h}{\nabla _h}q} \right\rangle '$ and the vertical moisture advection term $- \left\langle {\omega {\partial _p}q} \right\rangle '$ can be further divided as

$$\begin{equation} - \lt {V_h}{\nabla _h}q{ \gt ^{\prime}} = - \langle \overline {{V_h}} {\partial _p}{\rm{q}}^{\prime} \rangle - \langle {V_h}^{\prime} {\partial _p}\bar{\rm{q}}\rangle - \langle {V_h}^{\prime} {\partial _p}{\rm{q}}^{\prime} \rangle ,\end{equation} \tag{ 3 }$$

$$\begin{equation} - \langle {\rm{\omega }}{\partial _p}{\rm{q}}\rangle ^{\prime} = - \langle \bar{\omega} {\partial _p}{\rm{q}}^{\prime} \rangle - \langle {\rm{\omega }}^{\prime} {\partial _p}\bar{\rm{q}}\rangle - \langle {\rm{\omega }}^{\prime} {\partial _p}{\rm{q}}^{\prime} \rangle ,\end{equation} \tag{ 4 }$$

and the first (second) term on the right-hand side of each equation is defined as the thermodynamic (dynamic) component of the horizontal/vertical moisture advection term, while the last term denotes the nonlinear component. The thermodynamic term is associated with the changes in specific humidity mainly directly induced by temperature changes as atmosphere circulations are fixed, while the dynamic term reflects the contribution of atmospheric circulation changes on precipitation.

2.2.2. Response to global warming

2.2.3. Definition of specific periods

CMIP6 models show increasing trends of annual mean precipitation throughout the 20th and 21th centuries over Central Asia under different scenarios (figure 1(b)). The increasing rates are comparable among four scenarios in the near term (figure 1(c)). Continued increases in annual mean precipitation during the 21th century can be seen for SSP3-7.0 and SSP5-8.5, while the trends turn to negative after 2080 for the other two scenarios. Thus, significant differences can be seen for the long-term changes under different scenarios. Compared with the present-day climate (1995–2014), the robust increase reaches 4.23 [2.60 to 7.36 for the interquartile range] %, 10.52 [5.05 to 13.36] %, 14.51 [8.11 to 16.91] %, 14.41 [9.58 to 21.26] % in long-term (2081–2100) for SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5, respectively. The model spread for precipitation changes is larger for higher emission scenarios, which can be partly attributed to the larger uncertainty of the changes in surface air temperature associated with larger radiation forcing (figure S1 (stacks.iop.org/ERL/15/054009/mmedia); Roe and Baker 2007, Zhou and Chen 2015). And a large heterogeneity of model uncertainty can be found as in other regions (Tabari and Willems 2018, Tabari et al 2019), with a larger model spread in the south of Central Asia under all scenarios (figure S2).

The sensitivities of precipitation to global mean surface air temperature increase under the four scenarios are given in figure 2. Annual mean precipitation exhibits robust increase with warming over most part of Central Asia, especially the Tianshan mountain and the northern part of Central Asia as in previous studies (Huang et al 2014). The regions with larger response rates of annual mean precipitation to global warming also have more precipitation in present-day climatology (figure 1(a)). The wet-getting-wetter pattern implies that the effect of increasing moisture on the CA rainfall may be dominant under global warming. A comparison between different scenarios reveals a decrease of response rate accompanied with higher emission as in previous studies (Wu et al 2010, Hegerl et al 2015). Area-averaged annual mean precipitation increases with warming at the rates of 5.93 [0.27 to 10.09] % K⁻¹, 4.22 [2.88 to 5.73] % K⁻¹, 4.46 [2.71 to 5.09] % K⁻¹ and 3.47 [2.23 to 4.76] % K⁻¹ for SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5, respectively. The results of SSP2-4.5 and SSP3-7.0 are close to each other as the annual mean precipitation changes under these two scenarios are indistinguishable until 2080 (figure 1(b)). Though the response rate is highest for SSP1-2.6, the areas covered by positive signals are smaller compared with other scenarios as insignificant negative responses can be found over the southeast of Uzbekistan and Turkmenistan (figure 2(a)).

Zoom In Zoom Out Reset image size

In addition to the changes in annual mean precipitation, changes in the annual cycle also have important effects on water safety in CA. We first examine the annual cycle of area-averaged climatological precipitation over the north of Central Asia (NCA, $48^\circ - 55^\circ N,45^\circ - 90^\circ E$) and south of Central Asia (SCA, $35^\circ - 48^\circ N,45^\circ - 75^\circ E$) in different periods under the four scenarios (figure 3). Modulated by the meridional shift of subtropical westerly jet, the precipitation in the NCA and SCA have distinct annual cycle (Schiemann et al 2008, Bothe et al 2012). Consistent with previous studies, two peaks occur in May-July and October-December for NCA and a peak occurs in March-May for SCA can be found in the present-day (black lines). Compared with historical simulations, projected increasing trend in spring and decreasing trend in summer can be found in both the two regions. The changes are more significant in the long-term under higher emission scenario. Besides the changes in the amount of precipitation, a shift of the first peak for NCA can be found for SSP2-4.5, SSP3-7.0 and SSP5-8.5 (figure 3(b)). From the near-term, to mid-term and to long-term, the peak in May-July gradually shifts to March-May, making the annual cycle of precipitation in NCA close to that in SCA. Due to the different annual cycle in these two domains in the present-day (figure 3, black lines), the changes in seasonal mean can result in enhanced seasonalities for both two sub-regions and a shift of the first peak for NCA.

Zoom In Zoom Out Reset image size

To further investigate the physics behind the changes in precipitation under global warming, the moist budget method is performed in the subsequent analysis. The responses of moisture budget terms to global warming are shown in figures 4. Only the results of SSP5-8.5 are shown as the changes under the high emission scenario are the most significant. The changes in precipitation in NCA (figure 4(a), blue line) are similar to that in SCA (red line) with positive responses in winter and spring whereas negative in summer as revealed in section 3.2.

Zoom In Zoom Out Reset image size

The changes in evaporation coherently couple with the changes in precipitation except the cold winter (figures 4(a) and (b)), suggesting a positive feedback between precipitation and evaporation. The evaporation increases in all the seasons and reaches a peak in spring under global warming (figure 4(b)) consistent with a previous study based on the CMIP5 simulations (Yu et al 2018). The wetting trend (0.075 [0.032 to 0.087] mm day⁻¹ K⁻¹ and 0.020 [−0.027 to 0.052] mm day⁻¹ K⁻¹ for NCA and SCA, respectively) in spring is balanced by the increasing of evaporation (0.078 [0.050 to 0.086] mm day⁻¹ K⁻¹ and 0.041 [−0.017 to 0.056] mm day⁻¹ K⁻¹ for NCA and SCA, respectively). As for summer, the drying trend (−0.021 [−0.073 to −0.007] mm day⁻¹ K⁻¹ and −0.019 [−0.062 to −0.007] mm day⁻¹ K⁻¹ for NCA and SCA, respectively) is dominated by the decreasing trend of vertical moisture advection (−0.046 [−0.096 to −0.037] mm day⁻¹ K⁻¹ and −0.044 [−0.081 to −0.017] mm day⁻¹ K⁻¹ for NCA and SCA, respectively), while the contribution of evaporation and horizontal advection term can be neglected. Besides, the evaporation in spring and the vertical moisture advection in summer during the long-term can be statistically distinguished from that in preset-day at the 5 % level based on Student-t test, showing the changes are robust.

As a vast semiarid to arid region, Central Asia is dominated by descending motion and low-level divergence in summer (Yatagai and Yasunari 1998, Guan et al 2019). So the increase of atmospheric water vapor due to global warming, the thermodynamic effect, can result in the decrease of vertical moisture advection (figures 4(e) and (f),−0.018 [−0.050 to −0.009] mm day⁻¹ K⁻¹ and −0.082 [−0.093 to −0.064] mm day⁻¹ K⁻¹ for NCA and SCA, respectively) in summer, which is the so-call 'dry-getting-drier' mechanism. The larger thermodynamic drying effect in SCA relative to that in NCA corresponds with stronger descending motion in SCA (figure not shown). In contrast, the contributions of dynamic effects are different in these two domains, which contributes to the drying trend in NCA by −0.019 [−0.070 to −0.008] mm day⁻¹ K⁻¹ whereas partly offsets the drying trend in SCA by 0.029 [0.013 to 0.041] mm day⁻¹ K⁻¹. Both the dynamic and thermodynamic components for present-day and long-term are statistical distinguished at the 5 % level. Summer precipitation over Central Asia is modulated by Subtropical Westerly Jet (SWJ) and weakened SWJ can result in less precipitation in NCA and more precipitation in SCA (Zhao et al 2014, 2018). Thus, the different dynamic effects between NCA and SCA may be associated with the changes in SWJ which needs further study.