Comparative Analysis of the Mechanisms of Intensified Summer Warming over Europe−West Asia and Northeast Asia since the Mid-1990s through a Process-based Decomposition Method

The surface air temperature (SAT) is an important climate parameter of Earth’s system. Not only can it reflect the thermal conditions of the ground, but also has an influence on other physical variables, such as the distribution of the circulation field (Levine and Schneider, 2011; Power and Kociuba, 2011) and the intensity of evaporation and precipitation (Held and Soden, 2006; Donat et al., 2016). The distribution and evolution of SAT is therefore very important in climate studies. Previous studies have found that the SAT in Europe−West Asia and Northeast Asia has experienced a remarkably amplified increase since the mid-1990s, relative to other regions in Eurasia (Sutton and Dong, 2012; Stainforth et al., 2013; Chen and Lu, 2014; Dong et al., 2016, 2017; Chen et al., 2017; Hong et al., 2017), but the reasons for the amplification remain unclear. In view of the effects of changes in SAT on populations and the economy, investigations into such amplified warming are of social significance.

Chen and Lu (2014) pointed out that the decadal shift of summer Northeast Asian SAT around the mid-1990s was a remote response to the enhanced rainfall in South China. Sutton and Dong (2012) suggested that the Atlantic Ocean is the key driver for European climate change since the mid-1990s, and that the current climate will persist as long as the positive phase of Atlantic Multidecadal Variability (AMV) persists. Dong et al. (2016, 2017) utilized the atmospheric component of the global climate model, HadGEM3, to quantify relative roles of sea surface temperature (SST)/sea ice extent (SIE), anthropogenic greenhouse gases (GHGs) and aerosols, and found that 76% (62.2%±13.0%) of the area-averaged summer surface warming over Northeast Asia (Western Europe) can be explained by the changes in SST/SIE, with the remaining 24% (37.8%±13.6%) interpreted by direct impacts of changes in GHGs and aerosols. However, these works focused on several separate areas without considering their connection.

Based on an observational analysis, Hong et al. (2017) proposed that the intensified warming over Europe−West Asia and Northeast Asia is closely related to the phase shift in the AMV. The AMV impacts the SAT by modulating the decadal change in the Silk Road pattern (SRP)—the quasi-zonally propagating teleconnection wave train across the Mediterranean, West Asia, Central Asia and East Asia observed during the summer months (Lu et al., 2002; Enomoto et al., 2003; Kosaka et al., 2012; Hong and Lu, 2016). The SRP is considered to be the Eurasian portion of the circumglobal teleconnection (CGT) pattern, which is located along the northern hemispheric westerly jet during summer (Ding and Wang, 2005). The CGT pattern is excited by the Indian summer monsoon on the intraseasonal (Ding and Wang, 2007) or interannual time scale (Ding and Wang, 2005; Lin, 2009; Ding et al., 2011), and is associated with the AMV on the interdecadal time scale (Lin et al., 2016; Si and Ding, 2016; Wu et al., 2016). The SRP mainly occurs on intraseasonal (Ding and Wang, 2007) and interannual time scales (Lu et al., 2002; Enomoto et al., 2003). However, it also occurs on the decadal timescale, and the decadal variability explains about 30%−50% of its total variance (Hong et al., 2017; Wang et al., 2017). The SRP triggers wave-train-like descending−ascending−descending motion along its propagation path, leading to different changes in SAT over different regions. Given the limited observational records used in Hong et al. (2017), Sun et al. (2019) verified the role of the AMV in shaping the non-uniform warming in Eurasia using three atmospheric general circulation models forced by AMV SST anomalies. However, the conclusion of Hong et al. (2017) and Sun et al. (2019) contained uncertainty. As a dynamical wave-train, the SRP can effectively impact SAT through generating ascending or descending motion along its propagation path. However, observed SAT anomalies are the result of various dynamic and thermodynamic processes, and thermodynamic feedback processes related to the SRP can also affect the SAT. For example, the SRP-produced ascending or descending motion can cause increased or decreased cloud cover, which can feed back to the net longwave or shortwave radiative flux, leading to changes in SAT. In addition, SRP-induced atmospheric circulation anomalies can cause changes in horizontal heat transport, and anomalous warm temperature advection is a contributor to positive SAT anomalies over Europe and Northeast Asia (Sun et al., 2019). Therefore, it is necessary to diagnose the specific physical processes responsible for the enhanced warming in Europe−West Asia and Northeast Asia. Only when the warming over these two regions results from similar physical processes, can it be further verified that they may have a common origin—the AMV-associated SRP wave-train. This is conducive to predictions of East Asian and European climate, since the AMV provides a source of predictability on the decadal time scale (Monerie et al., 2018; Wu et al., 2019).

In this study, we adopt the coupled surface−atmosphere climate feedback−response analysis method (CFRAM) (Cai and Lu, 2009; Lu and Cai, 2009) to perform the physical diagnostic analysis due to its advantage of resolving processes. In the framework of CFRAM, the local SAT can be decomposed into partial temperature changes due to various radiative and non-radiative processes, the individual contributions of which to the total change in SAT can be quantitatively calculated. Thus, we can diagnose the effects of different physical processes and gain a better understanding of the sources of temperature changes between two climate states. Here, we aim to use CFRAM to diagnose the enhanced warming over Europe−West Asia and Northeast Asia, to find out whether the amplification over these two regions originates from similar physical processes and to further validate the role of the SRP in warming these two regions.

The rest of this paper is organized as follows. Section 2 describes the data used in this study and gives a brief introduction to CFRAM. Section 3 reports the trends in SAT over the Eurasian continent since 1979. Section 4 considers the total and partial changes in SAT resulting from the individual processes decomposed by CFRAM. The underlying processes of the primary contributors are analyzed in section 5, and a summary of our conclusions is presented in section 6.

The variables used in CFRAM include the incident shortwave radiation at the top of the atmosphere (TOA), the surface albedo, the surface pressure, the cloud cover, the cloud liquid water/ice water content, the specific humidity, the ozone mixing ratio, the sensible and latent heat fluxes, the air temperature, and the surface temperature. All the data apart from the surface temperature were obtained from the ERA-Interim dataset (Dee et al., 2011). ERA-Interim covers the period from 1979 to present with a horizontal resolution of 1.5°×1.5° and 37 vertical pressure levels ranging from 1000 to 1 hPa. Given this study is based on the observed amplified warming, the observational land surface temperature from the Climate Research Unit (CRU) dataset, version 3.24 (Harris et al., 2014), is used in the CFRAM surface layer. The temperature measurements in CRU are assumed to be at a height of 2 m, but they bear substantial similarities with the ERA-Interim skin temperature in terms of both spatial pattern and area-averaged magnitude (not shown). In addition, the NOAA/NCEP GHCN CAMS SAT dataset (Fan and van den Dool, 2008) and University of Delaware SAT data, version 5.01 (Willmott and Matsuura, 2001), were also adopted to validate the robustness of result. Since human activities (especially the combustion of fossil fuels) influence SAT, CO₂ is shown due to its high concentrations and substantial contribution to global warming (IPCC, 1995; Myhre et al., 1998). The CO₂ concentrations were derived from the meteorological station on Ascension Island recorded by NOAA’s Earth System Research Laboratory Global Monitoring Division (www.esrl.noaa.gov/gmd/dv/data/). The CO₂ concentration was 350.4 ppm during the time period 1979−1996 and 381.2 ppm during the time period 1997−2015.

The monthly mean evaporation, geopotential and horizontal winds (u, v) for the time period 1979−2015 from the ERA-Interim dataset were used to explain certain partial temperature changes. SST anomaly data from 1856 to 2016 were obtained from the Kaplan Extended Sea Surface Temperature dataset, which has a resolution of 5° × 5° (Kaplan et al., 1998). The AMV index is defined as the annual mean SST anomaly in the North Atlantic basin (0°−60°N, 75°−7.5° W) (Enfield et al., 2001; Wang et al., 2009). The linear trend in SST was removed, and a nine-point running mean was used to obtain the decadal components when calculating the AMV index. Summer refers to June−July−August (JJA).

CFRAM is based on the energy balance within the atmosphere−surface column. It consists of 37 atmospheric levels and one surface layer. The difference in the total energy at a given horizontal grid point between two time-mean climate states can be written as

where Δ represents the difference between the later decades (1997−2015) and the earlier decades (1979−1996). The three terms from left to right in Eq. (1) are the difference in the convergence of the shortwave radiation flux (ΔS), the difference in the divergence of the longwave radiation flux (ΔR), and the difference in the convergence of the total energy flux due to non-radiative processes (ΔQ_{non-radiative}).

The perturbation of radiative energy is regarded as roughly linear by assuming that the interactions among the various radiative feedbacks can be neglected. Therefore, the perturbed radiative energy in Eq. (1) can be linearly expressed as

where the superscripts tisr, α, w, o₃, c and co₂ represent the solar radiation at the TOA, the surface albedo, the amount of water vapor, the ozone concentration, the amount of cloud, and CO₂ concentration, respectively. ΔT is the total change in temperature at the Earth’s surface and atmospheric layers. ${{\partial {R}}}/{{\partial {T}}}$ and ${{\partial {R}}}/{{\partial {T}}}\Delta {T}$ represent the Planck feedback matrix (see Lu and Cai, 2009) and the ΔT-induced difference in divergence of the longwave radiation energy flux, respectively. All the radiative terms in Eqs. (2) and (3) can be calculated using the longwave and shortwave radiation fluxes derived from the Fu−Liou radiative transfer model (Fu and Liou, 1992, 1993). The changes in solar radiation at the TOA, ozone, and the CO₂ concentration are considered as external forcing (Ext) (Lu and Cai, 2009).

The energy perturbation due to non-radiative processes (ΔQ_{non-radiative}) is linearly decomposed into the energy perturbation resulting from changes in the surface sensible heat flux (ΔQ_SH), the surface latent heat flux (ΔQ_LH), the surface dynamical processes (ΔQ_{dyn_sfc}), and the atmospheric dynamical processes (ΔQ_{dyn_atm}):

In Eq. (4), ΔQ_SH and ΔQ_LH are derived directly from the ERA-Interim dataset. ΔQ_{dyn_sfc} and ΔQ_{dyn_atm} are estimated as residuals, implying that the energy perturbations that cannot be explained by other known processes can be attributed to these two dynamical processes. ΔQ_SH, ΔQ_LH and ΔQ_{dyn_sfc} have a zero value in the atmospheric layers, but non-zero values in the surface layer. By contrast, ΔQ_{dyn_atm} has a non-zero value in the atmospheric layers and a zero value in the surface layer.

Substituting Eqs. (2)−(4) into Eq. (1), rearranging the terms, and multiplying both sides of the resultant equation by ${\left({{{\partial {R}}}/{{\partial {T}}}} \right)^{ - 1}}$, we obtain

Equation (5) indicates that the total change in temperature between the two climate states (1997−2015 minus 1979−1996) can be divided into 10 partial temperature changes due to (from left to right) the CO₂ content, the solar radiation at the TOA, the surface albedo, the amount of water vapor, the ozone concentration, the amount of cloud, the surface sensible heat, the surface latent heat, and surface and atmospheric dynamical processes. The surface and atmospheric dynamical processes are collectively referred to as dynamical processes. Since we aim to diagnose the surface temperature change after the mid-1990s over Europe−West Asia and Northeast Asia, the focus is on the surface layer.

Figure 1 shows the time series of the region-averaged summer SAT anomalies over three regions: Europe−West Asia (20°−60°N, 10°−60°E), Central Asia (35°−55°N, 60°−90°E), and Northeast Asia (35°−55°N, 90°−120°E). The domain sizes of these areas are the same as defined in Hong et al. (2017) and are outlined by the black boxes in Fig. 2. Given possible uncertainty in different datasets, several datasets including CRU SAT, NOAA/NCEP GHCN CAMS SAT, and University of Delaware SAT version 5.01 are adopted. The results from these datasets agree with each other with a warming shift around the mid-1990s. For CRU (NOAA/NCEP GHCN CAMS, University of Delaware), the warming rates in Europe−West Asia, Central Asia and Northeast Asia are 0.25 (0.28, 0.25) °C (10 yr)⁻¹, 0.06 (0.16, 0.04) °C (10 yr)⁻¹ and 0.30 (0.42, 0.26) °C (10 yr)⁻¹, respectively. Thus, the warming trends in Europe−West Asia and Northeast Asia are much stronger than in Central Asia. The trends in Europe−West Asia and Northeast Asia are significant at the 99% confidence level based on the Mann−Kendall test (Mann, 1945; Hirsch et al., 1982; Gan, 1998), which is widely used in meteorology to assess the significance of trends in temperature, precipitation and runoff (Jaagus, 2006; Wang, 2009; Kumar and Jain, 2010). The results are in agreement with previous studies that illustrated remarkably amplified warming over Europe−West Asia and Northeast Asia but much weaker warming over Central Asia since the mid-1990s (Dong et al., 2016; Hong et al., 2017).

Figure 1.  Time series of the area-averaged summer (June−July−August) SAT anomalies (units: °C) for the period 1979−2015 (thin lines) and 9-year running averages (thick lines) over Europe−West Asia (20°−60°N, 10°−60°E, red lines), Central Asia (35°−55°N, 60°−90°E, green lines) and Northeast Asia (35°−55°N, 90°−120°E, blue lines) after removal of the long-term warming trend. The thick black line is the AMV index (AMVI). Vertical dotted gray lines indicate the phase turning points of AMV and SAT anomalies. CA, Central Asia; E-WA, Europe−West Asia; NEA, Northeast Asia.

Figure 2.  Spatial distribution of the linear trend in the summer (June−July−August) mean CRU SAT over the Eurasian continent during the period 1979−2015 after removal of the long-term warming trend of 1948−2015 (units: °C yr⁻¹). The areas marked with diagonal lines indicate significance at the 95% confidence level based on the Mann−Kendal trend test method. The three boxes represent, from left to right, Europe−West Asia (20°−60°N, 10°−60°E), Central Asia (35°−55°N, 60°−90°E) and Northeast Asia (35°−55°N, 90°−120°E).

Figure 2 displays the spatial distribution of the warming trend for the CRU SAT from 1979 to 2015. It shows that the Eurasian continent has a warming trend. The domains with the most significant warming are Europe−West Asia and Northeast Asia, with a rate of over 0.3°C (10 yr)⁻¹ over most of these regions and a maximum of over 0.5°C (10 yr)⁻¹. This is consistent with the fact that the shift in amplitude of the summer SAT over Europe−West Asia and Northeast Asia is greater than that over Central Asia.

Previous studies have shown that the decadal variability of the SAT over Europe−West Asia and Northeast Asia is closely related to the AMV and the AMV-associated SRP, and that radiative processes may play an important role (Hong et al., 2017; Wang et al., 2017; Sun et al., 2019). The relationship between the SAT anomalies and the AMV can be seen in Fig. 1. The SAT anomalies over Europe−West Asia and Northeast Asia vary in phase with the AMV, indicating the potential role of AMV in modulating SAT over these two regions. A longer data series through 1948−2015 shows a similar relationship and has another shift around 1962 (not shown). This implies that the warming over Europe−West Asia and Northeast Asia may have the same genesis and source. But, whether this is the case in reality is unclear. We aim to determine this issue by carrying out a diagnostic analysis based on CFRAM.

Figure 3 illustrates the composite differences in the CRU SAT between 1997−2015 and 1979−1996. Similar to the pattern of the warming trend (Fig. 2 vs. Fig. 3), warm SAT changes dominate over most of the Eurasian continent. The most significant warming is located in Europe−West Asia and Northeast Asia, consistent with previous results (Zhu et al., 2011, 2012; Sutton and Dong, 2012; Chen and Lu, 2014; Dong et al., 2016, 2017; Hong et al., 2017). Then, we adopt the surface layer of CFRAM to diagnose the amplified warming over these two regions. CFRAM can be used for air temperature budget diagnosis, in which the temperature anomalies can be separated into different components due to different processes. In other words, these components are additive and their sum is approximately equal to the observed temperature anomaly. Figure 4a shows that the CFRAM-derived total change in SAT is similar to the observation in terms of both the amplitude and spatial distribution (Fig. 4a vs. Fig. 3). This confirms that using CFRAM to decompose the total change in temperature into partial temperature changes due to different processes is reasonable.

Figure 3.  Mean differences in the summer (June−July−August) SAT (1997−2015 minus 1979−1996) (shading; units: °C). The areas marked with diagonal lines indicate significance at the 95% confidence level. The two boxes represent Europe−West Asia (20°−60°N, 10°−60°E) and Northeast Asia (35°−55°N, 90°−120°E).

Figure 4.  The CFRAM-derived (a) summer mean total surface temperature change (units: K) between 1997−2015 and 1979−1996 and the partial surface temperature change (units: K) due to (b) the surface albedo, (c) the CO₂ concentration, (d) the incident solar radiation at the TOA, (e) the ozone concentration, (f) the water vapor content, (g) the clouds, (h) the surface latent heat flux, (i) the surface sensible heat flux, and (j) dynamical processes. The two boxes in each panel are the same as in Fig. 3.

Figures 4b-g and 4h-j represent the partial temperature changes induced by radiative processes and non-radiative processes, respectively. Those associated with non-radiative processes are generally much stronger than those associated with radiative processes (Figs. 4b-g vs. Figs. 4h-j). The change in surface albedo dominates the change in SAT over the Tibetan Plateau and the northern polar region, where snow cover may play an important role. The changes in SAT due to surface albedo over Europe−West Asia and Northeast Asia are much weaker (Fig. 4b). Increasing CO₂ concentrations have contributed to uniform warming over the Eurasian continent, indicating that CO₂ is indeed responsible for global warming via the greenhouse effect (Fig. 4c). By contrast, the long-term changes in both the solar radiation at the TOA and the concentration of ozone reduce the SAT, with a weak rate of −0.1 K over Northeast Asia and most of Europe (Figs. 4d and e). The partial temperature change due to water vapor content is positive in Europe−West Asia, South Asia and most of Northeast Asia (Fig. 4f). The magnitude of the cloud-induced temperature change is much stronger than that induced by other radiative processes (Fig. 4g vs. Figs. 4b-f). The clouds contribute warming with a value exceeding 1 K over most of Europe−West Asia and Northeast Asia, indicating the importance of cloud forcing in warming these two regions.

Figure 4h shows that the changes in latent heat flux contribute to warming over Northeast Asia but cooling over northern Europe, whereas the changes in sensible heat flux induce cooling over both Europe−West Asia and Northeast Asia (Fig. 4i). The partial contributions from latent heat and sensible heat fluxes tend to compensate for each other (Figs. 4h and i). The dynamical processes induce positive changes in SAT over Europe−West Asia and Northeast Asia, suggesting that warming over these two regions might be related to dynamical processes (Fig. 4j).

To quantitatively estimate the relative contributions of various physical processes to the total change in temperature over Europe−West Asia and Northeast Asia, we calculated the pattern amplitude projection (PAP) coefficients defined in Eq. (6) (Park et al., 2012; Deng et al., 2013):

where $A$ is the area of Europe−West Asia or Northeast Asia (shown as the two boxes in Fig. 3), a is the mean radius of the Earth, and $\phi $ and $\lambda $ represent the latitude and longitude, respectively. ΔT_SAT and ΔT_i are the total changes in SAT and partial temperature changes due to various processes at the given grid point, respectively. The sum of all the PAPs is exactly equal to the area-averaged total change in temperature.

Figure 5 shows the PAPs of the nine individual processes, the external forcing processes, the radiative processes, and the non-radiative processes for Europe−West Asia and Northeast Asia, where the most remarkable warming can be observed. Figure 5a illustrates that nearly two-thirds of the area-averaged change in SAT over Europe−West Asia (1.00 K) is contributed from radiative processes (0.62 K), of which the CO₂ concentration, the clouds, and the water vapor content are the main drivers, contributing increases of 0.13, 0.40 and 0.24 K to the total change in SAT, respectively. The impact of dynamical processes (0.76 K) is partly counteracted by the impact of sensible heat flux (−0.45 K), inducing the non-radiative processes to only contribute 0.38 K to the total change in SAT.

Figure 5.  PAPs of partial temperature changes due to nine individual processes, external forcing processes (CO₂ concentration + incident solar radiation at the TOA + O₃ concentration), radiative processes (CO₂ concentration + incident solar radiation at the TOA + surface albedo + water vapor content + O₃ concentration + clouds) and non-radiative processes (surface latent heat flux + surface sensible heat flux + dynamic processes) for (a) Europe−West Asia and (b) Northeast Asia. The number in the top-right of each panel indicates the area-averaged summer mean total temperature change (units: K). Vertical dashed gray lines are used to separate nine individual processes and sum of several specific processes.

Figure 5b shows the PAPs for Northeast Asia. The change in SAT over Northeast Asia (1.02 K) is dominated by radiative processes (0.98 K), of which the cloud term is the main contributor (0.85 K) and the CO₂ concentration, surface albedo and water vapor content contribute 0.12, 0.02 and 0.07 K, respectively. Although the surface latent heat and dynamical processes individually generate increases of 0.85 and 0.60 K, these warming changes are mostly counteracted by the cooling effect of the surface sensible heat (−0.98 K), leading to a much weaker contribution from the non-radiative processes (0.05 K).

In summary, the influential factors for the amplified warming over Europe−West Asia and Northeast Asia bear a great similarity, with the radiative processes being the major contributor for both regions. The radiative processes are dominated by the clouds, followed by the CO₂ concentration and water vapor content, accounting for 62.4% and 95.1% of the warming over Europe−West Asia and Northeast Asia, respectively. So, to some extent, the amplified warming over Europe−West Asia and Northeast Asia has a common dynamical origin.

The above analysis shows that the water vapor content, clouds, surface sensible heat flux, surface latent heat flux and dynamical processes all play important parts in the changes in summer SAT over Europe−West Asia and Northeast Asia. Next, we explore the underlying causes for the partial temperature changes due to these five physical processes.

As an important GHG, water vapor not only absorbs the longwave radiation emitted by the Earth’s surface, but also re-emits longwave radiation that heats the ground. Figure 6a shows the vertically integrated difference in specific humidity from the Earth’s surface to 300 hPa, the spatial distribution of which matches well with the difference in surface heating rate and the difference in SAT due to the water vapor content (Fig. 6a vs. Fig. 4f and Fig. 6b). The positive (negative) difference in specific humidity corresponds to the positive (negative) difference in the surface heating rate and the positive (negative) difference in SAT. Furthermore, the change in SAT associated with the water vapor content is mainly a result of the longwave radiative effect, which is counteracted by the shortwave heating effect (Fig. 6b vs. Figs. 6c and d). After the mid-1990s, the water vapor content increased in Europe−West Asia and most of Northeast Asia, leading to an increased longwave heating rate at the Earth’s surface and an enhanced greenhouse effect. Ultimately, it brought warming in both Europe−West Asia and Northeast Asia, consistent with the positive water vapor PAPs in Fig. 5.

Figure 6.  (a) Mean change in the summer tropospheric specific humidity (units: kg m⁻²) between 1997−2015 and 1979−1996 from the surface to 300 hPa and the CFRAM-derived (b) total difference in heating rate (units: W m⁻²), (c) difference in shortwave heating rate (units: W m⁻²) and (d) difference in longwave heating rate (units: W m⁻²) at the Earth’s surface. The heating rates in (b−d) are due to changes in the water vapor. The two boxes in each panel are the same as in Fig. 3.

A change in cloud cover can bring about two opposite cloud radiative effects on the change in SAT because the increased (decreased) cloud fraction will not only obstruct (strengthen) the downward shortwave radiation, but also enhance (weaken) the absorbed and re-emitted longwave radiation (Ramanathan et al., 1989; Wang and Zhao, 1994; Gao et al., 1998). The change in the total cloud fraction corresponds well to the changes in temperature and changes in surface heating rate due to cloud (Fig. 7a vs. Fig. 4g and Fig. 7b). The long-term change in the total cloud fraction since the mid-1990s features a reduction over Europe−West Asia and Northeast Asia (Fig.7a), which is mainly the contribution from middle- and low-level clouds (Fig. 7a vs. Figs. 8a-c). Previous studies indicate that the low-level and middle-level clouds combine a small greenhouse effect with a generally high albedo and thus contribute significantly to the net cooling role of clouds in Earth’s climate (Ramanathan et al., 1989). The reduced low-level and middle-level clouds weaken the albedo, and lead the shortwave cloud radiative effect to be the dominant force of the cloud-induced temperature change, which can also be seen from Figs. 7b-d. After the mid-1990s, the middle- and low-level clouds decreased over both Europe−West Asia and Northeast Asia, which produced increased net radiative flux at the surface and led to a positive change in the SAT in both regions. This is in agreement with the positive cloud PAPs in Fig. 5.

Figure 7.  (a) Summer mean change in total cloud fraction (units: fraction) between 1997−2015 and 1979−1996 and the CFRAM-derived (b) difference in surface cloud heating rate (units: W m⁻²), (c) difference in surface shortwave cloud heating rate (units: W m⁻²) and (f) difference in surface longwave cloud heating rate (units: W m⁻²). The two boxes in each panel are the same as in Fig. 3.

Figure 8.  Summer mean change in the (a) fraction of high-level clouds (400−50 hPa; units: fraction), (b) fraction of middle-level clouds (700−400 hPa; units: fraction) and (c) fraction of low-level clouds (below 700 hPa; units: fraction) between 1997−2015 and 1979−1996. Dotted areas are significant at the 95% confidence level based on the t-test. The two boxes in each panel are the same as in Fig. 3.

Figures 9a and b display the long-term changes in surface sensible heat flux and surface latent heat flux (upward positive), respectively. The signs of the changes in the variables are the opposite to their corresponding changes in partial SAT because a positive (negative) upward net heat flux implies an increased (decreased) release of heat from the surface, resulting in a cooling (warming) change in the SAT (Figs. 4h and i vs. Figs. 9a and b). The partial temperature change induced by the latent heat flux can be explained by the change in evaporation. The region with an increased (decreased) rate of evaporation is consistent with a decreased (increased) SAT due to more (less) heat loss from surface. The positive changes in the surface sensible heat flux weakened the warming over both Europe−West Asia and Northeast Asia, in agreement with the negative sensible heat flux PAPs shown in Fig. 5. By contrast, the area-averaged changes in the surface latent heat are negative over these two regions, corresponding to their positive contribution in Fig. 5.

Figure 9.  Summer mean change in (a) surface sensible heat flux (units: W m⁻²), (b) surface latent heat flux (units: W m⁻²) and (c) evaporation (units: mm) between 1997−2015 and 1979−1996. Statistically significant changes at the 95% confidence level based on the t-test are dotted. The two boxes in each panel are the same as in Fig. 3.

The dynamics term derived by CFRAM is estimated as residuals, including the surface dynamics, atmospheric dynamics, and other processes not explicitly included in the CFRAM analysis (e.g., aerosol effect). Previous studies have shown that the atmospheric circulation plays an important role in warming Europe−West Asia and Northeast Asia (Chen and Lu, 2014; Hong et al., 2017; Lee et al., 2017) and AMV can influence the atmospheric circulation at the decadal time scale (Hong et al., 2017; Wang et al., 2017; Sun et al., 2019) Thus, we speculate that the warming due to dynamical processes may partly come from the AMV-induced dynamics.

Figure 10 shows the change in 200-hPa geopotential height and wind and the anomalies regressed against the AMV index. These results can be used to investigate the mechanism by which AMV influences the change in SAT through atmospheric dynamics. Both the change in the 200-hPa circulation field and AMV-related circulation field features an SRP-like pattern with positive geopotential height anomalies and anticyclonic circulation anomalies over Europe−West Asia and Northeast Asia, and negative geopotential height anomalies and cyclonic circulation over Central Asia at 200 hPa (Fig. 10a vs. Fig. 10b). Besides, Sun et al. (2019) validated that this AMV-related SRP-like pattern can cause amplified warming over Europe−West Asia and Northeast Asia in terms of dynamical processes by producing warm horizontal advection and descending motion. These results imply AMV may partly contribute to the dynamics-induced positive SAT changes.

Figure 10.  (a) Summer mean change in geopotential height (shading; units: m) and horizontal winds (vectors; units: m s⁻¹) between 1997−2015 and 1979−1996. (b) Summer mean change in geopotential height anomalies (shading; units: m) and horizontal wind anomalies (vectors; units: m s⁻¹) regressed against the normalized AMV index. The two boxes in each panel are the same as in Fig. 3.

In addition to a dynamical influence, the AMV-related circulation has a radiative influence on changes in SAT. The positive geopotential height brings reduced cloud cover, decreased precipitation and increased solar radiation over both Europe−West Asia and Northeast Asia (Sun et al., 2019). With the dynamical influence and radiative influence combined, the AMV-related SRP contributes to about 30%−50% of the warming over Europe−West Asia and Northeast Asia (Hong et al., 2017; Wang et al., 2017; Sun et al., 2019). Therefore, AMV may play an important role in warming Europe−West Asia and Northeast Asia.

The Eurasian continent has experienced widespread warming since the mid-1990s, with the greatest warming in Europe−West Asia and Northeast Asia. Hong et al. (2017) used observational datasets to show that the remarkable warming in these two regions may originate from the same dynamic SRP, the decadal variation of which is modulated by AMV. Sun et al. (2019), based on ensemble experiments with three atmospheric general circulation models forced by AMV SST anomalies, showed that SRP-induced cloud radiative effects may exert an important influence on this remarkable warming. Here, we adopted CFRAM to diagnose the relative contributions of various processes to the amplified warming over these two regions. Our main conclusions are as follows.

The amplified warming in both Europe−West Asia and Northeast Asia is mainly determined by radiative processes, which contribute 0.62 and 0.98 K to the area-averaged change in SAT over Europe−West Asia (1.00 K) and Northeast Asia (1.02 K), respectively. Further, the partial temperature changes due to radiative processes are mainly a result of changes in the clouds, which accounts for 0.40 K over Europe−West Asia and 0.85 K over Northeast Asia. Clouds primarily exert an influence via shortwave radiation rather than longwave radiation. The latter weakens the SAT. The total cloudiness decreased over Europe−West Asia and Northeast Asia after the mid-1990s, corresponding to an increase in shortwave radiation and a decrease in longwave radiation reaching the ground. Because the effect of shortwave radiation is stronger than longwave radiation, this causes an increase in surface temperature in these two regions. Given that the cloud data in the reanalysis contain great uncertainties, we used cloud cover and cloud liquid water/ice water content from ERA5 [the fifth-generation ECMWF atmospheric reanalysis with many improvements compared with ERA-Interim (Hersbach and Dee, 2016)], and from MERRA-2 (Modern-Era Retrospective analysis for Research and Applications, Version 2)—the latest atmospheric reanalysis of the modern satellite era produced by NASA’s Global Modeling and Assimilation Office (Gelaro et al., 2017). The results of ERA5 and MERRA-2 feature a very similar pattern to Fig. 4g (not shown). In ERA5, the changes in clouds contribute 0.35 K to the warming over Europe−West Asia and 0.82 K to the warming over Northeast Asia, which are slightly smaller than ERA-Interim. The latter yields values of 0.40 K and 0.85 K, respectively. This implies the robustness of the role played by clouds. The cloud-induced changes in SAT from MERRA-2 are weaker in Europe−West Asia, with an amplitude of 0.19 K, but stronger in Northeast Asia, with an amplitude of 1.14 K. The modest difference between ERA-Interim, ERA5 and MERRA-2 may come from the different microphysics and cumulus parameterizations. However, their overall agreement indicates that the cloud term plays an important role in the warming over Europe−West Asia, and especially Northeast Asia. The concentration of CO₂ and the water vapor content make positive contributions to the enhanced warming, which account for the changes in SAT of 0.13 and 0.24 K over Europe−West Asia and 0.13 and 0.07 K over Northeast Asia, individually. The effect of water vapor content can primarily be attributed to the longwave radiation. After the mid-1990s, the water vapor content in the troposphere increased over Europe−West Asia and Northeast Asia, along with increased longwave radiation reaching the ground, leading to surface warming there.

Both the surface latent heat and dynamical processes brought positive changes in SAT. The former is closely related to surface evaporation, and the latter is partly associated with the phase shift in AMV since the mid-1990s (Sutton and Dong, 2012). There was a significant positive geopotential height and anticyclonic circulation over Europe−West Asia and Northeast Asia during the positive phase of AMV, resulting in intensified warming. But, the effects of the surface latent heat and dynamical processes are offset by the cooling effect of the surface sensible heat flux, resulting in a weak positive contribution of the non-radiative processes.

The results of this study indicate that the amplified warming over Europe−West Asia and Northeast Asia are produced by similar physical processes, in which clouds play a crucial role. The SRP caused cloud cover to decrease over both Europe−West Asia and Northeast Asia after the mid-1990s. The reduced cloud cover led to amplified warming over these two regions through increased radiative flux reaching the ground, suggesting a substantial role of dynamical descending motion. The SRP initiates enhanced subsidence in Europe−West Asia and Northeast Asia during the positive phase of AMV. This indicates that the SRP is indeed the common dynamical origin of the remarkably amplified warming in these two regions.

This study validates the mechanisms proposed in previous observational and model studies (Hong et al., 2017; Sun et al., 2019), but has a weakness. Here, we only emphasize the role of AMV in the SRP. However, the SRP is also associated with changes in the Indian summer monsoon (ISM) and SST in the Pacific Ocean (Zhang and Delworth, 2007; Kamae et al., 2017; Monerie et al., 2019). In other words, the ISM and SST in the Pacific Ocean can also influence the SRP. For instance, Ding and Wang (2005) suggested that ISM-induced diabatic heating, which is affected by the upstream SRP, can in turn play an active role in exciting the SRP downstream. Some other studies suggest the influence of Indian rainfall anomalies on the atmospheric teleconnection pattern (Ding and Wang, 2007; Sun et al., 2010; Chen and Huang, 2012; Greatbatch et al., 2013). Furthermore, Ding et al. (2011) proposed that ENSO has a modulation effect on the CGT by impacting on tropical monsoonal heat sources. Hong and Lu (2016) suggested negative (positive) SST anomalies in the tropical central and eastern Pacific in the preceding spring lead to significantly cooler (warmer) tropical tropospheric temperatures in summer. The tropical tropospheric temperature anomalies may affect the SRP through the meridional displacement of the Asian jet. From Fig. 11, it can be seen that the SST anomalies regressed onto the AMV index closely resemble the SST difference between the two periods, with nearly half the magnitude, indicating the AMV’s modulation effect on the decadal variation of SST. Significant warm SST anomalies dominate both the North Atlantic and the North Pacific, whereas significant cooling ones with smaller amplitude occur in the tropical central and eastern Pacific. The significant changes in SST over the Pacific Ocean during 1979−2015 illustrate that it is not exclusive, although we only focus on the AMV’s modulation effect on the SRP. Thus, more in-depth and comprehensive studies are needed in the future.

Figure 11.  (a) Summer mean change in SST (shading; units: °C) between 1997−2015 and 1979−1996. (b) Summer mean change in SST anomalies (shading; units: °C) regressed against the normalized AMV index. The areas marked with diagonal lines indicate significance at the 95% confidence level. Panels (a) and (b) are for NOAA ERSST.v5; (c) and (d) are for the SST dataset of the Hadley Centre.

Then, CFRAM cannot provide a clear explanation of the dynamic processes. Whether other physical processes join to modulate the radiative processes influencing SAT over Europe−West Asia and Northeast Asia is unclear. Besides, it is well known that the sea ice over the Arctic has been decreasing substantially along with the phase shift of the AMV in the mid-1990s (Park and Latif, 2008; Yu et al., 2017, 2019). Wu et al. (2013) proposed that the winter sea-ice concentration anomalies to the west/southwest of Greenland are significantly correlated with following-summer atmospheric circulation over the Eurasian middle and high latitudes. Vihma (2014) suggested that the Artic sea ice decline can be expected to affect atmospheric circulation and weather patterns, not only in the marine Arctic, but also further south. That the sea ice reduction over the Arctic plays a role cannot be excluded. These aspects warrant in-depth analysis in future work.

Acknowledgments. This work was jointly supported by the National Key Research and Development Program of China (Grant Nos. 2018YFA0606403 and 2015CB453202) and the National Natural Science Foundation of China (Grant Nos. 41790473 and 41421004).