A mechanism of spring Barents Sea ice effect on the extreme summer droughts in northeastern China

The frequency of extreme drought events in northeastern China (NEC) has increased since the 2000s, and such a decadal anomalous trend may lead to significant stress on agriculture and economic development. The correlation between Arctic sea ice loss in spring and extreme summer droughts over NEC was investigated. The results show that the loss of sea ice over the Barents Sea in spring is associated with extreme droughts and positive height anomalies over NEC in summer. The physical processes include two pathways. First, sea ice loss from the Barents Sea to the Kara Sea results in reducing baroclinicity over the ice loss region but increasing baroclinicity over the ice melting region, which is favorable to the wave ridge over northern Europe and negative-phase Summer North Atlantic Oscillation (SNAO). One wave train originates from negative-phase SNAO over North Atlantic–Europe and spreads to central Europe, central Asia, and NEC. Second, another wave motion flux originates from the Barents–Kara Sea propagating eastward, and then disperses southward to NEC. Both wave trains lead to anomalous anticyclonic circulation and westward subtropical high, which favors descending motion and less water vapor flux, thereby contributing to extreme drought.


Introduction
Drought is characterized by below-normal precipitation, which has become an increasing concern in recent years. Under global warming conditions, severe drought events in northern China have occurred more frequently , and they can significantly damage the development of society, economy, and agriculture. Since the 1980s, the frequency of severe droughts over northeastern China (NEC) and the duration of droughts have been increasing (Liang et al. 2011;Yu et al. 2014). These characteristics have expanded southward from the semiarid boundary of northern China in recent decades, and the grassland region of NEC has been decreasing at 2.8% per year (Ma and Fu 2006). In 2016, NEC suffered a severe drought in summer, leading to an economic loss of 15 billion yuan (Li et al. 2018). Moreover, the frequency and duration of extreme drought in NEC are increasing (Wang et al. 2016;Zhang et al. 2009). As NEC is an important economic region in northern China, the drought variation and causes of extreme drought should be studied.
From a water balance viewpoint, less precipitation can significantly affect drought occurrence (Yang et al. 2012) as it is directly linked with anomalous circulations. Previous research indicated that extreme droughts in NEC are dominantly influenced by the East Asian summer monsoon (Huang et al. 2007;Wu et al. 2009a, b). Weakened Asian monsoon can increase the drought frequency in NEC because of inadequate water vapor transport (Zhang and Zhou 2015), which leads to southern floods and northern droughts in east China.
Upon modulation of the monsoon circulation position, the positive-phase Pacific-Japan (PJ) pattern (Nitta 1987) exhibits an anticyclonic anomaly over northern China in drought years. In addition, the decadal variation of NEC precipitation is related to the phase shift of the Pacific Decadal Oscillation (PDO) (Han et al. 2015) because negative-phase PDO can reduce zonal sea-land thermal contrast, leading to positive water vapor transport anomalies. Furthermore, the interannual variation of precipitation over NEC is closely related to tropical sea surface temperature (SST) anomalies (Han et al. 2017;Sun and Wang 2012). For instance, the location and intensity of subtropical high are influenced by El Niño-Southern Oscillation (ENSO) (Wang et al. 2008), and Indian Ocean Dipole has impact on the meridional displacement of the East Asian jet in the boreal summer (Qu and Huang 2012).
Westerly circulations interact with monsoon circulations, thereby modulating the convergence region and water vapor flux, and contributing to extreme droughts and floods (Du et al. 2020;Zhang et al. 2018a, b). Previous studies suggest that droughts in NEC correspond to westerly anomalies and low-frequency waves (Zhang et al. 2020a, b). Since the twenty-first century, the mid-latitude wave train named the Silk Road Pattern has intensified, and it propagates along the Asian westerly jet in the upper troposphere, strengthening the meridional circulation of stationary wave and anticyclonic disturbance over NEC (Lu et al. 2002;Hong et al. 2017;Zhang et al. 2019). Upstream of such low-frequency waves, the Summer North Atlantic Oscillation (SNAO) has a smaller spatial extent and is located farther north. Du et al. (2020) studied the impact of the eastward shift in the negative-phase SNAO on extreme drought over northern China in summer. The SNAO experienced a pattern change after 2000 as the southern center started shifting eastward (Du et al. 2020;Zhang et al. 2020a, b). In addition, NAO is not only related to the internal dynamic process of the atmosphere, but also to external forces such as the El Niño-Southern Oscillation (Wu and Zhang 2015), North Atlantic SST (Kushnir et al. 2002;Pan 2005), and Arctic sea ice (Petoukhov and Semenov 2010;Wu et al. 2016).
Previous studies have revealed the potential effect of external factors on summer droughts over NEC, which are related to Atlantic Multidecadal Oscillation (AMO) (Zhang et al. 2020a, b), ENSO event (Dai 2013), and snow cover over Eurasia (Wu and Kirtman 2007), while few studies have focused on the influence of Arctic sea ice. In fact, the loss of Arctic sea ice is an important potential driver, and it can force mid-latitude climate changes over Eurasia (Guo et al. 2014;Li and Wang 2013). For example, Nakamura et al. (2019) has pointed out the cooling anomalies over the Eurasian continent is closely related to the land process induced by the loss of sea ice. Zhang et al. (2020a, b) have pointed out that the increase in European summer heat waves is closely linked to reduction of the spring Arctic sea ice via the excitation of the anomalous Eurasian wave train. Wu et al. (2016) have reported that the Arctic sea ice in Norwegian Sea and Barents Sea in the preceding winter is intimately connected with interannual variations of East Asian spring precipitation. In addition, some studies have indicated that dipole rainfall pattern over eastern China in August associated with negative interannual variability of the sea ice over Barents-Kara Sea in July and August (Shen et al. 2019).
Since 2000, the Barents Sea has been one of the most dramatic regions of Arctic sea ice loss, with smaller sea ice covers in autumn. The reduced sea ice in the Barents-Kara Sea during autumn-winter is considered a driver of anomalous weather and climate over mid-latitude Eurasia (Petoukhov and Semenov 2010;Kim et al. 2014). Some studies mentioned that the decline in the Barents Sea ice can affect the Siberia High, which would strengthen the East Asian winter monsoon (Inoue et al. 2012;Wu et al. 1999). Tang et al. (2014) revealed that the reduction of Arctic sea ice may lead to decline in the meridional temperature gradient between polar and mid-latitude regions, and this decline could cause cold winter extremes in the northern hemisphere. Recently, He et al. (2017) observed that the decline of sea ice over the Barents Sea in June led to an anomalous rainfall pattern over eastern China through mid-latitude wave trains. Zhang et al. (2018a, b) observed that variations in the Ural blocking in summer displayed a statistically significant association with a persistent spring-summer sea ice pattern in the Barents-Kara Sea.
In addition to summer and autumn, Arctic sea ice loss is also significant in spring. However, there are few studies on the possible connections between spring sea ice and extreme summer climate events over China. Nevertheless, Li et al. (2018) observed that the sea ice loss over the Barents Sea in March might influence hot drought events over NEC in summer, with a significant causal relationship between them. Considering the Arctic sea ice loss in spring, this study aims to answer two issues. First, we aim to identify the correlation between the spring Barents Sea ice and extreme droughts in NEC and the related westerly circulation anomalies. Second, we aim to investigate the mechanism through which the spring Barents Sea ice affects extreme droughts over NEC in August. By exploring these two issues, we aim to enable the prediction of extreme drought events for disaster defense and ecological protection.

Datasets
Atmospheric circulation fields with horizontal resolution of 1° × 1° were obtained from ERA-Interim reanalysis data (http:// apps. ecmwf. int/ datas ets/). Monthly variables including geopotential height, zonal and meridional wind components, and mean sea level pressure were used.
Monthly Arctic sea ice concentration (SIC) datasets were obtained from the Hadley Centre, at a horizontal resolution of 1° × 1° (Rayner et al. 2003). The monthly mean SST data with horizontal resolution of 2° × 2° were based on the Extended Reconstruction SST version 5 (ERSST 5) dataset, which was available from the National Oceanic and Atmospheric Administration website (Smith and Reynolds 2003). The observation precipitation dataset, with a resolution of 2.5° × 2.5°, was obtained from the Global Precipitation Climatology Project (GPCP). The monthly mean NAO indices were acquired from the Climate Prediction Center website (ftp:// ftp. cpc. ncep. noaa. gov/ cwlin ks).
The monthly self-calibrating Palmer drought severity index (sc-PDSI, Dai 2011a) is a meteorological drought index that considers both water supply and demand, and it reflects the soil moisture deficit or surplus. Moreover, the sc-PDSI index is more scientific and objective than the original PDSI because it calibrates the PDSI by using local coefficients (Dai 2011b). In addition, the Penman-Monteith method is used to calculate potential evapotranspiration, because it is considered to be the most physical based and reliable method. The sc-PDSI data with Penman-Monteith equation used in this study were obtained from the Climate Data Guide website (https:// clima tedat aguide. ucar. edu/ clima te-data/ palmer-droug ht-sever ityindex-pdsi) at a 2.5° × 2.5° horizontal resolution. Moderate drought is sc-PDSI ≤ − 2, severe drought is sc-PDSI ≤ − 3, extreme drought is sc-PDSI ≤ − 4. The extreme drought is defined as an sc-PDSI lower than − 3 in this study.
The SIC index (SICI) is defined as the spatial-weighted SIC averaged in spring over the Barents Sea (70° N-77° N, 30° E-60° E). The underlying surface turns into seawater due to the loss of sea ice. Therefore, the BSSTI is defined as the spatial-weighted sea surface temperature in August over Barents Sea (70° N-77° N, 30° E-60° E). The KSSTI is calculated as the spatial-weighted sea surface temperature in August over Kara Sea (70° N-77° N, 65° E-90° E). The BKSSTI is the average of BSSTI and KSSTI in August.

Methods
The wave activity flux (TNF; Takaya and Nakamura 2001) was used to diagnose the horizontal wave propagation associated with a teleconnection pattern on a basic flow zonally varying in the middle-upper troposphere. Based on the Wentzel-Kramers-Brillouin (WKB) approximation, the flux is independent of wave phase and parallel to the local group velocity of stationary Rossby waves. The flux in the pressure (P) coordinate can be expressed as follows: where u and v are the horizontal winds, Ψ � is the streamfunction, R 0 is the gas constant, N 2 is the buoyancy frequency (Brunt-Väisälä frequency), H 0 is the atmospheric scale height, and T is the temperature. In Eq. (1), the background field includes the zonal non-uniform base flow and meridional wind field. Therefore it is suitable for a mid-high latitude circulation field analysis.
The Eady growth rate (EGR) was calculated to determine the atmospheric baroclinicity (Zhang et al. 2018a, b) as shown in Eq. (2).
where f is the Coriolis parameter, N is the Brunt-Väisälä frequency, u is the zonal wind, and z is the height. The EGR at 700 hPa was calculated to clarify the atmospheric baroclinicity response to underlying surface variation.
The Mann-Kendall test is a non-parametric statistical test method, which can clarify the start time of the mutation. The two curves of UF and UB have an intersection point, and the intersection point is between the critical lines, then the time corresponding to the intersection point is mutation time. Statistical techniques employed in this study included correlation, composite, and regression analyses. The statistical significance tests were computed using the Student's two-tailed t test.

Characteristics of Arctic sea ice from spring to summer and drought distribution in NEC
In this study, we focus on the transitional spring season, defined as MAM. In MAM (Fig. 1a), the Barents Sea is dominated by abundant sea ice, and the SIC in the north of Arctic exceeds 90%. From spring to summer, sea ice gradually melts eastward and northward from the Barents Sea ( Fig. 1b-d). By June, the sea ice has melted, except for the northern part of the Barents Sea region, and it nearly disappears over the Kara Sea in August. The frequency of droughts (sc-PDSI ≤ − 2) and extreme droughts (sc-PDSI ≤ − 3) over NEC in 1979-2017 are shown in Fig. 1e, f. The value indicates a high frequency of droughts and extreme droughts over NEC. Extreme droughts are mostly located in the southern part of NEC. Figure 1g, h show the drought spatial distribution characteristics in two periods, which were determined by the Mann-Kendall statistical test (Fig. 3b). Compared with the period from 1979-1997, the drought classification and affected areas clearly increased during 1998-2017.
Previous studies have discussed the relationship between autumn and winter Barents Sea ice and climate anomalies across East Asia (Fan et al. 2017;Wu et al. 1999). However, the variation of sea ice depends on the selected season. Therefore, it is necessary to identify the crucial sea ice areas Drought spatial distribution over NEC in August: g 1979August: g -1997August: g and h 1998August: g -2017 in spring. Figure 2a indicates the leading mode of the EOF for the spring sea ice anomalies over the Arctic, and the variance of this leading mode is 34.3%. The EOF mode clearly reflects the variation of spring sea ice over the Arctic region, and there is a considerable negative sea ice center over the Barents Sea, indicating that the variation of spring sea ice is more sensitive in the Barents Sea than in other regions. The PC1 shifted to a predominantly positive anomaly after 1990, indicating that the spring Barents Sea ice decreased remarkably during this period. Figure 2b displays the time series of SICI (red line) in spring from 1979 to 2017, which shows a decreased trend. The correlation between PC1 and SICI was significant, at − 0.96, indicating that the Barents Sea ice is more representative in the Barents Sea than in other Arctic areas.

Relationship between Barents Sea ice and droughts over NEC
Despite the known high temperatures and serious precipitation deficiencies over NEC in 2016, these events might not occur with extreme droughts, and temperatures and precipitation cannot fully indicate the classification of drought. In order to better illustrated the characteristics of drought over NEC, we employ the meteorological drought index sc-PDSI. As shown in Fig. 3a, the time series of the sc-PDSI over NEC in August (42° N-50° N, 110° E-130° E) showed interdecadal variation at around 1998 based on the 9-year moving average. The trend of drought over NEC intensified in 1979-2017, and extreme drought events (sc-PDSI ≤ − 3) occurred ten times from 1979 to 2017. Extreme drought events occurred eight times after 1998. Figure 3b shows the Mann-Kendall test of sc-PDSI, and it is clearly that the significant abrupt points of decadal changes of sc-PDSI began in 1998. The abrupt time point of the sc-PDSI was consistent with the results based on the 9-year moving average. Figure 3c displays the standardized values of SICI and sc-PDSI over NEC from 1979 to 2017, and their variations are more in phase after 1998. Li et al. (2018) have shown that the Barents Sea ice in March was significantly correlated with severe hot drought events over NEC after 1997, and this time point is also consistent with the time period when the frequency of extreme drought events increased. We calculated the correlation coefficients between SICI and sc-PDSI, and correlation coefficient (0.58) was obtained for 1998-2017 which is significant at the 95% confidence level, indicating that drought events are more likely to occur when the loss of Barents Sea ice occurs in spring.
To document the spatial distribution of the correlation during two periods, Fig. 4a, b show the correlation between SICI and sc-PDSI over NEC. It is clearly that sc-PDSI anomalies over NEC display a strong positive correlation with SICI after 1998, but there was no significant correlation over NEC for 1979NEC for -1996. In this latter period, the correlation between the sc-PDSI of NEC and the Barents Sea ice was significant, and there were positive correlation coefficients over the Barents Sea. In contrast, few areas over the Barents Sea were related to the sc-PDSI of NEC in 1979-1996.
As described above, the drought events over NEC in summer were closely related to the loss of spring sea ice over the Barents Sea. In this context, the mechanisms through which the Barents Sea ice affects the extreme drought events over NEC in summer are still unclear. It was worth discussing the mechanisms of the Barents Sea ice affecting the extreme drought over NEC. To provide more details on the regional climate change related to the spring sea ice anomaly over the Barents Sea, regression maps of SIC, albedo (AL), net shortwave radiation (SW), and Eady growth rate (EGR) with regards to inverted SICI are shown in Fig. 5a-e. A reduction in the spring sea ice over the Barents Sea (Fig. 5a) led to fewer surface albedo anomalies over the Barents Sea and surrounding areas, including the Kara Sea (Fig. 5c), indicating the possibility of more absorption of incoming shortwave flux (Fig. 5d). This shows that the net shortwave radiation over the Barents Sea increases and accelerates the melting of the sea ice in spring. Sea ice changes the radiation state of the ocean surface through albedo feedback, leading to anomalies in the thermodynamic conditions of the sea surface and ocean-atmosphere heat exchange (Deser and Teng 2008;Dethloff et al. 2006). Therefore, there is an increase in the EGR over the Barents Sea and its surroundings due to the melting of sea ice (Fig. 5e). Meanwhile, the underlying conditions change from sea ice to seawater, resulting in enhanced atmospheric baroclinicity. In addition, spring snow depth over Eurasian is significantly correlated with sea ice over Barents Sea, and the increased baroclinicity in the southern land of the Barents and Kara seas may be related to the decrease in Eurasian spring snow cover (Fan et al. 2017;Li et al. 2018). As the melting continues, the sea ice over the Barents and Kara seas become even smaller in subsequent months, leading to a broad area of seawater (Fig. 5b). Thus, we focused on the regression map of SW, EGR, sea level pressure (SLP), and horizontal wind in June with regards to the SST over Barents Sea in June. Figure 5d shows the atmospheric circulation anomalies related to the loss of sea ice. A negative SLP center can be observed over the Barents Sea, and low surface pressures were attributed to the reduction of Barents Sea ice, with higher SST. The loss of sea ice decreases cold air accumulation and increases temperature in the lower troposphere, which could reduce the SLP . Previous studies have also reported that numerical experiments can simulate the low-pressure structure in SLP resulting from sea ice loss (Blüthgen et al. 2012).In June, the anomalous southwesterly wind blows to the Kara Sea under the influence of low-level wind fields, leading to warm air from south and warm seawater from Barents Sea to accelerate the melting of sea ice in Kara Sea (Fig. 5f). Therefore, the melting of sea ice in the Kara Sea corresponds to increased baroclinicity of atmosphere (Fig. 5h). Due to the underlying surface of Barents Sea has turned into seawater in June, and the underlying surface did not change from June to August resulting in decreased baroclinicity of atmosphere. As described above, the decreased sea ice causes anomalies in regional atmospheric signals, which are linked to surface state dependence. Previous studies have reported that the amplitude of the regional geopotential height field is influenced by the Arctic Sea ice loss, and this influence depends on the change of pre-existing surface state conditions (Overland et al. 2016;Shepherd 2014). To determine the response of the atmosphere to the loss of sea ice in the following months, typical low sea ice years were selected based on a standardized SICI of less than − 0.5 in order to obtain more samples. Therefore, the years with low sea ice were 2000, 2006, 2007, 2008, 2011, 2012, 2013, 2014, 2015, 2016 and 2017. Figure 6 presents monthly composite mean anomalies of geopotential height at 500 hPa from June to August. For low sea ice years, the negative geopotential height over Barents Sea moves eastward to Kara Sea. In summer, the melted Barents Sea ice reduces the meridional temperature gradient between the Barents Sea and Europe, and the lower atmosphere baroclinicity over the Barents Sea could cause Greenland high to expand southeast, leading to anomalous block circulation. Therefore, anomalous positive geopotential height forms over Europe in July and August, and the upstream areas depict the characteristics of negative-phase SNAO with the southern center shifting eastward in summer (Fig. 6b, c). In fact, the anomalous positive geopotential height forms over Europe could strengthen the climatological ridge, which also strengthen the climatological trough in the upstream via downstream influence. Therefore, the anomalous cyclonic over the southern center of SNAO pattern is enhanced due to the strengthening of climatological trough. Meanwhile, the anomalous positive geopotential forms over Greenland in low sea ice years, and the positive geopotential height over Greenland and the negative geopotential height over Europe-North Atlantic region are SNAO pattern. Hence, the enhanced northern activity center of SNAO pattern also lead to the enhanced southern activity center of SNAO pattern. This configuration reveals the strengthening of climatological ridge and trough, which could enhance meridional circulation and lowfrequency waves. Therefore, anticyclone anomalies would occur in August over NEC. The change in surface properties (albedo, surface net shortwave radiation, etc.) enhances the atmospheric baroclinicity due to reduced sea ice, and following the initial baroclinic stage of adjustment, the response becomes barotropic and amplified in space and magnitude. The adjustment balance is reached a few months later, and is characterized by a barotropic structure resembling the NAO pattern (Deser et al. 2000(Deser et al. , 2007Ferreira and Frankignoul 2005). A few previous studies also indicate that negative sea ice anomalies have a strong impact on atmospheric circulation (Alexander et al. 2004;Wang and Overland. 2009;Ogi et al. 2010;Mori et al. 2014). In recent studies, Du et al. (2020) and Wang et al. (2018) demonstrated that the negative-phase of the SNAO may be a driver of extreme events in northern China, implying that the upstream SNAO may be crucial for the occurrence of extreme drought over NEC. Figure 7a shows the leading mode of the EOF for geopotential height anomalies at 500 hPa in August, with a variance of 24.4%. The EOF mode shows a clear positive and negative anomaly pattern from north to south, similar to the distribution of the geopotential height anomalies over the North Atlantic-Europe region , and e Eady growth rate (day −1 ). Regression map of sea level pressure (Pa) and horizontal wind (m/s; vectors) at 850 hPa (f) with regards to SST over Barents Sea in June; g net shortwave radiation at surface (W/m 2 ), and h Eady growth rate (day −1 ). Gray and black dots indicate significant regression coefficients at the 90 and 95% confidence levels (Fig. 6c). The latitudinal pattern of the geopotential height anomaly indicates an eastward shift of the SNAO pattern due to a low-pressure anomaly over the west coast of Europe. As shown in Fig. 7b, PC1 changed from negative to positive at around 2000, which corresponded to the phase shift of SNAO. Additionally, PC1 presented a significant correlation of − 0.7 with SNAO at the 99% confidence level. In August, the negative-phase SNAO with the southern center shifting eastward over North Atlantic region, and the sea ice melts into seawater over Barents-Kara Sea. The correlation coefficients between BSSTI, KSSTI, BKSSTI, SICI, SNAO, and PC1 in July and August of 1998-2017 are presented in Table 1. We confirmed that the August SNAO index was highly correlated with the SICI, and their correlation coefficients exceeded 0.5. Meanwhile, the correlation between the SNAO index and KSSTI and BKSSIT were also significant, which indicates that the underlying surface variation of the sea ice melting could affect the SNAO phase. However, the correlation between the SNAO index and BSSTI was not significant likely because the Barents Sea ice mostly melts in June (Fig. 1b). Additionally, the underlying surface over the Barents Sea is the seawater from June to August, whereas the SST evolution Fig. 6 Composite differences of geopotential height (gpm; shaded) at 500 hPa (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) between the low sea ice years and the climatology in a June, b July, and c August. Green and red contours represent averaged geopotential height of 5500-5700 gpm before and after 1998, respectively, with an interval of 50 gpm (a) Z500, June (b) Z500, July (c) Z500, August at high latitudes has less influence on the later period of the warm season (Qiao et al. 2020).
NAO is a force that can cause circulation and wave anomalies, especially in winter and summer. For instance, the positive (negative) phase of NAO is favorable for a shallower East Asian trough, which can weaken (strengthen) the East Asian winter monsoon (Feng et al. 2018;Qiao and Feng 2016). Wang et al. (2018) reported that SNAO can affect rainfall anomalies over northern China by altering the thermal forcing of the Tibetan Plateau. Figure 8 shows the spatial correlation distribution of SNAO and sc-PDSI, and significant positive correlation in NEC indicates that drought events are more likely to occur when SNAO is in negativephase. Du et al. (2020) demonstrated that negative-phase SNAO could have impacted extreme droughts over northern China. Therefore, the frequency of extreme drought events over NEC may be related to the shift of SNAO to a predominantly negative-phase after 1998.

Atmospheric circulation and wave pattern anomaly associated with extreme droughts
The teleconnection patterns over the North Atlantic-European region and East Asia are shown in Fig. 9. As reported by Watanabe (2004) and Wang et al. (2018), NAO is tied to East Asian climate variability by stationary Rossby waves. Consequently, the influence of sea ice on the SNAO pattern could be one of the pathways leading to extreme droughts over NEC. The regressions of streamfunction against SICI and SSTI (Fig. 9b, d, f, and h) show evident wave trains extending to downstream regions. One of the wave trains originates from North Atlantic-Europe and spreads to central Europe, central Asia, and NEC. Another weakened wave train extends from the Barents-Kara Sea and then propagates eastward and southward to NEC, and this pattern matches the Polar/Eurasia (POL) teleconnection that can influence atmospheric circulation over Eurasia including NEC (Gao et al. 2015). The apparent divergence in the wave activity flux appears at the negative anomaly center of the streamfunction, and it corresponds to the southern center of the SNAO and the Barents-Kara Sea, indicating there are two wave sources although the wave energy dispersion of Barents-Kara Sea is weak. The wave activity flux diverges from the source region along two paths and eventually converges over NEC.As a result, the positive streamfunction anomaly is formed at upper levels over NEC. It could be seen from Fig. 9 that the wave train patterns are also obvious in the middle atmosphere (500 hPa), and eastward shift of the negative-phase SNAO pattern over North  Atlantic-Europe region. Under the control of anticyclonic circulation over NEC, extreme drought events are more likely to occur. However, the Barents Sea ice mostly melted in June, and the underlying surface over Barents Sea is the seawater from June to August. It can be seen from regression map that Barents Sea SST contributes to the strengthening of the European ridge (Fig. 9a). The loss of sea ice signal can last from spring to summer, leading to a large amount of open water over the Barents-Kara Sea. We investigate the possible relationships between decreased sea ice and the atmospheric circulations over NEC. The climate state provides a helpful reference for our discussion of extreme drought. As shown in Fig. 10a, southerly winds from the South China Sea and southwesterly wind influenced by subtropical high prevail over NEC. Correspondingly, the larger quantities of water vapor (300-1000 hPa) are mainly transported from the South China Sea and western Pacific to NEC (Fig. 10b). The regression circulation maps of summer with regard to inverted SICI and BKSSTI in 1998-2017 are shown in Fig. 10 to clarify the influence of SICI and BKSSTI on extreme droughts over NEC. There is an anomalous anticyclonic center of upper level over NEC (Fig. 10c, d), accompanied by easterly anomalies at 40° N, which decrease the strength of the westerly jet. A positive geopotential height anomaly over NEC at 500 hPa (Fig. 10e, f) and the weakened jet in the upper atmosphere (Fig. 10c), which favors the west extension of subtropical high, can lead to the transport of water vapor to the west. Meanwhile, strong northeasterly winds in low-level winds was not benefit to the northward transport of water vapor (Fig. 10g, h). The draw cross section in Fig. 10j, k indicate that the quasi-barotropic downward motion over  NEC is also significant, which favors the adiabatic heating over NEC. Therefore, this configuration is unfavorable for the convergence of water vapor in NEC (Fig. 10h, j), thereby leading to precipitation deficiency (Fig. 10g, h).

Model results
To provide more robust evidence for the interpretation of the teleconnection between the reduction in spring sea ice over the Barents Sea and the downstream atmospheric circulation in August, we conduct numerical experiments to examine the atmospheric response to the spring sea ice variability in the Barents Sea. We use the version 5 of Community Atmosphere Model (CAM5) to design two numerical experiments, including a control experiment and a sensitive experiment. The CAM5 has a finite volume grid of 1.9° × 2.5° with 26 hybrid sigma pressure levels. In the control run, the sea ice concentration and SST are prescribed as boundary conditions in the mode, and the other external variables are fixed. In the sensitive run, the sea ice is reduced over Barents-Kara Sea from spring to summer, while other months are prescribed by climatological sea ice, and the boundary condition of SST is set to − 1.8 °C where SIC equals zero. The control run and sensitive are both repeated 50 times, with different initial condition, integrated from 1 March to 31 August. We focus on the difference between the sensitivity and control experiments in August to reveal lagged responses to the decreased SIC in spring. Model-simulated differences in 200 hPa horizontal wind, 500 hPa geopotential height, 850 hPa horizontal wind and precipitation between the sensitive and control experiments are displayed in Fig. 11. The response of 200 hPa winds shows that the NEC and its surroundings affected by significant anomalous anticyclonic, accompanied by easterly anomalies at 42° N, which decrease the strength of the westerly jet (Fig. 11a). The 500 hPa geopotential height is clearly characterized by the two wave trains originating from SNAO and Barents-Kara Sea to NEC, which is similar to the regression maps of geopotential height (Fig. 9g), and the positive geopotential height anomaly and weakened westerly jet over NEC favors the west extension of subtropical high. What's more, strong northeasterly winds in the low-level winds (Fig. 11c), and this was unfavorable for water vapor transportation from south to north under the climatologically typical southwesterly. Duo to the combined influences, it is unfavorable for the convergence of water vapor in NEC (Fig. 11d). Therefore, the anomalous circulation configuration created favorable conditions for extreme drought events over NEC.

Summary and discussion
Motivated by extreme droughts over northeast China, which have exhibited a decadal increase since the mid-1990s, this study explored its association with Arctic sea ice loss, as this sea ice has significantly decreased beyond the decadal timescale. The results show the enhanced impact of spring SIC over the Barents Sea on subsequent extreme summer droughts over NEC after 1998, and a possible mechanism was suggested. The relationship between spring SIC and subsequent extreme summer droughts presented an interdecadal variation, and the correlation coefficients between SICI and sc-PDSI over NEC were − 0.2 and 0.58 for 1979-1998 and 1998-2017, respectively. The reduced spring SIC over the Barents Sea corresponded with positive geopotential height anomalies over NEC during summer. The results indicate that the anomalous anticyclonic circulation patterns and westward subtropical high, which was accompanied by significantly descend motion and less water vapor, led to decreased precipitations and increased droughts.
The spring SIC loss over the Barents enhanced the extreme summer droughts after 1998 largely through two pathways. Sea ice melted into seawater from spring to summer, and the reduced spring SIC memories were sustained by SST during summer. Firstly, the sea ice loss from Barents Sea to Kara Sea results in increased atmospheric baroclinicity over sea ice melting region, and following the initial baroclinic stage of adjustment, the adjustment balance stage was characterized by negative-phase NAO pattern in summer. What's more, the TNF pattern suggested that the wave train originated from SNAO and spread to central Europe, central Asia and NEC. Secondly, another weakened wave train extended from Barents-Kara Sea and then propagated eastward and southward to NEC, similarly to Polar/Eurasia (POL) teleconnection. This associated wave train enhanced the anticyclonic anomaly over NEC, establishing a pronounced link with extreme droughts over NEC.
However, SIC loss is one of the factors that influence extreme drought events over NEC on the decadal timescale. Previous studies have investigated that the variations of precipitation and temperature over NEC is associated with PDO and AMO (Li et al. 2019;Han et al. 2015). Thus, the spring SIC over the Barents Sea could be a driver and potential predictor of extreme droughts over NEC. In this context, extreme drought events over NEC occurred on interdecadal change after 1998, matching the shift of AMO (Zhang et al. 2020a, b). AMO drives the variability in the European climate at decadal timescales (Delworth and Mann 2000;Guan et al. 2019;Knight et al. 2006). Positive-phase AMO caused rapid warming over the North Atlantic, which is responsible for the loss of Arctic sea ice over the Barents Sea (Day et al. 2012). Some studies have verified that AMO modulates the long-term loss of Arctic sea ice and influences summer precipitation over northern Europe (Miles et al. 2014;Screen 2013). In addition, Qian et al. (2014) suggested that AMO affects the decadal variations in summer droughts over eastern China. Zhang et al. (2020a, b) suggested that positivephase AMO can favor the east extension of negative SNAO. Li et al. (2018) reported that AMO and hot droughts over NEC have a close statistical relationship and are highly correlated with SIC over Barents Sea. Therefore, the interdecadal relation between AMO, SIC over the Barents, and extreme droughts over NEC should be further investigated in future studies. On the interannual time scale, Li et al. (2018) also have pointed out that SIC in March has close relationship with spring snow cover over Eurasia in April, which further influence the climate over East Asia in early summer. Some studies have proved that snow cover and soil moisture anomalies over Eurasia in spring lead to warm land, which amplifies stationary Rossby waves in summer via memory effects of the land processes (Matsumura et al. 2014;Sato and Nakamura 2019). In addition, Fig. 10e shows the spatial distribution of the 500 hPa geopotential height anomalies related to the inverted SICI is similar to the PJ pattern, indicating that the spring SIC loss over the Barents Sea has a certain modulation effect on the PJ wave train. We further verified that the correlation coefficient between the SICI and PJ pattern index is − 0.45, which is statistically significant at the 99% confidence level. The corresponding mechanisms of SIC over the Barents Sea affecting PJ wave trains on the interannual time scale should also be investigated in future studies.