1 Introduction

Teleconnection patterns, which usually accompany anomalous jet streams with their anomalous locations or strength, are involved in climate variabilities of atmospheric circulations and their associated extreme weather (e.g., Rodwell and Hoskins 1996, 2001; Wakabayashi and Kawamura 2004). Wakabayashi and Kawamura (2004) and Tachibana et al. (2004) show teleconnections lying across northern Eurasia cause abnormal weather in East Asia in summer. For example, abnormal heat wave in northern Eurasia occurred in early summer 2018. The meandering of polar front jet associated with the propagation of Rossby waves across northern Eurasia have been pointed out as one of the main factors for that abnormal summer (Kornhuber et al. 2019; Harada et al. 2020). Yasunari et al. (2021) pointed that circum-Arctic atmospheric wave patterns accompanying warm anticyclonic anomalies over Europe and Siberia often induce large-scale wildfires around the Arctic in recent summers. Teleconnections tend to originate where heat of condensation in convection provides an anomalous heat source in the middle troposphere over the tropics (e.g., Hoskins and Karoly 1981; Enomoto et al. 2003). The best-known teleconnections originate in areas of vigorous convective activity over the ocean, which yields large amounts of condensation heat. For example, the remote influences of the El Niño–Southern Oscillation (e.g., Alexander et al. 2002), the Pacific-North American teleconnection pattern (e.g., Mori and Watanabe 2008), the boreal summer intraseasonal oscillation (e.g., Moon et al. 2013), and the Pacific-Japan pattern (Nitta 1987) have been extensively studied. On the other hand, little attention has been paid to the influences of major convection centers over the land upon remote extreme weather. Rodwell and Hoskins (2001) shows teleconnection originating from land rainfall overall, but they did not examine specific remote extreme weather. Although the amount of heating by convection is often smaller over the land than over the ocean, updraft could develop more intensely over the land, where the ground surface is easier to warm because its specific heat is smaller than that of seawater. Therefore, the ratio of convective rainfall to stratiform rainfall is high over the land, and cloud tops can reach the upper troposphere more easily than over the ocean (e. g., Schumacher and Houze 2003; Liu and Zipser 2005). There is cooling in lower levels in stratiform rainfall, while in convective rainfall, there is a heat source throughout the troposphere (Schumacher et al. 2004). Such a heat source can be an effective wave source of teleconnections in mid and high latitudes, even when the amount of heat involved is smaller.

This study focused on convection over the Sahel region, the semi-arid southern edge of the Sahara Desert in North Africa. The Sahel owes its distinct rainy (June–September) and dry (October–May) seasons to the West African monsoon (e.g., Nicholson 2013). Convection in the rainy season is extremely deep (Kodera et al. 2019), and it fluctuates widely from year to year (e.g., Nicholson and Palao 1993). Furthermore, the Sahel region is close to the climatological subtropical westerly jet, which can play a role in a waveguide for a teleconnection pattern (Hoskins and Ambrizzi 1993). These factors favor the upper atmosphere over the Sahel as a wave source that may drive a teleconnection with remote mid-latitude locations.

Some studies have suggested that interannual variations in the Sahel convection affect European climate through meandering of the jet (e.g., Black et al. 2004; Cassou et al. 2005; Gaetani et al. 2011). Jet meandering induced by the Sahel convection may propagate beyond Europe owing to a stationary Rossby wave. However, no studies have considered possible teleconnections from the Sahel to more distant regions, although remote influences upon the Sahel rainfall from elsewhere have been extensively examined (e.g., Folland et al. 1986; Munemoto and Tachibana 2012; Diawara et al. 2016). In this study we used statistical analyses and simple numerical experiments to address the possible existence of a semi-circumglobal teleconnection from the Sahel to eastern Eurasia that arises during anomalous heat production by deep convection over the Sahel. Given that the Pacific-Japan teleconnection pattern is known to propagate from low latitudes over the Philippines through Japan to North America (Nitta 1987), we consider it possible that the Sahel convection excites a teleconnection halfway around the world.

2 Data and methods

We used the monthly Interpolated Outgoing Longwave Radiation (OLR) dataset from the US National Oceanic and Atmospheric Administration (NOAA) (Liebmann and Smith 1996) as an index of convective activity. This dataset is based on satellite observation, which corresponds to cloud-top temperature in the tropics. The OLR data set is thus widely used as a proxy of tropical convective activity (e.g., Wheeler and Kiladis 1999; Lawrence and Webster 2002). Other atmospheric variables such as geopotential height, vertical velocity, velocity potential, diabatic heating rate, and convective heating rate were obtained from the Japanese 55-year reanalysis (JRA-55) (Kobayashi et al. 2015; Harada et al. 2016). We calculated a global climatological map of convective heating rate, averaged from 1000 to 100 hPa altitude, to assess the strength of the Sahel convection relative to other regions (Fig. 1). Following a previous study (Gaetani et al. 2011), we defined the Sahel as the domain bounded by longitude 20° W–40° E and latitude 10° N–20° N. This region, which is slightly north of the center of strong convective heating, was selected because its interannual variation is greater than at the center (Munemoto and Tachibana 2012; Diawara et al. 2016). However, the results did not strongly depend on the choice of area. The monthly mean Sahelian OLR index is the area-averaged value for the Sahel region. The index values shown in Fig. 2 are linearly detrended, standardized, and multiplied by − 1 such that a positive value signifies convective activity that is stronger than normal. We performed a linear regression analysis of large-scale detrended atmospheric fields with the Sahelian OLR index during the rainy season (June through September) of each year from 1979 to 2016. Anomalies based on our regression analyses are referred to as “regressed anomalies” hereafter. The t test was performed for the regressed anomalies and the null hypothesis was rejected when the statistical significance was less than 10%. Although signatures of a wave train pattern appeared in all 4 months, this paper shows the results for September, when the pattern was clearest.

Fig. 1
figure 1

Climatology of convective heating rates in September (1981–2010 mean). Heating rates for land and sea are shown by color shades and contours, respectively, for visibility. The interval of the shades and contours is 5 K day–1

Fig. 2
figure 2

The Sahelian OLR index (blue bars) and the index of our proposed semi-circumglobal teleconnection pattern (Eq. 1; red dots) in September of 1979–2016. See the text for definitions of these indexes. A linear trend has been removed by linear regression. Note that the sign of the OLR index is reversed, such that a positive value signifies stronger than usual convective activity over the Sahel region

To investigate whether anomalous heat sources from Sahelian convection are capable of driving wave-train patterns, we performed a numerical experiment using a linear baroclinic model (LBM; Watanabe and Kimoto 2000) to compute the steady response of atmospheric fields to a heat source associated with convective activity over the Sahel. The model was run in T42L20 (horizontal resolution of 2.8° × 2.8° and 20 vertical layers). A heating anomaly reaching 0.5 K day−1 at its center was applied to the upper troposphere over the western Sahel. Detailed description of the LBM is in Supplemental Information.

3 Results

3.1 Influence of the Sahel convection on large-scale atmospheric circulation

The climatological map of convective heating rate in September (Fig. 1) shows that the Sahel region ranks among the world’s largest areas of convective activity over land. A regression analysis based on the Sahelian OLR index (Fig. 2) was used to calculate the resulting anomalies of OLR and geopotential height at 300 hPa in September (Fig. 3). Figure 3a shows that negative OLR anomalies appear strongly throughout the Sahel region since the region includes the area of the OLR index. There are also significant OLR anomalies in the tropical western Pacific and South America. Interpretation on the signature of the tropical Pacific is described in the last paragraph of this subsection. Figure 3b shows regression anomalies of geopotential height at 300 hPa in the same manner. The cyclonic-anticyclonic anomaly near Western Europe is in agreement with a Sahel-excited pattern reported in a previous study (Gaetani et al. 2011). In addition, a second cyclonic anomaly appears over Siberia and an anticyclonic anomaly appears to the north of Japan. These anomalies constitute a wave train pattern that extends from Europe to East Asia. When the convective activity over the Sahel is anomalously strong, high-pressure anomalies in the upper troposphere cover the Canary Islands, Scandinavia, and northeast Asia, and low-pressure anomalies cover the British Isles and the polar region. To indicate the scale of this anomaly pattern, we refer to it as a semi-circumglobal teleconnection pattern. We define the index of this teleconnection pattern from anomalies at four action centers shown in Fig. 3b as follows:

$$\begin{aligned} semi - circumglobal teleconnection index & = \Bigg[ {aveZ_{300} \left( {10.00^\circ -13.75^\circ W, 30.00^\circ -33.75^\circ N} \right)} \\ & \quad - aveZ_{300} \left( {16.25^\circ -20.00^\circ W, 48.75^\circ -52.50^\circ N} \right) \\ & \quad + aveZ_{300} \left( {6.25^\circ -10.00^\circ E, 65.00^\circ -68.75^\circ N} \right) \\ & \quad + aveZ_{300} \left( {50.00^\circ -53.75^\circ W, 66.25^\circ -70.00^\circ N} \right) \\ & \quad - aveZ_{300} \left( {105.00^\circ -108.75^\circ E, 77.50^\circ -81.25^\circ N} \right) \\ & \quad + aveZ_{300} \left( {121.25^\circ -125.00E, 50.00^\circ -53.75^\circ N } \right) \Bigg]/6 \\ \end{aligned}$$
(1)
Fig. 3
figure 3

Maps showing contours of anomalies in a OLR (W m−2) and b geopotential height at 300 hPa (m) regressed against the Sahelian OLR index in September, when convective activity over the Sahel is especially strong. Color shading indicates regions where values satisfy less than 10% 5%, and 1% levels of statistical significance by t-test. The red rectangle in a outlines the area where the Sahel OLR index is calculated. Six purple rectangles in b outlines the area where the semi-circumuglobal teleconnection index is calculated. Vectors in b indicate wave activity flux (m2 s–2)

where aveZ300 indicate area averaged values in defined rectangles of geopotential height at 300-hPa. The rectangles are 10.00°–13.75° W, 30.00°–33.75° N for the Canary Islands, 16.25°–20.00° W, 48.75°–52.50° N for the British Isles, 6.25°–10.00° E, 65.00°–68.75° N for Scandinavia, 50.00°–53.75° W, 66.25°–70.00° N for Greenland, 105.00°–108.75° E, 77.50°–81.25° N for the polar region and 121.25°–125.00° E, 50.00°–53.75° N for East Asia. The index is overlaid with red dots in Fig. 2. The correlation coefficient of 0.66 between the semi-circumglobal teleconnection index and the OLR index, which satisfies less than 1% levels of statistical significance by t test, indicated that the variations in the two indexes are synchronized in most years.

To visualize the propagation of the stationary Rossby wave, we plotted the vectors of wave activity flux (Takaya and Nakamura 2001) in Fig. 3, which clearly trace Rossby wave propagation from Europe to East Asia. This transcontinental wavy pattern over northern Siberia accords with the conditions favoring the appearance of the Okhotsk high (Nakamura and Fukamachi 2004; Tachibana et al. 2004).

We further examined the possibility that the tropical ocean influences this teleconnection, as suggested by a significant correlation between the Sahelian OLR index and some areas of the tropical Pacific Ocean and Atlantic Ocean (Fig. S1, e.g., Rodríiguez-Fonseca et al. 2015). However, we confirmed that the semi-circumglobal teleconnection pattern still persisted (Figs. S2a and S2b) when these oceanic effects were removed by a linear regression method (e.g., Kodera et al. 1999). Detailed description of these analysis is in Supplemental Information. We also examined linear regressed sea surface temperature (SST) and OLR fields with the semi-circumglobal teleconnection index (Eq. 1) by a similar manner to the Sahelian OLR index. The regressed maps did not show any significant signals in tropical-subtropical regions other than the Sahel with the exception of OLR in the tropical western Pacific (not shown). This signal also appears in Fig. 3a. Coherent variations between Sahel and the tropical western Pacific implies that Sahel convection is partially strengthened by or strengthens the tropical western Pacific through the Walker circulation. The coherent variations with the tropical Pacific is seen over specific parts of the Sahel (Fig. S3). Thus, the tropical western Pacific does not directly excite the semi-circumglobal teleconnection pattern, neither does the SST in the western tropical Pacific. Therefore, the relationship between convective activity over the Sahel and the wave train pattern is not caused by the tropical ocean alone.

3.2 Mechanism of formation of anticyclonic anomaly northwest of the Sahel

Whereas the previous subsection delineated the propagation of a Rossby wave from Europe to East Asia, here we discuss a possible mechanism of the formation of the anticyclonic anomaly over the Azores, the starting point of the wave propagation: a combination of the Matsuno-Gill pattern (Matsuno 1966; Gill 1980) and vorticity forcing. First, we consider the longitudinally averaged atmospheric vertical fields over the Sahel. Figure 4 shows anomalous positive diabatic heating and updrafts in the middle troposphere above 600 hPa, forming a classic Matsuno-Gill pattern. Figure 5 shows that regressed anomalies of the stream function are anticyclonic in the upper troposphere (300 hPa) and cyclonic in the lower troposphere (850 hPa) over the region northwest of the Sahel. This pattern is consistent with the Matsuno-Gill pattern when the heat source is located off the equator. Figure 6 shows a global plot of regressed anomalies of velocity potential and divergent wind at 200 hPa. That depicts a northwestward divergent wind north of the Sahel. Interpreting this feature in terms of the mechanism of advection of absolute vorticity (e.g. Sardeshmukh and Hoskins 1988), this northwestward wind is transporting a negative vorticity anomaly from the equatorial region to the subtropics. The calculation of Rossby wave source by Sardeshmukh and Hoskins (1988) supported that the negative vorticity advection is a main factor in the northwest of the Sahel region (not shown). Thus, the anticyclonic anomaly is generated in the upper troposphere northwest of the Sahel shown in Fig. 5b. Since some studies have already shown the influence of Sahel convection upon European climate through meandering of the jet (e.g., Black et al. 2004; Cassou et al. 2005; Gaetani et al. 2011), the Sahel-Europe teleconnection can be acceptable.

Fig. 4
figure 4

Latitude–height cross sections of the Sahel model region showing contours of longitudinally averaged values (20° W–40° E) of a diabatic heating (K day–1) and b vertical velocity (Pa s–1) regressed against the Sahelian OLR index in September. Color shading indicates regions where values satisfy less than 10% 5%, and 1% levels of statistical significance by t test. Note that positive values signify positive heating in a and downdraft anomalies in b

Fig. 5
figure 5

Regional map showing contours of anomalies in stream functions (× 105 m2 s–1) at a 300 hPa and b 850 hPa regressed against the Sahelian OLR index in September. Color shading indicates regions where values satisfy less than 10% 5%, and 1% levels of statistical significance by t test

Fig. 6
figure 6

Map showing contours of anomalies in velocity potential (× 105 m2 s–1) at 200 hPa regressed against the Sahelian OLR index in September. Blue and red colors indicate regions of divergence and convergence, respectively, where values satisfy less than 10% 5%, and 1% levels of statistical significance by t test. Vectors indicate divergent winds (m s–1)

3.3 Linear baroclinic model (LBM) experiment

We conducted a numerical experiment using a LBM to verify whether the anomalous heat source due to convection over the Sahel could drive teleconnection patterns. The results of our regression analysis using the Sahelian OLR index showed that diabatic heating reaches its maximum over the western Sahel in the middle troposphere at 400 hPa (Fig. 7a, b). We then used this regressed anomaly pattern as the heat forcing in the LBM, which reached a maximum value of about 0.5 K day–1 at the center of the anomaly (Fig. 7c, d). The model experiment yielded a clear wave-like response emitted from the Sahel region at the 300-hPa level (Fig. 8), although its amplitude is smaller than the result of the regression analysis (Fig. 3). Both results show an anticyclonic anomaly northwest of the Sahel, a cyclonic anomaly in southwestern Europe, and an anticyclonic anomaly in northern Europe. This consistency implies that convective activity over the Sahel can trigger the formation of a wave train pattern over Europe. There are other significant large heating areas over the Atlantic Ocean and the East Sahel (Fig. 7a). However, we could not confirm that these heat sources drive the wave train pattern by the additional LBM experiments in which heat forcings are given over these two areas (not shown).

Fig. 7
figure 7

a Regional map of regressed diabatic heating rate (K day–1) at 400 hPa (contours and shading). Hatching indicates areas where values satisfy less than 10% level of statistical significance by t test. b Vertical profile at 0°E, 13°N, the center of the large positive heat signature in a. c and d same as a and b but for the heating rate given in a LBM experiment

Fig. 8
figure 8

Steady-state response to the heating anomaly in the LBM experiment. Contours and colors indicate anomalies of 300-hPa geopotential height (m), and vectors indicate wave activity flux (m2 s–2)

There is a tiny transmission from North Atlantic towards Canary Islands in Fig. 3, which indicates that Sahelian convection may be affected by mid-latitude wave activity. However, there is a certain influence of Sahel upon the mid-latitude wave propagation in the LBM experiments. Actually, there may be a two-way interaction between the Sahelian convection and mid-latitude wave activity. This may amplify the Sahel convection that further influences upon the mid-latitude wave train.

4 Discussion

Our experiments suggest that the interannual variability of the Sahelian convection drives a semi-circumglobal teleconnection across Eurasia. The implication is that Sahelian convection may be as important as the oceanic convective activity that drives the Pacific-Japan and Pacific-North American teleconnections. Although the influence of continental convection is given less attention than the predominant influence of oceanic convection, the condensation heating over Sahel is exceptionally large for a continental area (see Fig. 1). As we described earlier, the relationship between the Sahelian convection and the wave train pattern is not caused by the tropical ocean alone. Here we explore the details of the proposed teleconnection.

Figure 9 shows the schematics of the proposed teleconnection. Enhanced Sahelian convective activity associated with heavy monsoon rainfall strengthens diabatic heating with updraft in the middle troposphere (Fig. 4). The anomalous heat source activates an anticyclonic anomaly in the upper troposphere northwest of the Sahel by the Matsuno-Gill mechanism (Fig. 5), like those seen in other subtropical areas over the ocean (e.g., Wang et al. 2000; Yang et al. 2007). In addition, a divergent wind driven by this heat source creates a northward wind component in the upper troposphere northwest of the Sahel that results in northward advection of absolute vorticity (Fig. 6). The wavy pattern from the Sahel to Europe is consistent with the results of a LBM that incorporates the anomalous heat source over the Sahel (Fig. 8). The wave’s arrival over Europe may act as a further wave source for the eastward propagation of stationary Rossby waves along the polar front jet stream over northern Eurasia (Fig. 3b).

Fig. 9
figure 9

Schematic diagram of proposed semi-circumglobal teleconnection driven by deep convection over the Sahel

We further conducted an additional LBM experiment in which negative vorticity forcing is given to the west of Norway (Figs. S4a and S4b) in order to examine whether anomalies over Europe can be a wave source of eastward propagation. Given negative vorticity corresponds to the anticyclonic anomaly shown in regression analysis (Fig. 3b) and the LBM experiment (Fig. 8). Responding atmospheric circulations show an anticyclonic anomaly west of Norway, a cyclonic anomaly in the polar region, and an anticyclonic anomaly over East Asia (Fig. S5). These are approximately consistent with the semi-circumglobal teleconnection pattern shown in Fig. 3b. The responding semi-sircumglobal wave pattern did not clearly appear as direct response to the Sahel convection (Fig. 8), probably because LBM does not include nonlinear feedback. Northward extension of divergent wind to Europe shown in Fig. 6 is a favorable condition for generating an additional Rossby wave source over Europe, because the divergent wind results in northward advection of absolute vorticity.

The propagation of the wave from the Sahel to East Asia may rely on the polar jet, which is known to act as a waveguide in the propagation of a Rossby wave packet (e.g., Nakamura and Fukamachi 2004; Tachibana et al. 2010). Wave propagation eastward along the polar jet over northern Eurasia is occasionally seen in summer when the polar jet stream is strong (e.g., Ogi et al. 2005; Ogasawara and Kawamura 2007; Otomi et al. 2013; Sato and Nakamura 2019). Sato and Nakamura (2019) shows atmosphere–land interactions over northern Eurasia amplify the transcontinental wave. Otomi et al. (2013) showed that such a transcontinental wave accounted for the abnormally hot summer of 2010 in Russia and East Asia. They also showed that tropical convective activity associated with anomalous oceanic temperatures in the Atlantic Ocean played a role in activating that transcontinental wave. Given Sahel’s position east of the Atlantic Ocean, the Sahel convection may play a similar role. Thus, we argue that anomalous convection over the Sahel is able to activate a transcontinental wave that produces abnormal weather in Russia and even East Asia. Active Sahel convection that coincides with stronger-than-normal polar jets would strengthen the propagation of the wave to East Asia because strong polar jets act as a waveguide. Similar eastward wave propagation excited by a tropical heat source is recognized over the mid-latitude North Pacific Ocean (e.g., Jin and Hoskins 1995; Nitta and Hu 1996).

We moreover conducted additional LBM experiments with two types of background states in order to examine dependence on the strength of the polar jets. One is the state when the polar jet is strong and the other is weak. We used the Arctic Oscillation (AO) Index defined by Ogi et al. (2004) as an index of the strength of the polar jet. We defined that background mean fields of the top 10 years of the AO index as the strong jet state, and those of the bottom 10 years as the weak jet state. The given heating rate is the same as in Fig. 7c, d. The results imply that the wave train pattern is more likely to propagate eastward when the background polar jet is strong (Fig. 10a) than when it is weak (Fig. 10b).

Fig. 10
figure 10

Steady-state response to the heating anomaly in LBM experiments under two background states; a strong polar jets and b weak polar jets. The given heating rate is the same as in Fig. 7c, d. Contours and colors indicate anomalies of 300-hPa geopotential height (m), and vectors indicate wave activity flux (m2 s–2)

Small amplitude in the LBM experiment shown in Fig. 8 can be partially attributed to the underestimation of diabatic heat in the reanalysis data set over the tropical convective regions. Kodera et al. (2019) shows that convection in Sahel is extremely deep with strong updraft. Deep convection tends to occur in a narrow region. Another reason is that non-linear effects are not included in the LBM. These may be also responsible for slight differences of anomaly centers of the LBM from those of regression anomalies.

Whereas we presented results for September, when this wave pattern is clearest, the wave pattern was also apparent in June, July and August. We suggest that September has a favorable condition for the propagation of Rossby waves in this region because the latitude belts of westerly wind shifts southward as the progress of the summer season, which is favorable for the wave propagation. Regression maps with the Sahelian OLR index in other months are in Fig. S6a–S6d for reference.

5 Summaries

In summary, this study has demonstrated the possible existence of a semi-circumglobal teleconnection from the Sahel to East Eurasia. The teleconnection relies on a causal link between two atmospheric waves, one a wave from the subtropics to mid-latitudes, the other a Rossby wave propagation along the polar jet. The details of that link are beyond the scope of this study, but should be examined in the future. It should be noted that the Sahel convection plays a role in amplifying an existing semi-circumglobal teleconnection.

We speculate that some extreme weather events in mid latitudes, such as the extreme hot summer in East Asia in 2018, might be influenced by abnormal convection in the Sahel. The wave train pattern clearly appeared in the Northern Hemisphere in that year (see Supplementary Information and Fig. S7), and convective activity over the Sahel was more active than the climatological mean (Fig. S8). It could be valuable to extend the analysis presented in this paper to other months and other abnormal years.

Referring again to Fig. 1, areas of strong cloud convection lie over other continental regions, such as South America and the Indochina Peninsula. It seems possible that deep convection over other subtropical continental regions may also exert remote influences upon extreme weather in extratropical localities. Further studies of the influence of convection in continental regions may provide us with an expanded comprehension of teleconnections.