4 Summary and discussion

A multiple linear regression analysis has been performed to explore the 11?year solar cycle influence on mean?sea?level pressure and frequency of blocking over the Atlantic/European sector. The 3–4 year lagged solar response signal in the 1870–2010 record of DJF mean?sea?level pressure, first identified by Gray et al. (2013), has been confirmed in a much longer reconstructed time series spanning the period 1660–2010 (350 years; approximately 32 solar cycles; see Figure 4(b)). These results confirm that there is a tendency for positive NAO anomalies to follow solar maxima and negative NAO anomalies to follow solar minima. The signal peaks at a lag of ∼4 years with a maximum amplitude greater than 2.5 hPa and 99% statistical significance. Apparent inconsistencies of results from previous studies have been resolved and stem primarily from the lagged nature of the response and the differences between early?winter and late?winter responses.

Analysis of the individual months that contribute to the 1870–2010 DJF?averaged signal (Figure 5) shows two main features: (i) a 2–4 year lagged response in early winter whose peak amplitude is greater than +3 hPa over the Azores region with 95% statistical significance at 2–3 year lags in December, and (ii) a 0–2 year lagged response in late winter whose peak amplitude is greater than +4 hPa over southern Europe with 99% statistical significance and greater than −4.5 hPa over Iceland with 95% statistical significance at 2?year lag. The early?winter signal tends to dominate the DJF average, resulting in the observed 3–4 year lagged DJF response.

The results suggest that two different mechanisms may be operating and these are discussed in terms of the current proposed mechanisms for solar influence. The so?called ‘top?down’ influence from the atmosphere via heating effects in the stratosphere is suggested to explain the 0–2 year lagged response in late winter. This mechanism operates throughout the winter but maximises in late winter with an almost?immediate impact at the surface. The influence of this sustained top?down forcing of the NAO during each winter also has an influence on the sea?surface temperatures (SSTs). The SST anomaly present at the end of each winter is subducted below the mixed layer from where it can re?emerge in the following winter, thus reinforcing the NAO signal each year so that the peak amplitude occurs at lags of approximately one?quarter solar cycle, i.e. 3–4 years. The impact of this signal is likely to be most evident soon after it re?emerges, in early winter.

While both the early?winter and late?winter solar responses project onto the NAO, they each show distinct patterns that can potentially be used for model validation purposes and thus help to clarify the mechanisms. In particular, the early?winter signal 3–4 years after solar maximum is dominated by positive pressure anomalies over the southern region of the NAO, i.e. over the Atlantic and centred over the Azores. In contrast, the late?winter signal at 0–2 years after solar maximum shows both a negative response over the northern region of the NAO, centred over Iceland, as well as a positive response that is positioned further south and eastward, centred over southern Europe.

A corresponding analysis was performed to examine the 11?year solar signal in frequency of blocking events over the North Atlantic and Europe. The analysis confirmed previous results of Woollings et al. (2010) that showed increased DJF blocking frequency around periods of solar minimum, although these results should be treated with caution since the analysed data span only 58 years. The maximum response (with 99% statistical significance; see Figure 8(b)) was found to occur over Iceland at 1?year lag, i.e. it does not display the 3–4 year lag seen in the SLP results. The DJF?averaged response was found to come primarily from the late?winter (JF) response, with no statistically significant influence seen in December. This suggests that the early?winter influence on the NAO via ocean feedbacks described above, which presumably influence the storm track, has little influence on the frequency of blocking. The late?winter influence lends support to earlier studies that suggest a stratosphere–blocking interaction (Woollings et al., 2010) since the stratosphere also responds to solar forcing almost immediately (Gray et al., 2013; Mitchell et al., 2015). However, we note that such short response lags (and the additional uncertainty in lag?times introduced by uncertainty in which solar index is best employed) mean that it is not possible to categorically distinguish cause from effect using only observational data. While the late?winter surface signal in SLP and blocking may be a response to top?down stratospheric forcing, it is also possible that the stratosphere could simply be responding to the change in blocking frequency caused by some other influence mechanism, since blocking events are associated with increased wave propagation into the stratosphere and have been identified as precursors to disturbances of the stratospheric vortex in winter. Well?designed model experiments are needed to clarify this.

The 11?year solar signal response was also compared with other known influences on blocking frequency over the Atlantic/European sector, namely from ENSO and the Atlantic Multidecadal Oscillation (AMO). The 11?year solar signal was found to be as large in amplitude as the ENSO and AMO signals and the region showing 99% statistical significance was larger than either (Figure 10). When blocking events occur over Iceland the effect on European temperatures can be particularly acute, so the potential to improve seasonal forecasting for European winters by taking account of the influence from the solar cycle is noted.