Extreme Temperature Diary- Wednesday January 17, 2024/ Main Topic: Correcting the News Media- No Evidence of Increased Cold Outbreaks Due to Recent Arctic Warming

No detectable trend in mid-latitude cold extremes during the recent period of Arctic amplification | Communications Earth & Environment (nature.com)

No detectable trend in mid-latitude cold extremes during the recent period of Arctic amplification

Communications Earth & Environment volume 4, Article number: 341 (September 28, 2023)

 Cite this article


It is widely accepted that Arctic amplification—accelerated Arctic warming—will increasingly moderate cold air outbreaks to the mid-latitudes. Yet, an increasing number of recent studies also argue that Arctic amplification can contribute to more severe winter weather. Here we show that the temperature of cold extremes across the United States east of the Rockies, Northeast Asia and Europe have remained nearly constant over recent decades, in clear contrast to a robust Arctic warming trend. Analysis of trends in the frequency and magnitude of cold extremes is mixed across the US and Asia but with a clearer decreasing trend in occurrence across Europe, especially Southern Europe. This divergence between robust Arctic warming and no detectable trends in mid-latitude cold extremes highlights the need for a better understanding of the physical links between Arctic amplification and mid-latitude cold extremes.

Similar content being viewed by others

Spatial variations in the warming trend and the transition to more severe weather in midlatitudes

Article Open access 08 January 2021

Growing prevalence of heat over cold extremes with overall milder extremes and multiple successive events

Article Open access 01 April 2022

The increase in intensity and frequency of surface air temperature extremes throughout the western South Atlantic coast

Article Open access 25 April 2023


The Arctic has experienced accelerated warming, observed to be between two to four times greater in the Arctic relative to the rest of the globe1, which is known as Arctic amplification (AA). Coincident with part or all of the period of AA, regions of the mid-latitude continents have experienced a cooling trend, with the most pronounced and consistent winter cooling trends in the interior of both Eurasia and North America2,3,4. The simultaneous occurrence of a relatively warm Arctic and cold continents is referred to as the warm Arctic/cold continents pattern5. Whether the juxtaposition of the two opposite anomalies, a north-south dipole, are coincidental and physically unrelated or whether the two are dynamically linked whereby the rapidly warming Arctic is contributing, at least in part, to more severe winter weather in parts of Eurasia and/or North America, has been debated for at least a decade and remains unresolved3. However, in recent years, targeted community-wide efforts have been developed to work toward a consensus view6.

Despite the accelerated Arctic warming, there have been a surprising number of historic cold air outbreaks in the United States (US) and Asia in recent years, which may even be increasing47,8,9,10 in frequency. In both 2021 and 2022, two of the deadliest and costliest US natural disasters have been related to extreme cold and/or heavy snowfall11. These societally impactful extreme weather events have garnered much media attention12,13,14,15.

Overall, however, global warming is clearly reducing the number of cold extremes16. A common argument applied to mid-latitude cold events is that, given that the Arctic is the source region for cold extremes across the mid-latitudes, as the Arctic warms more rapidly than other regions of the globe, cold extremes will moderate or warm and occur less often17,18,19. A recent report from the National Academies of Science when assessing the change in frequency of all weather extremes, was most confident in the decrease of cold extremes20.

The recent events in the US and Asia suggest that there may be exceptions to the overall decreasing trend in cold extremes, and that the link between AA and cold extremes may not be straightforward and linear but rather more complex. In particular, the potential for AA to contribute to a dynamical cooling across the mid-latitudes that at least partially offsets radiational and thermodynamic warming of the Arctic needs to be considered. It has been shown that a local anomalous meridional circulation, especially blocking, plays an important role in bridging AA and mid-latitude extreme cold events21. It has been a recent topic of interest if AA can increase or amplify meridional circulations, especially blocking events.

There are two generalized theories as to how AA can contribute to an increased meridional circulation that delivers severe winter weather across the mid-latitudes. The first theory is that a weakening of the equator-to-North Pole temperature gradient results in a slackening of the westerly jet stream, which then favors more amplified and more slowly eastward propagating large scale- or Rossby-waves across the mid-latitudes. The more amplified and persistent flow contributes to more extreme weather in general, including more severe winter weather22. The second theory is that heterogenous heating across the Arctic forced by uneven melting sea ice and/or greater snowfall across the continental regions of the Arctic acts to amplify or persist high latitude blocking that forces a weaker stratospheric polar vortex23,24. In apparent support of this theory, the frequency of severe winter weather across the mid-latitudes increases following a relative weak and distorted stratospheric polar vortex2,4.

In an effort to better contextualize the relationship between AA and mid-latitude cold extremes, we conduct a comprehensive trend analysis starting before and during the period of AA. For our analysis we have computed trends spanning the winter season of both Arctic temperatures and extreme cold in the major population centers of the Northern Hemisphere mid-latitudes: the Eastern US, Northeast Asia, Northern Europe, and Southern Europe.


To test whether observational analysis shows that AA is contributing to a moderation or warming of cold extremes in the mid-latitudes, we computed temperature trends in both the daily cold extremes for the midlatitude regions and for the Arctic just prior to observed mid-latitude cold extremes. For each region, trends in both the number of events and the temperature of events are compared to the corresponding Arctic values for the period extending through the present and starting from three different starting points: 1960, 1990, and 2000. This allows a comparison of trend behavior over the longer period with the more recent period of AA. Since there is some disagreement about the start date of AA, with some studies arguing an earlier start date circa 19902,25 and some a later start date circa 200026,27, we computed trends for both AA periods. Assuming that it takes a few days for the cold air to travel from the Arctic to the mid-latitudes and following the results of Li et al.28 who found that that the strongest high latitude circulation anomalies and associated cold air advection are observed three days prior to the greatest Eurasian cooling related to Arctic warming, we used a three-day lead time to compute Arctic temperatures; however, our results are not sensitive to choice of lead time. All cold extreme events are plotted as a series in sequential order. We then compute the winter trends starting in 1960 (decades earlier than the observed start of AA), and then we recompute the trends limited to the more recent period of AA, for both the longer and shorter definitions.

We define a cold extreme when the December, January and February (DJF) daily average temperature falls below the 5th percentile of the daily average temperature distribution for DJF (see Methods). As for many extreme weather events there is no single standard definition of cold extremes. Thompson and Wallace29 and Tang et al.30 define cold extremes as when the daily minimum temperature drops below the threshold of 1.5 standard deviations of the seasonal mean. Johnson et al.31 define a cold extreme when the maximum temperature anomaly based on the seasonal mean for a given day falls below the 10th percentile of the maximum temperature anomaly distribution of 5-day rolling means. van Oldenborgh32 defined a cold extreme as the lowest minimum temperature at any given location. At the end of our study, we repeated the analysis with other previously used definitions of cold extremes and did not find that our conclusions are sensitive to a specific definition.

For this analysis, we define four mid-latitude regions and an Arctic region. The Arctic region is defined as the area poleward of 70°N, to closely match the region of accelerated warming in Fig. 1 of the analysis of Rantanen et al.1. We also analyze four mid-latitude regions to compute regional average temperatures and cold extremes, with the regions chosen based on societally impactful cold events and their relationship to high-latitude variability. The regions are the Central-Eastern US region (labeled CEUS), the Northeast Asia region (labeled SSNC for Southern Siberia/Northern China), Northern Europe (NEUR) and Southern Europe (SEUR), outlined with boxes in Fig. 1. The teleconnection index that is most closely related with extreme winter weather in the mid-latitudes is the Arctic Oscillation (AO)29 where the negative polarity is associated with an increase in severe winter weather including below- to well below-normal temperatures widespread across the mid-latitudes. This index is also the teleconnection pattern most closely associated with stratospheric polar vortex variability33, where a weak stratospheric polar vortex favors the negative polarity of the AO and widespread below-normal temperatures across the mid-latitudes34. In Fig. 1 we show the regression of the negative polarity of the North Atlantic Oscillation (NAO), which we treat as equivalent to the AO, and Northern Hemisphere surface temperatures. The three mid-latitude regions where cold air outbreaks are most strongly related to a negative NAO or AO are the US east of the Rockies, East Asia and Northern Europe (CEUS, SSNC, and NEUR, respectively). Siberian surface temperature is also highly sensitive to NAO or AO variability but is of less interest because of its northern latitude and sparse population.

Fig. 1: Eastern US, Northern Europe, and East Asia all share cold temperatures during the negative polarity of the NAO.

We also analyze a fourth region where cold air outbreaks are less common, Southern Europe (SEUR), for comparison. This is one region in the mid-latitudes where below-normal temperatures are not associated with a negative AO or NAO and a weak stratospheric polar vortex. Additionally, as will be shown, it is in clear contrast from the other three analyzed regions in that it has a consistent trend for all three periods analyzed in this study and in previous studies when mid-latitude cooling trends were more widespread and of greater magnitude (see Fig. 2 of Cohen et al.2).

Fig. 2: Arctic and mid-latitude temperature trends have been consistent since 1960 but in parts, diverge during AA.

To provide seasonal context, the DJF seasonal mean surface temperature trends and the time series of winter surface temperature trends for each of the five regions analyzed in this study (CEUS, SSNC, NEUR, SEUR and Arctic) are shown in Fig. 2. All trends are shown for all three periods: since 1960, 1990 and 2000. For the DJF seasonal mean, all four mid-latitude regions are warming but the Arctic is warming at least twice as fast as the mid-latitude regions since 1960 and 1990 consistent with the results of Cohen et al.2. However, since 2000, the warming trend in the Arctic has slowed while the warming in the mid-latitudes has accelerated and consequently, Arctic warming is comparable to the warming trend in the mid-latitudes since the beginning of the twenty-first century.

We then establish consistency with previous work by repeating the analysis of Cohen et al.2 that showed since 1960 the entire Arctic and Northern Hemisphere (NH) extratropical land regions were warming using the Climate Research Unit (CRU) land station temperature version five (CRUTEM5) dataset derived from station data only to exclude any bias introduced by reanalysis. Here too, all land regions of the Northern Hemisphere are warming since 1960 (shown in Fig. 3). The NH trend analysis is consistent with the idea that global warming contributes universally to warming winter surface temperatures. Next, we repeat the trend analysis for both periods of AA since 1990 and 2000. Similar to the analysis in Cohen et al.2, over the period of AA cooling trends appear in the continental region of the mid-latitudes but are of smaller extent and magnitude than shown in Cohen et al.2. Interestingly, the cooling trend in North America has strengthened in the most recent period while it has weakened in Eurasia. It could be argued that the North American cooling since 2000 is more impressive than previous similar winter trend analysis during AA shown in Fig. 1 of Cohen et al.35 and Fig. 2 of Cohen et al.2. In contrast, Eurasian cooling is weaker and contracted compared to earlier winter trend analysis.

Fig. 3: Mid-latitude winter temperatures have been warming since 1960 but are mixed since 1990.

For the remainder of the trend analysis, we use the fifth-generation ECMWF atmospheric (ERA5) reanalysis. The advantage of ERA5 is that coverage of the Arctic Ocean is complete in contrast to CRUTEM5 which is limited to land points only and the main focus of our analysis is to relate AA to mid-latitude cold extremes. For the period since 1960, ERA5 also exhibits universal warming across the Northern Hemisphere (Fig. 3). For the two AA periods, a general pattern of amplified Arctic warming with some cooling in the mid-latitudes is evident but the cooling is less in ERA5 when compared to CRU. It is beyond the scope of this study, but it is possible that ERA5 introduces a warm bias in the latter years of the record in the mid-latitudes that weakens the observed cooling trends.

Next, in Figs. 47, we present the trends in cold extreme events for each of the midlatitude regions analyzed in comparison to the Arctic, in terms of both event number and temperature, for all three periods. In Fig. 4, we plot the annual frequency and area average surface temperature for the US east of the Rockies (CEUS; Fig. 1 blue box) for the 5% coldest DJF days of the entire period (color-coded by decade) and the Arctic (Fig. 1 bold black circle) surface temperature from three days earlier than the date of the observed cold extreme. Consistent with the universal warming of the NH there is a statistically significant decreasing number of extreme cold days in the US since 1960. Though there is a warming trend in the coldest days of the year, the trend is not statistically significant in contrast to the warming trend in Arctic temperatures, which is statistically significant. We then repeat the analysis limited to the two AA periods. The Arctic is warming for both AA periods but is only statistically significant for the longer period. In contrast, the trend in CEUS cold extremes frequency is decreasing for the longer period but increasing for the shorter period, although neither trend is statistically significant. Similarly, the temperature trend of CEUS cold extremes is mixed for both AA periods. The only statistically significant trend in cold extremes for CEUS, is a statistically significant cooling trend for the shortest period since 2000. There is a clear divergence in the trends of Arctic temperatures, which are robustly warming, and cold extremes in the US, which exhibit no statistically significant trends during the period of AA and even for the longer period since 1960.

Fig. 4: A robust warming trend is observed in the Arctic but no consistent trend is found for Eastern US cold extremes.

Fig. 5: A robust warming trend is observed in the Arctic but no consistent trend for Northeast Asia cold extremes.

Fig. 7: Southern Europe is experiencing a notable decrease in the frequency of cold extremes.

In Fig. 5, we repeat the analysis of Fig. 4 but for Northeast Asia (SSNC; Fig. 1 orange box) and the Arctic. Though there has been a decreasing number of extreme cold days in the SSNC region since 1960, it is not statistically significant. Over the period of AA, the frequency of extreme cold days is mixed for Northeastern Asia with the frequency of extreme cold days increasing for the longer period of AA but decreasing for the shorter AA period, reversed of what is found the Eastern US. However, similar to the CEUS, the trend is not statistically significant. For both AA periods, there is a very slight negative trend in the temperature of coldest days of the year but again neither trend is statistically significant. Consistent with what was shown for North America, there is a clear divergence between the trends in Arctic temperatures, which are robustly warming, and cold extremes in Northeast Asia, which exhibit no statistically significant trends over the period of AA.

To complete the analysis of those regions that experience frequent cold air outbreaks across the Northern Hemisphere during the negative polarity of the NAO or AO and/or a weak stratospheric polar vortex, we show the same analysis for Northern Europe (NEUR; Fig. 1 green box) in Fig. 6, despite that previous cooling trends in Europe (e.g., Cohen et al.2) have all but disappeared with the inclusion of the most recent winters, in contrast to North America and Asia, as shown in Fig. 3. Cold extremes in Europe also lack a warming trend not only during both AA periods, but even starting in 1960, though the frequency is decreasing at a rate similar to the US. Also like the US for the two periods of AA, the temperature trend of cold extremes is mixed between the two AA periods. However, in contrast to the US and Asia, the frequency of cold extremes is decreasing for all periods, especially since 2000, though it is only statistically significant for the longer period since 1960.

For comparison to the three mid-latitude regions where cold air outbreaks are more common, we analyze the temperature trends of cold extremes in Southern Europe (SEUR) in Fig. 7. In contrast to the other three regions analyzed, SEUR region experiences less frequent Arctic outbreaks and does not exhibit a strong relationship between cold extremes and either a weak stratospheric polar vortex or a negative AO or NAO. Unlike the US, Asia and even Northern Europe, Southern Europe is the only region experiencing a statistically significant decrease in the frequency of cold extremes during the two AA periods, which based on our analysis, is accelerating. Interestingly, though, it is also experiencing a negative trend in cold extremes for all three periods. This negative trend in cold extremes is largely due to the recent winters of 2012 and 2018. The severity of the cold in the 2000s was last matched in the 1960s for Southern Europe.

All trends shown in Figs. 47 are summarized in Table 1 for ease of comparison. In Supplementary Table 1 we compute an alternate temperature trend of extreme cold per decade, showing that our analysis is not sensitive to whether we compute the trend by event or over time. What is immediately apparent from Table 1 (and Supplementary Table 1) is that the robust Arctic warming trend stands in contrast to all four mid-latitude regions where trends are mixed no detectable warming trends are found with the exception of Northeast Asia for the longest period since 1960.

No detectable trend in mid-latitude cold extremes during the recent period of Arctic amplification | Communications Earth & Environment (nature.com)

Leave a Reply

Your email address will not be published. Required fields are marked *