LETTERS Decreased frequency of North Atlantic polar lows associated with future climate warming Matthias Zahn 1,2 & Hans von Storch 2,3 Every winter, the high-latitude oceans are struck by severe storms that are considerably smaller than the weather-dominating syn- optic depressions 1 . Accompanied by strong winds and heavy pre- cipitation, these often explosively developing mesoscale cyclones— termed polar lows 1 —constitute a threat to offshore activities such as shipping or oil and gas exploitation. Yet owing to their small scale, polar lows are poorly represented in the observational and global reanalysis data 2 often used for climatological investigations of atmospheric features and cannot be assessed in coarse-resolution global simulations of possible future climates. Here we show that in a future anthropogenically warmed climate, the frequency of polar lows is projected to decline. We used a series of regional climate model simulations to downscale a set of global climate change scenarios 3 from the Intergovernmental Panel of Climate Change. In this process, we first simulated the formation of polar low sys- tems in the North Atlantic and then counted the individual cases. A previous study 4 using NCEP/NCAR re-analysis data 5 revealed that polar low frequency from 1948 to 2005 did not systematically change. Now, in projections for the end of the twenty-first century, we found a significantly lower number of polar lows and a north- ward shift of their mean genesis region in response to elevated atmospheric greenhouse gas concentration. This change can be related to changes in the North Atlantic sea surface temperature and mid-troposphere temperature; the latter is found to rise faster than the former so that the resulting stability is increased, hinder- ing the formation or intensification of polar lows. Our results pro- vide a rare example of a climate change effect in which a type of extreme weather is likely to decrease, rather than increase. Polar lows develop as disturbances in low-level air flows of (for example) orographic or baroclinic origin 1,6,7 , which later amplify by convective processes 1,8,9 under the influence of large vertical temper- ature gradients. Oceanic energy loss elicited by latent and sensible heat fluxes can influence Greenland Deep Water formation and thus deep ocean circulation 10 . Typical large-scale conditions favouring polar low development are cold air outbreaks, in which cold and dry air masses originating from ice-covered regions are advected towards the open ocean. It is therefore likely that changes in large-scale parameters that favour polar low development will also influence the frequency of polar lows and can thus affect the strategy of offshore activities in the Arctic and Subarctic regions. Changes in the frequency of small-scale atmos- pheric systems are usually derived from large-scale proxy data, as has been done for thunderstorms in the USA 11 or for polar lows in the North Atlantic 12,13 . Studies based on simulated individual mesoscale storms are lacking so far. An analysis of a dynamical downscaling effort, in which a regional climate model was processing NCEP/NCAR re-analysis data 5 for the period 1948–2006, found that indeed many polar lows were realistically formed 4 . This simulation, henceforth named REA, generated an average number of 56 polar lows and high inter-seasonal variability (standard deviation 613), but no long-term trend in the number of seasonal polar low occurrences. A comparison of this model-based climatology with observed data is difficult because of the inhomo- geneous and often subjective character of past weather analysis, in particular when smaller-scale features such as polar lows are con- cerned. However, comparison with analyses using satellite data for a 1 Environmental Systems Science Centre, University of Reading, 3 Earley Gate, Reading, Berkshire RG6 6AL, UK. 2 Institute for Coastal Research /System Analysis and Modelling, GKSS- Research Centre, Max-Planck-Strasse 1, D-21502 Geesthacht, Germany. 3 Meteorological Institute, University of Hamburg, Bundesstrasse 55, D-20146 Hamburg, Germany. C20 A2 B1 0 0 5 10 15 20 25 30 35 40 45 50 10 20 30 40 50 60 70 80 90 100 Percentage frequency Number of polar lows per polar low season A1B a b 0 2 4 6 8 10 12 14 Jan Mar May Jul Sep Nov Average number of polar lows per month Month Ref. 15 Ref. 14 REA C20 B1 A1B A2 Figure 1 | Number of polar lows per polar low season and the seasonal cycle. a, Cumulative frequency of the number of polar lows resulting from the IPCC C20 scenario (red) and future scenarios B1 (green), A1B (yellow) and A2 (blue). Vertical error bars denote means and horizontal error bars denote the standard deviation. b, Average number of polar lows per month from observations 14,15 (orange and grey), from REA 4 (black) and from the IPCC scenarios (red, green, yellow and blue). Vol 467 | 16 September 2010 | doi:10.1038/nature09388 309 Macmillan Publishers Limited. All rights reserved ©2010