Historical weather extremes in the “Twentieth Century Reanalysis”

Meteorological or climatological extremes are rare and hence studying them requires long meteorological data sets. Moreover, for addressing the underlying atmospheric processes, detailed three-dimensional data are desired. Until recently the two requirements were incompatible as long meteorological series were only available for a few locations, whereas detailed 3-dimensional data sets such as reanalyses were limited to the past few decades. In 2011, the “Twentieth Century Reanalysis” (20CR) was released, a 6-hourly global atmospheric data set covering the past 140 years, thus combining the two properties. The collection of short papers in this volume contains case studies of individual extreme events in the 20CR data set. In this overview paper we introduce the first six cases and summarise some common findings. All of the events are represented in 20CR in a physically consistent way, allowing further meteorological interpretations and process studies. Also, for most of the events, the magnitudes are underestimated in the ensemble mean. Possible causes are addressed. For interpreting extrema it may be necessary to address individual ensemble members. Also, the density of observations underlying 20CR should be considered. Finally, we point to problems in wind speeds over the Arctic and the northern North Pacific in 20CR prior to the 1950s.

Arctic Winds in the “Twentieth Century Reanalysis”

Climate in the European part of the Arctic underwent a rapid warming between the 1910s and the 1930s. Previous studies have addressed the role of atmospheric circulation in this period based on geopotential height fields because observations of upper-level winds in the Arctic are rare. Here we analyse winds over the Arctic and specifically over Spitsbergen in the “Twentieth Century Reanalyses” (20CR). We compare in situ upper-air wind measurements performed in 1912 and 1913 in Spitsbergen with six-hourly 20CR data. Furthermore, we compare monthly-to-seasonal 20CR winds at 700 hPa over the European Arctic with statistically reconstructed winds at 3 km altitude. Finally, we analyse long-term trends in Arctic winds in 20CR. The general agreement between observed upper-air winds and 20CR on the day-to-day scale is rather poor, which is not surprising given the paucity of observations in the Arctic at that time that constrain 20CR. In contrast, the seasonally averaged winds (which represent a larger spatial scale) in 20CR compare well with statistically reconstructed winds. The analysis of long term near-surface wind time series in 20CR shows arguably artificial trends from 1871 to around the 1950s over sparsely observed regions, particularly oceanic regions. Densely observed regions such as Europe or the USA show no such trends. This analysis shows that great care needs to be taken when working with 20CR in the Arctic and other sparsely observed regions.