Analysing and forecasting the weather of early June 1944
Weather forecasts critical to the success of the D-Day landings of 6 June 1944 were made on the nights of 3/4 and 4/5 June 1944:
The first of these two forecasts, presented to General Eisenhower by his meteorological advisor, Group Captain J.M. Stagg, led to the postponement of the invasion planned for 5 June; the second enabled Eisenhower to make the decision to go ahead on the following day.
One of the forecasters involved, Lawrence Hogben, writing in the Royal Meteorological Society’s magazine Weather in June 1994 recalled how three separate teams, from the Met Office, the Royal Navy and the US Air Force, first made separate forecasts and then sought consensus - an early example of what today we refer to as ensemble forecasting. On the evening of 3 June the teams initially split two-to-one in favour of conditions leading to postponement; the following evening it was initially a two-to-one split in favour of conditions that would allow the invasion to proceed. Demanding military requirements, stormy weather in the Atlantic and associated fronts moving up the English Channel combined to make forecasting far from easy, and decisions were finely balanced.
Reconstructing the weather conditions for those critical days
Today, the science and technology of weather forecasting is far removed from that of sixty years ago. Forecasting beyond a few hours ahead nowadays relies on automated procedures for processing digitized observations to give a snapshot of the state of the atmosphere. This is used to initiate a computerised model of the atmosphere that projects the information forward in time to form the forecast. The initial snapshot, known as the “analysis”, itself depends on the atmospheric model; it is constructed by blending the latest observations with a “background” forecast initiated from the analysis made a few hours earlier. As time goes by, this process of “data assimilation” produces a picture of how the weather has evolved, built up by the succession of analyses.
The European Centre for Medium-Range Weather Forecasts (ECMWF), and sister organisations in Japan and the USA, have been active in programmes of “reanalysis” in which the observations taken over several decades are reprocessed using modern data assimilation systems to produce a complete historical record of the atmosphere, for use in a wide range of meteorological and climatological studies. In particular, ECMWF has recently completed ERA-40 (http://www.ecmwf.int/research/era/), a reanalysis that provides a three-hourly record of conditions back to 1957, when the atmospheric observing system was enhanced in preparation for the International Geophysical Year. The ERA-40 system was also run for an earlier spell to provide a depiction of meteorological events suurounding the disastrous North-Sea storm and flooding on the night of 31 January to 1 February 1953. (ERA40 Report Series - Report no 10) An earlier reanalysis carried out by the US National Weather Service went back as far as 1948.
We have now stepped back a further few years, and applied a slightly updated version of the ERA-40 data assimilation system to reconstruct the weather of May and June 1944. This has not been a comprehensive study; instead we simply made use of existing tools and a quite limited number of readily available observations. The results nevertheless agree in broad terms with the historical record insofar as we have examined it, and serve to illustrate several aspects of weather conditions and the forecasts leading up to D-Day.
The observations that have been assimilated were obtained from the archives of the US National Weather Service, and were supplied to us by Jack Woollen, to whom we express our thanks. The observations do not represent all that are potentially available; rather they comprise only those directly available in a suitable digitized format. The two main types of measurement are illustrated in Figures 1 and 2. The maps show the distributions of data available for the six-hour period between 0900 and 1500UTC on 3 June 1944, the observations that were blended with the background forecast to produce the analysis for 12UTC on that day. Figure 1 shows the locations of the land stations and ships which provided measurements of near-surface conditions, and Figure 2 the locations where vertical profiles of wind speed and direction were available from tracking so-called pilot balloons. In addition to these observations, ascents from four radiosonde balloons were also available, providing temperature as well as wind information. All of these, however, were located in the southern hemisphere. Click here to see a complete set of data coverage maps for the period 3-6 June. Click here to see today’s data coverage.
Three comments may be made on these maps. The first is that with some effort almost certainly a better coverage of surface observations could be obtained, as it is highly likely that additional records exist in many countries, albeit not in appropriate digitized form. There plentiful balloon data but an absence of surface data from the
The third comment on these maps relates to the nature of the observational coverage of the North Atlantic. Essentially, the observations describe conditions at its lateral and lower boundaries. The numerous pilot-balloon ascents, available for each six-hourly analysis, provide control of the data assimilation around much of the periphery of the North Atlantic, while the observations from ships provide control at the surface. The assimilating model of the atmosphere completes the picture.
Figure 1. Distribution of surface synoptic observations for 12UTC 3 June 1944
Figure 2. Distribution of wind observations from pilot balloons for 12UTC 3 June 1944
The weather situation, 3-6 June, 1944
Click here to see a complete set of analysis maps for the period.
The wind at a height of 10m above the surface and low-level cloud from the 12UTC analysis for 3 June is shown in Figure 3. An extensive area of south-westerly flow and low cloud can be seen ahead of a major Atlantic depression centred to the west of the map area. It was on the evening of this day that the forecasters, after earlier uncertainty, predicted that the cold front associated with this depression would move into the Channel, preceded by south-westerly wind and low cloud sufficient for the landing on 5 June to be postponed.
Figure 3. Low-level cloud and wind at 10m for 12UTC 3 June 1944
The corresponding map for 12UTC 4 June presented in Figure 4 shows stronger winds and more extensive cloud in the Channel than 24 hours earlier. Also evident, however, in the west of the map area is an indication of clearer, calmer weather. The forecasters at the time concluded that the clearance would extend to the Normandy beaches by 6 June, and that weather conditions, though by no means ideal, would be such as to allow the invasion to proceed.
Figure 4. Low-level cloud and wind at 10m for 12UTC 4 June 1944
Conditions for each of the analysis hours of 5 June are shown in Figure 5. Winds have strengthened ahead of the next frontal system that moves in from the south-west approaches of the Channel, and there is extensive low cloud. With the decision to invade on 6 June having already been taken, the first ships in fact had to set out from southern
Corresponding maps of the analyses for 6 June, D-Day, are shown in Fig. 6. Moderate westerly winds persist over the Channel throughout the day, strengthening a little over time. There is low cloud over the invasion beaches in the morning, but it clears by afternoon.
Figure 6. Low-level cloud and wind at 10m for 6 June 1944, at 00UTC (upper left), 06UTC (upper right), 12UTC (lower left) and 18 UTC (lower right)
The photographic record provides clear confirmation of this general picture. The two images of the early morning approach to Omaha Beach shown in the upper frames of Figure 7 are among many that reveal overcast conditions, and the drift of smoke along the beach indicates a wind from the west or west-north-west. The lower images show sunnier conditions for a convoy further from the beaches, and clear skies over Omaha Beach in the afternoon of June 6. The Met Office’s first synoptic report from Normandy, from Sword Beach at 1300UTC, was of mainly sunny weather and a northwesterly wind at Force 4 (http://london.iwm.org.uk/upload/package/4/dday/pdfs/DDayWeather.pdf). This and other evidence suggests that the analyses rather overestimate the extent of low cloud, and may have the wind a little too much from the west.
The early-morning approach to Omaha Beach
En route to the Normandy beaches Omaha Beach in the afternoon
Forecasts from 12UTC 3 June
Sets of forecasts have been made from the analyses produced for early June 1944. Figure 8 shows a selection of forecast maps from the analysis for 12UTC 3 June made using the same version of the atmospheric model as used in the data assimilation process. It shows the 42-hour forecast for 6UTC on the 5th, close to the time originally planned for the landings, and the longer range forecasts valid for 6, 12 and 18UTC on D-Day itself. It reproduces essentially the same picture as the analyses shown earlier, though with even a little more low cloud. The methods and supporting computing technology that we have today are capable, based on the limited observations available up to 15UTC on 3 June, of deducing not only that conditions were basically unfavourable for invasion on 5 June, but also that conditions offered the prospect of success for an invasion postponed by just one day.
Figure 8. Low-resolution forecasts at 42-, 66-, 72- and 78-hour range, starting 12UTC 3 June 1944
The reanalysis system that we have used to process the observations leading up to 6 June was not designed to provide the most detailed possible snapshot of conditions for a particular day. Instead, it was constrained to limit the computational cost of processing observations for more than four decades for the ERA-40 project. In particular, it used a lower horizontal resolution than that currently employed for day-to-day weather forecasting. This has implications for the capability of the analyses and forecasts to describe the sharpness of cloud features and local coastal effects.
We have thus also carried out a forecast from 12UTC 3 June using a higher resolution atmospheric model. The computational mesh length was reduced from about 125km to 25km, rather finer than the 40-km resolution of ECMWF’s current operational model. Figure 9 shows examples of forecasts from this high-resolution model; they are for the same times as shown for the low-resolution reanalysis model in Figure 8. The sharpness of the cloud boundary and change in wind direction at the cold front at 6UTC on 5 June, and the more distinctly banded structure of the cloud circulating closer to the centre of the Atlantic depression, are evident. The diminishing cloud band at 6UTC on 6 June is much more confined, with low cloud still over the invasion beaches but clearer skies over the Channel. The sky is clearer over the Normandy coast also by 12UTC on the 6th.
Figure 9. High-resolution forecasts at 42-, 66-, 72- and 78-hour range, starting 12UTC 3 June 1944
Click here to see a more complete set of forecast charts.
The sensitivity of the cloud forecasts to the resolution of the atmospheric model serves as a reminder that our reconstruction of the weather for early June 1944 is based on an atmospheric model that cannot be perfect, constrained by a set of observations that by today’s standards is highly limited. There are doubtless aspects of the picture that we have presented that are wrong, in detail at least, just as small inconsistencies emerge if one reads different historical accounts of events at the time. Doubtless also, a better reconstruction could be made by carrying out “data archaeology” to recover more of the original observations in suitable digitized form, and to assimilate these observations using a higher resolution version of the atmospheric model. Nevertheless, the exercise we have undertaken appears to provide a broadly correct view of the sequence of events in early June 1944. It has also, we hope, served to illustrate some of the challenge that was faced so successfully by Stagg and his teams of forecasters sixty years ago.
Complete set of D-Day reanalysis charts and forecasts
US Naval Historical Research Center
Today's data coverage charts
ERA40 Project Report Series
Met Office report on D-Day weather and forecasts