Background:

‘David’ was a damaging storm that developed on the night of 17 January 2018 as it approached the Republic of Ireland and then moved eastwards across England and Wales. At its strongest, inland gusts of 60–70 knots were observed across East Anglia between 05 and 07 Z, and then later that morning along the coastal strip of the Netherlands, where a peak gust of 78 knots was observed at Hoek van Holland, this was reported around the time that the pressure in the depression bottomed out at 975 hPa.

The storm had major impacts on the rail infrastructure of East Anglia, with all but 1 line affected by either fallen trees or downed power lines, and approximately 155,000 homes were left without power during the morning. The road network was also affected, but to a lesser extent. A brief closure of the Orwell bridge (A14) caused issues with traffic in Ipswich, and there were delays also on the M25 (with the closure of the QE2 bridge) as well as the M11, A47 and A11 due to overturned lorries across Cambridgeshire and Suffolk.

There was however much more disruption across other parts of north-western Europe. Amsterdam’s main aviation hub, Schiphol Airport, was closed during the storms peak with around 260 flights cancelled in total; the rail networks across both the Netherlands and throughout the North Rhine-Westphalia region in Germany were also badly affected by the strong winds. Germany was the worst hit area with the storm peaking in the middle of the day, causing 10 known deaths and damage estimated at between €1 bn and €2.6 bn.

An uncertain situation:

As is typical of these dynamic situations, small but subtle shifts in the position of the jet stream (near to, or over the UK - particularly as in this case where the jet max is >200 knots) in comparison to an approaching marked baroclinic zone had significant impacts on the timing of interaction and subsequently the predicted depth and track of the storm through the forecast period. There were therefore some quite dramatic swings in the predicted storm track, first north, and then back south, with all models predicting a vast swathe of solutions. The deeper solutions depended upon optimal coupling between the marked baroclinic zone and the jet stream, with its deepening in response to crossing to the cold side of the jet in the left exit region (large-scale ascent through divergence aloft). That said, the output from the Met Office UM suite was relatively consistent even at a long lead time, generally forming the deepest depression out of all the deterministic models, albeit moving away from these deeper solutions on the day before the storm developed. Indeed the guidance unit used this output, as well as that available from other centres, to produce this forecast chart for T+120 (see Figure 8) which verified remarkably closely with the observed depth and track of Storm David.

Uncertainty is often difficult to present clearly or quantify, particularly as simply meaning the various forecast pressure fields will often only provide an unhelpful bland signal, where extreme solutions that could be as likely, or even more likely, are smoothed out as each is usually located in a slightly different position – therefore below I have presented some of the output currently available for use by meteorologists in order to mitigate against this.

The CDF ('Cumulative Distribution Function') plot is designed to show not just the spread of solutions within the ensemble, but also how they compare with the ‘M’ (model) climate which is what the black line in Fig 3 represents. This can then give the forecaster an idea of how extreme an event is likely to be compared with climatology (comparing like-for-like using the same model biases, etc). More detail on the ECMWF ‘M-Climate’ can be found on their web-pages https://www.ecmwf.int/en/forecasts/charts/product-descriptions/Medium-range%20forecasts but it is effectively an ensemble of re-forecasts over a 5 week period centred around the week of interest ran over 32 days in 20 previous years with 9 staggered forecasts and 11 ensemble members in each set of forecasts. This gives approximately 1980 re-forecast fields and a very comprehensive spread from which to compare forecast data with to enable a judgement made over how extreme an event is likely to be.

To demonstrate the underlying uncertainty, Figure 5; below right, which has been annotated to show strength of wind in knots, the plot is a CDF provided by ECMWF showing the spread of predicted maximum gusts at Wattisham in Suffolk. This was taken from the 00 Z run on the day from ECMWF (at just T–5hr before the storm hit and would not have been available to meteorologists until after the storm had passed!) and as you can see on the latest run (coloured red) there is even then still some uncertainty with a spread from 56 knots to 72 knots, albeit a lot less than the previous few runs which suggested a maximum gust of anywhere from 40 knots up to 65 knots! The jump in the final run is also quite significant and serves only to highlight the high sensitivity of the model to the aforementioned changes described above.

Each of the past 10 runs (over 5 days) is then overlaid against the black ‘M-climate’ in increasingly bright colours on a Cumulative Distribution Function (CDF) chart. In general, steep CDFs are indicative of high confidence, i.e. small spread, and shallow curves with long tails represent low confidence – as in this case in earlier runs. Note also the large departure from the M-climate, with the final run lying almost entirely beyond the 99 percentile and therefore indicative of an ‘extreme’ event.

Presented in Figure 7, also below; are the CDFs for Amsterdam (Netherlands) and Bernburg (Germany) which show similar forecast evolutions – note also the large steps between each successive run in the Bernburg CDF which is indicative of a significant change to the forecast, likely occurring because of the sudden much more southerly track forecast as the event approached. Also presented in Figure 9, below; is the output from the UKV taken from the 00 Z run on the 17th January 2018 which shows how well the Met Office suite did with predicting the expected strength and areas most likely to be affected by the worst of the winds associated with Storm David.

Rapid development:

Initially a rather innocuous shallow depression with central pressure of 1007 hPa located out at 51 N, 40 W at midnight on 17th January 2018, Storm David rapidly deepened on its approach to Ireland during the evening and by midnight on the 18th January 2018 (see Figure 1; right - for detailed successive ASXX of development on 17/18 January 2018 and Figure 10 for its track, as produced by DWD) it had deepened to around 981 hPa, a fall of 26 hPa in 24 hrs - a weather 'bomb', as it moved into the Irish Sea. At its peak over the southern North Sea it was estimated to have deepened to around 975 hPa shortly before moving into the coast of the Netherlands - not especially unusual (or particularly notable) in its depth for the time of year - however the region through which it went through its explosive cyclogenesis phase was a lot further south than typical and hence as a newly formed depression its most active and dangerous phase coincided with much more populous areas than would be typically expected.

The swathe of strongest winds associated with the storm developed in a relatively narrow corridor of approximately 100 nm in length. The sounding for Essen (Figure 11, 12 Z on 18 January 2018) typifies the boundary layer profile observed in this region – with a well-mixed and unstable DALR (Dry Adiabatic Lapse Rate) profile up to 2500 FT lying beneath a capped moist layer which had an inversion at approximately 7000 FT. Supportive of efficient downward transfer of momentum, this may have been enhanced further by much denser cold and dry air which was being pulled into the southwest of the depression, as well as a marked isallobaric surge (see Figure 3, left) of rising atmospheric pressure in the left jet entrance region.

Figure 3 analyses the change in pressure tendency across the southern half of the UK, this analysis can be useful at determining the isallobaric wind, a departure from geostrophic flow, with a magnitude proportional to the isallobaric gradient directed from higher values of pressure tendency to low. In essence producing outflow from regions of strongly increasing pressure (positive isallobars) to inflow in regions of rapidly falling pressure (negative isallobars), where the tightest gradient is indicative to a significant ageostrophic component to the observed wind.

Also evident from water vapour imagery was the development of an incursion of very dry air over the developing storm, indicative of stratospheric air, which itself has much higher potential vorticity, descending into the troposphere - with the wind profiler output from Wattisham (Figure 12) possibly corroborating this through the observed slow descent of a drop-off in wind data, indicative of the presence of very dry air, as the storm approached. Figure 3 (above): Isallobaric analysis at 06 Z on 18 January 2018

Conclusions:

Storm David was a notable winter windstorm which produced winds inland across East Anglia that had not been recorded since October 2002; its development was incredibly finely balanced on the exact interaction between a strong jet stream and a marked baroclinic zone with consequent unusually high levels of uncertainty over its predicted track and depth. The UM suite in particular handled the storm reasonably well with this highly dynamic situation pushing the UM and the ECMWF IFS to their limits, and given this uncertainty timely warnings were issued by the Guidance Unit well in advance of the storm. Whilst the exact depth was slightly underestimated in advance of the storm (with knock-on effects on storm naming etc.) this blog aims to show how this uncertain and rapidly changing situation is one of the most challenging situations for meteorologists, and also to suggest that CDF output from MOGREPs could be of immense value.

During the evening skies cleared as expected from the model output, though the temperature fell much faster than expected - this due to the fact that the topography surrounding Wattisham favours rapid cooling in a slack west to southwest flow. As the temperature fell away the visibility also decreased and matched well the fall in visibility predicted from the adapted Middle Wallop technique (see forecast and observations in Figure 6), with fog forming around 0040 Z at a temperature of just below 7C.

This event demonstrates how by using a mix of empirical techniques, model output and local forecasting skill a reasonable forecast can be drawn up providing good advice to the customer - in this event the customer was able to carry out night flying operations ahead of the poorer conditions that developed by 21 Z.