CS2 Snow Case Study: 12th January 2017

Despite five (by recent standards) pretty brutal winters spent on Open Road between 2007 and 2012, I still enjoy forecasting snow. It is a welcome change from the grey skies and curtains of depressingly familiar drizzle; however it appears less frequently today, and although always difficult to call correctly, it is guaranteed to cause headlines and give forecasters a headache due to snow forecasting almost exclusively associated with marginal events today. So not only does its sudden arrival mean a quick dusting down of the tried-and-tested forecasting techniques, its fragile dependence on temperature, precipitation rate etc. mean these techniques and our own experience are tested more than ever! One such event is now discussed in more detail below…

Developing frontal wave

This event was picked up a few days before on the 9th January, with predictions of a fairly flat and innocuous looking frontal wave to develop rapidly into a vigorous depression as it transited through the English Channel on the 12th (see Figure 1; right, showing the position of the wave at 12 Z). This became rather more worrying when the Euro 4 then predicted t hat it would deepen whilst over southeast England along with a simultaneous cold Arctic plunge digging south into it, this then potentially leading to a spell of snow and blizzard conditions with strong N to NW winds on the northern flank of the depression. Neither GFS nor ECMWF suggested such a deep low and so low confidence was afforded to the solution provided by the Euro 4 and to a lesser extent the UKMO Global. A yellow NSWWS alert was issued to account for the potential impact (though low probability) of the event and all eyes awaited the next run.

There are numerous empirical techniques available to help forecasters to predict whether precipitation is likely to fall as rain, sleet or snow - but the most appropriate in this case was to use the wet bulb freezing level. This is because by doing this you take into account latent cooling effects, which can - especially in these types of events - dictate the level to which snow is likely to fall. For example, even with a wet bulb zero level at the upper limit of the 1000–2000 ft (300–600 m) range, persistent moderate precipitation can quite rapidly transition to sleet at the surface through cooling of the underlying air by melting snow. With wet bulb zero levels of below around 1200 ft precipitation is quite likely to rapidly change over to snow.

A technique that is also used widely is Boyden’s technique (see Figure 9); this looks specifically at the layer of the atmosphere between 1000 and 850 hPa (otherwise known as the partial thickness, essentially the lowest 1.5km or 5000 ft) and its accuracy is highly dependent on the boundary layer being homogenous / well mixed – which is not always the case. Therefore this technique tends to perform better in convective setups and/or where strong flow predominates allowing a well-mixed boundary layer. The technique does not capture the subtleties of the temperature profile within the boundary layer (i.e. a cold undercut) and is therefore not considered as useful in this specific case, nor other cases like it where the precipitation is driven dynamically.

Figure 6: Forecast MSLP from ECMWF T+18 (left) and GFS T+6 (right) for 18 Z on 12th January 2017

Figure 4: Observations at 18 Z (left) compared with Euro 4 prediction (right) from 06 Z model run

Snow day

The 12 January arrived and less than 12 hours before the event there still remained a rather marked disparity between the major computer model output from UKMO, GFS and ECMWF leading to lower than usual confidence over the forecast. The former still much more developmental than the other centres, however all now developed a low and predicted a spell of wet snow over southeast England during the early evening on the north-western flank of the depression. At Wattisham, the duty meteorologist (in accordance with guidance) anticipated that rain would arrive by midday and gradually change over to sleet and eventually snow after 16 Z as a combination of cold advection and intensity driven cooling acted to lower the snow level, and this was issued to the customer as the most likely forecast.

Figure 1: ASXX for 12 Z on Thursday 12 January 2017

Figure 2: ASXX for 00 Z on Friday 13 January 2017 - showing the now mature depression (980hPa) moving quickly eastwards into the Netherlands/Belgium/NW Germany

Figure 5: Forecast of snow lying from E4 (top row) and UKV (bottom row) on 12th January 2017

Figure 7: Map showing where Twitter users are reporting snow in the UK at 1730 Z

Figure 8: Observation sequence from Wattisham (1330-2300 Z on 12 January 2017)

In Figure 3 (below) are the radiosonde soundings from Larkhill for 10 Z, 14 Z and 16 Z respectively. The first two show the wet bulb freezing level relatively high at around 2500 ft AMSL, suggesting snow down to around 1300 ft AMSL – this alone not enough to bring snow to the surface – though note the isothermal that is forming on the 14 Z ascent. By 16 Z, marked cold advection has taken place with the wet bulb freezing level now down to just over 1000 ft and very likely to lead to snow at the surface (indeed sleet was reported by Larkhill at 16 Z and snow by 17 Z) – note also that the Boyden 80% height is at the wet bulb freezing level, and is considered much too optimistic (or pessimistic, depending on your point of view!)

Figure 3: Radiosonde ascents from Larkhill for 10 Z (left), 14 Z (centre) and 16 Z (right) showing the progression to which cooling of the boundary layer took place through the day

A marked cloud head was evident on IR (Infrared) imagery showing cyclogenesis underway, which accelerated through the afternoon and evening as the wave moved through the channel, as forecast. The Euro 4 had perhaps over deepened the low (see Figure 4, which compares the 06 Z run with the observations for 18 Z), though regardless a significant depression was starting to form with a large corridor of precipitation on the poleward side associated with a thermally strong warm front. A plot of isallobars (pressure change/tendency), showed where the largest pressure falls were and where the depression was moving towards at 18 Z, as well as an indication of a marked isallobaric 'surge' in the southwestern quadrant of the depression. This surge marked an area where there were significant gusts in excess of 60 knots during the evening. The isallobaric wind describes the ageostrophic component driven by local rates of change in pressure and is proportional to the isallobaric gradient, directed from positive values of pressure tendency (i.e. rising) to low.

Snow or no snow?

It definitely looked likely in the slack flow to the north of the depression where intensifying precipitation is moving north against a slow ebbing away of the milder air with steady cold advection taking place from the northwest. Using the NWP output and observations from the Larkhill radiosonde soundings, a forecast for snow seems to be a fairly safe call now - the question next is will it accumulate? If so, where was it most likely to accumulate?
Things that needed to be taken into account included:

The initial depth temperature (if this is significantly higher than 3C - MOD to heavy snow is required to settle), which leads nicely to...
Precipitation intensity
and finally.. how much precipitation expected over the period of interest.

The Euro 4, perhaps with its overly deep depression, was keen to suggest falls of between 5 and 10 cm widely across East Anglia - whereas GFS and ECWMF (see Figure 6) both suggested that amounts would be relatively negligible, with perhaps 0-1cm likely across the areas where the heaviest precipitation was likely, i.e. across Essex and parts of Kent. Note that there were large variation in the snow amounts suggested by the post processed Euro 4 and UKV (see Figure 5) and confidence remained low over the amounts of snow likely due to uncertainty in the development of the depression, and therefore its track and the region likely to see the heaviest precipitation.

What happened?

The wave depression developed upon entering the English Channel during the late morning just south of Cornwall, and had deepened to a low of approximately 992 hPa by 15 Z close to Alderney, with an arc of precipitation - falling entirely as rain - extending to the northeast of the depression across central southern England and into East Anglia. By 18 Z, the depression had deepened significantly to 985 hPa and precipitation had widely turned to snow across Essex, London, Kent, Berkshire, Surrey and into parts of Hampshire and Wiltshire with moderate snow falling in and around Essex. One of the benefits of social media is that it allows information about rapidly changing situations to be quickly communicated, and indeed a quick check of '#uksnow' (see Figure 7) on Twitter quickly confirmed the extent of snow falling. With the depression centred slightly further south (approximately by 50-80NM) and slightly shallower than predicted by the Euro 4, the heaviest falls were actually across southern Essex, and into North Kent where reports of up to 2 cm lying snow were given - far below the expectations of the post processed output - but significant nonetheless.

Meanwhile at Wattisham (for observation sequence, see Figure 8) the rain increased in intensity and changed to sleet around 1630 Z with a temperature of 2C. Whilst there was a period of snow as the wet bulb freezing level was lowered (around 1745 Z onwards), this through a combination of precipitation intensity and steady cold advection from the northwest, the precipitation was lighter than expected and so ended up mainly a mix of rain and snow; consequently this meant that the snow failed to settle at the airfield apart from briefly in a heavier spell around 1750 Z.

Although in the end the event was less severe than expected, uncertainty surrounding the depth and track of the low, and the evidence of what happened within the belt of heavier precipitation across Essex demonstrates that the warning was justified. Events like this (high impact but low certainty) can often be quite challenging and lead to a 'bust' forecast when only a few miles away it would have been correct, and this case demonstrates that to get the best results will always take a combination of experience and judgement to challenge and reconcile NWP output in order to provide the best possible forecast to the customer.

Figure 9: Boyden's snow forecasting technique