The provision of weather forecasts for the London Olympic and Paralympic Games in 2012 offered the opportunity for the Met Office to accelerate the transition to operations of several advanced numerical modeling capabilities and to demonstrate their performance to external scientists. It was also an event that captured public interest, providing an opportunity to educate and build trust in the weather forecasting enterprise in the United Kingdom and beyond. The baseline NWP guidance for the duration of the Olympic Games came from three main configurations of the Met Office Unified Model: global 25-km deterministic, North Atlantic/Europe 18-km ensemble, and U.K. 1.5-km deterministic. The advanced capabilities demonstrated during the Olympic Games consisted of a rapid-update hourly cycle of a 1.5-km grid length configuration for the southern United Kingdom using four-dimensional variational data assimilation (4D-Var) and enhanced observations; a 2.2-km grid length U.K. ensemble; a 333-m grid length configuration of the Unified Model and 250-m configuration of the Simulating Waves Nearshore (SWAN) ocean wave model for Weymouth Bay; and a 12-km grid length configuration of Air Quality in the Unified Model with prognostic aerosols and chemistry. Despite their different levels of maturity, each of the new capabilities provided useful additional guidance to Met Office weather advisors, contributing to an outstanding service to the Olympic Games organizers and the public. The website provided layered access to information about the science and to selected real-time products, substantially raising the profile of Met Office weather forecasting research among the United Kingdom and overseas public.

The provision and performance of advanced numerical modeling capabilities during the 2012 London Olympic and Paralympic Games are discussed.

Since the Atlanta Olympic Games in 1996 (Rothfusz et al. 1998), a tradition has developed of using the meteorological requirements of the Olympic Games to support a program of technology acceleration and transfer activities. In Sydney, this was opened to international participation as a World Meteorological Organization (WMO) Forecasting Demonstration Project (FDP) that took advantage of enhanced observing capabilities and an experimental forecaster workstation to demonstrate the capabilities of several precipitation and severe storm nowcasting systems (Keenan et al. 2003). In Beijing in 2008, there was, again, a greatly enhanced observing network and the China Meteorological Administration implemented several new numerical weather prediction (NWP) models and forecasting systems. The opportunity was taken to hold both a FDP on nowcasting systems and a Research and Development Project (RDP) on short-range mesoscale ensembles, to both of which several international teams contributed (Wilson et al. 2010; Duan et al. 2012). During the Winter Games at Whistler in 2010, Environment Canada again implemented a greatly enhanced observing system and used new capabilities to support an RDP in winter hazard nowcasting (Joe et al. 2010). With this background, there was an expectation that some form of demonstration project might run during the London 2012 Olympic Games. However, the Met Office was in a different position from its predecessors in that only very limited enhancements to the observing network were required to support the Games and an existing forecasting production system would be used, which did not have the capability to support partners' forecasting systems. It was concluded that a more appropriate and affordable option would be to accelerate selected areas of research so as to deliver a web-based technology demonstrator of some new advances in weather forecasting science, both to provide additional guidance to forecasters and to share their performance with the international science community.

The requirement for weather information in support of the Games consisted of three parts:

  • the requirement of the public, especially overseas visitors, expressed through the Public Weather Service Customer Group (Met Office 2013a), for generic forecasts for their location, for locations of events (including the torch relay), and to support travel plans before, during, and after the games;

  • the contractual requirement of the London Organizing Committee of the Olympic Games (LOCOG) to support management of the games, especially the opening and closing ceremonies, and those sports with specific weather sensitivities; and

  • the requirement of the emergency response community, expressed through the Public Weather Service Customer Group, to provide information about hazardous weather conditions that might cause a civil emergency or that might affect their response to such an emergency.

The Met Office operational response to these was as follows:

  • The enhanced public requirement was primarily met through the website and mobile phone apps, giving access to automated forecasts for 5,000 locations in the United Kingdom up to 5 days ahead (Met Office 2013b,c).

  • The LOCOG requirement was primarily met through briefings by teams of forecast advisors at Olympic Park, Eton Dorney, and Weymouth (see Fig. 1 for locations), who used NWP and nowcast products supplemented with expert interpretation and judgment by the Chief Forecasters at Exeter, and by their own interpretation in the light of specific local sensitivities. These teams were volunteer Met Office forecasters with appropriate experience who underwent specific training in the critical weather requirements of the Games' events.

  • The emergency response requirement was primarily met through the National Severe Weather Warning Service, (Met Office 2013d) operated by expert forecast advisors at Exeter under the guidance of the Chief Forecaster, and delivered both through the “Hazard Manager” web service (Met Office 2013e) and by dedicated mobile Public Weather Service Advisors (Met Office 2013f) in London and in the U.K. regions, as required.

Fig. 1.

Locations referred to in the text.

Fig. 1.

Locations referred to in the text.

In recent years, following operational implementation of the 1.5-km grid length U.K. configuration (UKV) of the Met Office Unified Model (MetUM) (Davies et al. 2005; Tang et al. 2012) for short-range forecasting, the Met Office has redefined its forecasting production system. Most public forecasts are now issued without human intervention (Met Office 2013b), while high-impact weather forecasts and warnings are delivered through forecast advisors who interpret and communicate weather information specific to the needs of the recipient. The Met Office's program of further development of automated short-range weather forecasting is focused on

  1. improving the performance of the UKV model, primarily through improvements to data assimilation and the use of ensemble prediction; and

  2. coupling the UKV model to impacts models, including land surface hydrology, the oceans, atmospheric chemistry, and human systems such as transport, buildings, and drainage, in partnership with several leading scientific institutes in the United Kingdom under the umbrella of the Natural Hazards Partnership (Met Office 2013g).

From this program, four new science capabilities were chosen for acceleration to deliver enhanced guidance and products during the Games:

  • The Air Quality in the Unified Model (AQUM) forecast system was put into operational use to add daily air quality index (DAQI) forecasts to the automated forecasts for the 5,000 sites. In addition to site-specific forecasts, U.K. maps of this index were displayed on the Olympic Showcase website. The 5-day forecast capability provided guidance to forecast advisors supporting LOCOG and emergency responders. In addition, this made a 5-day forecast of air quality available to the general public for the first time (no other organizations produce U.K. forecasts this far in advance) as an aid to their planning.

  • A very high-resolution NWP [333-m grid length (UK300)] model and ocean wave (250-m grid length) model for Weymouth Bay able to predict the influence of the Isle of Portland on wind and wave conditions to the east, presented in the Olympic Showcase, and used by the games' forecast advisors at Weymouth in their briefings to the organizers and sailing teams.

  • A 1.5-km grid length hourly rapid update NWP cycle for southern part of the United Kingdom, incorporating additional observations and an advanced data assimilation capability, from the Nowcasting Demonstration Project (NDP), aimed at improving forecasts for the next few hours ahead. This was described in the Olympic Showcase, and some examples were provided in the technical pages for external scientists. Live forecast data were also used as additional information by the forecast advisors, particularly on those occasions when convective storms threatened the main Olympic Park.

  • A 2.2-km grid length ensemble prediction system for the United Kingdom, nested in the existing operational global and regional ensemble systems, to provide information on the range of outcomes that should be considered when making weather-dependent decisions. A new range of ensemble presentation products was developed for evaluation during the games, focusing on the chances of the occurrence of high-impact weather, particularly low or high winds, heavy precipitation, and high temperatures.

Each of these systems was already under development when preparations for the Games started and the main new requirement for the games was to create the online Olympic Showcase to explain the science and to display the results. A summary of operational and experimental model configurations is given in Table 1.

Table 1.

Atmospheric model configurations for the 2012 London Olympic Games. All configurations have 70 levels. Those indicated “high top” have a raised model top with lower resolution in the troposphere.

Atmospheric model configurations for the 2012 London Olympic Games. All configurations have 70 levels. Those indicated “high top” have a raised model top with lower resolution in the troposphere.
Atmospheric model configurations for the 2012 London Olympic Games. All configurations have 70 levels. Those indicated “high top” have a raised model top with lower resolution in the troposphere.

The object of this paper is to provide an overview of the four showcase capabilities in the context of their use during the games. More detailed papers (Table 2) will describe the science behind them and present rigorous evaluations. In the following sections, we first provide an overview of the new science, its implementation, and the products that were made available, both to the public and to forecast advisors, and then we give a brief indication of performance, based on anecdotal feedback and on a single case study. We conclude with some general observations on overall outcomes.

Table 2.

Projected papers on detailed performance of the Olympics Showcase capabilities.

Projected papers on detailed performance of the Olympics Showcase capabilities.
Projected papers on detailed performance of the Olympics Showcase capabilities.

THE NEW CAPABILITIES, THEIR IMPLEMENTATION, AND DELIVERY OF PRODUCTS.

The aim of delivering experimental forecasts to a diverse range of users, without creating confusion with the official forecasts, provided a major challenge. This was addressed by creating a dedicated hierarchy of web pages, within the Met Office website (Met Office 2012), presenting the information in a layered manner with detailed real-time information available at the lower levels to those with the technical knowledge to make use of it. At the top level, primary links from operational forecasts for the Games' venues were provided to web pages giving background on the influence of the weather on particular sports. The second-level pages contained outline descriptions of Met Office forecasting capabilities, including the new advances. Each of these was linked to more detailed, scientific descriptions of the new capabilities and finally to pages offering real-time outputs in a variety of formats. Multiple navigation routes were provided for different audiences, so that, for instance, scientists could access real-time model outputs directly through the “Invent” tab, normally used for beta testing of new services.

Daily air quality index forecasts.

The Met Office has produced air quality forecasts for the British Broadcasting Corporation (BBC) for many years by feeding meteorological forecasts into its Lagrangian dispersion model, the Numerical Atmospheric Modeling Environment (NAME; Jones et al. 2007). Work to integrate atmospheric composition modeling into the MetUM has been underway for several years as part of the climate research program in partnership with the Universities of Cambridge and Leeds in the United Kingdom Chemistry and Aerosol (UKCA) project (Morgenstern et al. 2009). This capability has now been further developed to permit gas phase chemistry and aerosol modeling in a limited area forecast configuration of the MetUM—the AQUM model (Savage et al. 2013). The domain covers northwestern Europe at a 12-km grid length; lateral boundary conditions for meteorology are taken from the MetUM global model forecast and those for chemistry and aerosols from the Monitoring Atmospheric Composition and Climate (MACC) global model forecast (Flemming et al. 2009) operated by the European Centre for Medium Range Weather Forecasting (ECMWF). AQUM is routinely verified against U.K. air quality observations and has been shown to be competitive with other world-leading regional air quality models (Savage et al. 2013). Following discussions with stakeholders, the decision was taken to accelerate the transition to operational use of AQUM, making this a part of the core forecasting service to the public for the Games with automated daily outputs available as part of site-specific forecast displays on the Met Office and BBC websites. In addition, for the Olympics Showcase, maps of predicted DAQI out to 5 days ahead were provided. For consistency with established dose response relationships, the current U.K. standard for reporting air quality is defined by the U.K. Department for Environment, Forestry, and Rural Affairs (DEFRA), specifically as a daily index, which is computed from specified average concentrations of key pollutants with reference to dose limits set by the European Union's Ambient Air Quality Directive (DEFRA 2013).

Very high-resolution forecasts for Weymouth Bay.

A key operational requirement for the games was to predict the wind distribution in Weymouth Bay in support of the sailing competitions. The limits of acceptable sailing conditions are very stringent, excluding extended calms as well as dangerously strong winds or high waves. A key determinant of the wind regime in Weymouth Bay is the Isle of Portland, a 100-m high peninsula with vertical sea cliffs extending 5 km south from the mainland into the English Channel. In summer, heating of the land in shallow pressure gradients leads to complex patterns of sea and land breezes, while in strong wind conditions the peninsula creates a wake. None of these effects is resolved in the UKV model that represents Portland by just two grid points. To understand the influence these effects might have on the sailing events in the Games, a climate study was carried out using a 333-m configuration of the MetUM (UK300) to downscale historical analyses. Although few observations were available for validation, results were consistent with subjective understanding of the wind patterns in this area. It was therefore decided to use the same configuration in real time to provide wind guidance to the Weymouth Bay forecast advisors.

The Isle of Portland also has a major impact on the ocean wave climate of Weymouth Bay. In addition to its sheltering effect, variations in the sea bed bathymetry in the vicinity of the bay affect steepness and dissipation of wind-generated waves. For longer period waves, such features also lead to the focusing of wave energy, and in cases of very long period swell in the English Channel, these effects can have severe impacts for beach erosion and coastal flooding (e.g., Draper and Bownass 1983). At the time of the Olympics, the Met Office ran an operational 12-km grid length “North Atlantic European” domain wave model based on WAVEWATCH III (Tolman 2009). This relatively coarse configuration did not represent Portland at all and had not been configured for tidal waters, so a 250-m grid length configuration of the Simulating Waves Nearshore model (SWAN; see Booij et al. 1999) was employed. While it would have been attractive to couple this to winds from the UK300 model, it was felt that the risk of system failure would be increased and that the benefits to wave growth in the nearshore region would be minimal. It was therefore driven with wind forecasts from the UKV model, wave boundary conditions from the 12-km wave model, and variable water levels from the Continental Shelf, version 3 (CS3) storm surge model (Flather et al. 1998).

In displaying model results, the aim was to communicate the overall nature of the wind and wave conditions in the bay as a function of time and space and then to enable more detailed inspection if required. A combined presentation focusing on the central part of the computational grids was used, as shown in Fig. 2, with animation of the forecast time range and with options to select specific lead times and to display wind or waves separately.

Fig. 2.

The 27-h forecast of winds (half barb = 5 kt; full barb = 10 kt; flag = 50 kt) and waves (color contours, m) from the Weymouth Bay models for 1200 UTC 2 Aug 2012. Note the sheltering effect of the Portland on the left.

Fig. 2.

The 27-h forecast of winds (half barb = 5 kt; full barb = 10 kt; flag = 50 kt) and waves (color contours, m) from the Weymouth Bay models for 1200 UTC 2 Aug 2012. Note the sheltering effect of the Portland on the left.

Hourly NWP forecast updates.

In 2011, the Met Office implemented the UKV model as its primary guidance for U.K. short-range weather forecasts to 36 h. It has been very successful, particularly in its prediction of the realistic structure of rain systems that are either convectively generated or have convection embedded (Lean et al. 2008). Much of this success has been achieved through the interaction of free subsynoptic-scale weather structure, passed through the boundaries from the North Atlantic and Europe (NAE) model (see Table 1), with finer-scale topographic detail resolved by the UKV. Additional benefit comes from continuous cycling 3-hourly 3D-variational data assimilation (3D-Var; Lorenc et al. 2000; Renshaw and Francis 2011) plus latent heat nudging (Jones and Macpherson 1997). However, at analysis time, this system cannot match the observed convective-scale weather, largely due to the remaining limitations in the data assimilation scheme.

For forecasting of such small-scale weather features up to about six hours ahead, the Met Office is currently dependent on radar- and satellite-based nowcasting methods blended with NWP (Golding 1998; Wilson et al. 2010). These methods use persistence or extrapolation of the observed pattern of the weather variable of interest to correct the first six hours of the NWP forecast. For precipitation, we use the blended version of Short-Term Ensemble Prediction System (STEPS) in which a multiplicative-scale cascade is used to separate the treatment of scales with different predictability horizons (Bowler et al. 2006). While this system has provided the primary guidance for operational flood warnings in recent years, it is not able to predict the very short-range evolution (initiation and dissipation) of convective storms.

To achieve this goal, Met Office research has been focused on developing a frequently updated NWP data assimilation capability to improve the match to observations at scales below about 50 km, particularly for precipitation (Dixon et al. 2009; Ballard et al. 2012a,b). This is very challenging and requires the use of enhanced observing systems that sample the atmosphere at kilometer scale and at subhourly frequencies, together with advanced data assimilation techniques.

The hourly nowcast NWP cycle used for the 2012 Olympic Games was the culmination of many years' work under the NDP in developing a four-dimensional variational data assimilation (4D-Var) configuration (Rawlins et al. 2007) for convective-scale NWP and in using new observations, especially Doppler radar (Ballard et al. 2012a; Simonin et al. 2014). A full description will be submitted for separate publication, so here we focus on the key features that distinguish it from the larger-scale operational schemes. The demonstrated configuration addressed several of the key features required of a convective-scale data assimilation and prediction system. It had a fast hourly cycle, delivering a 12-h forecast in 15 min, so that products were available within one hour of the nominal observation time at the center of the assimilation window. Most of the remaining time was spent collecting the observations. The 1-h assimilation window used observations that reported many times during the window, including radar wind data every 10 min and radar-derived surface precipitation rates and satellite-derived 3D cloud analyses (Golding 1998) every 15 min. Background error covariances were calculated from differences between 6- and 3-h forecasts, and derived length scales vary from a few kilometers to a few tens of kilometers. They are much smaller than the values used in the UKV 3D-Var, which were derived from differences between 24- and 12-h forecasts.

High-resolution U.K. ensemble.

Ensemble prediction systems are well established as a means of predicting the range of possible forecasts, the confidence of specific medium-range forecasts, and of providing probabilities of exceedance of key thresholds of forecast fields (e.g., the probability of exceeding more than 25 mm of rain in a day). Prior to the Games, the Met Office had access to two established ensemble prediction systems: the ECMWF global ensemble (Molteni et al. 1996) running to 10 days ahead, twice a day, with a grid spacing of about 33 km at U.K. latitudes (a coarser-resolution product is available to 15 days); and the Met Office Global and Regional Ensemble Prediction System (MOGREPS; Bowler et al. 2008), consisting of nested global and regional ensembles running to 3 days ahead, also twice a day, with grid spacings of 60 and 18 km, respectively, at U.K. latitudes. While the ECMWF ensemble has 51 members, the Met Office ensemble contained 24 members at that time.

Early developments in ensemble prediction were focused on medium-range prediction of surface pressure and 500-hPa height, relying heavily on the predictability characteristics of low Rossby number midlatitude weather systems (Toth and Kalnay 1993; Molteni et al. 1996). Recent work, for example, at the National Oceanic and Atmospheric Administration (NOAA) Hazardous Weather Testbed (Clark et al. 2012), has begun to explore the application of ensembles to convection-permitting NWP.

For the Games, the MOGREPS system was switched to a 6-hourly cycle with an additional nested domain, MOGREPS-UK, covering the United Kingdom at 2.2-km resolution. Forecasts were made to 36 hours ahead with a control and 11 perturbed members in each cycle. MOGREPS-UK was set up as a “downscaling” ensemble so as to avoid any complications of mismatched boundary conditions. Thus, initial state, boundary conditions, and perturbations were all interpolated from the regional component of the MOGREPS (MOGREPS-R) outputs. A consequence of the lack of convective-scale data assimilation is a spinup period for convective-scale detail in the forecast. For this reason, forecasts were only presented for lead times of six hours or more.

In light of feedback indicating resistance to or lack of understanding of probability language (e.g., Morss et al. 2008), efforts were made to describe probability information in user-friendly language and extensive “frequently answered questions” were provided through a help facility on the web pages. Three main types of product were prepared: maps of probabilities of exceedance of thresholds of significance to the public, generic site-specific time series forecasts showing the expected range of outcomes, and specially designed products for the rowing and sailing venues. It was recognized that products obtained by combining outputs from 12 members would not adequately span the space of uncertainty for variables with strong gradients such as precipitation. However, if the ensemble perturbation strategy was working well, the main deviations in the large-scale evolution should be well captured. Threshold exceedance outputs were therefore postprocessed to allow for the residual spatial uncertainty (Germann and Zawadzki 2004; Roberts and Lean 2008). Based on precipitation verification results, a sampling domain of 30 × 30 km2 around each grid square was used for this purpose. In this way, maps were generated for a variety of thresholds of temperature, precipitation rate, and wind speed. Figure 3 shows two complementary approaches to communicating a maximum temperature forecast: the left panel shows the raw ensemble-mean maximum temperature at each grid point, while the right panel shows the postprocessed probability of exceeding the 30°C threshold, relevant to risk assessment for high-intensity exercise. An example of a threshold precipitation forecast during the games is shown in the performance section below (Fig. 12). Figure 4 shows two alternative approaches to communicating wind forecast information for a point location: using a wind rose or a meteogram. These were used by forecasters at the sailing events at Weymouth and the rowing at Eton Dorney. A unique product was generated for the rowing course at Eton Dorney (Fig. 5), taking account of the orientation of the course. This display shows the course in its true orientation, with the along-course and cross-course component wind speed forecasts superimposed for a range of lead times. In the example shown, a headwind in the late afternoon changes through the evening: first being indeterminate for a period, then becoming a weak tail wind. At the same time, the direction of the crosswind changes. This was of significance to competitors due to the subtle differences in shelter between the upwind and downwind lanes of the course. Use of the box-and-whisker plot enables competitors to see low probability solutions that might have high impact, as well as focusing on the range of more probable solutions in the center of the distribution.

Fig. 3.

Forecasts of ensemble-mean maximum temperature (°C) and neighborhood-processed probability of exceeding 30°C from 0300 UTC 25 Jul 2012 for the period 1000–2200 UTC on the same day.

Fig. 3.

Forecasts of ensemble-mean maximum temperature (°C) and neighborhood-processed probability of exceeding 30°C from 0300 UTC 25 Jul 2012 for the period 1000–2200 UTC on the same day.

Fig. 4.

Wind displays for specific locations. (top) Ensemble wind rose, showing probabilities of wind speed and direction categories for a specific time; (bottom) forecast time series of wind speed, displayed as median line with probability box and whiskers.

Fig. 4.

Wind displays for specific locations. (top) Ensemble wind rose, showing probabilities of wind speed and direction categories for a specific time; (bottom) forecast time series of wind speed, displayed as median line with probability box and whiskers.

Fig. 5.

Box-and-whisker plots of headwind and crosswind forecast sequences for the rowing lake, displayed to give an intuitive interpretation of the impact.

Fig. 5.

Box-and-whisker plots of headwind and crosswind forecast sequences for the rowing lake, displayed to give an intuitive interpretation of the impact.

PERFORMANCE.

Detailed papers are in preparation on performance of the ensemble and hourly update cycle, so the results shown here are preliminary and for illustration only.

Air quality forecasting system.

Air quality forecasts are routinely verified by the Met Office in near real time using observations from the U.K. Automatic and Urban Network (Stacey 2012). There were very few days of settled anticyclonic weather during the Olympic Games period, so poor air quality was not a significant issue. The exceptions were in the run up to the Olympics opening ceremony and the Paralympics closing ceremony.

Case study

In the run up to the Olympics opening ceremony, starting on around 20 July, a buildup of ozone and secondary inorganic aerosol (mostly nitrate and sulfate aerosol) occurred over western Europe, extending into the United Kingdom. By 25 July, ozone levels in southeastern England and East Anglia had risen to values sufficiently above the 160 μg m−3 8-h rolling-mean threshold to give “high” values of the DAQI. AQUM predicted the magnitude and spatial variation of these elevated concentrations well. Ozone levels continued to rise and the episode peaked on 26 July, with a large area of central and southern England recording high ozone concentrations, giving the observed DAQI pattern shown in Fig. 6.

Fig. 6.

Measured DAQI for 26 Jul at rural and urban background sites in the United Kingdom Automatic Urban and Rural Network. Note that not all sites measure ozone; some adjacent sites may appear inconsistent for this reason.

Fig. 6.

Measured DAQI for 26 Jul at rural and urban background sites in the United Kingdom Automatic Urban and Rural Network. Note that not all sites measure ozone; some adjacent sites may appear inconsistent for this reason.

The AQUM forecast generated on 25 July and valid for 26 July is shown in Fig. 7. The spatial distribution and magnitude of the forecast DAQI agreed well with the observed values, in particular capturing the “moderate” values (DAQI 4–6) in greater London and the high values in central and southern England, extending toward the southwest. On 27 July (the day of the opening ceremony) ozone levels reduced, with just a narrow strip of moderate DAQI, extending from the southern English Midlands to the Hampshire coast; values of DAQI in greater London returned to “low” values. This timely improvement in air quality for the day of the opening ceremony was predicted well by AQUM two days in advance (not shown).

Fig. 7.

Forecast DAQI on 0000 UTC 25 Jul 2012 valid for 26 Jul. (See www.metoffice.gov.uk/guide/weather/air-quality#quality-index-banding for a definition of the thresholds used for each DAQI band.)

Fig. 7.

Forecast DAQI on 0000 UTC 25 Jul 2012 valid for 26 Jul. (See www.metoffice.gov.uk/guide/weather/air-quality#quality-index-banding for a definition of the thresholds used for each DAQI band.)

Model predictions for the particulate matter component of air quality over U.K. land points (not shown) were generally good, but there were some excessively high values predicted over the sea, indicating a continuing need to refine the modeling of pollutants under stable conditions in a shallow boundary layer.

Feedback on the forecasts from the public was positive and the showcase was useful in highlighting some issues concerning the communication of air quality forecasts. In particular, the use of an “air quality index,” (as defined by DEFRA) where high values actually mean “very poor air quality,” was found confusing by some members of the public; a number of respondents pointed out that this convention is more consistent with the term “air pollution index” rather than air quality index. The issue was subsequently raised with other stakeholders and discussions regarding the adoption of a more consistent terminology are ongoing.

Very high-resolution forecasts for Weymouth Bay.

Few observations were available to verify the Weymouth Bay forecasts, so we can only provide anecdotal evidence. There was no evidence of boundary distortions corrupting the forecasts, despite the small domain, and feedback from forecasters on the performance of the Weymouth Bay models was very positive. The model wind forecasts provided critical input to decision making on two days of light winds, one during the Olympic Games and one day during the Paralympic Games. On both occasions, the model correctly identified a period in the middle of the day when winds would fall calm at locations close to or within the sailing courses. Both LOCOG and the competitors were highly complimentary about the accuracy of the forecasting for the sailing events, and the forecasters have attributed much of this to their use of products from the UK300 model. However, it was also noted that the wind forecast was sensitive to any errors in the cloud forecast over land, due to its effects on sea breeze development. Implementation of such high-resolution models is not currently a priority due to their high computer cost and limited geographical coverage. However, the positive experience during the games highlights their potential benefits, especially in the prediction of topographically controlled weather in light wind conditions.

Case study

On 9 August, U.K. weather was dominated by a building anticyclone centered in the Irish Sea with the remnants of a weak front across southern England during the morning (not shown). As a result of low winds, racing was postponed until the following day. Using global MetUM output, forecasters identified early in the week that light winds would be a problem for the sailing events on this day. For preparation of the morning briefing, the forecaster had available the operational 2100 UTC UKV model and the 0000 UTC UK300 model. The 0000 UTC UK300 (Fig. 8b) indicated light southeasterly winds across Weymouth Bay with the potential for southwesterly sea breezes to develop along the coast. The weakness of the sea breeze appears to have been related to suppressed solar heating due to cloud along the weak front just inland from the coast (not shown), which was well represented in this run. Observations from the buoy in the bay, from Portland Bill and the harbor wall wind systems (Fig. 8a) indicated that the southeasterly surface wind became south-southeasterly by early afternoon, with speed increasing to around 6 knots (kt; 1 kt = 0.51 m s−1) between 1200 and 1300 UTC, but this was not sustained—the forecast advisor informed the organizers that this was unlikely to persist for any length of time. Meanwhile, a southwesterly sea breeze was observed at the Sailing Academy on Portland from late morning. This was seen to make some progress across the harbor, but did not reach the harbor wall until nearly 1600 UTC, when it was weak. One of the forecast advisors went to the Nothe and observed the weak southwesterly sea breeze arrive there around 1500 UTC, but it was very weak and not sustained.

Fig. 8.

(a) Surface observations and (b) 14-h model forecast of wind speed (shaded, kt) and direction (for speeds > 2 kt) for 1400 UTC 9 Aug 2012.

Fig. 8.

(a) Surface observations and (b) 14-h model forecast of wind speed (shaded, kt) and direction (for speeds > 2 kt) for 1400 UTC 9 Aug 2012.

Hourly NWP forecast updates.

This showcase demonstrated a major step toward achieving more accurate nowcast performance for precipitation from a NWP system and was of particular interest to external scientists. Inevitably the T + 0, T + 1 forecasts could not capture the level of detail present in the STEPS system (which used the radar-derived rain rates themselves as its analysis). However, a key user requirement for nowcasts is for a reliable monotonic trend toward the observed convective-scale precipitation as the forecast lead time reduces, and this was often seen. Detailed study of the results (S. P. Ballard et al. 2013, personal communication) has shown that the system was able to provide an improved forecast out to six hours compared to the STEPS system and latest available UKV forecast on many occasions. Sufficient computer power to make this system operational for the whole United Kingdom will not be available until 2015, and the results of the showcase are being used to direct the goals of research aiming at an enhanced capability on that time scale.

Case study

Several Olympic events, including the women's marathon were affected by intense convective storms across England and Wales on 5 August. The rainfall distribution is illustrated by the radar-derived hourly accumulation up to 1100 UTC in Fig. 9 [for an introduction to the capabilities of the U.K. radar network, see Kitchen and Illingworth (2011)], showing intense rain cells in southwestern England and southern Wales (where there was flash flooding) and a cluster of cells near and to the northwest of London. NDP forecasts for 1100 UTC, shown in Fig. 10, gave a good indication of the main locations of heavy rain, and its skill at capturing local detail improved as forecast range decreased. These comparisons indicate that the hourly assimilation process successfully adjusted the model evolution toward the true solution, with no evidence of noise or spinup/down problems. However, when compared with radar extrapolation (not shown), the deviation from reality, as defined by radar-derived rain rates, at T + 0–1 remains a disadvantage. Figure 10 also illustrates the tendency, common to all convective-scale configurations of the MetUM, for the NDP to miss areas of light precipitation and to produce localized over-intense precipitation rates (as in northern Wales; cf. Figs. 9 and 10). This probably indicates the need for better parameterization of turbulent mixing between clouds and the environment in convection-permitting models (Lean et al. 2008).

Fig. 9.

Radar-derived rain accumulation (mm) for 1000–1100 UTC 5 Aug 2012.

Fig. 9.

Radar-derived rain accumulation (mm) for 1000–1100 UTC 5 Aug 2012.

Fig. 10.

NDP forecasts of rain accumulation (mm) for 1000–1100 UTC 5 Aug 2012. (left to right) Forecasts from the 1000, 0900, and 0800 UTC cycles.

Fig. 10.

NDP forecasts of rain accumulation (mm) for 1000–1100 UTC 5 Aug 2012. (left to right) Forecasts from the 1000, 0900, and 0800 UTC cycles.

High-resolution U.K. ensemble.

Feedback on this showcase was very positive, both from international colleagues and from the Games' forecasters. Particular successes were the performance of the ensemble in some flood-generating precipitation events in June, prior to the Games, and the rowing products during the Games themselves. As with large-scale ensembles, the members of an individual ensemble run showed a tendency to cluster around the control run, leading to jumps between subsequent ensemble outputs. For this reason and to increase the sampling, processing has since been extended to combine two successive ensemble runs, making 24 members, while retaining the 30-km spatial reprocessing. This improves the ensemble spread and reduces run-to-run jumpiness, but at the cost of introducing older forecasts. Future work will also address inclusion of a convective-scale initial state and initial perturbations and model perturbations. MOGREPS-UK has subsequently become fully operational during 2013.

Case study

The day 29 July was the first Sunday of the Olympic Games and was the day of the women's cycling race. Global 3-day forecasts had correctly raised the expectation of showers affecting England on this day, but the 24-h forecast ensemble provided an opportunity for identifying which counties would be affected and when. A vigorous depression was centered over northeastern Scotland (not shown) with showers covering the whole of the United Kingdom. From 1100 UTC, heavy showers developed over the southern half of England, as a trough moved eastward. Some of these showers formed into bands that moved downwind. The heaviest showers developed over Hampshire and propagated northeast through London and into East Anglia where they reached their maximum intensity during 1300–1500 UTC. Figure 11 shows the radar accumulation for the period when the storms passed through London, with 8–16-mm hourly accumulations in a band southwest of London and 16–32-mm accumulations to the east of London. Figure 12 shows forecasts of “intense rain during a period,” which strictly represents the peak model rain rate. However, for the purposes of this comparison, peak rates and peak hourly accumulations are sufficiently similar. The 1500 UTC forecast from the previous afternoon (Fig. 12a) picked out the southern Hampshire-to-Essex track as the highest risk with a core probability of 60% in southern Hampshire and the London area, consistent with the observations. It also had areas of 40% probability over northern Norfolk, where intense storms occurred later and in northwestern England where no intense storms occurred. Looking at the detail in the forecast from 0300 UTC, we find that there is a timing error of about 2 h. Allowing for that difference, Fig. 12b shows a very good match to reality, though with much lower probabilities of 40%–60% over south Hampshire and 20%–40% over London. One of the challenges of developing operational products from this ensemble will be to find a way of communicating the appropriate level of confidence for these hourly outputs with maximum probabilities that rarely exceed 40%.

Fig. 11.

Radar-derived rain accumulation (mm) for 1100–1200 UTC 29 Jul 2012.

Fig. 11.

Radar-derived rain accumulation (mm) for 1100–1200 UTC 29 Jul 2012.

Fig. 12.

(left) Probability of rain rate exceeding 16 mm h−1 (described as “intense rain”) sometime during the period 0600–2400 UTC 29 July 2012 from a forecast initiated at 1500 UTC on the previous day. (right) Probability of intense rain sometime during the period 1300–1400 UTC from a forecast initiated at 0300 UTC on the same day.

Fig. 12.

(left) Probability of rain rate exceeding 16 mm h−1 (described as “intense rain”) sometime during the period 0600–2400 UTC 29 July 2012 from a forecast initiated at 1500 UTC on the previous day. (right) Probability of intense rain sometime during the period 1300–1400 UTC from a forecast initiated at 0300 UTC on the same day.

OUTCOMES.

The decision to focus on showcasing the latest research through the Met Office's Olympic web portal was vindicated: the work was affordable, it raised considerable public and international interest in the research and its operational application, and it enabled a much more flexible approach to the work of providing detailed forecasts to the Games organizers and visitors. There were 30,000 visits to the Olympic Showcase website, a much larger number than would normally be the case for Met Office research pages. The layered approach appears to have worked well. Large numbers of visitors to the explanatory pages obtained information on the scope and excellence of Met Office science, while a much smaller number were able to judge for themselves the value of specific forecast products. The separation of operational from experimental products was achieved successfully.

The new products were referred to by forecasters as additional guidance on those occasions when difficult decisions needed to be taken. The Olympics were characterized by typical changeable British summer weather. Key decisions related to heavy rain and thunderstorms on the two Sundays, which affected the road cycling in Surrey and the tennis at Wimbledon. The winds, light, strong, and fluctuating in direction, provided challenges for the forecasting teams at Eton Dorney and Weymouth. However, they correctly forecast days on which sailing was unlikely at Weymouth due to the lack of wind, aided by the Weymouth Bay model, and the days when the wind direction and/or speed would be a problem for the rowing at Eton Dorney and more generally when strong winds would cause problems for the infrastructure at other venues. The Paralympics was a more settled period, but considerations of air quality and high temperatures were an issue for the marathons on the final Sunday. Feedback indicates that the additional information was useful on many of these occasions.

Subsequent to the Games, both AQUM and MOGREPS-UK have become established components of the operational production suite. Forecasts of the daily air quality index remain part of the regular site-specific forecasts on the Met Office website. Aspects of the NDP data assimilation scheme have been incorporated in the UKV 3D-Var, but full implementation awaits the next computer upgrade.

Delivery of the showcase depended on developments to the website, Doppler upgrades to the network radars, and installation of the IBM P7 supercomputer upgrade. Technical issues caused delays to each of these, complicating the planning. At the start, achievement of the high level of commitment required from a diverse range of scientists and technical experts was difficult. Research could not be prioritized over operations or commercial outputs, and showcasing was difficult to prioritize over research. These issues evaporated as the Games approached and enthusiasm grew. During the leadup and during the games themselves, a very high level of commitment was obtained.

ACKNOWLEDGMENTS

Delivery of the Olympic Showcase involved a large number of people across many parts of the Met Office and in itself was just a small part of the overall effort devoted to the games. We would like to acknowledge the contributions of all of these people, without whom the Olympic Showcase could not have been successful. We would also like to acknowledge the detailed and helpful comments provided by the reviewers.

REFERENCES

REFERENCES
Ballard
,
S. P.
,
Z.
Li
,
D.
Simonin
,
H.
Buttery
,
C.
Charlton-Perez
,
N.
Gaussiat
, and
L.
Hawkness-Smith
,
2012a
:
Use of radar data in NWP-based nowcasting in the Met Office
.
Weather Radar and Hydrology
,
R. J.
Moore
,
S. J.
Cole
, and
A. J.
Illingworth
,
Eds.
,
IAHS Publ.
351
,
336
341
.
Ballard
,
S. P.
,
and Coauthors
,
2012b
:
Convective scale data assimilation and nowcasting
.
Proc. ECMWF Seminar on Data Assimilation for Atmosphere and Ocean
,
Reading, United Kingdom
,
ECMWF
,
265
300
.
Booij
,
N.
,
R. C.
Ris
, and
L. H.
Holthuijsen
,
1999
:
A third-generation wave model for coastal regions: 1. Model description and validation
.
J. Geophys. Res.
,
104
(
C4
),
7649
7666
.
Bowler
,
N. E.
,
C. E.
Pierce
, and
A. W.
Seed
,
2006
:
STEPS: A probabilistic precipitation forecasting scheme which merges an extrapolation nowcast with downscaled NWP
.
Quart. J. Roy. Meteor. Soc.
,
132
,
2127
2155
.
Bowler
,
N. E.
,
A.
Arribas
,
K.
Mylne
,
K. B.
Robertson
, and
S. E.
Beare
,
2008
:
The MOGREPS short-range ensemble prediction system
.
Quart. J. Roy. Meteor. Soc.
,
134
,
703
722
.
Clark
,
A. J.
,
and Coauthors
,
2012
:
An overview of the 2010 Hazardous Weather Testbed Experimental Forecast Program Spring Experiment
.
Bull. Amer. Meteor. Soc.
,
93
,
55
74
.
Davies
,
T.
,
M. J. P.
Cullen
,
A. J.
Malcolm
,
M. H.
Mawson
,
A.
Staniforth
,
A. A.
White
, and
N.
Wood
,
2005
:
A new dynamical core for the Met Office's global and regional modelling of the atmosphere
.
Quart. J. Roy. Meteor. Soc.
,
131
,
1759
1782
.
DEFRA
,
cited 2013
:
UK and EU air quality policy context
.
Dixon
,
M.
,
Z. H.
Li
,
H.
Lean
,
N.
Roberts
, and
S.
Ballard
,
2009
:
Impact of data assimilation on forecasting convection over the United Kingdom using a high-resolution version of the Met Office Unified Model
.
Mon. Wea. Rev.
,
137
,
1562
1584
.
Draper
,
L.
, and
T. M.
Bownass
,
1983
:
Wave devastation behind Chesil Beach
.
Weather
,
38
,
346
352
.
Duan
,
Y.
,
and Coauthors
,
2012
:
An overview of the Beijing 2008 Olympics Research and Development Project (B08RDP)
.
Bull. Amer. Meteor. Soc.
,
93
,
381
403
.
Flather
,
R. A.
,
and Coauthors
,
1998
:
Status of UK operational storm surge forecasting for the NW European shelf
.
Ann. Geophys.
,
16
(
Suppl. II
),
C779
.
Flemming
,
J.
,
A.
Inness
,
H.
Flentje
,
V.
Huijnen
,
P.
Moinat
,
M. G.
Schultz
, and
O.
Stein
,
2009
:
Coupling global chemistry transport models to ECMWF's integrated forecast system
.
Geosci. Model Dev.
,
2
,
253
265
,
doi:10.5194/gmd-2-253-2009
.
Germann
,
U.
, and
I.
Zawadzki
,
2004
:
Scale dependence of the predictability of precipitation from continental radar images. Part II: Probability forecasts
.
J. Appl. Meteor.
,
43
,
74
89
.
Golding
,
B. W.
,
1998
:
Nimrod: A system for generating automated very short range forecasts
.
Meteor. Appl.
,
5
,
1
16
.
Joe
,
P.
,
and Coauthors
,
2010
:
Weather services, science advances and the Vancouver 2010 Olympic and Paralympic winter games
.
Bull. Amer. Meteor. Soc.
,
91
,
31
36
.
Jones
,
A.
,
D.
Thomson
,
M.
Hort
, and
B.
Devenish
,
2007
:
The UK Met Office's next-generation atmospheric dispersion model, NAME III
.
Air Pollution Modelling and Its Application XVII
,
C.
Borrego
and
A.-L.
Norman
,
Eds
.,
Springer
,
580
589
.
Jones
,
C. D.
, and
B.
Macpherson
,
1997
:
A latent heat nudging scheme for the assimilation of precipitation data into an operational mesoscale model
.
Meteor. Appl.
,
4
,
269
277
.
Keenan
,
T.
,
and Coauthors
,
2003
:
The Sydney 2000 World Weather Research Programme Forecast Demonstration Project: Overview and current status
.
Bull. Amer. Meteor. Soc.
,
84
,
1041
1054
.
Kitchen
,
M.
, and
A.
Illingworth
,
2011
:
From observations to forecasts—Part 13: The UK weather radar network—Past, present and future
.
Weather
,
11
,
291
297
.
Lean
,
H. W.
,
P. A.
Clark
,
M.
Dixon
,
N. M.
Roberts
,
A.
Fitch
,
R.
Forbes
, and
C.
Halliwell
,
2008
:
Characteristics of high-resolution versions of the Met Office Unified Model for forecasting convection over the United Kingdom
.
Mon. Wea. Rev.
,
136
,
3408
3424
.
Lorenc
,
A. C.
,
and Coauthors
,
2000
:
The Met. Office global three-dimensional variational data assimilation scheme
.
Quart. J. Roy. Meteor. Soc.
,
126
,
2991
3012
.
Met Office
,
cited 2012
:
Enhanced weather capability for the Paralympic games
.
Met Office
,
cited 2013a
:
Public weather service customer group
.
Met Office
,
cited 2013b
:
Farnham weather forecast
.
Met Office
,
cited 2013c
:
Weather services
.
Met Office
,
cited 2013d
:
National severe weather warnings—United Kingdom
.
Met Office
,
cited 2013e
:
Hazard and risk management
.
Met Office
,
cited 2013f
:
Public weather service advisors
.
Met Office
,
cited 2013g
:
Natural Hazards Partnership
.
[Available online at www.metoffice.gov.uk/nhp.]
Molteni
,
F.
,
R.
Buizza
,
T. N.
Palmer
, and
T.
Petroliagis
,
1996
:
The new ECMWF ensemble prediction system: Methodology and validation
.
Quart. J. Roy. Meteor. Soc.
,
122
,
73
119
.
Morgenstern
,
O.
,
P.
Braesicke
,
F. M.
O'Connor
,
A. C.
Bushell
,
C. E.
Johnson
,
S. M.
Osprey
, and
J. A.
Pyle
,
2009
:
Evaluation of the new UKCA climate-composition model—Part 1: The stratosphere
.
Geosci. Model Dev.
,
2
,
43
57
.
Morss
,
R. E.
,
J. L.
Demuth
, and
J. K.
Lazo
,
2008
:
Communicating uncertainty in weather forecasts: A survey of the U.S. public
.
Wea. Forecasting
,
23
,
974
991
.
Rawlins
,
F. R.
,
S. P.
Ballard
,
K. R.
Bovis
,
A. M.
Clayton
,
D.
Li
,
G. W.
Inverarity
,
A. C.
Lorenc
, and
T. J.
Payne
,
2007
:
The Met Office global 4-dimensional data assimilation system
.
Quart. J. Roy. Meteor. Soc.
,
133
,
347
362
.
Renshaw
,
R.
, and
P. N.
Francis
,
2011
:
Variational assimilation of cloud fraction in the operational Met Office Unified Model
.
Quart. J. Roy. Meteor. Soc.
,
137
,
1963
1974
.
Roberts
,
N. M.
, and
H. W.
Lean
,
2008
:
Scale-selective verification of rainfall accumulations from high-resolution forecasts of convective events
.
Mon. Wea. Rev.
,
136
,
78
97
.
Rothfusz
,
L. P.
,
M. R.
McLaughlin
, and
S. K.
Rinard
,
1998
:
An overview of NWS weather support for the XXVI Olympiad
.
Bull. Amer. Meteor. Soc.
,
79
,
845
860
.
Savage
,
N. H.
,
P.
Agnew
,
L. S.
Davis
,
C.
Ordóñez
,
R.
Thorpe
,
C. E.
Johnson
,
F. M.
O'Connor
, and
M.
Dalvi
,
2013
:
Air quality modelling using the Met Office Unified Model (AQUM OS24-26): Model description and initial evaluation
.
Geosci. Model Dev.
,
6
,
353
372
,
doi:10.5194/gmd-6-353-2013
.
Simonin
,
D.
,
S. P.
Ballard
, and
Z.
Li
,
2014
:
Doppler radar radial wind assimilation using an hourly cycling 3D-Var with a 1.5 km resolution version of the Met Office Unified Model for nowcasting
.
Quart. J. Roy. Meteor. Soc.
,
in press, doi:10.1002/qj.2298
.
Stacey
,
S. E. B.
,
2012
:
QA/QC data ratification report for the automatic urban and rural network, October–December 2011, and annual report 2011
.
Tech Rep. AEAT/ENV3284
,
31
Tang
,
Y.
,
H. W.
Lean
, and
J.
Bornemann
,
2012
:
The benefits of the Met Office variable resolution NWP model for forecasting convection
.
Meteor. Appl.
,
20
,
417
426
,
doi:10.1002/met.1300
.
Tolman
,
H. L.
,
2009
:
User manual and system documentation of WAVEWATCH III version 3.14
.
NOAA/NWS/NCEP/MMAB Tech. Note 276
,
194
pp
.
Toth
,
Z.
, and
E.
Kalnay
,
1993
:
Ensemble forecast at NMC: Practical aspects
.
Wea. Forecasting
,
8
,
378
398
.
Wilson
,
J.
,
Y.
Feng
,
M.
Chen
, and
R. D.
Roberts
,
2010
:
Nowcasting challenges during the Beijing Olympics: Successes, failures, and implications for future nowcasting systems
.
Wea. Forecasting
,
25
,
1691
1714
.