Skip booklet index and go to page content

The Toronto 2015 Pan and Parapan American Games Experience

8. Forecast Services and Prediction

This section describes the forecasting and modelling at the heart of the operation of ECCC’s services in support of the Games. The modelling and prediction is a culmination of all of the foundational work that took place with the installation of the Mesonet and its associated technological infrastructure. With data flowing from instruments and equipment, the most significant piece then became prediction of daily weather conditions for each of the venues during the Games. The role and setup of the OSPC are described along with the products and services provided from within. The intricacies of various modelling tools are examined in sections 8.5.1 to 8.5.3.

8.1 The Role of the Ontario Storm Prediction Centre in Support of the Games

The Ontario Storm Prediction Centre (OSPC) is an ECCC weather forecast office based in Toronto that produces forecast products for all of Ontario, the National Capital Region, the Great Lakes and the Ontario portion of the St. Lawrence River. It is a 24/7 operation, staffed by a team of about 30 meteorologists.

In everyday operations, the OSPC produces weather forecasts (regularly issued forecasts and special alerts, including weather watches, warnings and advisories) for the general public and marine communities in Ontario. It also provides weather information and support to a variety of governmental agencies and emergency management organisations.

The OSPC provided specialized weather forecast products in support of the Games, starting with the Pan Am torch relay run prior to the Games and ending with the closing ceremonies of the Parapan Am Games. A new point forecasting methodology was used to produce venue- (or venue cluster) specific alerts and forecasts. These products and a daily Thunderstorm Outlook graphic product were provided by the OSPC starting on July 6 and continuing throughout the Games. Additionally, route-specific forecasts were produced during torch relays that preceded both the Pan and Parapan American Games.

The objective with ECCC’s point forecasting methodology was to use the Mesonet in the Games footprint to provide venue-specific forecasts and alerts to our partners that could be used in decision making for Games operations. For example, if showers or thunderstorms were entering the region from northwest to southeast, the objective was to offer as much notification as possible of the impending precipitation and storm, and to identify which venues, if any, could be affected or if the storm would miss the location. This was especially important for outdoor venues. Organizers would then have the necessary weather information for Games operations and scheduling decisions, which would ultimately benefit the athletes, organizers’ schedules and live media coverage of the events.

To support the program, two forecast production desks were created in the OSPC that were dedicated to Games forecasting, requiring eight full-time meteorologists. These forecasters were drawn from within the OSPC, the PASPC in Winnipeg, Manitoba, the National Lab for Nowcasting and Remote Sensing, and from the Workstation Development and Innovation team. The forecast team relied on the expanded Mesonet of weather observations, high-resolution weather forecast model data produced especially for the Games, enhancements made to the regular forecast production software tools (i.e., “NinJo” and “SCRIBE”) in order to produce the venue alerts and forecasts, and a specially configured computer workstation, ISW.

In order to disseminate the forecast products to the end users, a number of platforms were created (including the Ocean Networks Weather Portal [see Section 9.5], EC Alert Me [see Section 9.6] and the Web-based “Datamart”). However, the meteorologists relied on the ECCC weather briefers at the MOC and the UCC to disseminate and provide interpretation of their forecast products to the most critical users (see sections 9.1 and 9.4).

Below are further details on OSPC operations, including the high-resolution weather models and tools used to create the specialized forecasts produced for the Games.

8.2 Ontario Storm Prediction Centre Operations for the 2015 Games

The OSPC was expanded to include an additional four desks (the “quad”), two of which were Research Support Desks (RSDs), staffed by ECCC’s research meteorologists (see Section 11.9 for further detail). The other two desks were staffed by meteorologists dedicated to producing the Games-specific forecast products, and providing weather information and support for Games-related activities.

The “quad” desks were physically arranged so that they facilitated easy communication and information sharing between the Games forecasters and the research meteorologists at the RSDs. Furthermore, the desks were adjacent to the main OSPC forecast desks and thus allowed for quick collaboration, coordination and consultation with those meteorologists.

Each of the operational desks was equipped with an ISW that allowed the meteorologists to access the Mesonet’s observation data, output from the specialized high-resolution weather and air quality prediction models, and the specially configured Games forecast production software. Additionally, a large monitoring screen was mounted above the Games quad, to provide situational awareness information to the forecasters.

Long-standing working relationships with the United States National Weather Service (NWS) were renewed and strengthened as three staff rotated visits to the OSPC during the Games. This was for the purposes of observation and information sharing, which greatly benefited both organizations.

8.3 Venue-specific Forecasting and Alerts

Venue-specific forecasts were issued three times per day (amended as conditions warranted) and were valid for seven days. The forecasts included sky condition, precipitation, temperature, humidity, wind speed and direction, and obstructions to visibility. The Games marine forecast for the sailing venue in Toronto Harbour was issued three times per day during the Pan Am Games (no marine forecasts were required during the Parapan Am Games). Forecasts of wind speed and direction, wave height, obstructions to visibility and precipitation were valid for five days. Neither the venue-specific public forecasts nor the marine forecast used sport-specific alerting criteria. Alerts were issued as required and were based on existing and approved ECCC criteria. Alert types included: fog advisories, heat warnings, rainfall warnings, severe thunderstorm watches and warnings, smog and air health advisories, tornado watches and warnings, generic weather advisory warnings, wind warnings, and special weather statements.

In terms of alert types, alert criteria and target lead times, the program mirrored that of ECCC’s regular summer season public alerting program for southern Ontario. The differences, as with the regular forecast products, were that the Games alerts were issued for specific venue locations rather than larger regional areas, and the electronic distribution of the alert products was limited to specific decision-makers and emergency managers/responders associated directly with the Games (e.g., TO2015, sports organizations, UCC).

8.4 Torch Relay Forecasts

The Pan Am Torch Relay commenced May 30, 2015, in Toronto, travelled across the country and concluded in Toronto on July 10 at the Pan Am opening ceremonies. A five-day Parapan Torch Relay started on August 3 in Ottawa and Niagara Falls, Ontario (two torches), crossed Ontario and ended in Toronto with a joint lighting of the cauldron during the Parapan opening ceremony on August 7. The Pan Am torch run required forecast and alerting support from multiple forecasts offices in the country, while the OSPC was the sole forecast source for the Parapan Torch Relay. For both relays, a three-day forecast was provided to the TO2015 Torch Relay team for locations along the torch route. In addition to a text forecast of sky condition, precipitation, maximum temperature, wind speed and direction, and obstruction to visibility, the forecast also highlighted any significant risks and potential impacts that the weather might have on the relays and associated ceremonies. Torch relay officials also had access to telephone consultations with the OSPC meteorologists as required.

Figure 15.

Photo of Environment and Climate Change Canada forecasters at the Ontario Storm Prediction Centre (three standing, three working at computer workstations).
Photo: © Hong Lin

8.5 Weather, Wave and Air Quality Prediction Modelling

The next subsections describe the three types of experimental prediction models that were used extensively during the Games. The first describes the high-resolution urban-scale weather prediction model, which led to a clearer visualization of meteorological features in and around the City of Toronto. The second describes the high-resolution wave modelling that took place for Lake Ontario, which produced higher wave model resolution and more accurate representation of significant wave information, especially in near-shore areas. Finally, the last subsection describes the high-resolution air quality model. This model not only provided predictions for future air quality, but also enabled a visualization of air pollution, to better understand urban pollutant sources that impact on the health of surrounding communities.

8.5.1 High-resolution Urban Scale Weather Prediction Model

Numerical weather prediction (NWP) is an important tool for operational weather forecasting. The NWP systems currently used operationally at ECCC have grid spacing on the order of 10 to 50 km. Higher horizontal resolution (i.e., smaller grid spacing) is required for improved representation of critical physical processes occurring at the surface and in different levels of the atmosphere. This leads to a better resolution of atmospheric circulations and meteorological features such as lake breezes and urban heat islands that are directly linked with high-resolution surface features like cities. Improved detection of lake breezes can in turn improve the prediction of air quality, thunderstorms, lightning and other severe weather events such as tornadoes. Similarly, higher-resolution model predictions help to better define the urban heat island effect where city temperatures are often warmer than in surrounding rural areas. Complex cloud microphysics is also found to perform better in higher-resolution atmospheric models, with potential for improved forecast of severe precipitation events.

The Games footprint was over an area where lake breezes from the lower Great Lakes occur frequently during the summer, and the majority of the Games venues were in an urban environment. Experimental versions of ECCC’s operational Global Environmental Multiscale (GEM) atmospheric model were therefore implemented as guidance for our forecasters and briefing teams during the Games. The high-resolution models were used in addition to the standard operational forecast models to support the venue-specific weather forecast and alerting program. Another primary objective in running the experimental models was to evaluate the quality of the numerical forecasts produced by finer spatial-scale versions of the operational GEM and to prepare for future operational implementations of these types of systems.

A 2.5-km GEM model, currently run 4 times per day for 48-hour forecasts at CMC, was the first experimental tool made available to forecasters during the Games. Two other versions of GEM, more experimental and run specifically for the Games, were integrated once per day for 24 hours with grid spacing of 1 km and 250 m. All these model versions feature state-of-the-art configurations of numerical, dynamical and physical processes, including complex representation of clouds, precipitation and atmospheric radiation. They are also able to directly represent the impact of urban surfaces on the atmosphere and have a better representation of land surface conditions (e.g., surface temperature, soil moisture) and water temperature over Lake Ontario. Outputs from the 1-km and 250-m experimental NWP systems were designed to enhance weather forecasts during the Games and to support applications related to weather and health services (see Section 12). In a sample 250-m Urban Scale model output graphic shown below, the Universal Thermal Climate Index (UTCI) is predicted for early afternoon on July 28, 2015. When the UTCI is above 38°C, there is very strong heat stress on the human body. Strong heat stress would be experienced with the UTCI in the range of 32°C to 38°C, and moderate heat stress in the range of 26°C to 32°C. The figure highlights that the strongest heat discomfort is over the most urbanized regions of Toronto, with the greatest heat discomfort located a few kilometres north, due to the lake breeze’s northward advection of the warmest air.

Information was provided on standard meteorological variables such as temperature, winds and precipitation, but heat stress indices were also made available for health applications (e.g., Humidex, UTCI and Wet Bulb Globe Temperature [WBGT]). Most of these products were made available digitally through the CMC Datamart or visually through an experimental Web mapping service (map graphics).

Results obtained from the experimental GEM run during the Games were found to compare favourably against the GEM and other forecast model products that are currently operational at ECCC. Subjective evaluation of several case studies of strong precipitation and winds, intense heat, and lake breezes that occurred during the Games reveals a clear advantage of using high-resolution (km-scale and sub-km-scale) grid spacing. Preliminary results from a more objective approach based on the entire set of model integrations for summer 2015 already indicate a clear benefit (in a statistical manner) of using the experimental models compared with current CMC operational models.

In a continuous effort to make guidance from numerical atmospheric models more accurate as well as more precise spatially, and to expand its use to a wider range of applications, the implementation of the Pan and Parapan American high-resolution GEM configurations could be considered as the basis for the next generation of CMC’s short-range local forecasts. Further evaluation of the high-resolution numerical weather prediction system over the southern Ontario Games region and period will help identify and correct model weaknesses and deficiencies. Deployment of similar systems for other Canadian regions and cities is to be expected in the next two or three years.

Figure 16. Model predicted fields from the Urban GEM-LAM model with 250-m grid spacing

Map of the Greater Toronto Area (see long description below)

Description

Map of the Greater Toronto Area showing model predicted fields from the Urban GEM-LAM model with 250-m grid spacing, for July 28, 2015, at 18:00 UTC (1:00 p.m. local time). Wind vectors at 10 m above canopy level (knots) at the model grid points are superimposed with shaded areas of the Universal Thermal Climate Index at 2 m above ground level (°C). The UTCI scale is in 1°C intervals from 21°C to 40°C, with wind barb categories 0.1, 3.5, 7.0, 10.5 and 14.0 knots. The highest UTCI values are over land (in particular over central Toronto with maximum values of 40°C), and the lowest values are over Lake Ontario (in particular near-shore east of Toronto with minimum values of 21°C).

8.5.2 Lake Ontario Wave Modelling

During the Games, a number of open-water events were hosted in the inner harbour and south of the Toronto Islands, an area where the competition or logistics could be affected by high winds, fog, waves, severe weather or heat stress. Given these sensitivities, ECCC provided near-shore marine forecasts to inform the TO2015 Games competition management, athletes, volunteers and spectators about winds, waves and weather for the day. Wave models provide valuable guidance in the development of these forecasts. The Games gave us a unique opportunity not only to use our current wave modelling system but also to test and validate the next generation of wave forecasting systems currently under development at ECCC.

The current Great Lakes wave prediction system is a single deterministic model at 5-km resolution with 3-hourly output driven by the winds of the 10-km GEM model. Each lake is covered by its own grid, excluding Lake Michigan. An experimental high-resolution model that was tested for the Games was instructive for accelerating and informing upgrades to the Great Lakes wave modelling system. As part of the Mesonet built-in support of the Games, ECCC deployed two additional wave buoys in Lake Ontario just south of the Toronto Islands (see sections 5.4 and 5.5). Data from the buoys was used in the validation of the wave models.

Three wave modelling systems were used during the Games with horizontal grid resolution of 2.5 km, 1 km and 250 m. The deterministic wave prediction system at 250-m resolution over western Lake Ontario was deployed specifically in support of the Games. The system produced 24-hour forecasts using input from the 250-m resolution urban scale atmospheric model, described in Section 8.5.1.

The higher wave model resolution improves parameterizations, and more accurate representation of coast or shoreline leads to a better representation of significant wave information, especially in near-shore areas. In particular, the 250-m resolution grid over western Lake Ontario can begin to resolve small areas such as Toronto's Inner and Outer Harbours and realistically depict “obstacles,” for example, the Port Lands and Toronto Islands, which provide a shadowing effect to waves in Humber Bay when winds are from the east. The model also allows for a detailed depiction of waves driven by features such as lake breezes and thunderstorms. An example of the effect of a fast-moving line of thunderstorms on the wave field is shown in Figure 17.

Analysis of the model performance during the Games has shown that the new wave forecasting systems performed better than the system currently in place. In particular, the 250-m resolution wave model performed the best of the three models. The OSPC forecasters also provided a positive evaluation of the new wave modelling system.

Figure 17. Simulated fields for the 250-m grid resolution WaveWatch III model

Map of western Lake Ontario (see long description below)

Description

Map of western Lake Ontario showing simulated fields for the 250-m grid resolution WaveWatch III model forced by the winds of the Urban GEM-LAM model at the same resolution on August 3, 2015, at 05:00 UTC (1:00 a.m. local time). Areas of significant wave height (m) are shaded at 0.1 m intervals from 0.0 to 1.0 m. Wind barbs at 10 m above surface at the model grid points (each full barb: 5 m/s) are depicted. Maximum wave heights are up to 0.8 m in small areas of western and southern Lake Ontario, with maximum wind speeds in these same areas of 30 knots from the west.

8.5.3 High-resolution Air Quality Modelling for the Games Demonstration

A regional air quality model integrates our understanding of air quality by simulating pollutant emissions, chemical transformation, meteorology and deposition processes. Models not only provide predictions of future air quality, but also enable a visualization of air pollution, thus helping to interpret how urban pollutant sources are impacting the health of surrounding communities.

The Games provided an opportunity for ECCC to showcase the development and application of its next-generation air quality model, GEM-MACH, now called version 2. This also provided a unique opportunity to collaborate with other research groups within ECCC.

The Global Environmental Multiscale – Model of Atmospheric Chemistry (GEM-MACH) is the tool that combines the current state of the atmosphere and our understanding of air quality and weather to generate air quality forecasts. Forecasters use this model guidance to help them make predictions of the Air Quality Health Index (AQHI) for communities across Canada. Of great interest to CMC and the Storm Prediction Centres is the development of higher-resolution forecasts and how the predictive skill compares with the current operational model versions. The primary objective of this real-time demonstration project was to compare the performance of the 10-km grid-spaced operational GEM-MACH model (version 1.5.1) with the next-generation GEM-MACH version 2 model run at a finer grid spacing of 2.5 km. An improved model performance of GEM-MACH version 2 would provide justification for CMC to consider future operational use of a high-resolution air quality model in highly populated urbanized areas of Canada, such as southern Ontario.

There were three main GEM-MACH version 2 development steps that were required before the Games. First, model scripts were developed with a new, more user-friendly, graphical environment to launch the model. Second, the chemistry routines from the current operational GEM-MACH needed to be ported to the high-resolution GEM-MACH, which is now based on the numerical weather prediction model GEM version 4. Prior GEM-MACH versions, including the current operational GEM-MACH, are based on the operational weather forecast GEM version 3. GEM version 4 has several advantages over GEM version 3, such as more physically based cloud physics schemes. The third development step was to create high-resolution pollutant emissions for the GTA. The spatial mapping of on-road traffic emissions, as input into the gridded air quality model, was improved by allocating emissions using the most recent Canadian road network.

The high-resolution version of the next-generation GEM-MACH model was run in real-time forecast mode from June 1 to September 20, 2015, providing 24-hour AQHI forecasts for the Games sporting venue locations. These experimental forecasts were produced once per day, in comparison with the twice per day issuance of operational model forecasts. The forecast products were placed on several data portals for dissemination. Daily forecast briefings were given by ECCC’s air quality research scientists to provide guidance in selecting a daily measurement plan for the CRUISER air quality mobile laboratory (see Section 11.4). The OSPC forecasters had access to all of the operational and experimental high-resolution air quality model products for guidance in their AQHI forecasting and alerting.

Time series of real-time air quality observations for the past 24 hours and 24-hour pollutant predictions were made available to the OSPC forecasters. The real-time information was generated for observational air quality stations in the GTA, with predictions available for these locations and the Games sport venue locations. The image files from the model forecasts were uploaded to a data portal for dissemination to the wider community (OSPC forecasters, ECCC research collaborators and university researchers, Health Canada and Ontario Public Health Units). Maps of surface pollutant concentrations were also created and placed on the Science and Technology ECPASS data portal (see Section 11.10) so that outside users could manipulate the model output using Google Earth.

Figure 18. Model predicted PM2.5 mass concentration

Map of the Greater Toronto Area (see long description below)

Description

Map of the Greater Toronto Area and northwestern New York State, showing shaded areas of model predicted PM2.5 mass concentration for July 13, 2015, at 4:00 p.m. local time. The areas are shaded at 2 µg/m3 intervals up to 10 µg/m3 and then every 5 µg/m3 up to 50 µg/m3. Fine particulate matter (PM2.5) pollution concentrations are generally 20-30 µg/m3 over land in Ontario, but there are areas of higher values over land in Ontario and the U.S.:

  • the highest concentrations are over 50 µg/m3 southwest of Shelburne, Ontario and near Medina, New York in the U.S. (south of central Lake Ontario)
  • up to 40 µg/m3 in Ontario long the lake-breeze front over Hamilton, in an area north of Arthur, and northwest of Guelph

The lowest values range from 8 to 10 µg/m3 over Lake Ontario and in an area north of Waterloo, and 4 to 6 µg/m3 over parts of northwestern New York State.

Wind arrows are depicted at the model grid points, indicating the direction and general strength of the wind. Winds are generally from the east to northeast over Lake Ontario, shifting to the south over land away from the Lake, and increasing in speed.

A suite of new graphical products and model evaluation tools were developed that can be run in near real-time to yield immediate results on model performance. The performance of the operational forecast model was compared with the high-resolution forecast model. The results demonstrated that better model performance was achieved with the new high-resolution GEM-MACH version 2, helping to identify locations and periods when poor air quality was expected.

The high-resolution GEM-MACH forecast was also sent to the University of Toronto AirSensors website. Through an agreement with ECCC, the University of Toronto developed miniature air quality sensors and deployed dozens of these sensors across the GTA, many near schools (see Section 12.3.3). A graphical user interface was developed to compare the GEM-MACH model results with the real-time air sensor output.

A number of air quality research projects have been developed using both the high-resolution GEM-MACH model and the special air quality observations collected by the Mesonet and by special science demonstration projects (see Section 11) in summer 2015. Two studies in particular are investigating the impacts of the urban environment and lake breezes on air quality. Asphalt surfaces in the urban environment increase surface temperatures (i.e., the urban heat island effect). Under stagnant atmospheric conditions, pollutant concentrations and circulations can be significantly altered over cities and the suburban environments. Research scientists will be studying such a case that occurred on July 28, 2015. Another study will investigate the effects of the Great Lakes lake breezes on air quality. Higher concentrations of pollutants, such as ozone (O3), are often observed along the convergence zone of lake-breeze fronts. Lake-breeze fronts were sampled by ECCC’s mobile air quality laboratory (CRUISER) (see Section 11.4) on July 24, 2015, during the Games.

An additional study and test of the model performance will investigate the impacts of urban pollution in rural communities. Urban emissions are often hot spots for air pollution, but a longer period of sustained wind from a given direction can result in rural communities being impacted by urban sources. One such case was observed during the Games on the afternoon and evening of July 13, 2015, at Parry Sound, approximately 200 km north of Toronto, when high values of O3 were observed over several hours, as the Toronto urban pollution plume interacted with the lake breeze from Georgian Bay.

Date modified: