Air Quality Forecasting - Stage 2

01 Prepare your plan for Stage 2

The goal of Stage 2 is to develop a basic operational air-quality forecasting capability focused on routine public communication, awareness of air-quality conditions, and preparedness for pollution events (e.g., see Section 5 of the US Environmental Protection Agency’s Guidelines for Developing an Air Quality Forecasting Program; also see as an example from the World Meteorological Organization (WMO) Global Air Quality Forecasting and Information System Implementation Plan). In this stage, focus on using existing data and tools to meet early forecasting objectives, rather than building a sophisticated forecasting enterprise. Since this may be the first time your jurisdiction has undertaken air quality forecasting, first understand current in-country and external capabilities (e.g., see an example review from the United States).  Specifically, understand which monitoring data are available, what emissions information exists, which meteorological forecasts are available, which external forecast products already cover the jurisdiction, and what technical expertise is available locally or through partners.

Form a small working group along with representatives from Air Quality Monitoring, Emissions Inventory, Source Attribution, Public Engagement & Communication, Decision Support, and Health Impact Assessment teams. In this group, review the data and forecast products to help define the initial basic forecasting activity appropriate for your jurisdiction’s needs and current capacity. For example, the system could initially be used to provide useful public information, such as raising awareness of elevated pollution episodes, supporting basic health advisories (e.g., warning certain subpopulations about adverse conditions), or improving preparedness during recurring pollution episodes (e.g., during wildfires, dust storms, pollution-trapping thermal inversions, agricultural burning events, etc.). See Chapter 3 of the US EPA Guidelines for more information on understanding forecast user needs.

Contact your national or regional meteorological service early in the planning process to understand not only which meteorological forecasts and data products are available, but also whether those institutions have any air quality forecasting capabilities already in place (e.g., see Chapter 2 of the WMO’s Training Materials and Best Practices for Chemical Weather/Air Quality Forecasting).  Even without such air quality forecasting capabilities, the national meteorological service may already have operational forecasting systems, severe weather alert procedures, or partnerships with the WMO that can provide useful synergies for air quality forecasting. Where national expertise is limited, contact the WMO ​​(e.g., through the Global Atmospheric Watch program), regional forecasting centers, universities, or national research centers to identify training opportunities, modeling expertise, or examples of air quality forecasting from similar countries.

During this initial planning, identify (i) the most important pollutants to cover in this jurisdiction (e.g., informed by Air Quality Monitoring), (ii) the geographic area(s) to be covered (e.g., informed by Health and Environmental Impact Assessment findings), and (iii) intended users of forecast products (e.g., the general public, Decision Support colleagues, other government agencies, etc.).  Consider focusing on specific pollutants, such as PM2.5, ozone, desert dust, smoke from seasonal wildfires or agricultural burning, or stagnation events.  Forecast areas could range from a single urban center to a larger regional airshed.

As part of a practical Stage 2 plan,  estimate the budget, training, staffing, and data needs required to begin activities.  At a minimum, identify staff responsible for:  coordinating the forecasting activity; reviewing monitoring and meteorological data; interpreting external forecast products; preparing daily or episodic forecast summaries; coordinating with the national and regional meteorological services; communicating forecasts and health messages to the public; and documenting forecast performance. For a basic budget, consider staff headcount and time, training, internet access, data management, computers or servers, access to forecast products, visualization tools, communications channels, and technical support from universities, consultants, WMO, or other partners.

For more detailed information on air quality forecasting science and forecast planning, with examples from implementation in the Global South (Accra, Addis Ababa, Santiago, and Lima), the US EPA Megacities Partnership provides a wealth of information.  These include air quality training, multilingual planning templates and presentations, and lists of globally available datasets. 

02 Review forecasting guidance and good practice exampless

After establishing a Stage 2 plan and associated resources, begin forecasting by reviewing examples from comparable cities, countries, and regional and global forecasting centers (e.g., WMO, CAMS, NASA GEOS-CF, Monash University).  For example, examine published literature and technical guidance documents, standard operating procedures (SOPs), and examples of forecast products from other operational forecasting centers to understand good practices better and avoid common challenges (e.g., see this example PM2.5 forecast from Ghana and see other Global South examples from the US EPA Megacities Partnership). Familiarize all staff with the fundamentals of air quality forecasting (e.g., relationships between meteorology and pollution accumulation; concepts such as forecast uncertainty, model lead times, and differences between observational (e.g., from Air Quality Monitoring) and forecast data products.  For example, see this overview training from the US EPA and the Megacities Partnership trainings. Be sure that staff understands how to appropriately communicate the limitations of forecasts to decision-makers and the public.  

03 Examine and adapt external forecast products for country circumstances

At this stage, external forecast products can provide an effective starting point for jurisdictions with limited technical capacity. Examine global and regional forecasting systems such as the Copernicus Atmosphere Monitoring Service (CAMS), NASA GEOS-CF, NOAA’s smoke (biomass burning) forecasting products, WMO’s Sand and Dust Storm Warning Advisory and Assessment System (SDS-WAS), and other regional smoke or dust forecasting products, which can provide valuable operational information. Use these as initial tools while local forecasting capacity is being developed.

CAMS provides global and regional forecasts of atmospheric composition, including aerosols, ozone, nitrogen dioxide, carbon monoxide, dust, and fire-related pollution, and can be useful for understanding large-scale pollution patterns, regional transport, and multi-day pollution episodes. NASA GEOS-CF provides near-real-time global forecasts of atmospheric composition, including ozone, particulate matter, nitrogen dioxide, carbon monoxide, sulfur dioxide, and other species, and can be useful for interpreting how meteorology may influence the buildup, dispersion, or transport of pollutants. NOAA’s smoke forecasting products can provide information on the likely transport and surface impacts of wildfire or biomass-burning smoke, where such events are relevant to the jurisdiction. WMO SDS-WAS provides dust forecast products and regional coordination for sand and dust storm warning systems, and may be especially useful for jurisdictions affected by desert dust or long-range dust transport.    

Evaluate which external products are most suitable for your geographic location, pollutants of concern, operational resources, and communication needs. Compare these products with available local monitoring data to understand how well they represent local conditions. Global products can provide useful context, but they may not fully capture local sources, complex terrain, street-scale gradients, or locally specific emissions. Where external systems provide source-sector, tagged-tracer, sensitivity, or regional-transport information, these outputs may also help interpret the probable drivers of a forecast event. However, do not assume that a concentration forecast by itself identifies the sources responsible for the predicted concentrations. Formal attribution generally requires complementary analyses using emissions information, receptor modeling, trajectories, source tagging, emissions-sensitivity simulations, or other methods described in the Source Attribution guidance. 

04 Formalize relationships with meteorological services 

In Step 2, Air Quality Forecasting teams interfaced with national, regional, and global meteorological organizations to understand existing capabilities, data, and tools applicable to air quality forecasting (e.g., see Section 5 of the US EPA guidelines). In this step, formalize relationships with these organizations. Establish strong coordination with national and regional meteorological and hydrological agencies.  Coordinate with these agencies to obtain timely access to quality-controlled, quality-assured (QA/QC) weather forecasts, observations, and technical expertise. In many cases, meteorological agencies already produce weather forecasts that support basic air quality forecasts (e.g., during dust storms, wildfire smoke events, and stagnation episodes).  

Additionally, because meteorological organizations may already provide public weather advisories and alerts, work with these agencies to establish analogous communications for air quality events.  When meteorological institutions are already providing air-quality-relevant forecasts (e.g., during stagnation events), coordinate with your meteorological counterparts to clarify institutional roles based on the Air Quality Forecasting objectives established in Step 2. Regular communication with your meteorological counterparts can support capacity-building for both organizations. See Section 4.1 of the WMO GAFIS plan for examples on how to implement coordination with meteorological agencies.  

05 Formalize the relationship with the air quality monitoring team

Air Quality Forecasting requires timely, quality-assured observational data from air quality monitoring systems. Work closely with the air quality monitoring team in your jurisdiction to fully assess the availability, completeness, representativeness, and operational reliability of the monitoring network, and determine whether the available observations are sufficient to support forecasting. This review should formalize a long-lasting coordination with the air quality monitoring team, since the forecasting function will depend on the monitoring network to understand current conditions, evaluate forecasts, and identify recurring local pollution patterns.

Different levels of air quality data may be useful depending on what is available. See this US EPA training for an introduction to air monitoring and measurements. Regulatory-grade continuous monitors provide the most reliable information for operational use and forecast evaluation. Manual filter-based measurements, where available, can help characterize longer-term patterns and provide context for particulate matter episodes. Low-cost sensor data may also be useful, particularly when sensors have been co-located, calibrated, and quality-checked against reference instruments. Chemically speciated particulate matter measurements, black carbon observations, and other source-sensitive tracers may not be available at the temporal frequency required for daily forecast initialization, but they can support retrospective interpretation of recurring forecast errors and pollution episodes. Coordinate with Source Attribution colleagues to ensure these datasets are collected and managed in ways that support both receptor modeling and forecast-system evaluation. Even where only limited local data are available, observations from nearby cities, regional networks, satellite products, or global data platforms can provide useful context while local monitoring capacity is strengthened.

Utilize validated operational data wherever possible. Coordinate closely with the monitoring staff to ensure that data quality assurance and quality control procedures are appropriate for operational use. Real-time or near-real-time access to monitoring observations can substantially improve forecast interpretation and communication, especially during rapidly changing pollution episodes.

In addition to local observations, use external satellite products, regional observations, fire detection products, aerosol optical depth products, dust monitoring systems, and meteorological reanalysis datasets. For example, OpenAQ can help identify publicly available air quality observations from local, national, and international sources; NASA Worldview can support visual review of smoke, dust, and aerosol patterns; NOAA fire and smoke products can help identify fire activity and smoke transport; CAMS and NASA GEOS-CF provide regional and global forecast context (e.g., see NASA ARSET training); and WMO SDS-WAS provides information relevant to sand and dust storm forecasting. These products can help identify regional pollution transport, wildfire smoke plumes, dust outbreaks, and large-scale haze events.

Conduct a review of available data and document the most important data gaps limiting forecasting performance, such as insufficient monitoring coverage, limited representative siting, unreliable telemetry, insufficient QA/QC procedures, limited meteorological observations, or lack of access to operational forecast products. Communicate findings to your air quality monitoring colleagues to identify practical near-term improvements and to inform future investments in monitoring and forecasting capacity. Communication and coordination with the monitoring team should be maintained going forward.  

06 Establish a basic forecast production process

Air quality forecasting begins by combining trusted external forecast products with local interpretation and judgment. Establish daily workflows that include: reviewing meteorological conditions (including forward- and backward-trajectories, transport pathways, and other information on likely pollution source regions); evaluating external forecast products and observations (e.g., from satellites); comparing air quality monitoring data and meteorological data with these external data sources and expected conditions; and preparing forecast summaries. Develop basic Standard Operating Procedures (SOPs) (see this example from India) and operational checklists for consistent forecast production, data review, communication, and recordkeeping. Local expertise and understanding of weather and seasonal pollution patterns are critical for interpreting external forecast products; local knowledge of, for example, traffic patterns, agricultural burning cycles, dust events, local industrial operations, or recurring weather patterns can improve the relevance of forecasts. Prepare simple forecast products, including maps and other visualizations that are understandable to non-technical users, and use them to communicate expected air quality (see this list of resources for visualizing air quality monitoring data developed by the California Air Resources Board). 

At Stage 2, use existing publicly available global Air Quality Index (AQI) maps as simple interim forecast products. Development or adaptation of a locally appropriate AQI is covered in Stage 3 of the Air Quality Forecasting Guidance, in conjunction with the Air Quality Monitoring Guidance and the Public Engagement & Communication Guidance.​     ​  

07 Communicate forecasts to the public

Communicate forecast information in clear, concise, and accessible language appropriate for the audience. Use forecast products that explain expected pollution conditions and recommend appropriate responses (e.g., see Chapter 10 from the WMO Sand and Dust Storms Compendium). Referring to the Public Engagement and Communication Guidance Stage 3, Steps 2 and 3, develop procedures for communicating elevated pollution events associated with health risks, especially for vulnerable populations (e.g., children, older adults, those with respiratory and cardiovascular disease). The key to communication is its consistency across government agencies and the public. Develop channels for forecasting product distribution that are accessible to your constituents (e.g., websites, mobile apps, SMS systems, broadcasts, social media). These channels should be appropriate for local conditions (e.g., local internet access, literacy, languages, etc.). In all cases, communicate uncertainty associated with the forecast in an accessible manner (e.g., forecast accuracy can vary depending on pollutant type, weather conditions, and the time horizon over which the forecast is prepared).   

08 Build staff capacity

Focus initial air quality forecasting staff training programs on practical operational forecasting concepts and tailor them to the staff’s technical capacity and the jurisdiction's needs (e.g., WMO training materials, US EPA Megacities Partnership trainings, and example courses from Africa). Staff should become familiar with commonly used forecast tools and models, satellite products, operational dashboards, visualization software and systems, and techniques for forecast interpretation (see NASA ARSET GEOS-CF training and CAMS training). Training should also emphasize communication and interpretation of forecast uncertainty. Strengthen staff expertise through participation in regional workshops, peer exchanges, online training programs, and collaborations with universities or other regional and global forecasting centers. Capacity building will be ongoing to support staff in understanding meteorology and air pollution processes, interpreting observational data, and communicating operationally.  

09 Begin routine forecast evaluation

Begin comparing your forecasted pollutant conditions, based on external forecasting products and local input, with observed concentrations (see Chapter 9 of the WMO training material and examples from scientific literature such as Kondragunta et al., 2008). Attempt to identify recurring patterns in forecast performance (e.g., biases in particular locations or times). Simple bias tracking, identification of missed high-pollution events, review of false alarms, and assessment of forecast consistency during seasonal pollution episodes can help your team improve forecasts and better understand the relationship between external forecast products and local conditions. Conduct regular, structured reviews after major pollution episodes to identify lessons learned and opportunities for improvement. In all cases, document the results of forecast evaluation and use them to guide future investments in staff training and infrastructure, as well as to inform operational changes.