Air Quality Forecasting - Stage 5

01 Secure permanent national mandates and governance 

By Stage 5, secure and operate the air quality forecasting system under formal legal authority, including defined institutional mandates, governance structures, and operational responsibilities.  These responsibilities should be integrated into other aspects of air quality management and relevant national frameworks, such as emergency preparedness structures.  Formally document all roles and responsibilities across agencies, ministries, and jurisdictions with clear avenues for coordination with meteorological agencies, environmental authorities, public health institutions, emergency management and transport agencies, and regional governments, as applicable.  Write and maintain a multi-year strategic plan defining long-term operational objectives, priorities for modernization and staffing, infrastructure development, and partnerships.  An example strategic plan for the Global Air Quality Forecasting Information System is available from the WMO. 

02 Sustain a fully operational production system

At Stage 5, forecasting operations are a critical government function; they operate the forecasting system continuously or on schedules appropriate to jurisdictional needs (e.g., see this example from Canada).  Build redundancy, disaster recovery, backup communications, and cybersecurity protections into the system owing to its criticality.  Conduct continuous monitoring of operational uptime/downtime, forecast delivery performance, communication reliability, and system availability.  Ensure that data-sharing systems that provide timely access to forecast inputs, outputs, observations, and evaluation datasets are made available to authorized users and, where appropriate, to the public and research community to maintain transparency.   

03 Integrate forecasting into regulatory and planning systems

Ensure that air quality forecasting is integrated into government functions, including regulatory programs, urban planning, climate strategies, infrastructure planning, and long-term environmental management by Stage 5. Forecast products should support, for example,  regulatory actions and compliance, public health interventions (especially during severe pollution episodes), transport planning, infrastructure design, climate adaptation planning, land-use decisions, and pollution mitigation strategies.  For regulatory actions and compliance, forecasts can help agencies anticipate likely pollutant exceedances. Advanced warning or health-relevant pollution episodes can be provided to health agencies, schools, hospitals, and emergency managers. For transport planning, forecasts can inform congestion management and advisories when vehicle activity is expected to worsen air quality. For infrastructure design, climate adaptation planning, and land-use decisions, forecast products and analyses of historical forecast performance can help identify areas repeatedly affected by smoke, dust, heat-pollution events, stagnation, or elevated exposure, so that siting, design, zoning, and resilience measures better account for air quality risk. Forecast evidence and analysis of historical forecasts support strategic planning and help to evaluate the effectiveness of air quality policies. See Ghude et al., 2024, for an example of an air pollution forecast system being used to evaluate local emission abatement policy scenarios in Delhi. 

04 Maintain leading-edge science and technology

It is critical that, by Stage 5, your air quality forecasting system works both operationally and at the leading edge of science and technology.  Evaluate and integrate emerging scientific methods, new observational systems, and analytical technologies into your forecasting system.  For example, routinely assess next-gen models, remote sensing systems (e.g., satellites), and data assimilation approaches.  Intensive field studies, including mobile observations, supersite campaigns, and aircraft data, can help diagnose forecast errors and improve understanding of model processes that impact pollution events.  Increasingly, Artificial Intelligence (AI) and Machine Learning (ML) tools can support forecasting functions such as pattern recognition, data processing, bias correction, and decision-support applications.  Document such AI/ML systems, including their limitations, and evaluate their reproducibility (e.g., see the ECMWF example documentation).  Ensure that any changes to forecasting systems are made incrementally, with strong documentation standards, reproducible results, no adverse impact on forecast metrics, and full transparency. 

05 Strengthen partnerships across government, academia, and industry

To help remain at the leading edge of science and technology, seek to establish strong collaboration across operational agencies (e.g., meteorological services), universities, research organizations, regional forecasting initiatives, and the private sector. Such partnerships can foster technical innovation, workforce development, and scientific collaboration. Establishing working relationships with regional and global forecasting initiatives can further help harmonize forecasting methods, facilitate benchmarking, and improve response to major transboundary pollution events (e.g., dust storms and biomass-burning plumes). First, evaluate innovations gleaned through partnerships as pilot projects before deploying them operationally (e.g., see this example pilot project from the US). An example of a partnership between academia and the EPA is available here.

06 Deliver advanced public and sector services

At this stage, focus on developing and providing ​​​​​​tailored, user-focused forecast products, such as multi-hazard compound heat-pollution event forecasts (Huang et al., 2026, Zhang et al., 2022), for decision support across economic sectors (maritime, aviation, etc.) and communities (the general public and vulnerable populations). Examples of such products include smoke, dust, or ozone pollution early-warning products for emergency management; visibility advisories for aviation or maritime industries; or dashboards for industry and regulators that indicate periods when prevailing conditions increase the likelihood of regulatory exceeding pollutant concentrations (potentially necessitating temporary emissions controls).

Increasingly, strive for higher spatial resolution in the forecast to support health impact assessment with exposure guidance and strategies.  To develop the most relevant data products, continuously engage with stakeholders to improve product accessibility and usability (see Public Engagement & Communications Guidance, Stage 5). 

07 Maintain rigorous verification and transparency systems

Advanced air quality forecasting systems should strive for transparency across all their components. To achieve this, maintain comprehensive verification and accountability systems that are published regularly.  For example, publish performance and operational metrics, as well as technical documentation, and public dashboards with links to accessible data archives. These materials, including documentation of methodological changes, known limitations, and operational challenges, should be available for periodic internal and external technical reviews and independent audits. This helps maintain the scientific credibility and public trust in forecasting systems. See Pagano et al., 2025, for an example from Numerical Weather Prediction (NWP). 

08 Operate an annual review and modernization cycle

Alongside formal annual reviews that cover operational performance, governance, technical systems, staffing, user needs, and strategic priorities, implement schedules for continuous modernization. Examine new forecast models, workflows, software systems, communication platforms, and training programs for incorporation into your operations.  These modernizations should be prioritized according to evolving science, user needs, team experience, and technological developments. Beyond the formal annual modernization review process, foster a culture of continuous operational improvement, innovation, collaboration (both internal and external), and scientific integrity. See Li et al., 2025, for an example of forecast model documentation and updates from NOAA. 

09 Establish in-house analytical and R&D capacity for continuous innovation 

By the end of Stage 5, maintain dedicated in-house analytical and research and development (R&D) capabilities that support continuous innovation.  To achieve this, ensure you have dedicated specialists in place across atmospheric science, operational forecasting, software engineering, data analytics, artificial intelligence and machine learning, and data science. Your R&D team should pilot, for example, forecasting approaches, automated workflows, and advances in atmospheric science and modeling.  As these pilots yield useful results, they should be incrementally incorporated into operational mode, always with sufficient documentation and verification.  Having strong internal teams ensures you can respond rapidly to emerging operational challenges and sustain excellence in long-term forecasting.

With a strong R&D team in place, contribute your advances in modeling, data analysis, and other relevant techniques to the broader scientific community through peer-reviewed publication.  Participate in regional and global scientific workshops and initiatives, and lend your expertise to initiatives that advance air quality management in other jurisdictions.  

Finally, recognize that historical relationships between weather, emissions, and air quality may change over time, particularly in the face of climate change (Lai et al., 2025). Climate change can increase the frequency or intensity of events that affect air quality, including wildfires, heat waves, drought, sand and dust storms, and stagnation events. These changes can shift pollution baselines, alter the timing and severity of seasonal episodes, and make past forecast performance less representative of future conditions. As part of continuous improvement, periodically review whether forecast assumptions, emissions inputs, episode definitions, and alert thresholds remain appropriate as climate conditions change, and coordinate with meteorological, climate, emergency management, and public health agencies where climate-sensitive air quality risks are increasing.