Skip to main content

Smog in Shanghai, China

CityAQ: A Pilot to Combine Local Monitoring Data with GEOS-CF Model Outputs to Develop City-Scale Operational Forecasts



Context: Cities around the world struggle to find the financial and human capital resources to know more about the levels, sources, and impacts of air pollution. Many of the places with the highest level of pollution have the least information to work with to know their air quality and manage it accordingly. Free and publicly-available air quality data from satellites, along with low-cost sensor systems and output from air quality models, have the potential to augment high-quality, regulatory-grade data in countries with in situ monitoring networks and provide much needed air quality information in countries without them. By combining the strengths of these multiple monitoring technologies, an integrated air pollution network has the potential to provide free/low cost and publicly-accessible air quality information to health and air quality managers in cities around the world.

Brief Description: The World Resources Institute (WRI) and the NASA Goddard Space Flight Center (GSFC) are working together to pilot a scalable model for developing tools using various technologies that provide air quality information to city health and air quality managers.

Air Quality Forecasts: We are combining locally available air quality monitoring information with the outputs of NASA’s global GEOS Composition Forecast model (GEOS-CF) to develop optimized air quality forecasts that sub-national health and air quality managers can use. The NASA Global Modeling and Assimilation Office (GMAO) currently produces the GEOS Composition Forecasts to provide estimates and forecasts of the concentration of various pollutants at 25 x 25 km2 resolution near the ground and throughout the atmosphere. The model is run once a day to simulate and forecast air quality with inputs from an interpolated dynamic emissions inventory, a model of atmospheric chemistry, knowledge of past meteorology and forecasted future meteorology. Outputs are available in various time frequencies online, including 24-hour “daily” average forecasts are visualized on WRI’s Resource Watch platform. The approach to produce the CityAQ forecasts relies on a machine learning algorithm that optimally combines the GEOS-CF model forecasts and the local monitoring data.

Satellite Data: Global satellites have observed air pollution around the world for several decades and there are many exciting new and upcoming satellites. These data are being used by the air quality and human health communities in numerous ways. We are working to integrate satellite data into the standard operating procedures of city governments through the use of various NASA webtools and training resources, such as the ARSET program.

We are working with 10 cities around the world to pilot the methodology, identify user needs for analysis and user interface, and to aid in air pollution messaging to the public. Furthermore, we are developing an operational plan for extending the CityAQ forecast methodology to more cities and for creating additional analytical layers such as health warnings, source insights, or others as identified by participating cities. For example, we have developed a health-based air quality index (HAQI) for messaging air quality information to the public in a simple to understand way.

Project Goal: The project has four goals:

  • To provide participating cities with useful forecasts that they can use to anticipate air quality events, communicate with stakeholders, and manage local interventions more effectively.
  • To test and refine a methodology for combining locally held information with globally consistent analysis to offer new local tools and, in the future, potentially improve global analysis by aggregating fragmented local data into programmatically useful datasets.
  • To develop a methodology, which may include the development of web tools, to make the GEOS-CF forecasts usable by all world cities.
  • To test and refine a scalable approach for engaging with users to co-create air quality tools that leverage and extend existing scientific analysis.


  • Continue to run the GEOS-CF forecast.
  • Refine and apply the machine learning algorithm to local monitoring data and model outputs to develop correction factors to generate locally corrected forecasts for participating cities.
  • Advise on technical aspects of data and workflow to ensure cost-effective management of data coming in from cities and estimates being returned to city users.
  • Overall strategy, project management, and documentation.
  • Identifying, recruiting, and supporting participating cities to engage with the NASA team and each other.
  • Supporting participating cities in identifying and preparing relevant data for incorporation into the locally corrected forecasts.
  • Qualitative assessment of user needs and use cases for the locally corrected forecasts.
  • Initial visualization of locally corrected forecasting outputs on the Resource Watch platform.
  • Ingest identified city monitoring data into OpenAQ platform as the retrieval point for combining it with GEOS-CF.
  • Work with Development Seed to ensure that ingested data are addressable for developing the correction factors and stream of forecasts.
Development Seed
  • Programming workflow to draw local monitoring data from OpenAQ, combine it with GEOS-CF model outputs, and return the combined forecasts to city users through an API that is accessible to city users and Resource Watch.
  • Participate in 4 whole group meetings on: Inception, Monitoring Data and Quality Control, Locally Corrected Forecasts, User Interface Design.
  • Participate in meetings with WRI to follow up
  • Make locally available QA/QC monitoring data available to OpenAQ team for ingestion into the platform.
  • Provide feedback to WRI team on use cases and desirable user interfaces and additional analysis around air quality forecasts.


Stakeholder Partners: In discussions with various world cities.