Data-driven Ensemble Forecasts of Hazardous Air Quality Events over North America

Although air quality continues to improve in the United States, the frequency of high-impact extreme events, such as wildfires and dust storms, has increased rapidly in the past decades and is projected to rise further in response to climate change. Hazardous events can impose drastic impacts on the society, including adverse health effects, life and property losses. Consequently, HAQAST always sees a spike in the demand for timely information from health and emergency managers during these crises, exemplified by the California Camp Fire and Hawaii Volcano eruption in 2018. The goal of this proposed study is to improve our collective predictability of these high-impact events through emission data assimilation (EDA) and multi-model ensemble forecasting, in order to mitigate harmful effects on human health and the economy.
We assemble a team of Earth scientists, air quality and health experts and managers to perform the following tasks:
a) Using NASA satellite data to improve dust and fire emissions. We will use newly developed emission data assimilation techniques to address model uncertainties during two hazardous events: “killer” dust storms and high-opacity wildfires. For “killer” dust storms, we will combine new crop data with MODIS NDVI to best constrain cropland dust sources, and then use VIIRS Albedo to improve surface roughness for better prediction of these small-scale dust storms. For large smoky wildfires, inverse modeling with MODIS MAIAC AOD and TROPOMI CO will be used to correct underestimated wildfire emissions and FRP due to smoke/cloud effects. Improved emission techniques and data will be provided to five forecasting models operated by three federal agencies (NOAA, NASA, and NRL) for ensemble forecasting.
b)North America Ensemble Forecasting. We propose to develop an ensemble forecast of dust and fire by leveraging two operational/research programs: the National Air Quality Forecast Capability (NAQFC) and the International Cooperative for Aerosol Prediction (ICAP). Two regional forecasts (HYSPLIT and CMAQ) from NAQFC and three global forecasts from NASA, NOAA and NRL will the initial members of the ensemble. Satellite, sub-orbital and ground observations will be used to evaluate model performance over North America in order to determine weighing factors and optimize the ensemble.
c)End user applications. We will work with stakeholders to prepare customized data packages for three applications: 1) producing air quality metrics for the City Health Dashboard, a health initiative serving 750 largest US cities; 2) providing real-time ensemble dust forecasting as a pilot project for World Meteorology Organization (WMO) Pan-American Sand & Dust Storm Early Warning Advisory; and 3) working with WHO Pan-American Health Organization (PAHO) to provide forecasting and observations of wildfires and air quality to its member countries. Other applications, such as real-time mapping of cropland wind erosion and dust prediction for emission control, are also requested by federal and local agencies.

Fig. Conceptual design of the project tasks.