Assimilation of VIIRS Aerosol Optical Depth Information in the RAP and HRRR System to Improve Smoke, Visibility, and Weather Forecasts

3D smoke distribution forecasts will become operational as part of the rapid refresh (RAP) version 5 and high-resolution rapid refresh (HRRR) version 4 analysis and forecast systems scheduled to become operational at NCEP in 2020. Smoke fields are initialized in these systems by specifying emissions and fire plume rise derived from fire radiative power (FRP) data from the VIIRS instruments on the Suomi-NPP and NOAA-20, and the MODIS instrument onboard the Terra and AQUA polar orbiting satellites. The resulting smoke forecasts are valuable for air quality forecasting, aviation, and many other uses. In addition to smoke distribution forecasts, HRRR smoke fields are used in the diagnosis of surface visibility in HRRR forecasts. Case studies further indicate that allowing the smoke to interact with the radiation scheme also improves surface temperature forecasts when smoke is present. The feedback of smoke on radiation is implemented in the experimental HRRRX and RAPX.

An important aspect of the current implementation of HRRR-Smoke is that in addition to simulated emissions of new smoke specified by the VIIRS FRP empirical relation, the existing smoke field carries over from run to run (forecast smoke from one run provides an initial field for the next). While this cycling of the smoke field is very beneficial, errors in forecasts of lingering smoke can accumulate over time. To remedy this shortcoming, a smoke-filtered aerosol optical depth (AOD) product developed by NESDIS can be assimilated in a variational framework with the HRRR’s 1-hour smoke forecast supplying the background. Test cases are selected from the 2018 fire season, during which many large wildfires occurred in the northwestern US and during which the WE-CAN field campaign also took place. Smoke forecasts with and without AOD assimilation are quantitatively verified and subjectively evaluated: plots of modeled smoke are compared with satellite imagery from GOES ABI, and visibility forecasts with and without AOD assimilation are verified against mesoscale visibility analyses. Finally, comparisons are made between smoke forecasts and surface PM2.5 observations in the northwestern US, which are dominated by smoke in the summer months. Improvements to the HRRR’s smoke analyses through AOD assimilation will lead to improved forecasts not just of the smoke distribution itself, but also of surface visibility and weather (via simulated smoke-meteorology interactions).