HIGH RESOLUTION RAPID REFRESH COUPLED WITH SMOKE (HRRR- SMOKE): REAL-TIME AIR QUALITY MODELING SYSTEM AND ITS APPLICATION TO CASE STUDIES Ravan Ahmadov 1,2 (ravan.ahmadov@noaa.gov) Acknowledgement: G. Grell 2, E. James 1,2, C. Alexander 2, S.Benjamin 2, B.Jamison 1,2, T.Alcott 2, J. Stewart 1,2, S. Freitas 3, G. Pereira 3, 4, I. Csiszar 5, M. Tsidulko 8, B. Pierce 6, S. McKeen 1,2, S. Peckham 7, L. Deanes 8, A. Edman 10, M. Goldberg 11, B. Sjoberg 11 JPSS proving ground and risk reduction program 1 Cooperative Institute for Research in Environmental Sciences, University of Colorado at Boulder, Boulder, CO, USA 2 Earth System Research Laboratory, NOAA, Boulder, CO, USA 3 NASA Goddard Space Flight Center & USRA/GESTAR, Greenbelt, MD, USA 4 Federal University of São João del-rei, MG, Brazil 5 Center for Satellite Applications and Research, NOAA/NESDIS, College Park, MD, USA, 6 Advanced Satellite Products Branch, Center for Satellite Applications and Research, NOAA/NESDIS, Madison, WI, USA 7 Cold Regions Research and Engineering Laboratory, US Army Corps of Engineers, Hanover, NH, USA 8 I.M. Systems Group, Inc, Rockville, MD, USA 9 National Weather Service, NOAA, USA 10 now at Penn State, University Park, PA, USA 11 NOAA's Joint Polar Satellite System Program Office 4 th Biannual Western Modeling Workshop September 8, 2017
Introduction High-Resolution Rapid Refresh (HRRR) is a numerical weather prediction system running operationally at the National Weather Service and in real time at NOAA Earth System Research Laboratory/ Global Systems Division (NOAA/ESRL/GSD). The model is run at 3km resolution over the CONUS domain with an hourly update cycle. At present mostly offline air quality models with relatively coarser resolution are used for smoke forecast. Some of these models do not simulate plume rise. A very few air quality models use the satellite Fire Radiative Power (FRP) data to estimate wildfire emissions and plume rise. The goal of this project Towards the Inclusion of VIIRS Fire Products into the HRRR Real-Time Forecasts funded by the JPSS PGRR program is to include the VIIRS products like FRP data into a coupled air quality model (HRRR-Smoke), in order to improve the numerical prediction of fire emissions and smoke dispersion in forecast models used at NOAA, and also to improve weather forecasting. The HRRR-Smoke model configuration is based on the HRRR model with added smoke tracer emitted as fine particulate matter by biomass burning emissions (including simulation of plume rise by the model). 2
HRRR-Smoke model The primary advantages of the HRRR-Smoke modeling system: High spatial resolution to allow simulation of mesoscale flows and smoke dispersion over complex terrain. Full coupling between meteorology and smoke: feedback of smoke on predicted radiation, cloudiness, and precipitation (using double moment microphysics). Biomass burning emissions and inline plume rise parameterization based on the satellite FRP data. A rapidly updating data assimilation cycle for meteorology; HRRR-Smoke uses meteorological input data prepared by the GSI data assimilation system and boundary conditions from another weather forecast model: Rapid Refresh (RAP). The forecast lead time is 36 hours. Four times a day (00, 06, 12 and 18UTC) a new forecast starts. Operational weather forecast models at NWS: RAP (white), 13km resolution HRRR model CONUS domain (green), 3km resolution (https://rapidrefresh.noaa.gov/)
Mapping the VIIRS FRP data to the HRRR-Smoke CONUS grid The clustering procedure performs a combination of all detected fires from VIIRS according to the model spatial resolution and grid configuration. Averaged VIIRS FRP data mapped over 3x3km HRRR CONUS grid pixels for July 19, 2017 4
Workflow of the real-time HRRR-Smoke modeling system Real-time VIIRS FRP data from the NESDIS ftp server Processing the FRP data on NOAA HPC in real-time (mapping, calculating fire size over the HRRR domain) Calculating fire emissions, and other parameters using the prep_chem tool Initial and boundary conditions for meteorology from the experimental RAP and HRRR weather forecast models run on NOAA HPC HRRR-Smoke real-time runs on NOAA HPC, 36hr forecasts Post processing of the HRRR-Smoke output using Unified Post Processor The experimental HRRR-Smoke started in June, 2016. The system is run 4 times a day at 00, 06, 12 and 18 UTC. It takes ~4 hours to complete entire cycle. Forecast plots are posted as simulations progress. Two NOAA HPC systems (Jet and Theia) are used. Multiple real-time datasets (meteorological observations, GFS model output, satellite ) are used. The forecast lead time is 36 hours. Providing output files in GRIB2 format to various users Plotting of the FRP and smoke fields and posting on the GSD web-site
The real-time HRRR-Smoke web-site for public access (rapidrefresh.noaa.gov/hrrr/hrrrsmoke/) 6
Smoke forecast for July 19, 2017 (rapidrefresh.noaa.gov/hrrr/hrrrsmoke/) This plot shows simulated fire emitted fine particulate matter (PM2.5 or fire smoke) concentrations and wind at the first model level (~8m above ground). The following plot shows forecast of near-surface fire smoke for July 19, 8pm EDT over the CONUS and its subdomain. This forecast is based on the model simulation of 24 hours from the model initialization time, which is 8pm EDT, July 18, 2017. 7
Fire plume rise dynamics Paugam et al., ACP, 2016 8
Smoldering and flaming emissions in HRRR-Smoke To calculate plume rise we need to know heat flux. The traditional approach in WRF-Chem to calculate plume rise: Use constant fire released heat flux numbers for a given land use class, e.g. Tropical Forest: min and max heat flux = 30, 80 kw/m2 New approach used in HRRR-Smoke: Heat flux ~ FRP/ burnt_area FRP measured by satellites Burnt_area is determined by using fire size M.Bela et al. (WRF-Chem tutorial) 9
Smoke concentration along WE x-section Forecast for June 27, 4pm EDT 10
11
1) VISUALIZATION OF FIRE RADIATIVE POWER (FRP) FROM VIIRS SATELLITE DATA (AUGUST 17-31, 2015) www.guardian.co.uk Active fires from the Suomi NPP VIIRS: Product status and first evaluation results, Csiszar et al., JGR, 2013/
AIR QUALITY MONITORING SITES ACROSS THE WESTERN US www.epa.gov/aqs
Mean Bias = -48.4 ug/m 3 RMSE = 289.8 ug/m 3
r =.49 Mean Bias = -16.8 ug/m 3 RMSE = 45.8 ug/m 3
HRRR-Smoke simulated vertically integrated aerosol concentrations and aerosol optical depth from VIIRS for August 27, 2015 The model does NOT assimilate the VIIRS AOD data! Modeled vertically integrated aerosol concentrations VIIRS AOD VIIRS data are also very useful for independent model verification!
VIIRS smoke mask and HRRR-Smoke forecast for vertically integrated smoke, July 28 2016 VIIRS AOD with smoke mask HRRR-Smoke VIIRS AOD data could be used in model evaluations 20
Qualitative comparisons of the GOES-R images with the HRRR-Smoke forecasts The Detliwer fire GOES-R image, 00:30 UTC, July 20 21 Vertically integrated smoke forecast 01UTC, July 20
Recent megafire in Central Valley, CA (http://www.latimes.com) 22
Simulating smoke feedback (direct and indirect) in HRRR-Smoke using double moment Thompson microphysics scheme 23 Adapted from J.Fast (WRF-Chem tutorial)
HRRR-Smoke weather prediction with and w/o smoke feedback on meteorology Difference in shortwave radiation between HRRR-Smoke simulations without smoke feedback and feedback included, 8pm EDT, August 20, 2016 Sensitivity study using HRRR-Smoke for August 12-20, 2016 Temperature bias verification of 18 hour model forecast using radiosounding measurements in the US Mean temperature bias No feedback With feedback Difference Including smoke feedback on meteorology reduces forecasted temperature bias.
Recent updates to the HRRR-Smoke modeling system Starting August, 2017: HRRR-Smoke is ingesting FRP data from the MODIS Terra and Aqua satellites in addition to VIIRS HRRR-Smoke preprocessor is using FEER (Fire Energetics and Emissions Research) emission coefficients [Ichoku and Ellison, 2014] to estimate the biomass burning emissions Smoke feedback on radiation and microphysics is included in the real-time HRRR-Smoke 25
https://rapidrefresh.noaa.gov/hrrr/hrrrsmoke/ GOES-R 26
Concluding remarks and future plans The real-time HRRR-Smoke modeling system provides an online smoke forecasting tool, which is used by incident meteorologists, air quality agencies, researchers and the public. The model can help us to improve weather forecasting by accounting for smoke impact on meteorology. The real-time VIIRS provides valuable datasets (AOD) for modeling fire emissions and model verification. Adding visibility product to the forecast output; Transitioning our smoke parameterization to NOAA s future NGGPS global modeling system; Using HRRR-Smoke (including NGGPS-Smoke in future) for upcoming intensive field campaigns in the US (FIREX, FIRE- Chem, WE-CAN) and further verification and refinement of the model using future aircraft measurements (e.g. parameterization of injection heights and emission factors); Use high frequency GOES-R FRP data for rapid update of fire detection and emissions; Synergy with the GOES-R initiated fire and smoke related activities; More rigorous evaluations are necessary. How best we can validate the smoke models? 27
THANK YOU FOR YOUR ATTENTION! QUESTIONS OR COMMENTS? 28