260 likes | 277 Views
An overview and evaluation of the real-time air quality forecasting at NCDAQ. This presentation discusses the modeling platform, forecasting methods, sample images, and future work.
E N D
An Overview and Evaluation of the Real-Time Air QualityForecasting at NCDAQ 2014 National Air Quality Conference Presented By Nick WitcraftNC Division Of Air Quality
NCDAQ Modeling Platform Linux cluster originally purposed for SIP modeling SMOKE, CMAQ 1 head node, 8 six-core dual CPU nodes Use 1 node per model run, up to 4 WRF and 4 CMAQ runs Number of forecast runs is reduced during regulatory modeling efforts. 2014 NAQFC
Real-Time Modeling NAM/GFS + 1/12 SSTs Initial and Boundary Conditions WRF MCIP ‘Canned’ Emissions CMAQ Initialize from previous run Web Images and Extractions
NCDAQ RT-AQ Model • WRF v3.51 provides the meteorology • National 36km/ Eastern US 12km domains • 4 ‘flavors’ of WRF
WRF Domains 36km 12km
WRf Model Output • Many different images produced internally • Will work on uploading some key variables to the web
Model external web page http://www.ncair.org/airaware/forecast/model
CMAQ output “mask” • Max 1 hour and 8 hour ozone is pulled from each forecast region • We forecast for regions, not individual monitors! • Now monitor based, with varying grid cell buffers
Southeast Masks • North Carolina – AVL, HKY, CLT, GSO, RDU, FAY, RWI • Georgia – ATL • South Carolina – AGS, CAE, CHS, FLO, GSP, UZA • Virginia – NoVA, FRD, RIC, WIN, SHEN, RNK, ORF • I can add more areas if interested. All I need is monitor lat-lon for each forecast area.
Sample mask – Triangle Region AIRS LAT LON CellBuffer NAME 371830014 35.856098 -78.574203 2 Millbrook 371830016 35.585 -78.7947 1 Fuquay 370630099 35.892502 -78.8769 2 BurdenEPA 370630015 35.9919 -78.8964 2 DurhamArm 370770001 36.1408 -78.7692 1 Butner 371010002 35.5 -78.4375 1 JoCo 370690001 36.0975 -78.4636 1 Franklntn 370370004 35.7572 -79.1597 1 Pittsboro 379999999 35.94 -79.07 1 ChapelH 379999999 35.779385 -78.356044 1 Wendell 379999999 35.528507 -78.324734 1 Smithfield-Selma HS 379999999 36.088436 -78.295541 1 Louisberg 379999999 36.359513 -78.520331 1 Oxford-Hendrsn
2013 12Z PerformanceCharlotte 36km runs in parentheses () *B and C runs did not run the entire summer
Case Study – June/July 2012 Ozone • Biggest ozone event in the last several years.
CLT – June/July 2012 Ozone The Model followed general ozone trend fairly well; slight high bias
Summary NCDAQ model had general over predictions. Model usually was ‘believable’. NOAA had better overall performance than NCDAQ models EGU emissions may be off; CEM data shows NC coal power plants at 30% capacity…NOAA may be picking up on this by using recent CEM data as basis for their EGU emissions PM2.5 forecast trends are usually pretty good. Need to do more analysis on PM2.5 NCDAQ model not always available (advantage NOAA); at minimum we try to keep 36km models running. We need to keep the NOAA model going given its reliability and good performance! 24
Future Work • Try to develop EGU emissions based on last years CEM data • Start extracting PM2.5, look at performance • Put extracted Ozone and PM2.5 Netcdf files on the web • Put some WRF images on the web • Transition to another web server. Current server has connection issues.
Questions? • Nick Witcraft • NCDAQ Attainment Planning Branch • Meteorologist II – Air Quality Meteorologist • 919-707-8484 • Nick.witcraft@ncdenr.org • Elliot Tardif • Bradley McLamb • www.ncair.org