You are here:
APPLICATION OF BIAS AND ADJUSTMENT TECHNIQUES TO THE ETA-CMAQ AIR QUALITY FORECAST
Citation:
KANG, D., R. MATHUR, AND S.T. RAO. APPLICATION OF BIAS AND ADJUSTMENT TECHNIQUES TO THE ETA-CMAQ AIR QUALITY FORECAST. Presented at 5th Annual CMAS Models-3 User's Conference, Chapel Hill, NC, October 16 - 18, 2006.
Impact/Purpose:
The objectives of this task include: (1) to continuously evaluate and analyze the forecast results to provide diagnostic information on model performance and inadequacies to guide further evolution and refinements to the CMAQ model, and (2) extending the utility of the daily air quality forecast model data being produced by NOAA's National Weather Service (NWS) as part of a NOAA/EPA collaboration in air quality forecasting, to EPA mission-oriented activities. These objectives include developing and maintaining a long-term database of air quality modeling results (ozone and PM2.5), performing periodic analysis and assessments using the data, and making the air quality database available and accessible to States, Regions, RPO's and others to use as input data for regional/local scale air quality modeling for policy/regulatory purposes.
Description:
The current air quality forecast system, based on linking NOAA's Eta meteorological model with EPA's Community Multiscale Air Quality (CMAQ) model, consistently overpredicts surface ozone concentrations, but simulates its day-to-day variability quite well. The ability of bias correction techniques in improving the accuracy of the forecasts at discrete monitoring locations is investigated through their application to archived real-time surface level ozone forecasts from the Eta-CMAQ air quality forecasting modeling system. In particular, two bias correction techniques, namely the Hybrid Forecast and the Kalman Filter Predictor approaches, are applied to model forecast data during July to September 2005 period to examine if the corrected forecasts can improve forecast accuracy in both discrete and categorical measures. Hybrid Forecast is based on today's observation and the change between today's and tomorrow's model forecasts to correct tomorrow's forecasts, while the Kalman Filter Predictor Bias Correction is a recursive algorithm to optimally estimate bias correction term from previous measurements and forecasts. Preliminary results have shown that both Hybrid Forecasts and Kalman Filter Bias Correction Forecasts can significantly reduce forecast errors compared with original model forecasts. However, the impact on categorical metrics such as hit rates and false alarm ratios varies with time and locations. A detailed analysis of this research will be presented.