Analog-Based Postprocessing Methods for Air Quality Forecasting
Two new postprocessing methods based on analogs are proposed to reduce the systematic and random errors of air quality prediction. The analog of a forecast for a given location and time is defined as a past prediction that matches selected features of the current forecast. The first method is the weighted average of the observations that verified when the best analogs were valid (AN). The second method consists in applying a postprocessing algorithm inspired by the Kalman filter (KF) to AN (KFAN). The AN and KFAN are tested for ground level ozone and PM2.5 0–48 h predictions from the Community Multiscale Air Quality (CMAQ) model, with observations from 1602 surface stations from the EPA AirNow network over the continental United States for a 1-year period. Preliminary results of the new methods include a large reduction of the systematic and random errors of the direct model output, with an increase of the correlation between observations and predictions at all forecast lead times.
KeywordsKalman Filter Observational Error Good Analog Ground Level Ozone Chemistry Transport Model