Assimilation of PM ground measurements: Looking for optimal settings for the PM forecasts

Abstract

PM10 forecasts for Europe and the Netherlands are part of the daily LOTOS-EUROS products, next to ozone forecasts. So far, ozone ground observations and OMI NO2 column observations were assimilated in the forecasting system, using an Ensemble Kalman filter. Previous studies have shown that assimilation of daily PM10 ground observations improve the PM forecasts. Ideally, ozone, NO2, AOD and PM10 can be assimilated within one analysis. If we want to combine ozone and PM assimilation in one system, assimilation of hourly PM observations is needed and the model system was extended by assimilation of PM10 and PM2.5 ground observations. We have investigated optimal settings of the system, taking several aspects into account. First of all, a complication with regard to ozone and NO2 is that LOTOS-EUROS underestimates the PM concentration. This was previously accounted for by a bias correction, but can more elegantly be solved by introducing a new species of unknown composition which is allowed to vary independently. Secondly, hourly PM observations can have relatively large errors and are sometimes influenced by very local conditions. Therefore, sensitivity experiments were done to investigate the sensitivity to and optimal settings for the correlation length between stations, error constraints on observations, noise factors on the emissions and correlation timescales. We present the lessons learnt and the current performance of the system.