Six-day PM10 air quality forecasts for the Netherlands with the chemistry transport model Lotos-Euros

In this work we study the ability of the chemistry transport model Lotos-Euros to forecast, with a range of six days, PM10 concentrations and exceedances thereof. For both rural and non-rural locations in The Netherlands and based on one year of data, model performance does not deteriorate up to a lead time of three days. Thereafter the PM10 forecast is increasingly affected by uncertainty in the meteorological forecast. However, up to a lead time of 6 days the forecast still has skill, beats persistence and complies with several performance criteria. The correlation between forecast and observations is between 0.66 and 0.70 for the first half of the forecast and remains above 0.54 until the end of the forecast range. Exceedances of the PM10 concentration over thresholds are also forecasted with reasonable skill up to a forecast range of three days, after which a gradual deterioration sets in. The stability of the forecast displays the same behaviour. Up to a lead time of three days, the forecast remains reasonably stable with more than 80% of forecasted exceedances still present in all later shorter-term forecasts for the same date. Because exceedances can be forecasted with considerable skill a number of days in advance, the forecast can be used for applications that require a range of a few days, such as outdoor activities and the scheduling and implementation of short-term emission reduction measures.

► An operational system for forecasting air quality with a range of 6 days was set up.
► The quality of PM10 forecasts was evaluated using different metrics and benchmarks.
► Generated PM10 forecasts do not deteriorate up to a lead time of 3 days.
► These forecasts still display skill up to a lead time of 6 days.