On air quality forecasts

Air quality forecasts are based on chemical-transport modeling of pollutants in the atmosphere. For these forecasts, it is crucial to know amount of the emissions  and detailed forecasts of meteorological elements - such as temperature, wind direction and speed in individual layers of the atmosphere and precipitation. Emissions, which are variable in space and time, are in some cases calculated on the basis of measured data - this is how data from very large sources (stacks) are available. More often, however, they are calculated on the basis of emission factors from specific activities and the quantification of their activity (predicted traffic flows, fuel consumption for heating, soil fertilization, etc.). We often only know the total amount of emissions per year for a given sector and a given country from national emission inventories. In this case, we need to spatially distribute emissions based on the assumed distribution of sources, which is done for example using spatial data on land use - arable land, built-up area, farm buildings, industrial sites, etc. The CAMS service provides such disagregated emissions for nearly entire territory of Europe. The temporal distribution of emissions also needs to be estimated - some facilities, especially large pollution sources, work almost constantly throughout the year, other sources - for example, related to heating - only during the winter. Cars also have higher emissions in the cold season, mainly due to cold starts, traffic density also changes during the day and week. Thus, the temporal distribution of emissions is approximated in the models by the average predicted daily and annual profile. In the case of local heating, the annual profile (how emissions change during the days of the year) is calculated based on the average daily temperature (more heating is done on colder days) and the daily profile is considered with two maxima: morning and evening.

Air quality forecasts are strongly dependent on the quality of emission inputs.The predicted concentration may not coincide with the measured data if the emission source is not included in the input data, has an incorrect location, or its abundance at the given time is much higher or lower than expected. Further uncertainty is introduced into the forecasts by the uncertainty of the forecast meteorological fields, especially the precipitation and the temperature stratification of the atmosphere, which determines the stability and thus the dispersive conditions of the atmosphere. Due to the computational complexity, chemical-transport models count on a horizontal resolution of several kilometers - its outputs are therefore average concentrations in a relatively large area, therefore the model often significantly underestimates the concentrations in the territory of smaller villages and towns, or near busy roads, for example. Air quality forecasts are mainly used to determine the trend of pollutant concentrations (improvement/deterioration of air quality, approximate determination of the source of pollution (local, cross-border, desert dust), to determine expected dispersion conditions and the direction of spread of pollutants.

Thanks to the CAMS National Collaboration Program - Slovakia project, we are making available on the SHMÚ website forecasts from two CMAQ and CAMS EU models for basic pollutants PM10, PM2.5, O3 and NO2. The CMAQ model is run on the SHMÚ HPC3 supercomputer, and CAMS EU forecasts are downloaded from the European Center for Medium-Range Weather Forecasts ECMWF. The advantage of the CMAQ model is a more detailed horizontal resolution of 2 km compared to the 8 km resolution of the CAMS model. On the other hand, the EU CAMS forecast is based on the median of the forecasts of 11 state-of-the-art numerical air quality models developed in Europe, which eliminates systematic and random model errors. At SHMÚ we evaluate both forecasts and continuously update the input data (emissions, spatial data) for the CMAQ model.