The forecast algorithm is simulating expected concentrations at monitoring sites. However, the user can define weighed combinations of different monitoring stations in order to estimate a forecast that is valid for restricted geographical areas such as city centre, suburban areas etc.
A Non-complex Method for the User The EnviMan Forecaster is extremely simple to set up and to work with - no need for maps or emission data, just monitoring data and a weather forecast.
Basic Principles and Input Data Requirements The EnviMan Forecaster is based on an adaptive statistical procedure.
1. The basic model estimation Historical data of Air Quality is analysed according to a classification of day types such as working days, Saturdays and Sundays. The 'social component' of the Air Quality is determined and reduced from the original data. The remaining part of the data is assumed to be partly weather dependent and partly stochastic (random). A statistical method is applied and the weather depending on variation is parameterised as a function of the weather at different time scales.
2. The adaptive model estimation The basic model is applied for a period immediately preceding the date and hour when the forecast is to be issued. Pattern corrections and offset corrections are applied and the final forecast equation is set-up.
3. The forecast application The weather forecast is interpolated to hourly values and the final forecasting equation is applied.
Expected Forecast Accuracy The accuracy of the forecast depends on:
1. The quality of the historical data It is of vital importance that historical data should be validated. The type and quality of the measuring technique as well as the location of the monitoring station will affect the prediction quality.
2. The quality of the weather forecast During periods, the quality of weather forecast can be extremely important for the quality of the pollutant forecast.
3. Pollutant type Certain pollutants are easier to forecast than others.
It is expected that the EnviMan Forecaster should substantially improve the air quality forecast when comparing with manual methods. The typical precession should imply a correlation with observed air quality measurements between 0.7 and 0.9.