PM10 Prognosis Model

Society's willingness to use our limited resources responsibly has grown significantly in the past years. In 2005, new maximum values for fine particles were set in the 22nd German Federal Emission Protection Directive (22. BImSchV). Since then, air quality has been an issue of special attention for the general public.

However, there are still some challenges which yet have to be mastered concerning air quality and emissions, such clearly identifying the proportions of natural and anthropogenic emissions, quantifying the efficacy of air quality control measures, and assessing the meteorological and location-specific propagation conditions of emissions.

At the moment, communities and municipalites are in a complicated position: on the one hand, they are bound by national law to comply with European air quality standards by implementing appropriate measures. On the other hand, they neither have any objective criteria nor sufficient knowledge to reliably evaluate emission levels and order appropriate air quality control measures in their respective field of authority.

The methods of PM10 data screening and PM10 emission prognosis developed at the Fraunhofer IVI aim to give support in this difficult matter.

Comprehensive data sets acquired by the automatic measuring stations of the air monitoring network about the most important pollutants, including PM10 fine particles, are the basis for the institute's model-based evaluation methods. The data are recorded in high resolution and close time intervals over a very long period of time.

Screening functions are based on powerful methods from the field of sinal analysis and allow the extraction of statistically relevant interrelations from large sets of data, such as the interrelation between dynamically changing traffic flows and local emission levels or between weather conditions and emission peaks.

The screening functions developed at the IVI  

• eliminate redundant process information from large sets of measuring data,
• take into account signals that change over time, such as typical hydrographs and their characteristic changes, and
• reduce air quality monitoring to a few statistically significant signals.

As a result, specifically targeted statistical evaluations can be performed.

The evaluations showed that especially meteorology and location have a major impact on emission levels. Unfavorable meteorological conditions, such as long periods without any kind of precipitation and thermal inversion, have a much larger influence on fine particle emissions than changes in traffic.

Based on the findings from screening the data of over 30 measuring stations, a prognosis model was developed at the Fraunhofer IVI. It is also based on the recorded long-term emission data sets and uses appropriately structured neural networks to generate a prognosis about emission levels.

The model offers a forecast of the average PM10 concentration during the two following days under consideration of the weather conditions. At the same time, it demonstrates the realistic reduction potential of traffic control measures. Under certain weather and propagation conditions, exceeding PM10 limit values can only be delayed for a short time and avoided only in threshold level situations.

The PM10 emission model developed by the Fraunhofer IVI was in practical use with Saxony's State Office for the Environment, Agriculture and Geology (Sächsisches Landesamt für Umwelt, Landwirtschaft und Geologie LfULG) for several years.