Environmental Engineering Reference
In-Depth Information
3.7 Smog Forecasting in The Netherlands Using
Assimilation of Ground-Based and Satellite
Observations
Astrid Manders
1,3
, Suzanne Calabretta-Jongen
2
, Henk Eskes
2
, Martijn
Schaap
1
, Renske Timmermans
1
, Arjo Segers
1
, Ferd Sauter
3
, and Daan Swart
3
1
TNO Unit Environment, Health and Safety, P.O. Box 80015, 3508 TA Utrecht,
The Netherlands
2
KNMI , P.O. Box 201, 3730 AE De Bilt, The Netherlands
3
RIVM, P.O. Box 1, 3720 BA Bilthoven, The Netherlands
Abstract
The capability of the regional chemistry transport model LOTOS-
EUROS to forecast air quality in the Netherlands has been tested. For PM10, the
free-running LOTOS-EUROS model outperforms the time correlation of the
Dutch operational statistical air quality forecasting model. But when a bias
correction is applied also the abolute concentrations are modelled better than for
the statistical model. For ozone, data assimilation of ground-level concentrations
in The Netherlands, Belgium and Germany was applied using ensemble Kalman
filtering. The data assimilation improved the initial conditions and the one-day
forecast. However, the model has difficulties in forecasting extreme ozone con-
centrations. Work is under way to improve on this and to include ground-based
PM observations and satellite observations of PM and NO2 in the assimilation
process.
Keywords
Air quality forecasting, data assimilation, ozone, PM10
1. Introduction
In the Netherlands, smog episodes due to enhanced concentrations of particulate
matter or ozone occur occasionally. In such cases, the public is informed, so that
people can adapt their behaviour. But already for modestly increased concentrations,
sensitive people would like to be aware of the situation. Therefore, it is important
to have an accurate smog forecast. Presently, two statistical models are used in the
Netherlands. These models predict the daily average PM10 concentration and the
ozone maximum concentration on observation locations, based on the weather and
observed concentrations of the day before and on the weather prediction. The
advantage of these models is that they are relatively simple, do not have a bias and
