Smog forecasting in the Netherlands is a legal task of RIVM, performed in commission of and
paid by the Dutch Ministry of Environment. The current smog forecast is based on hourly
measurements in the Dutch Air Quality Monitoring Network, temperature forecasts by the
Dutch Met Office KNMI, and a statistical model PROZON maintained by RIVM.
This project aims to develop and demonstrate a new instrument to improve the existing smog
forecasts in the Netherlands. It extends the chemical transport model LOTOS-EUROS with
Ensemble Kalman Filter data assimilation to improve the forecast performance of the free
running model. Assimilated data include surface ozone data from ground based networks in
the Netherlands and neighbouring countries, and tropospheric nitrogen dioxide column data
over Western Europe measured by OMI on board of NASA’s AURA satellite.
LOTOS-EUROS is an operational Chemical Transport Model that is used, maintained and
developed further by a consortium of Dutch research institutes: TNO, RIVM, PBL and
KNMI. Model results and validation studies have been reported in literature and the model
has participated in model comparisons for ozone and aerosols.
The groundbased data used in this project are acquired by the member states under the EU
“Directive on ambient air quality and cleaner air for Europe”. Member states are obliged
under EU legislation to perform these measurements and report the results to the European
Union. The directive will be extended to include near-real-time data in the near future. RIVM
operates the Dutch air quality monitoring network and contributes its data to the EU. It has
access to all similar European data at no cost.
The model has been adapted, and the necessary real time data streams have been
implemented. Ground based data are assimilated and used in the forecast. Satellite data can
technically be assimilated and used in the forecast but suffer from discrepancies between
observations and model. Some flaws remain in the stream of European scale ground based
data, but these are outside the scope of this project and will be resolved soon.
The performance of the new system has been validated using a historical dataset, and results
were compared to the statistical model PROZON. Results show that the new tool shows
higher spatial and temporal resolution (13 km, 1 hour), and provides the new capability to
generate maps and movies of the evolution of air quality over a large part of Western Europe,
including a good quality nowcast. The quality of its one-day forecast is better than PROZON
with respect to timing issues: in a changing situation PROZON is often seen to lag behind,
where SmogProg is in better synchronisation with the observations. On the other hand,
SmogProg at this stage of development still has considerable difficulty to predict next day’s
absolute smog levels correctly, and is generally underestimating levels above 150 µg/m3. As
the relevance of smog forecasting is closely linked to predicting correctly the crossing of the
European information and alarm thresholds of 180 and 240 µg/m3, this is still a major flaw. It
is however anticipated that the new model’s performance on this point can be improved
considerably in the coming months.
Due to its simplicity, the operational costs of the old tool PROZON are very low. As a
consequence, the new tool SmogProg will inevitably be more expensive to run on an
operational basis. A cost-benefit analysis should therefore focus on the question if the
additional benefits obtained outweigh the additional costs.
The additional benefits lie in a better scientific understanding of the smog situation and in
forecasts with better timing and higher spatial resolution, which enable sensitive people to
avoid adverse situations more effectively. While these are clear benefits for scientists,
policymakers and society, it is hard to express them directly in economic value. From
economic point of view the increased communication power to non-experts is considered the
most direct asset of the new system: Smog-induced healthcare costs amount to at least 100
million euro per year in the Netherlands. Part of these costs can be avoided by “smog-aware”
behaviour of the population. If the new tool is not considered cost-effective on the less
material grounds mentioned above, just a minor swing of the order of 0.1 % in health related
costs will make the new tool cost-effective on a pure economic basis.
DPJ Swart, S Jongen, H Eskes, M Schaap, RMA Timmermans, A Segers, AMM Manders, FJ Sauter, JPJ Berkhout, DE Lolkema. SmogProg: Towards operational smog forecasts using near-real-time satellite measurements