 |
| Titel |
Uncertainty calculation in the RIO air quality interpolation model and aggregation to yearly average and exceedance probability taking into account the temporal auto-correlation. |
| VerfasserIn |
Bino Maiheu, Nele Veldeman, Stijn Janssen, Frans Fierens, Elke Trimpeneers |
| Konferenz |
EGU General Assembly 2010
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 12 (2010) |
| Datensatznummer |
250036166
|
|
|
|
|
|
| Zusammenfassung |
| RIO is an operational air quality interpolation model developed by VITO and IRCEL-CELINE
and produces hourly maps for different pollutant concentrations such as O3, PM10 and NO2
measured in Belgium [1]. The RIO methodology consists of residual interpolation by
Ordinary Kriging of the residuals of the measured concentrations and pre-determined trend
functions which express the relation between land cover information derived from the
CORINE dataset and measured time-averaged concentrations [2]. RIO is an important tool
for the Flemish administration and is among others used to report, as is required by
each member state, on the air quality status in Flanders to the European Union. We
feel that a good estimate of the uncertainty of the yearly average concentration
maps and the probability of norm-exceedance are both as important as the values
themselves.
In this contribution we will discuss the uncertainties specific to the RIO methodology,
where we have both contributions from the Ordinary Kriging technique as well as the
trend functions. Especially the parameterisation of the uncertainty w.r.t. the trend
functions will be the key indicator for the degree of confidence the model puts into
using land cover information for spatial interpolation of pollutant concentrations.
Next, we will propose a method which enables us to calculate the uncertainty on the
yearly average concentrations as well as the number of exceedance days, taking into
account the temporal auto-correlation of the concentration fields. It is clear that the
autocorrelation will have a strong impact on the uncertainty estimation [3] of yearly averages.
The method we propose is based on a Monte Carlo technique that generates an
ensemble of interpolation maps with the correct temporal auto-correlation structure.
From a generated ensemble, the calculation of norm-exceedance probability at each
interpolation location becomes quite straightforward. A comparison with the ad-hoc
method proposed in [3], where the uncertainty on the number of exceedance days was
calculated from the range in number of exceedance days obtained when adding and
subtracting the uncertainty on the yearly average concentrations on a daily basis, learns
that the Monte Carlo method yields slightly higher uncertainties on the number of
exceedance days. In general both methods however deliver reasonably similar uncertainty
estimates.
We will present uncertainty maps for the PM10, NO2 and O3 yearly average
concentrations in 2008 derived over the Belgian territory as well as the exceedance
probability of the PM10 norm of 50 μg/m3.
This study [4] was sponsored by the Flemish Environmental Agency.
References
Website at : http://www.irceline.be
Janssen, S., et al., Spatial interpolation of air pollution measurements using
CORINE land cover data. Atmospheric Environment, 2008. 42: p. 4884-4903.
Denby, B., et al., Comparison of two data assimilation methods for assessing
PM10 exceedances on the European scale. Atmospheric Environment, 2008. 42:
p. 7122-7134.
Maiheu, B. et al, Bepaling van Onzekerheid in Interpolatiemodellen, VITO
Report 2009/RMA/R/249 (Dutch). |
|
|
|
|
|
|