Environmental Engineering Reference
In-Depth Information
4.14 Multi-model Versus EPS-Based Ensemble
of Atmospheric Dispersion Predictions:
A Quantitative Assessment
S. Galmarini
1
, S. Potempski
1
,
F. Bonnardot
2
, A. Jones
3
, and L. Robertson
4
1
European Commission - DG Joint Research Centre, Institute for Environment and
Sustainability, Via E. Fermi 2749, 21027, Ispra, VA, Italy
2
METEOFRANCE, Dir. Prod., Serv./Environ., 42 av. Coriolis, 31057 Toulouse, France
3
Met Office, FitzRoy Road, Exeter EX1 3PB, UK
4
Swedish Meteorological and Hydrological Institute (SMHI), SE-601 76 Norrköping, Sweden
Abstract
Several techniques have been developed over the last decade for the
ensemble treatment of atmospheric dispersion model predictions. Among them two
have received most of the attention, the multi-model and the Ensemble Prediction
System (EPS) modeling. In the paper we compare both approaches with the help of
statistical indicators, using simulations performed for ETEX-1 tracer experiment.
Both ensembles are also compared against measurement data.
1. Introduction
The main difference between multi-model and EPS-based ensembles is the way
how the ensemble members are created. The multi-model approach relies on
model simulations produced by different atmospheric dispersion models using
meteorological data produced by potentially different weather prediction systems.
The EPS-based ensemble is generated by running a single atmospheric dispersion
model with the ensemble weather prediction members. The difference between the
two methods is motivated by the different emphasis that each of them puts on
different aspects of model uncertainty and how probabilistic forecast should be
used. While the EPS-based method concentrates on the influence of various equally
probable weather scenarios on the dispersion produced by one dispersion model,
the multi-model considers different answers from multiple sources that include
both the uncertainty in the weather predictions and the one that originates from the
use of different modeling approached to atmospheric dispersion modeling. The case
analyzed is the ETEX-1 release chosen for the abundance of measurements
collected and the wide range of studies performed on it in the past. For the specific
case the ECMWF EPS system was re-run. The two ensembles have been treated
statistically and compared.
