The EuroDelta inter-comparison, Phase I Variability of model responses to emission changes L. Tarrasón, L. Rouil, P. Thunis, C. Cuvelier P. Roberts, L. White R. Bergström, B. Bessagnet M. Schaap, A. Graff and R. Stern
Main Objective The main objective of the EuroDelta inter comparison is to quantify the variability of modelled responses to emission changes, and use such information to determine the robustness of the source-receptor relationships currently provided to integrated assessment modelling supporting the design of emission control strategies over Europe.
Novel model inter-comparison that has allowed to: Establish the performance of the UNIFIED EMEP model with respect to other state-of-the-art regional-scale models Determine the range of confidence of modelled responses to emission changes at regional scale.
Publications so far Model validation for ozone with 2 additional models TM5 and DEMH, that participated in the EMEP model evaluation Van Loon et al. (2006) Evaluation of long-term ozone simulations from seven regional air quality models and their ensemble average, Atmos. Env.,41, 10, pp. 2083-2097 Evaluation the significance of the ensemble approach for air quality Vautard et al. (2006) Is regional air quality model diversity representative of uncertaintty for ozone simulation? Geosphys. Res. Lett., 33 L24818, doi:1029/2006GL27610
Eurodelta in numbers 5 CTM models: CHIMERE, REMCALGRID, MATCH, LOTOS, EMEP 20 (+) scenarios: 2010_CLE and scenarios from 2020CLE 1 base year: 2000 close to CAFE_BASELINE 1 common meteorological year: 1999 1 common emission inventory: Updated EMEP grid distribution 5 different boundary conditions: Global models and observations
List of scenarios for source-receptor calculations Base cases : 2000, 2010CLE, 2020CLE Consistent with the baseline scenarios defined in the CAFE programme
Novelty in EuroDelta intercomparison: Additional tests of model performance The performance of the models is established, by their ability to reproduce observed air concentrations and by their internal consistency in particular, concerning their ability to represent responses to emissions changes** Ozone responses to Nox control (50% red vs 25% red) The analysis allowed to identified errors in the deliveries of the data, as for instance in Model 5, which had provided data for the wrong scenarios
Scenario results for Ozone and PM2.5 The five models provide consistent results of the impact of emission reduction scenarios on PM2.5 and SOMO35 for year 2000, 2010 and 2020. Note that the variability in model responses is approximately constant, which implies also that the relative importance of the model variability increases as pollution levels decrease (Averages over ED countries)
Model variability is driven by systematic biases in the individual models Largest variability in cities than in rural areas, specially for ozone The variability between model responses can be explained in terms of different approaches in the treatment of emissions, the parametrization of vertical exchange and dry deposition processes. Differences in the treatment of biogenic sources and the choice of boundary conditions also play a role.
Model variability driven by systematic biases in the individual models (II) Differences in the treatment of emission of primary pollutants is an important reason why variability between models is largest over populated areas.
Model variability driven by systematic biases in the individual models (III) Source-receptor analysis Differences in the parametrisation of tritation effects is another mean reason for model variability
Model variability driven by systematic biases in the individual models (IV) For the relevant policy indicators, the effect from titration differences is less significant and in 2020, model variability around the ensemble is about (50-60%) for PM2.5 and SOMO35 Ppm.day
Where is the EMEP model with respect to the ensemble? (I) max ensemble EMEP min PM2.5 The EMEP model EMEP model is generally close to the ENSEMBLE both for base case and reduction scenarios in most of the countries
Where is the EMEP model with respect to the ensemble? (II) Example for transect for VOC control of French emissions
Transboundary fluxes (I) The transboundary transport calculations for the different models are mostly consistent for all models for PM2.5
Transboundary fluxes (II) Differences between modelled transboundary fluxes are largest for SOMO35 and ozone Comparison of transboundary calculations for Italian emission reductions for SOMO35. Note that the EMEP model, in green, is close to the ensemble.
Conclusions Systematic differences in the model formulation are the main reason for the variability of responses, and these are quite similar whatever the chemical regime under study. The model presently used to support policy development over Europe, the EMEP model, is robust compared to others whatever the scenario and pollutants studied. This is reassuring for the progress of the negotiations to control air pollution in Europe. The EuroDelta inter comparison illustrates the usefulness of an ensemble approach to test the robustness of model responses to emission changes and to help identifying the reason for systematic biases in the responses from different models
Recommendations for TFMM The identified main reasons for the variability of responses between models are: 1) Differences in the vertical exchange parametrisations: dilution of pollution in the surface layer 2) Differences in the parameterisation of tritation These differences imply that the largest variability between models takes place in urban areas. This has consequences for the interpretation of results from CITY DELTA specially for ozone, and thus requires further study by TFMM. Studies addressed to validate the modelled vertical profiles and to evaluate the reasons for different representations of titration should be prioritized under TFMM In addition, on-going results from ED Phase II, shows that the nitrate and ammonium results from the EMEP model are not close to the ensemble. Here is an area that needs further evaluation !