More than ten state-of-the-art regional air quality models have been applied as part of the Air Quality Model Evaluation International Initiative (AQMEII). These models were run by twenty independent groups in Europe and North America. Standardised modelling outputs over a full year (2006) from each group have been shared on the web-distributed ENSEMBLE system, which allows for statistical and ensemble analyses to be performed by each group. The estimated ground-level ozone mixing ratios from the models are collectively examined in an ensemble fashion and evaluated against a large set of observations from both continents. The scale of the exercise is unprecedented and offers a unique opportunity to investigate methodologies for generating skilful ensembles of regional air quality models outputs. Despite the remarkable progress of ensemble air quality modelling over the past decade, there are still outstanding questions regarding this technique. Among them, what is the best and most beneficial way to build an ensemble of members? And how should the optimum size of the ensemble be determined in order to capture data variability as well as keeping the error low? These questions are addressed here by looking at optimal ensemble size and quality of the members. The analysis carried out is based on systematic minimization of the model error and is important for performing diagnostic/probabilistic model evaluation. It is shown that the most commonly used multi-model approach, namely the average over all available members, can be outperformed by subsets of members optimally selected in terms of bias, error, and correlation. More importantly, this result does not strictly depend on the skill of the individual members, but may require the inclusion of low-ranking skill-score members. A clustering methodology is applied to discern among members and to build a skilful ensemble based on model association and data clustering, which makes no use of priori knowledge of model skill. Results show that, while the methodology needs further refinement, by optimally selecting the cluster distance and association criteria, this approach can be useful for model applications beyond those strictly related to model evaluation, such as air quality forecasting

continents. The focus for this particular evaluation was meteorological parameters relevant to air 51 quality processes such as transport and mixing, chemistry, and surface fluxes. The unprecedented 52 scale of the exercise (one year, two continents) allowed us to examine the general characteristics of 53 meteorological models' skill and uncertainty. In particular, we found that there was a large 54 variability between models or even model versions in predicting key parameters such as surface 55 shortwave radiation. We also found several systematic model biases such as wind speed 56 overestimations, particularly during stable conditions. We conclude that major challenges still remain 57 in the simulation of meteorology, such as nighttime meteorology and cloud/radiation processes, for 58 air quality simulation. 59 60 61 -3 -

ABSTRACT:In this short note we discuss a long-standing problem in modelling the atmospheric boundary layer (ABL) over complex terrain: namely, an excessive use of the Monin-Obukhov length scale L MO . This issue becomes increasingly relevant with the ever-increasing resolution of numerical weather-prediction and climate models, which typically use L MO in one way or another for parametrizing the surface layer, or at least for formulating the lower boundary conditions. Hence, inevitably, the models under-represent a significant part of the mesoscale flow variability.We focus here on the stable ABL over land: in particular, sloped cooled flows. However, a qualitatively similar reasoning applies to the corresponding unstable ABL. We show that for sufficiently stratified flows over moderately sloped surfaces, Monin-Obukhov scaling is inadequate for describing the basic ABL dynamics, which is often governed by katabatic and drainage flows.

Abstract. This paper introduces two changes of the turbulence parameterization for the EMEP (European Monitoring and Evaluation Programme) Eulerian air pollution model: the replacement of the Blackadar in stable and O'Brien in unstable turbulence formulations with an analytical vertical diffusion profile (K(z)) called Grisogono, and a different mixing height determination, based on a bulk Richardson number formulation (RiB). The operational or standard (STD) and proposed new parameterization for eddy diffusivity have been validated in all stability conditions against the observed daily surface nitrogen dioxide (NO2), sulphur dioxide (SO2) and sulphate (SO42−) concentrations at different EMEP stations during the year 2001. A moderate improvement in the correlation coefficient and bias for NO2 and SO2 and a slight improvement for sulphate is found for the most of the analyzed stations with the Grisogono K(z) scheme, which is recommended for further application due to its scientific and technical advantages. The newly extended approach for the mechanical eddy diffusivity is applied to the Large Eddy Simulation data focusing at the bulk properties of the neutral and stable atmospheric boundary layer. A summary and extension of the previous work on the empirical coefficients in neutral and stable conditions is provided with the recommendations to the further model development. Special emphasis is given to the representation of the ABL in order to capture the vertical transport and dispersion of the atmospheric air pollution. Two different schemes for the ABL height determination are evaluated against the radiosounding data in January and July 2001, and against the data from the Cabauw tower, the Netherlands, for the same year. The validation of the ABL parameterizations has shown that the EMEP model is able to reproduce spatial and temporal mixing height variability. Improvements are identified especially in stable conditions with the new ABL height scheme based on the RiB number.

Dispersion models require hourly values of the mixing height (H) that indicates the existence and vertical extent of turbulent mixing. Urban areas, which are usually industrial areas too, have H higher than rural areas, and commonly used methods for deriving H should not be applied under the same conditions as in homogeneous conditions. The bulk Richardson number (RiB) method was applied to determine H over Zagreb, Croatia. Impact of urban areas on the choice of critical values of bulk Richardson number (RiBc) was explored, and different values were used for convective boundary layer (CBL) and for stable boundary layer (SBL). Aire Limitee Adaptation Dynamique development InterNational (ALADIN), a limited area numerical weather prediction (NWP) model for short‐range 48‐h forecasts, was used to provide one set of input parameters. Another input set comes from radio soundings. The values of H, modelled and based on compared measurements, and the correlation coefficient as well as standard deviation and bias were calculated on a large data set to determine RiBc ranges applicable in urban areas. It is shown that RiB method can be used in urban areas, and that urban RiBc should have certain limitations despite of a wide spectrum of practical values used today. Significantly increased RiBc values in SBL were determined from the NWP and soundings data, which is the consequence of increased surface roughness in the urban area. The verification of ALADIN through the determination of H was also done. Decoupling from the surface in the very SBL was detected as a consequence of the flow cease resulting in RiB becoming very large.