Insights into the deterministic skill of air quality ensembles from the analysis of AQMEII data

Abstract

Abstract. Simulations from chemical weather models are subject to uncertainties in the input data (e.g. emission inventory, initial and boundary conditions) as well as those intrinsic to the model (e.g. physical parameterization, chemical mechanism). Multi-model ensembles can improve the forecast skill, provided that certain mathematical conditions are fulfilled. In this work, four ensemble methods were applied to two different datasets, and their performance was compared for ozone (O 3 ), nitrogen dioxide (NO 2 ) and particulate matter (PM 10 ). Apart from the unconditional ensemble average, the approach behind the other three methods relies on adding optimum weights to members or constraining the ensemble to those members that meet certain conditions in time or frequency domain. The two different datasets were created for the first and second phase of the Air Quality Model Evaluation International Initiative (AQMEII). The methods are evaluated against ground level observations collected from the EMEP (European Monitoring and Evaluation Programme) and AirBase databases. The goal of the study is to quantify to what extent we can extract predictable signals from an ensemble with superior skill over the single models and the ensemble mean. Verification statistics show that the deterministic models simulate better O 3 than NO 2 and PM 10 , linked to different levels of complexity in the represented processes. The unconditional ensemble mean achieves higher skill compared to each station's best deterministic model at no more than 60 % of the sites, indicating a combination of members with unbalanced skill difference and error dependence for the rest. The promotion of the right amount of accuracy and diversity within the ensemble results in an average additional skill of up to 31 % compared to using the full ensemble in an unconditional way. The skill improvements were higher for O 3 and lower for PM 10 , associated with the extent of potential changes in the joint distribution of accuracy and diversity in the ensembles. The skill enhancement was superior using the weighting scheme, but the training period required to acquire representative weights was longer compared to the sub-selecting schemes. Further development of the method is discussed in the conclusion.

Two-scale multi-model ensemble: is a hybrid ensemble of opportunity telling us more?

“…As presented in the past, the diversity of modelling approaches is the element that favours a better ensemble product (Kioutsioukis and Galmarini, 2014;Kioutsioukis et al, 2016). In this sense the combination of model results that focus on different scales and that account in a different form for the chemical mechanism has the potential to increase the value of an ensemble to which we will refer from now on as the hybrid ensemble.…”

“…In this study, we would also like to build upon the research performed in other multi-model ensembles over the years; rather than calculating only the classical model average or median ensemble (mme), we shall also calculate three ensembles based on the findings of Potempski and Galmarini (2009), Riccio et al (2012), Solazzo et al (2012aSolazzo et al ( , b, 2013, Galmarini et al (2013), and Kioutsioukis and Galmarini (2014). We shall therefore refer to the ensemble made by the optimal subset of models that produce the minimum RMSE as mmeS (Solazzo et al, 2012a, b); the ensemble produced by filtering measurements and all model results using the Kolmogorov-Zurbenko decomposition presented earlier and recombining the four components that best compare with the observed components into a new model set as kzFO ; and the optimally weighted combination as mmeW (Potempski and Galmarini, 2009;Kioutsioukis and Galmarini, 2014;Kioutsioukis et al, 2016). Figure 6 shows the effect of the various ensemble treatments for the two groups of models separately and presented as a Taylor diagram.…”

“…Figure 3 presents two classical skill scores for categorical events also applied by Kioutsioukis et al (2016), namely the probability of detection (POD) and false-alarm rate (FAR). The former represents the proportion of occurrences (e.g.…”

“…Therefore, meso-, regionaland global-scale model results are grouped in ensembles according to their scale of pertinence. In air quality studies, this has been the case for example in Fiore et al (2009), Solazzo et al (2012a, b), Kioutsioukis and Galmarini (2014), and Kioutsioukis et al (2016). Colette et al (2012) analysed, as part of an analysis of the exposure in Europe, results from an ensemble of opportunity of a total of six models, three of which were global and three regional.…”

Advanced error diagnostics of the CMAQ and Chimere modelling systems within the AQMEII3 model evaluation framework

“…The scientific problem of assessing the quality of a modelling system for air quality is tackled by Dennis et al (2010) who distinguish four complementary approaches to support model evaluation -operational, probabilistic, dynamic, and diagnostic -which are also the four founding pillars of AQMEII. Several studies performed under AQMEII have focused on the operational and probabilistic evaluation (Solazzo et al, 2012a(Solazzo et al, , b, 2013Im et al, 2015a, b;Appel et al, 2012;Vautard et al, 2012) and more recently efforts have been expanded to the diagnostic aspect (Hogrefe et al, 2014;Solazzo and Galmarini, 2016;Kioutsioukis et al, 2016;Solazzo et al, 2017).…”