Model output statistics (MOS) applied to Copernicus Atmospheric Monitoring Service (CAMS) O-3 forecasts: trade-offs between continuous and categorical skill scores

Air quality (AQ) forecasting systems are usually built upon physics-based numerical models that are affected by a number of uncertainty sources. In order to reduce forecast errors, first and foremost the bias, they are often coupled with model output statistics (MOS) modules. MOS methods are statistical techniques used to correct raw forecasts at surface monitoring station locations, where AQ observations are available. In this study, we investigate the extent to which AQ forecasts can be improved using a variety of MOS methods, including moving average, quantile mapping, Kalman filter, analogs and gradient boosting machine methods, and consider as well the persistence method as a reference. We apply our analysis to the Copernicus Atmospheric Monitoring Service (CAMS) regional ensemble median O-3 forecasts over the Iberian Peninsula during 2018-2019. A key aspect of our study is the evaluation, which is performed using a comprehensive set of continuous and categorical metrics at various timescales, along different lead times and using different meteorological input datasets. Our results show that O-3 forecasts can be substantially improved using such MOS corrections and that improvements go well beyond the correction of the systematic bias. Depending on the timescale and lead time, root mean square errors decreased from 20 %-40 % to 10 %-30 %, while Pearson correlation coefficients increased from 0.7-0.8 to 0.8-0.9. Although the improvement typically affects all lead times, some MOS methods appear more adversely impacted by the lead time. The MOS methods relying on meteorological data were found to provide relatively similar performance with two different meteorological inputs. Importantly, our results also clearly show the trade-offs between continuous and categorical skills and their dependencies on the MOS method. The most sophisticated MOS methods better reproduce O-3 mixing ratios overall, with the lowest errors and highest correlations. However, they are not necessarily the best in predicting the peak O-3 episodes, for which simpler MOS methods can achieve better results. Although the complex impact of MOS methods on the distribution of and variability in raw forecasts can only be comprehended through an extended set of complementary statistical metrics, our study shows that optimally implementing MOS in AQ forecast systems crucially requires selecting the appropriate skill score to be optimized for the forecast application of interest.