Interactive comment on “Application of WRF/Chem-MADRID and WRF/Polyphemus in Europe – Part 1: Model description and evaluation of meteorological predictions” by Y. Zhang et al

Major comments: 1) In this study, the authors try to compare an offline-couple model and an online-coupled model. However, the meteorological evaluation only focuses on WRF results which are only used for the offline model. Due to the feedbacks of air pollutants and complex interactions between meteorological variables with pollutants, the online-coupled model should have different meteorological predictions. The differences between meteorology predicted by offline and online models are what I expected to see in the manuscript. In fact, the authors only briefly discuss the difference in T2 and Precipitation between offline and online models in Page 17 Lines 10-20. If the meteorological differences due to aerosol feedbacks are only “decrease near surface temperature and precipitation”, I highly doubt about the merits of this study. It is only an evaluation of WRF outputs, and does not worth it to be published separately on ACP. Furthermore, without assessing the differences between the online and offline meteorology carefully, the comparison of offline model and online model in Part 2 tends to be less interesting: the differences in air pollutant concentrations are due to the air quality models unnecessarily they are offline or online. 2) Why mean bias (MB), the root mean squared error (RMSE), the normalized mean bias (NMB), the normalized mean error (NME), and correlation coefficients are used in this study? Emery et al. (2001) have proposed different statistics merits for different meteorological variables together with benchmarks. Using different statistics merits avoid a possible comparison with the suggested benchmarks and cannot give readers how good the performance is. 3) In addition to the variables evaluated in this study, there are other important meteorological variables having substantial effects to air quality simulations such as solar radiation, PBL height, or the vertical distributions of temperature, humidity, and wind fields. How about the model performance on these meteorological variables? 4) For the time series at specific sites, different selections of sites will lead to different conclusions. Also, from the time series, it is difficult to tell how close the predictions to the observation or which domain has better results. A statistics merit for each variable at each site will help the readers.