One concern from the reviewers was our use of the MYJ planetary boundary layer scheme. They pointed to evidence that the MYNN scheme was superior, and may improve our results, particularly in the timing of a lake breeze passage on June 18, 2016. We declined to re-run the simulation using the MYNN due to computing resources and time.
I have since re-ran the simulation with the MYNN and found the results are very similar with the MYJ. Below are time series graphs for the stations presented in the paper with the temperature and wind with the MYNN model run overlaid in magenta. The only change between the "WRF" and "MYNN" model data is that the MYNN uses the MYNN PBL scheme instead of the MYJ PBL scheme.
Color and station name denote the observed temperature and wind at the station location. Black dashed is the original WRF simulation, magenta dashed is the same simulation except uses MYNN boundary layer physics.
The cause of the observed delay in the lake breeze progression is likely caused by stronger meridional wind component below 5,000 meters (550 mb). In the vertical profiles below, you can see the opposing southerly winds at the Salt Lake City Airport were from the direct south at 10 m/s below an inversion layer at 5,000 m. The HRRR and WRF with MYNN PBL scheme had slightly weaker winds and more from the southwest. This subtle difference is likely the primary reason for the delayed lake breeze on the afternoon of 18 June 2016. The next question is, "how do we fix that subtlety in the model." I'm not sure I know the answer to that.
Just as a comparison between 3 km and 1 km domains: the lake breeze is much less "sharp" in the outermost domain, run at 3 km (right), than the inner domain run at 1 km (right).
|Domain 2 with 1 km grid spacing|
|Domain 1 with 3 km grid spacing|