Dear Dr. Garrett, We Have Addressed the Comments from Reviewers 1 and 2 and Incorporated Changes Into a Revised Manuscript. Plea
Total Page:16
File Type:pdf, Size:1020Kb
Dear Dr. Garrett, We have addressed the comments from Reviewers 1 and 2 and incorporated changes into a revised manuscript. Please find the referees’ comments, our point-by-point responses, and a marked up version of the manuscript below. The text in black are the replies uploaded to ACPD. During incorporation of the revisions, we change our minds on a few points. These changes are noted in red text, as are indications for where to find the related changes in the manuscript. Thank you for considering this manuscript for publication in ACP. Thanks, Natalie Wagenbrenner Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-761-RC1, 2016 ACPD © Author(s) 2016. CC-BY 3.0 License. Interactive comment Interactive comment on “Downscaling surface wind predictions from numerical weather prediction models in complex terrain with WindNinja” by N. S. Wagenbrenner et al. Anonymous Referee #1 Received and published: 21 February 2016 Review This is very well-written and well-constructed, scientifically sound paper that is of ben- efit to the scientific community. I don’t have any major issues with the paper as-is. However, some discussion of the following issues would strengthen the paper in my opinion: 1. For Figures 1, and 6-8 it would be really helpful to have the Big Southern Butte area Printer-friendly version zoomed in more. I have a hard time seeing what is going on in that region and then the rest of the plots are mostly empty with a few observations scattered downstream. Discussion paper What I would recommend would be to have a zoomed-in plot of the Big Southern Butte area. C1 2. On the same note, it would be helpful if sensor R2, R26 discussed in Fig. 4 were indicated on the map in Fig. 1. ACPD 3. Why is the only meteorological parameter discussed in this paper wind? Is there no parallel for downscaling for other critical fire parameter: relative humidity and temper- Interactive ature? For the general reader some discussion of this in the intro would be helpful for comment the more general audience of ACP. 4. I am under the impression that strongly-forced wind events are the most important factor for high fire spread, and therefore the strengths of this study play to that need. I think this point should be made somewhere, even if it is fairly obvious. 5. WRF-HRRR fields are now available to drive the 1.33 km model. Were any tests conducted to determine the improvement in WRF using HRRR analyses to drive the 1.33 km model instead of NARR? Could this be included as potential future work? 6. I would separate the discussion more clearly into ‘externally forced flows’ (large- scale winds) and locally-forced flows (upslope and downslope). The current text dis- cusses weak external forcing in many places; my opinion is that it would be better to distinguish the two cases by referring to ‘locally-forced’ more when the external forcing is weak. 7. Why was a ‘quasi-large-eddy simulation’ (200-400 m horizontal grid spacing) not conducted with WRF and compared to the other runs? This simply requires turning off the PBL scheme and a few other parameters (WRF-LES settings). A lack of numerical resources is a good argument. While the argument is made that these simulations are not feasible in real-time, the 1.33 km runs are also not feasible in real-time. If the LES run was conducted, then a comparison could be made such as ‘the downscaling Printer-friendly version is equivalently good as the quasi-LES run over the area where the terrain structure is explicitly simulated. There may be issues with CFL criteria with high resolution to- Discussion paper pography in the simulations, but many investigators are starting to use WRF as a LES model with the PBL turned off:::would be curious how these results would compare C2 to the downscaled. In any case, a sentence or two in future work discussing how this could be done in the future would be good. ACPD 8. The potential limitations of the slope flow parameterization and discussion of a com- panion paper is looking at this is discussed but in a rather scattered manner. If a sum- Interactive mary of the strengths and weaknesses of WindNinja were included in the introduction comment this would help make the findings in the body of the paper flow better. 9. The weakness with windninja simulating lee-side flows is an important point, as many fires occur on the ‘downslope’ side of mountain ranges. Does this weakness also happen during strongly forced conditions? It was not clear to me. Interactive comment on Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-761, 2016. Printer-friendly version Discussion paper C3 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-761-RC2, 2016 ACPD © Author(s) 2016. CC-BY 3.0 License. Interactive comment Interactive comment on “Downscaling surface wind predictions from numerical weather prediction models in complex terrain with WindNinja” by N. S. Wagenbrenner et al. Anonymous Referee #2 Received and published: 23 February 2016 The authors compare different WRF simulations and a diagnostic model against wind observations over an isolated mountain. Spacial emphasis is given to quantify the added value of downscaling the WRF simulations with the diagnostic model (Wind- Ninja). The topic is relevant to progress in our understanding of the ability of mesoscale simulations, and diagnostic models, to reproduce the wind over complex terrain. The observational dataset is quite unique. The article is well written and the conclusions are supported by the results. Perhaps it could be highlighted that none of the WRF Printer-friendly version simulations is able to represent the mountain. Only the simulation at 1.33 km of hori- zontal resolution starts to “see” something (fig. 1). More specific comments, all of them Discussion paper of minor character, are provided below. C1 I therefore recommend acceptance of the manuscript pending to minor revisions ACPD SPECIFIC COMMENTS 1. Page 5, line 9: Please define the FDM acronym. 2. Page 5, line 25: I believe you are using the Kain-Fritsch scheme to represent the Interactive cumulus in WRF. If so, please update the reference. comment 3. Page 6, line 4: The first model level of WRF-UWis 40m a.g.l. Later in the manuscript it is mentioned that these winds are interpolated to 3 m using a log profile. Is this inter- polation taking into account the atmospheric stability? Or the atmosphere is assumed to be neutral? I say this because the target height, 3 m a.g.l, is far from 40 m a.g.l. and the interpolation method could play a major role in the evaluation. 4. Page 7, line 7: The NAM model has the first level at 200 m and the horizontal resolution is 12 km. NAM is obviously not representing the mountain at all. RAP provides information of the winds in the valley. It would be informative to have some discussion of this here pointing out why NAM is used even though the mountain is not represented at all. 5. Page 13: liens 11-15: None of the WRF simulations have enough horizontal reso- lution to represent the mountain. Only at 1.3 km WRF starts to see something. The WRF predictability probably comes from the simulation of the flow in the Snake River Plain. It may be interesting to compute anomalies with respect to the mean flow and compare them with the equivalent observations to see if there is any benefit as a result of increasing the horizontal resolution. 6. Page 14: It will be good to show a spatial plot showing the under prediction of windward and ridge top locations. Printer-friendly version 7. Page 18, first paragraph Summary: You should mention that the horizontal resolu- Discussion paper tions are too coarse to represent the mountain in the WRF simulations. C2 Interactive comment on Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-761, 2016. ACPD Interactive comment Printer-friendly version Discussion paper C3 1 Reply to Reviewer 1 2 1. Figures 1 and 6-8 will be updated to show a zoomed in version of the butte. Zooming in on the butte 3 would be nice for the left panel of Figure 1, however, it’s not really possible for the other three panels, 4 since the butte is represented by just one or two pixels; therefore, we chose to leave these figures as is. 5 Additionally, these figures depict the domain extent used in our downscaling simulations and so there is 6 value in leaving extents in the figures as is. 7 2. R2 and R26 will be added to Figure 1. Added in Figure 1, p. 34. 8 3. The diagnostic model evaluated in this paper, WindNinja, is only designed to downscale the flow. 9 WindNinja includes physics for modeling the mechanical and thermal effects of the terrain on the flow 10 field. WindNinja is capable of interpolating other parameters (e.g., temperature and relative humidity) 11 to a finer grid, but does not provide any additional physics (e.g., conservation of energy) or 12 parameterizations to simulate terrain effects on these parameters. For these reasons, WindNinja does 13 not output additional downscaled weather parameters. Additionally, wind varies more spatially than 14 temperature and RH, so is more important to predict at a high resolution. Wind is known to often be the 15 driving environmental variable for wildfire spread and behavior. We will clarify these points in the 16 paper. Some discussion on this was added in lines 101-104.