Heuristic and Bayesian Tornado Prediction in Complex Terrain of Southern Wyoming
Round 1
Reviewer 1 Report
Title: Heuristic and Bayesian Tornado Prediction in Complex Terrain of Southern Wyoming
Author: Thomas A. Andretta
Recommendation: Major Revision
Summary: The study aims to propose a heuristic and bayesian approach to make a real-time tornado forecast where numerical weather prediction (NWP) models do not capture the spatial and temporal details. As a result, it is argued that the tornado can be predicted accurately by using “surface predictors that include moisture convergence axes associated with dryline segments and the Jacobian determinant of the horizontal wind illustrating flow shearing and stretching into the updrafts of the storms.” Proposed techniques were studied using dense real-time in-situ surface mesonetwork and dual-Doppler weather radar observations of Wyoming.
This paper has the merit of publishing in Meteorology because predicting accurate time and space of tornado occurrence, especially in complex topography, is a significant problem. However, the conceptualization of the paper lacks the main issues related to the accurate prediction of the tornado enclosing both time and space if this is the study’s primary goal, which is not indicated enough. On the one hand, if the main goal is the accurate prediction of the timing of the tornado, then using a dense real-time in-situ surface mesonetwork and dual-Doppler weather radar observations may not be enough from the previous time steps. On the other hand, if the primary purpose of the study is to determine the high possibility locations for tornado occurrences only based on seasonal event climatology, then four cases between the years 2018 and 2019 may not be enough for the determination of the tornado climatology where we need at least 30-year events. Furthermore, since the exact tornado occurrence locations were not provided in the graphs for the case studies, we may not evaluate the success of the proposed methodology.
Moreover, the tornado occurrence-related assumptions made in the study’s null hypothesis, “Horizontally thin sheets of saturated air within the 0- to 2-km AGL layer are stretched and sheared along moisture convergence axes associated with a dryline near one or more severe thunderstorms,” were not addressed convincingly.
I recommend Major Revision, and the recommendations for improving the manuscript are as follows:
Major Comments:
- The summary should be revised.
- Current operational prediction methodologies should be added to the Introduction.
- It should be discussed how much the method contributes to operational prediction.
- The aims of the article should be clearly stated.
- The reason for the null hypothesis should be discussed.
- The exact locations of the occurred tornadoes for each case should be included in the graphs.
- Graphs should be placed near the first mentioned paragraph.
- The importance of the radar Storm Relative Velocity at 0.5 degrees and Long Range Base Reflectivity at 0.5 degrees in tornado prediction should be explained in detail.
- The details of the Bayesian approach to finding the stretching and shearing probabilities should be provided.
- References should be revised.
Minor Comments:
- Include equation numbers and cite appropriately, please.
- Line 51: the reference number of the Jacobian Matrix number should be [7] instead of [8].
- Line 167: What is the moisture flux convergence amount for each case?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
I thought this was an interesting and novel approach to tornado forecasting.
Line 28-29: Can you clarify here? I think you are saying that 81% of tornadoes in this specific county occur from May through July?
Figures, generally speaking: It is difficult to interpret some of the features on the maps because there are a lot of features on them. Perhaps for Figures 2 through 9 the topographic background could be eliminated? I think it's harder to see the features when you have green shading on gray background (as in Figure 2) vs. what it might look like on a white background?
In addition, there is no legend for Figure 2 and Figure 9 - it is especially important for Figure 9 where the probabilities are presented. I can see that the contours are labeled when I really zoom in but a legend would make it much easier to interpret.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper is devoted to heuristic tornado forecasting based on the probabilities of seasonal climatology and mesoscale meteorological predictors using in-situ surface mesonetwork and dual-Doppler weather radar observations.
Four cases of high-altitude moderate to severe tornadoes in southern Wyoming are explored.
I have following comments and remarks:
1) Literature review on the research topic is very poor
2) It is declared that - heuristic methodology assumes that the lowest elevation scan of radar data (0.5 degrees) is representative enough to detect a low-level tornado vortex signature.
However it seems not very correct for distant targets, say 120 - 200 km from radar, where radar beam remotes from earth on 1 km and more. The distance parameter should also be accounted or at least the phrase defining the applicability of 0.5 deg.elev should be added. Also the optimal range of ground layer thicknesses to detect mesocyclone should be mentioned somewhere in the hypotheses;
3) It is indicated that the observation data grid has a resolution along the x and y axes of about 3 km. Can such a resolution "catch" a local pressure drop at which a mesocyclone and a tornado can form? One must also take into account the fact that the observation points themselves can be located much more than this distance, and the grid is obtained by interpolation of neighboring points.
4) it is also not entirely clear why the radar scanning angle of 0.5 degrees was chosen, and not the Cappi horizontal level or other radar product.
5) Despite the case study is very valuable itself, in the introduction and discussion, I would like to see briefly what methods of tornado forecasting and detection are existed, and the rationale for why the proposed method is better or in what way it complements them.
6) It is interesting how the results of the analysis and the criteria obtained on the basis of Bayes' theorem can change if several dozen cases of tornadoes are processed, including those under other orographic conditions?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
The article has been revised according to our comments. Of course, there are still questions. In particular, how will the criteria of the developed methodology behave when not 4 but 40 - 100 cases of tornadoes are analyzed? We have no other objections. We recommend the current version for publication.