How an uncertain short-wave perturbation on the North Atlantic wave guide affects the forecast of an intense Mediterranean cyclone (Medicane Zorbas)

Mediterranean cyclogenesis is known to be frequently linked to ridge building over the North Atlantic and subsequent anticyclonic Rossby wave breaking over Europe. But understanding of how this linkage affects the mediumrange forecast uncertainty of Mediterranean cyclones is limited, as previous predictability studies have mainly focused on the relatively rare cases of Mediterranean cyclogenesis preceded by upstream extratropical transition of tropical cyclones. This study exploits a European Centre for MediumRange Weather Forecasts (ECMWF) operational ensemble forecast with an uncertain potential vorticity (PV) streamer position over the Mediterranean that, 3 d after initialization, resulted in an uncertain development of the Mediterranean tropical-like cyclone (Medicane) Zorbas in September 2018. Later initializations showed substantially lower forecast uncertainties over the Mediterranean. An ad hoc clustering of the ensemble members according to the PV streamer position in the Mediterranean is used to study the upstream evolution of the synoptic to mesoscale forecast uncertainties. Cluster differences show that forecast uncertainties were amplified on the stratospheric side of a jet streak over the North Atlantic during the first day of the ensemble prediction. Subsequently, they propagated downstream and were further amplified within a short-wave perturbation along the wave guide, superimposed onto the large-scale Rossby wave pattern. After 3 d, the uncertainties reached the Mediterranean, where they resulted in a large spread in the position of the PV streamer. These uncertainties further translated into uncertainties in the position and thermal structure of the Mediterranean cyclone. In particular, the eastward displacement of the PV streamer in more than a third of the ensemble members resulted in a very different cyclone scenario. In this scenario, cyclogenesis occurred earlier than in the other members in connection to a pre-existing surface trough over the Levantine Sea. These cyclones did not develop the deep warm core typical of medicanes. It is proposed that the eastward-shifted cyclogenesis resulted in reduced values of low-level equivalent potential temperature in the cyclogenesis area. As a result, latent heating was not intense and deep enough to erode the upper-level PV anomaly and allow the formation of a deep warm core. The westward displacement led to surface cyclones that were too weak, and a medicane formed in only half of the members. The central, i.e. correct, PV streamer position resulted in the most accurate forecasts with a strong medicane in most members. This study is the first that explicitly investigates the impact of PV streamer position uncertainty for medicane development. Overall, results extend current knowledge of the role of upstream uncertainties in the medium-range predictability and unsteady forecast behavior of Mediterranean cyclones including medicanes.


Background
This is a massively improved manuscript. The authors have clearly taken the time that they needed to revise their approach to the study and present the material in a clear and logical way. The result is a document that is fun to read and easy to draw conclusions from.
I have one general concern regarding the robustness of the storyline for different (primarily earlier) initialization times, and a few specific questions that may require some adjustments to the analysis or text to address (notably Specific Comments # 28, 36, 37 and 59). However, once the authors have completed a set of relatively minor revisions in response to these remaining issues, I look forward to seeing this article in print. I truly appreciate the efforts that the authors have made to bring this manuscript up to a very worthy standard. [L57] Suggest simplifying to "This cyclone belonged to a special class …". 9.
[L62] The note that "sometimes they acquire the typical appearance of a hurricane" suggests that sometimes medicanes don't acquire hurricane-like characteristics. Given that medicanes are typically identified by morphology, this seems unlikely and therefore the statement should be strengthened. 11. [L73] Remove comma after "during ET". 12. [L97-107] This paragraph is too general to be particularly useful, and contains a defense of the experimental design that would be better placed in a data/methods section. It also disrupts the flow between the "study objectives" paragraph and the "outline" paragraph at the end of the introduction. I think that this paragraph should be removed. 13. [L131-L133] The TRMM mission ended in 2015, replaced by the GPM. Are the 2018 data used here processed by the legacy TRMM algorithm or the newer IMERG algorithm? The text here may be completely correct, but please confirm. 14.
[L138-L139] The idea of intersection points between a trajectory and a layer is a bit strange. I would ordinarily have thought of a trajectory line intersecting a 2D plane (e.g. the 325 K surface). Doesn't the use of the 322.5 K lower boundary for the layer simply imply that the points are defined as the locations at which the ascending WCB trajectories cross the 322.5 K surface? The Fig. 1 caption suggests that the simpler 325 K surface definition is used, at odds with the description here. The Fig. 2 caption suggests that the 322.5 K surface is used instead. I realize that the difference between these is relatively small, but consistency is desirable. 15. [L140] Suggest "around the cyclogenesis position". 16.
[L148] This use of the CPS does not seem to distinguish between convection-and seclusioninduced warm cores. Is there any risk that some of the members classified as medicanes by the CPS are the result of frontal seclusion, a process that is not associated with tropical transition? Might B be useful to distinguish between storms at the early stages of seclusion and those whose warm cores are diabatically generated? 17. [L172-L182] This is an excellent description of your testing technique: very well done. 18. [L184] Suggest reversing "provides first". 19. [Fig. 1] The PV streamer has already formed in (d), so the phrase "before the formation of the PV streamer over the Mediterranean" should be removed or replaced with something like, "during the development of the PV streamer". 20. [Fig. 2] The contour interval for SLP should be noted in the caption (it looks like 4 hPa). The fact that 500 hPa height is shown in yellow contours should be noted, as should the contour interval used for this field.
21. [Fig. 2] Suggest adding "322.5 K isentropic level" to avoid confusion with the equivalent potential temperature that serves as the departure threshold. 22.
[L206] Suggest inverting to read, "The PV streamer broke up at the time of cyclogenesis, resulting in the …". 23.
[L209] Suggest referring to Fig. 3a for locations to help with this discussion. 25. [L245] "Landfall" is usually written as a single word. 26. [L248] The phrase "about one-day period" is approximate and slightly strange. Was this an 18-h period? The exact length doesn't matter very much here, so the phrase could simply read "… prior to cyclogenesis, the initial cyclone intensification, and the formation …". 27. [Fig. 4] A greyscale bar should be included for the brightness temperatures. 28. [L267] To be "substantial" this spread would need to be larger than typical ensemble spread at these lead times. If such a typical spread is known, it would be useful to add this line to Fig. 5 for reference. If it is not (and is not readily computed), then this sentence could be restructured to focus on the decrease in spread in the medium range. 29. [L287] Suggest removing "a" before "substantial". not mean that moist diabatic processes are not relevant to the uncertainty. I've included a satellite retrieval (Fig. 1 below) that shows extensive high-topped cloud cover in the region. In addition to potential latent heating effects, these clouds will affect the radiative heating profile. If the clouds are handled differently by the members in the different composite groups, they could explain the differences in the solutions rather than "dry upper-tropospheric dynamics" (L338). Imagine, for example, that members in which the local SPPT coefficients suppress the radiative heating tendencies do not amplify the ridge while those in which the coefficients amplify the heating signal create a much more robust ridge. If this is truly the sensitive region for the Mediterranean streamer, this difference in heating could cause the eventual separation of the solutions that is observed. I haven't demonstrated that any of this is true, but the analysis in the manuscript does not rule out this possibility despite the assertions in the text. 37. [L346 and L362-367] This is a very 2D way of describing the tropopause evolution. What if you considered the "approach" of the high-PV contours as a steepening of the tropopause or development of a tropopause fold. This is of course necessarily related to the jet streak, but might give the readers a useful way to conceptualize the process that's promoting wave amplification in some members. For example, the approach of a northern stream PV perturbation towards the jet seems dynamically similar to the events discussed by Winters and Martin (2017 Fig. 1d. I think that's good because it's definitely a PV streamer rather than a trough at this time; however, the reference needs to be corrected here. Also, there is no S2 in Fig. 1c. Is it possible that the panel references are inverted here? 40. [L368-370] As in Comment #28, it would be useful to have a spread climatology in Fig. 5 for reference so that it is clear that this spread reduction is larger than would be expected by the decreasing lead time. 41. [L376-L379] There is also a notable difference in streamer tilt, with W more positively tilted and E notably more meridional (note the westward shift of the high-PV region over Eastern Europe in the E cluster). This suggests that the streamer may be in a slightly different stage of its life cycle in the different groups. 42. [Fig. 11] If it doesn't take too much effort, you could consider masking out significance regions that are too small to be meaningful. This would help to clean up the otherwise-beautiful plots a bit without any loss of important information. 43.
[L413] Remove dash after "upper-" (this is not a compound adjective). 45. [Fig. 12] I really like Fig. 12 and the associated discussion, but the precipitation panel should appear as its own figure (Fig. 13). There is no association of either axis between panels a and b with panel c, so placing them in a single figure is not beneficial. 46.
[L444] Change "much lower" to "much reduced" to avoid ambiguity with precipitation fluxes in the vertical. 51. [L445] Change "indicating much lower latent heating" to "indicative of a reduction in columnintegrated latent heating". 52. [L445-L446] Without more context or an analysis of why this precipitation bias exists, this does not seem like an "interesting side remark". I think that this sentence should be removed to maintain the focus of this discussion. 53. [L447] Suggest changing "pathway" to "evolution". 54.
[L447] Unless WCD uses the APA style, I do not think that the word following a colon should be capitalized. 55. [L451] Consider replacing "… this storyline and they are mostly …" with "… this archetype, mostly …". 56.
[L459] Suggest "… were used to assess how uncertainties originating in a short-wave perturbation on the North Atlantic wave guide influenced a downstream PV stream and, as a result, …". 57. [L461] Suggest inverting "appeared first" and changing "at" to "on".