| 31 Jul 2026
A Predictability Pathway from Franklin to Daniel: Linking Extratropical Transition, European Blocking, and Devastating Mediterranean Cyclone Daniel
Abstract. High-impact weather occurring during periods of low forecast skill can lead to severe socio-economic consequences. This study examines the predictability of the causal chain of events leading to the formation of Medicane Daniel and its associated flooding impacts over Greece on 5 September and Northeast Libya on 10–11 September 2023. Daniel’s formation was preceded by Hurricane Franklin’s extratropical transition over the North Atlantic on 1 September 2023, the development of an Omega-type block over central Europe on 3 September 2023, and the detachment of a PV streamer over the eastern Mediterranean which led to Daniel’s formation as a cut-off low on 5 September 2023. Using a novel block error metric, we investigate the predictability of these events using operational ECMWF and Met Office ensemble forecasts and ERA5 reanalysis.
Ensemble sensitivity analysis identifies Franklin’s location as a leading source of downstream uncertainty. Accurate prediction of Franklin’s track was critical for developing an upper-level ridge and European blocking. In ensemble members that best represented the European block, Franklin remained anchored to a downstream ridge, excited a Rossby wave packet over Europe, and initiated a wave breaking event that produced the cut-off leading to the development of Storm Daniel. Two distinct high error recurvature scenarios are identified across the forecasting systems, both of which lead to poor representation of blocking and subsequently to no Mediterranean cyclone development. The strength of Franklin's interaction with the midlatitude flow, and therefore the downstream impact of extratropical transition, is highly sensitive to small track errors at short forecast lead times. Together, these results demonstrate how small upstream errors can contribute to poor predictability of high-impact weather, reinforcing the need for improved representation of the physical processes that take place during tropical–extratropical interactions in operational forecasting to better protect vulnerable regions from impactful weather events.
This submission shows a lot of promise and makes some good arguments supporting the idea that small errors in the track and behavior of Franklin’s extratropical transition led to large errors downstream blocking and evolution of Daniel. This kind of case study is important and appropriate. However, there are issues that need addressing before publication. The central questions need to be answered in the conclusion and the second question needs to be addressed much more thoroughly. The choice of sensitivity measurement and questions about field significance testing must also be addressed. The rest of the review is broken into major and minor specific comments and a list of suggestions for making the figures clearer.
Major Comments:
The central questions enumerated in lines 96 – 98 are never explicitly answered. I suggest tying the existing discussion explicitly back to central question (i). Discussion of the forecast skill of the HIW over Europe is somewhat lacking, central question (ii). My understanding from your introduction is that the main impact from the period of study was heavy rainfall and severe flooding in Greece and Bulgaria but the relationship between the uncertainties in Franklin’s ET and the rainfall in Greece and Bulgaria is not. As I see it, your argument for the relationship between the position and timing of Franklin’s ET and the accuracy of the forecasted omega block is good and convincing (but could be highlighted more, see comment below). However, completing the investigation by examining the precipitation forecast skill’s relationship to the omega block and following the chain from ET to block to precipitation would really make this paper shine. I understand this an additional analysis but I think that checking the connection between the high and low ZALO members and precipitation skill would solidify your arguments and
Line 446 – I think the last sentence in this paragraph “It should be noted that Franklin’s initial…” is a very clear statement of the central idea of the paper, at least as I see it. This should really be given more discussion and be the main point supported by figure 10 in conjunction with figures 8 and 9.
Line 206 – I don’t see that any choice of r2min can be said to “…retain only reliable sensitivities…” The implementation of alpha damps the value of S for grid points with low correlations coefficients without removing them. It’s also unclear why 0.15 and not 0.25 or any other r2 value. I checked the reference but no thorough justification is provided there. I think readers might reasonably wonder at 0.15 being quite a low bar to clear for a sensitivity to be reliable. One way to address this concern would be to test for spurious correlations using a statistical test as in Ancell and Hakim (2007). I also think a discussion of the drawbacks or limitations of this mode of analysis would be very helpful for readers (e.g. How does it deal with non-linear responses if your assumption of linearity breaks down)
Ancell, B., and G. J. Hakim, 2007: Comparing Adjoint- and Ensemble-Sensitivity Analysis with Applications to Observation Targeting. Mon. Wea. Rev., 135, 4117–4134, https://doi.org/10.1175/2007MWR1904.1.
Line 352 – I don’t see the region of positive sensitivity you discuss. Is it covered up by the stars? Can you comment on the interpretability of one, presumably small area of sensitivity amongst many of similar size and magnitude throughout the Atlantic? I also have trouble locating a “defined dipole” around Franklin in figure 6b for similar reasons. In fact, without any statistical test on the significance of these sensitivity values I have trouble believing that much of the sensitivity of ZALO to PV320 is useful to interpret. The main signals I see occur in the gaps between the 2PVU contours not really near Franklin. I wonder if this whole section about the sensitivity to PV320 could be removed as I’m not sure what it brings that hasn’t already been discussed in other figures or the sensitivity to Z500?
Minor Comments:
Line 27 – ECMWF should be defined before first use
Line 38 – ET is already defined in line 30 as “extratropical transition”
Line 85 – I think you mean something about the relative frequency of RWB and ET making the attribution difficult? Could you clarify what is meant in the sentence beginning “Climatological studies have…”
Line 119 – Are there other predictability characteristics you mean besides poor predictability at longer than 96 hour lead times?
Line 121 – Is “this” the interaction of Hurricane Franklin with the waveguide or accurate block onset? The sentence in question has a lot going on, I recommend expanding to multiple sentences and being explicit.
Line 125 – I believe that the “E” in ERA5 stands in for ECMWF not just European.
Line 127 – I’d appreciate some additional comment on the choice to coarsen the horizontal resolution. Also, why are there parenthesis around 0.5?
Line 135 – Did you regrid the data from its native resolution to 0.5 degrees? If so, why? Some discussion of your regridding choices would be appreciated.
Line 159 – Is this constraint on Z’ found to be sufficiently large in some previous work you could cite?
Line 162 – I don’t think you need to tell readers that ZALO is similar to SAL and PAL when you don’t explain how SAL and PAL are formulated. It makes things a little confusing.
Line 174 – Do I understand correctly that the L in ZALO is directly affected by the choice of domain, through the length of the diagonal? If that’s the case, did you do any testing of different domain sizes to learn more about how sensitive L and ZALO are to the choice of domain?
Line 221 – I think you should add the word “both” after MOGREPS-G in this line to clarify that you mean each ensemble under and over predicts intensity rather than ECMWF ENS under predicts and MOGREPS-G overpredicts.
Line 223 – I’m not sure figure 2 clearly shows that L is “generally small” especially relative to Z error. If this is a point you want to make it would be helpful to show the components of ZALO on plots with the same scale of Y-Axis (you could center it differently depending on the spread of the individual metric)
Line 226 – the sentence beginning “In the combined ZALO metric…” could mean a couple of things about figure 2e. Can you be more explicit in what you mean here. For instance what does effective mean in this context? Also best and worst on what dimension? Do you mean best and worst for single ensemble and lead time? Separated along what dimension is another question. Does separation refer to ensemble spread or something else?
Line 228 – What do you think it implies about this metric, or the forecasting of blocks generally, that this is true for Area and Shape but appears not to be true for Intensity and Location? Particularly for Location this decrease appears to occur 24 hours earlier? Is the predictability barrier for certain aspects different?
Line 235 – To my eye location might experience its most rapid error convergence between 168 and 120. Is that wrong? Did you quantify error convergence in some way?
Line 246 – I’m not sure what “short lead times” refers to here. I think we’re still talking about figure 3 but lead time increases as we move down the y-axis. Additionally, towards the top of the graph, I can’t really tell if there is a difference between the ECMWF and Met Office forecasts of Franklin’s position which would be the “short lead-times” you refer to. Can you clarify the point you’re making here? I’m also not sure how the last clause of this sentence relies on the previous one.
Line 285 – I think Is this equivalent to saying that lower Z500 in areas of positive sensitivity is associated with lower ZALO? Is it also true that higher Z500 in areas of negative sensitivity is associated with lower ZALO? I think an explanation in these terms of the relationship between deviations from the ensemble mean Z500 and ZALO would be very helpful for readers to understand what you’ve done here without going back and forth between the figure and the text identifying features. Additionally, I think providing an interpretation of negative sensitivity would help unfamiliar readers understand this slightly unintuitive concept.
Line 292 – Should “ensembles” be “ensemble members”
Line 296 – I don’t quite follow the sentence starting “A wave-like pattern…” I don’t quite follow this sentence. You start by mentioning a wave-like pattern in the sensitivity field but I don’t see why that indicates something about the ensemble members with a good forecast of Franklin’s position (those with small errors). Isn’t the sensitivity field calculated across all ensemble members? The connection to the RWP is also a little tenuous. Can you explain how Rossby waves are expressed in the sensitivity field (if that’s your claim) or clarify the relationship you’re putting forward between the wave-like pattern in the sensitivity field and RWP. I can believe they’re related but I need a little deeper explanation of that relationship from you.
Line 315 – Similar question as above
Line 437 – the first sentence of this paragraph “Figure 10 shows how…” could be removed if you referenced figure 10 at the end of the following sentence.
Line 456 – replace “cyclone one” with “cyclone in one”
Line 474 – replace do with does
Line 478 – replace “determine cause” with “determine the causes”
Line 536 – I’d suggest starting this paragraph with “We have shown block error to be sensitive to…”
Figure Comments:
Figure 1 – It should be fairly obvious to most readers but I think it’s worth specifying 320K is a potential temperature, at least the first time.
Figure 3 – I think it would be easier on readers if you wrote out the full ECMWF EPS ad MOGREPS-G in the panel titles rather than EC and MO which are abbreviations that haven’t been defined yet.
Figure 3 – Why are there different time periods displayed between the ensembles? I strongly suggest standardizing them to facilitate comparison of the evolutions of these two ensemble prediction systems.
Figure 4 – Same comments as figure 3.
Figs 3,4,5,6 – Why are there no indications of ensemble members’ locations of Franklin after September 3rd? Are they not available or is it some other reason? I think they would be interesting and help complete the picture of these forecasts.
Figure 5,6 – You can’t see the IBTrACS Franklin star in panel b. Could you plot it on top of the blue best fit stars?
Figure 8,9 – Please note what the solid black contours represent in the caption and specify the 2PVU contour is in red (if that’s the case).