| 04 Jan 2026
Predictability of cyclones associated with heavy precipitation events in the Sahara
Abstract. Heavy precipitation events (HPEs) are a precious source of water in the Sahara, but often trigger devastating flooding. These events are strongly associated with surface cyclones, making accurate cyclone forecasting crucial for predicting hazards related to HPEs and their impacts. In this study, we investigate the predictability of HPE-associated cyclones across the Sahara and its drivers. We use ERA5 reanalysis and ECMWF initialized reforecasts between December 2000 and November 2020. Forecast skill on short-, medium-, and extended-range timescales is evaluated based on the overlapping areas of observed and forecasted cyclones over the Sahara. Results show that the lead time of skillful prediction is up to about 10 days. In winter, when cyclones are mainly located in the northern Sahara, forecast skill is higher for deeper cyclones. In summer, skill is higher for cyclones located in the southwestern Sahara. On short-range lead times, forecast skill is higher in winter, whereas on medium to extended lead times, skill is higher in summer and fall. Rossby wave patterns extending over the North Atlantic are associated with both high and low skill forecasts, highlighting a flow-dependent control on predictability over the Sahara and underscoring the need for more detailed investigation. These findings identify key controls and characteristics of skillful forecasts of cyclones that lead to HPEs in the Sahara on timescales of a few days to two weeks in advance. Understanding these variations across regions and seasons is key to improving the predictability of HPEs and their related impacts.
In this paper an attempt is made to investigate the predictability of high precipitation cases related to cyclones and the dynamic drivers. The topic is very interesting for the Mediterranean region weather and climate. I recognize the huge amount of processing data and the complexity of the methodology. However, I have some concerns:
•   lines 85-90: The method used to retain from 42,000 HPEs only around 12,500 cases in which cyclones 90 were associated with HPEs, «according to a Monte Carlo cyclone-association test performed in Armon et al. (2024) that determines whether HPEs occur closer to a cyclone than would be expected by chance, based on repeated comparisons withrandomly selected cyclone dates» is not clear to me. I think that the authors should be more specific.Â
•   Line 115: “ Cases where the detected cyclone is located at a distance ≥2000 km..”. How this threshold value dirived? It sems rather arbitrary
•   Line 125: how do you define cyclone mask? The MSLP does not measure  directly the intensity. The Laplacian of p is such a measure.
•   Line 146: Although the authors state that the 30% value is arbitrary, they should document this value (e.g based on operational experience, statistical analysis, sensitivity tests)
•   Line 142: why the lead time extends so far to 15.5 days?Â
•   Figure 2 presents an example of the four category verification methodogy. It is very confusing for me and I need more clarification.
•   Line 155: “The upper 40% and lower 40% of these events were classified as high- and low-skill..”. Again, how this percentage derived?Â
•   Line 157: why specifically at 12 UTC?Â
•   Figure 3: These four plots are very confusing and hard to follow. The authors might find another manner to present their findings. In the legend it is stated “The black solid line shows the average climatological frequency of cyclone coverage, computed as the weighted cyclone frequency at each grid point of each cyclone area.” What cyclones are considered? How they are identified?Â
•   Section 3.1 is very wordy and tiring in terms of numbers and parameters. I think that the focus should be a physical interpretation and discussion related to the implications of the findings