Nikolaos Mastrantonas, ECMWF Scientist, Diagnostics Team, Evaluation Section
Extreme Precipitation Events (EPEs) are associated with devastating consequences for our societies, economies, and environment. This applies also to the regions across the Mediterranean, an area of immense importance for our users and Member and Co-operating States. In fact, the magnitude, frequency, and associated damages of EPEs across the Mediterranean have increased over recent years, with even more worrying projections for the future. Thus, it is crucial to better understand EPEs and their drivers, so we can mitigate their impacts and increase the resilience of affected societies.
In our recent study (Mastrantonas et al., Int J. Climatol 2021), my co-authors and I provide new insights about EPEs over the Mediterranean. The work analyses the EPE characteristics in space and time, and quantifies their connection to large-scale weather variability, covering the entire region. The study utilises data from the ERA5 reanalysis dataset, and the EPEs are identified as instances exceeding the 99th percentile (P99) of the daily distribution of each grid cell.
Spatio-temporal connections
The Mediterranean has a clear divide in the seasonality of EPEs (Figure 1). Winter is the dominant season for the eastern Mediterranean, and autumn dominates the western parts. Summer is also crucial for the occurrence of EPEs in mountainous locations, for example the Alps and the Pyrenees. This result is associated with convective EPEs in these mountainous areas. Such convective EPEs are better represented in ERA5, compared to previous reanalysis datasets, due to the finer resolution of this product.
Figure 1: Season of highest (a) and 2nd highest (b) occurrence of EPEs per grid cell. Figure taken from Mastrantonas et al., 2021.
An important characteristic of EPEs is their degree of persistence, meaning how possible it is to have an extreme event just a few days after a previous extreme in the same area. High persistence can further exacerbate the negative consequences of these extremes. Our results show that such connections are very strong. For most locations in the Mediterranean 1 in every 10 EPEs occurs just the day after the previous EPE in the same area. This connection doubles (2 in every 10 EPEs) when a 7-day window is allowed.
Finally, orography can substantially modulate the simultaneous occurrence of EPEs even in distant areas. For example, in central-western Italy, 3 in every 10 EPEs happen simultaneously with EPEs over Montenegro and Croatia, a result attributed to the influence of the Apennine Mountains.
Connections to large-scale patterns
Daily anomalies of sea level pressure (SLP) and geopotential height at 500 hPa (Z500) were used to group weather variability over the Mediterranean into different classes. After an initial dimensionality reduction with Empirical Orthogonal Function, K-means clustering was implemented to derive 9 weather regimes of distinct atmospheric characteristics over the region (Figure 2).
Figure 2: Composites of the 9 weather regimes over the Mediterranean. Colour shading refers to SLP anomalies (hPa), and contours to Z500 anomalies (dam). Percentages indicate the climatological frequency of each weather regime. Figure taken from Mastrantonas et al., 2021.
These regimes are associated with unstable low-pressure systems such as cut‐off lows and troughs, or with stable anticyclonic conditions, such as ridges, extending over hundreds of kilometres. The importance of the Atlantic, and the storms generated over the ocean is very clear, as three of the nine clusters are associated with negative anomalies centred over/near the western Mediterranean, which is influenced strongly by the Atlantic (Figure 2 a-c).
The analysis showed that each of these nine groups is preferentially connected with the occurrence of EPEs at different subdomains of the Mediterranean (Figure 3). For example, the first three regimes (Atlantic Low, Biscay Low, Iberian Low) are linked with EPEs over western and central Mediterranean, while the Balkan Low is preferentially associated with EPEs over the Balkans and western Turkey.
In general, the probability of observing an EPE under the preferential cluster is over three times higher compared to the nominal probability of the studied extremes (P99-1% nominal probability). Especially for locations over mountainous and coastal regions, this probability is more than 6 times higher, for example for areas in northern Spain, the Atlas Mountains, and the Apennines (Figure 3b).
Figure 3: Conditional probability of EPEs for each weather regime. The subplots show only areas with statistically significant connections. The percentages in parenthesis refer to the percentage of grid cells that have statistically significant connections with each weather regime. The boxplots present the conditional probabilities of all grid cells with statistically significant connections. Figure taken from Mastrantonas et al., 2021.
Summarising the next steps
These findings are important because current numerical weather prediction models have challenges in predicting localised EPEs for more than a few days in advance. These models, nevertheless, can provide useful information about large-scale weather variability up to two to three weeks ahead. Thus, the identified strong connections of EPEs to large-scale patterns, together with their high temporal persistence and robust spatial links, can help us in making better predictions at longer lead-times.
Follow-up studies will quantify the benefits of using the derived nine patterns for extended-range predictions. The aim is to use this information for the development of new operational products addressing user needs, an aim in accordance with the ECMWF Strategy 2021-2030.
In the meantime, it would be fortunate if this study motivates additional research on the use of domain-specific weather regimes for analysing extreme events. In this regard, I hope that the publicly accessible scripts of this work will be helpful to interested researchers.
Acknowledgements
I would like to thank the co-authors of this work, Linus Magnusson, Florian Pappenberger, Pedro Herrera-Lormendez, and Jörg Matschullat, for their support/supervision. Scientific collaborations and exchange of ideas and knowledge are key for advancing our understanding. That said, many thanks to the rest of my colleagues at ECMWF and the CAFE consortium for the numerous discussions I had with many of them.
This work is part of the Climate Advanced Forecasting of sub-seasonal Extremes (CAFE) project. The project has received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 813844.
Follow Nikos on Twitter at @NikMastrantonas.
Top banner image: Sundry Photography/iStock/Getty Images Plus