Revisiting Cyclone Detection and Tracking Methods using ECMWF ERA5 dataset for climatological purposes in the Mediterranean Region

The main target of this study is to evaluate the Cyclone Detection and Tracking Methods (CDTM) using the ECMWF ERA5 dataset, state-of-the-art in reanalysis models, to identify the main cyclogenesis zones and cyclone tracks inside the Mediterranean region during a climatological period. Several studies based on ECMWF analysis and reanalysis (ERA40 and ERA Interim) datasets indicate a large divergence related to the average number of cyclones passing through the Mediterranean region by year. However, the majority agrees on the most important cyclogenesis areas, seasonality variation of the number of cyclones, and trends of cyclone track. In general, the differences between those methodologies concerns to the meteorological variable used to detect cyclones and the metric used to define its intensity. Nevertheless, spatial and temporal resolutions were fundamental to achieve the results, since the most advanced dataset used in the literature presented relatively low values ​​such as 1.125°x1.125° and 6h, respectively. Past studies reported that these values ​​were already high enough to produce numerical noises. Here, the geopotential height at 1000 hPa (Z1000) was used, with horizontal spatial resolution of 0.25°x0.25° and time resolution of 1h, to identify the local minima for each time step (hereafter, candidates), and filtering those with negative gradients of Z1000 within a radius of 1000 km to exclude candidates associated with thermal lows or geopotential troughs.
Following the literature, the domain for Mediterranean region was defined by the area within 9°W to 42°E, and 27°N to 48°N, where were considered only cyclones with at least one tracking point inside the domain. Also, the results were produced for the period 1979-2008 using two types of input data: (Model I) ERA5 data with resolutions reduced to 1.5°x1.5° and 6h, as well as the main previous studies; and (Model II) full-resolution ERA5 data. As expected, Model I results were very similar to those found in the literature in all aspects (number of cyclones, seasonal distribution, areas of cyclogenesis and tracks). On the other hand, since the use of higher resolution data provides greater spatiotemporal detailing of the climatological period, the results of Model II presented a total number of cyclones substantially higher than that of Model I (~25%), but still within the range described in the literature. The models indicated more frequent cyclones during the spring months with maximums in April (Model I) and May (Model II). An interesting point highlighted in other studies but not observed in their results, is an increase in cyclone frequency between August and October, captured in both Models I and II and more evident in Model II. An explanation is found in the greater number of short-life cyclones, which act in relatively narrow areas intangible to datasets with limited resolution.