A method to characterize the statistical extremes in marine biogeochemistry: the case of the Mediterranean chlorophyll

Biogeochemical extreme events are analysed for the special case of the surface chlorophyll in the Mediterranean open sea. The extremes are defined statistically as values over the 99th percentile threshold that are connected in space and time. A multidecadal simulation with daily output (1979-2012) was run to collect the statistics needed to carry out the study. The online coupled hydrodynamic-biogeochemical (MITgcm-BFM) model has a spatial resolution of 1/12°, with 75 vertical levels. The model was corroborated with available observations and other validated models. To account for the heterogeneity of the chlorophyll features and dynamics across the Mediterranean area, the definition of the extremes as peaks over the threshold was initially applied at each grid point and then refined to include in one extreme event all the points near in space with extremes at the same time. The macro-events were then characterized by a set of indexes, related to their spatio-temporal localisation, shape and impact on the local marine ecosystem due to their supply of high and/or anomalous biomass production. Following, the macro-events occurred in the Mediterranean area in the winter-spring months of the 1994-2012 period were described and classified using the introduced set of indexes. Overall, the macro-events occurred in each year during the winter months, especially in the northern part of the Mediterranean Sea. They have a wide spectrum of possible shapes, mainly corresponding to a duration shorter than 20 days and a radius contained within the larger Mediterranean mesoscale (200 km). The most persistent macro-events (uniformity index higher than 60%) are relatively small and short within the total spectrum, with an inverse proportionality between area and duration. The most severe macro-events (mean severity higher than 1.1 kgChl/km2/day in the first meter of depth) have intermediate area and duration; generally they are not very persistent, but sometimes they are very anomalous (anomaly higher than 15%). The most anomalous macro-events have typically medium or long duration (up to 84 days) and they can be also very large (up to the whole sub-basin scale), but not very persistent. In agreement with literature, no significant trends are observed in the macro-events indexes on the basin-scale. Some subdomains were analysed in detail: the North Western Mediterranean Sea (NWM), the Ionian Sea (ION), the South Adriatic Sea (ADS) and the Levantine Sea (LEV). The decreasing eastward gradient of the mean severity across the subdomains reflects the well known gradients of the mean surface chlorophyll. More than 70% of the macro-events of chlorophyll occurred in NWM and ADS are classified as events of high severity and high anomaly with respect to the medians of the distributions of the mean severity and anomaly on the whole domain. Overall, the most severe macro-events are in NWM. ADS is the subdomain with the most persistent macro-events, possibly ascribed to the spatial constraints of the topography and of the circulation structure. In ION and LEV more than the 50% of macro-events show low values of both indexes, even if hot spots of macro-events heavily affecting the ecosystem can be observed in the North Ionian Sea, close to the southern coasts of Sicily and in the Rhodes Gyre regions. Focusing on some representative examples of macro-events further investigated, we observed that the model is able to reproduce the main features of timing and location of the chlorophyll patterns observed by remote sensing and that the physical forcing anomalies with respect to climatology appear to play the major role as trigger of the macro-events. Finally, the selected macro-events are characterized by phytoplankton subgroups (BFM parametrization of picophytoplankton, flagellates, diatoms, large phytoplankton) that evolve in space and time in different ways in the subdomains.