DATE2022-05-12 17:31:30
TITLEEvolution of precipitation regimes in the Mediterranean region with climate change : from drought to extreme rainfalls
AUTHORSJulie André (1) ,Fabio Dandrea (1) ,Philippe Drobinski (1) ,Caroline Muller (2)
  1. 1) Laboratoire De Météorologie Dynamique, Paris (France) ,2) Institute Of Science And Technologie, Klosterneuburg (Austria)
ABSTRACTClimate change is known to have consequences on both mean and extreme precipitations, with potentially threatening impacts on societies. The Mediterranean region is particularly concerned, being a hotspot of temperature and precipitation changes: this region is expected to get much drier, but with more intense extremes. The research on precipitation evolution has mainly focused on either the absence of rain (drought, dry spells, …) or on the extreme rainfalls, letting the rest of the rain distribution quite overlooked. Looking at the whole distribution instead of small portions of it might help to get a broader and more coherent picture of the situation. In this presentation, we will look at how the whole rain distribution changes, in the Mediterranean region over the recent past. We are using observational data covering the whole 1950-2021 period with a daily resolution and at a spatial resolution of 0.25° (EOBS gridded data set). It enabled us to compute statistically significant trends of the rain percentiles values over the last 70 years. We found that the situation in the Mediterranean is more complex spatially speaking than the image of a homogeneous region with decreasing mean and increasing extremes. In fact, three main behaviors are observed: as expected regions with decreasing rain percentiles up to a given threshold and then an increasing distribution tail, but also regions with the whole distribution decreasing, and even a few other regions which have a behavior more representative of Northern Europe (the whole rain distribution increases). Finally we are able to reduce the information of the whole distribution trend to just two parameters, using a log-normal distribution model, which accounts very well for the three behaviors described above.