This dataset contains the return levels and parameters of the GEV and GPD extreme value distributions fitted to maximum surface daily wind speed and 5-day precipitation for every grid point in 52 of the EURO-CORDEX simulations. This data was produced as the primary objective part of the DEWPAD project, and internal project at the Bjerknes Centre for Climate Research, Bergen, Noway. The underlying model data used was from the high resolution (0.11 degree) EURO-CORDEX simulations which downscaled the CMIP5 climate simulations and the fields analysed were the maximum surface daily wind speed (sfcWindmax) and precipitation flux (pr). The precipitation was converted into a 5-day precipitation using a moving average. Four 30-year time slices were selected, as per the previous work of Outten and Sobolowski 2021, one at the end of the Historical simulations and three in the future simulations under RCP 8.5. These time slices were named, and covered the periods of 1976-2005 (Historical), 2011-2040 (Near Future), 2041-2070 (Mid-Future), and 2071-2100 (Far Future). For each of the 30-year time slices, the annual maxima and peaks-over-threshold approaches of Extreme Value Theory (EVT) were applied to both the extreme wind and 5-day precipitation. This resulted in a Generalised Extreme Value distribution (GEV) and a Generalised Pareto Distribution (GPD) being fitted to the annual maxima and exceedances respectively. Return levels for nine different return periods were calculated using both of the fitted distributions, specifically for the 2, 5, 10, 20, 30, 50, 70, 100, and 200-year return periods. For the peaks-over-threshold approach, the threshold was taken as the minimum of the annual maxima for each grid point separately. This method does not guarantee convergence in all locations and care must be taken to evaluate the quality of the return levels and fitted distributions when using the GPD results. This method was used due to the lack of a more reliable method to automate threshold selection, which is usually done through visual inspection of plots of Sample Mean Excess and parameter stabilities. The dataset is divided into extreme winds (sfcWindmax) and 5-day precipitation, with each further sub-divided into the four periods: Historical, Near Future, Mid-Future, and Far Future. For each of these, there are 52 NetCDF files, one for each model of EURO-CORDEX analysed. These files each contain the following variables: X = Number of grid points in the x direction, [x=412] Y = Number of grid points in the y direction, [y=424] Return levels = Return levels, values of the return periods for which return levels are assessed [return levels=9] Parameters = Names of the three parameters for the distributions, [parameters=3] Lat = Latitude, degrees north, [x,y] Lon = Longitude, degrees east, [x,y] Threshold = Threshold used in GPD in each grid point, m/s or mm accordingly, [x,y] Count of annual maxima = Number of annual maxima in each grid point, [x,y] Count of exceedances = Number of exceedances in each grid point, [x,y] Exceedance lambda = Parameter of Poisson distribution used in calculating return levels from GPD at each grid point, [x,y] GEV parameters = Parameters of the GEV distribution fitted at each grid point, [parameters, x, y] GEV return levels = Return levels based on the GEV distribution calculated at each grid point, [return levels, x, y] GPD parameters = Parameters of the GPD distribution fitted at each grid point, [parameters, x, y] GPD return levels = Return levels based on the GPD distribution calculated at each grid point, [return levels, x, y] The count of annual maxima should in theory be 30 at every grid point, since there are 30 years in each period analysed, however, in some locations in some models there were less than 30 annual maxima for the precipitation because those locations had no day with precipitation over 1mm in the entire year. These were primarily in the Sahara desert. The Exceedances lambda is equivalent to the average number of exceedances that occurred in each year. For more detailed information please contact the dataset creators.
