Menu Close

Vertical profiles of wind gust statistics from a regional reanalysis using multivariate extreme value theory

Pub­lished in Non­lin­ear Process­es in Geo­physics, Vol­ume 27, Issue 2, 2022:

Many appli­ca­tions require wind gust esti­mates at very dif­fer­ent atmos­pher­ic height lev­els. For exam­ple, the renew­able ener­gy sec­tor is inter­est­ed in wind and gust pre­dic­tions at the hub height of a wind pow­er plant. How­ev­er, numer­i­cal weath­er pre­dic­tion mod­els typ­i­cal­ly only derive esti­mates for wind gusts at the stan­dard mea­sure­ment height of 10 m above the land sur­face. Here, we present a sta­tis­ti­cal post-pro­cess­ing method to derive a con­di­tion­al dis­tri­b­u­tion for hourly peak wind speed as a func­tion of height. The con­di­tion­ing vari­ables are tak­en from the COSMO-REA6 region­al reanaly­sis. The post-pro­cess­ing method was trained using peak wind speed obser­va­tions at five ver­ti­cal lev­els between 10 and 250 m from the Ham­burg Weath­er Mast. The sta­tis­ti­cal post-pro­cess­ing method is based on a cen­sored gen­er­al­ized extreme val­ue (cGEV) dis­tri­b­u­tion with non-homo­ge­neous para­me­ters. We use a least absolute shrink­age and selec­tion oper­a­tor to select the most infor­ma­tive vari­ables. Ver­ti­cal vari­a­tions of the cGEV para­me­ters are approx­i­mat­ed using Legendre poly­no­mi­als, such that pre­dic­tions may be derived at any desired ver­ti­cal height. Fur­ther, the Pickands depen­dence func­tion is used to assess depen­den­cies between gusts at dif­fer­ent heights. The most impor­tant pre­dic­tors are the 10 m gust diag­nos­tic, the barotrop­ic and the baro­clin­ic mode of absolute hor­i­zon­tal wind speed, the mean absolute hor­i­zon­tal wind at 700 hPa, the sur­face pres­sure ten­den­cy, and the lift­ed index. Prop­er scores show improve­ments of up to 60 % with respect to cli­ma­tol­ogy, espe­cial­ly at high­er ver­ti­cal lev­els. The post-pro­cess­ing mod­el with a Legendre approx­i­ma­tion is able to pro­vide reli­able pre­dic­tions of gusts’ sta­tis­tics at non-observed inter­me­di­ate lev­els. The strength of depen­den­cy between gusts at dif­fer­ent lev­els is non-homo­ge­neous and strong­ly mod­u­lat­ed by the ver­ti­cal sta­bil­i­ty of the atmos­phere.

Authors: Julian Stein­heuer and Petra Friederichs

Link to arti­cle

 

Research Highlights