Evaluation and Comparison of Reanalyses

With regard to the appli­ca­tion of reanaly­sis data, a com­pre­hen­sive assess­ment of the qual­i­ty of the reanaly­sis is need­ed. There­fore, we are con­tin­u­ous­ly devel­op­ing an eval­u­a­tion frame­work that allows a thor­ough and time­ly eval­u­a­tion of reanaly­sis data with respect to key para­me­ters of inter­est for cli­mate mon­i­tor­ing and appli­ca­tions, espe­cial­ly in the field of renew­able ener­gies.

The focus of the frame­work is on the uti­liza­tion of unique inde­pen­dent obser­va­tions at atmos­pher­ic obser­va­to­ries (e.g. JOYCE and DWD Richard Ass­mann Obser­va­to­ry) or from the full reanaly­sis domain (e.g. ceilome­ter net­work) includ­ing an assess­ment of the mar­gin­al as well as the joint dis­tri­b­u­tion of a para­me­ter to assess its spe­cif­ic prop­er­ties. The eval­u­a­tion also com­pris­es the com­par­i­son with oth­er region­al and glob­al reanaly­ses to assess the per­for­mance of our reanaly­ses and deter­mine the added val­ue com­pared to exist­ing data sets.

The eval­u­a­tion does not only focus on aver­aged devi­a­tions, but includes an eval­u­a­tion of the entire dis­tri­b­u­tion includ­ing the extreme val­ue behav­iour, the rep­re­sen­ta­tion of spa­tial depen­den­cies or the spec­tral prop­er­ties of spa­tio-tem­po­ral fields. For this pur­pose we devel­op and use advanced sta­tis­ti­cal meth­ods e.g. based on extreme val­ue the­o­ry, cop­u­la or Fouri­er and wavelet spec­tra.