For the MODINI part, the MPI-ESM-LR was initialised for decadal retrospective forecasts
(hindcasts in the following) using windstress anomalies from reanalysis. This work was
published as part of an inter-comparison of different initialisation and ensemble generation
methods (Polkova et al., 2019). This work was based on the work by Thoma et al. (2015) from
MiKlip1, with some modifications, i.e. hindcasts were started 1961-2011 instead of 1990-2006
and another wind product was used (ERA-40 and ERA-Interim instead of NCEP-cfsr). Within
the inter-comparison project, the relatively simple and cheap method of MODINI was
competitive, but could not hold the big expectations triggered by Thoma et al. (2015). In
particular, hindcast skill was only improved over limited areas of the Pacific compared to the
standard MiKlip system, a downside we attribute to errors in the wind product, especially
before the satellite era.