WP2 - Ensemble Data Assimilation Regional Reanalysis Dataset [Months: 1-48]
The EURO4M project (2010-2014) has provided the core ‘deterministic’ European regional reanalysis system,
assimilating conventional, satellite and hydrological cycle (humidity, cloud, precipitation) observations into the Met
Office Unified Model (UM)’s advanced four-dimensional variational (4D-Var) data assimilation (Rawlins et al 2007).
The addition of the UERRA-MOGREPS-EU capability will provide consistent ensemble uncertainty estimates using
a 20-50 member, regional configuration of the operational MOGREPS-G system currently implemented at Met
Office for global operational probabilistic NWP.
Basic observation database will be from ECMWF MARS (ERA-CLIM), supplemented by high-resolution
conventional observations made available for regional reanalysis by partners within UERRA WP1. There will
be additional hydrological cycle observations suitable for high-resolution reanalysis, namely disaggregated
precipitation accumulations and surface/satellite cloud observations for the period of the reanalysis.
The Ensemble Variarional (EVDA) derived ensemble regional reanalysis will be evaluated deterministically through
a) Comparison of ensemble mean against independent, unassimilated observations, and b) Sanity check on quality
of forecast run from ensemble control analysis. Probabilistic evaluation of the quality of the ensemble reanalysis
will be provided via spread-skill matching, rank histograms, and Brier skill scores. Additional evaluation against
gridded observation datasets (e.g. E-OBS) and intercomparison with global (ERA-CLIM) and regional reanalysis
datasets will be performed within UERRA WP3].
The HARMONIE Data Assimilation system as developed and used within the HIRLAM and ALADIN consortia will
be implemented and optimised for the entire European area with surrounding sea areas at as high resolution as is
possible (11 km and at least 65 levels). It will be run over a 50 year period, from 1961, and serve as one member
of a multi-model reanalysis.
The physiographic properties will be derived or modelled to take the time evolution into account. Interaction with
the surface (soil and sea and ice) is very important for the near surface ECVs and requires special attention.
The data assimilation will be driven by the global ECMWF ERA-CLIM reanalysis and also use a large scale Jk
constraint (Dahlgren, 2011) to add large Atlantic scale information from ECMWF satellite assimilation into the 3DVAR
MF will use the 2D-analysis system MESCAN, developed during the EURO4M project with SMHI, to provide a
surface analysis for temperature, relative humidity, precipitation and wind. MF will downscale the HARMONIE 3DVAR
analysis as an input field or background for the 2D–analysis sysan ensemble surface analysis will be developed and evaluate on a shorter period (5 years) over Europe with MESCAN using uncertainties from task 2.1 and 2.2
and/or observation network and perturbed observations.tem MESCAN at 5.5 km. If possible additional surface
datasets from WP1 will be used.
Good quality data CM-SAF data sets exist for both Geostationary METEOSAT and AVHRR polar platforms. They
complements each others over the European area but an optimally gridded data set is needed for climate studies,
validation of models and solar energy potential. A 2D pan-European analysis of cloud fraction will be run with the
SMHI MESAN for 30 years, at 5.5 km resolution 1982-2013.
A hybrid ensemble data assimilation system will be implemented for the DWD NWP model COSMO. The system
will be comprised of a local ensemble transform Kalman filter component currently developed at the DWD and
an ensemble nudging component for continuous data assimilation between two Kalman Filter initializations. The
ensemble nudging will be based on the current nudging implementation in the COSMO model and make use of the
covariance structure given by the ensemble realizations. Perturbed observations will be nudged into the system
using an observation data set developed in the project.
An ensemble regional reanalyses using the combined data assimilation system will be carried out for a test period.
To show feasibility for the pre-satellite era, a probabilistic dataset will be used to compensate for missing satellite
data in this era.
The KF ensemble variational DA will be used with a 6-hour Kalman filter interval and continuous ensemble nudging
between two Kalman filter initializations. The target resolution for the ensemble is 12 km ensuring high resolution
uncertainty estimates for the European CORDEX domain (covering whole Europe). Boundary conditions will be
provided by the ERA-20C or NOAA 20-CR reanalyses.
The produced regional ensemble reanalyses data will be evaluated against independent observations, e.g. unused
satellite observations as used in the current HErZ regional reanalysis scheme. Probabilistic evaluation will contain
standard matches for ensemble reliability and/or resolution, e.g. spread-skill relation, rank histograms, Brier/CRPS
scores. Additional comparisons will be made against the high resolution deterministic HErZ regional reanalysis.
Extensive evaluation of the reanalysis ensemble will be performed within UERRA WP3.