Regional reanalysis

EURO4M produces high resolution regional reanalysis datasets for Europe. To this end the Met Office´s Unified Model (UM) has been ported to ECMWF computing infrastructure and coupled to their ERA global reanalysis system. The new 4DVar-based reanalysis system is demonstrated for an initial 3-month period during summer 2010. This period was chosen due to a number of significant European weather events that year (Russian heat wave, Polish floods, etc).

Maximum surface temperature on 7 July 2010
The maximum surface temperature on 7 July, 2010, during the Russian heatwave based on E-OBS (KNMI), ERA Interim, and the 4D-Var based reanalyses (UK-MO).

Complementary to this, the HIRLAM 3D-VAR re-analysis system is set up by SMHI and run first for 2009 and then for 1989-1992. It is for a large region covering Europe (EEA members) and most of the North Atlantic. The area can be seen below where the albedo is displayed. The resolution is 22 km horizontally and 60 levels vertically.

HIRLAM 3D-VAR re-analysis system by SMHI
The region covered by the HIRLAM 3D-VAR re-analysis system is shown in the map of surface albedo (Source: SMHI).

It is crucial to monitor the continual performance of the assimilation and this has been done from inspecting statistics of analysis increments and differences to ERA-Interim. Observation statistics (Observations - First Guess and Observations - Analysis) have been extracted and checked. The assimilation has been stable, as can be seen in an example for SYNOP pressures below:

Example SYNOP pressures
Statistics of the HIRLAM re-analysis sysstem: RMS (upper graph), bias (middle graph), and number of observations assimilated (lower graph) (Source: SMHI).

A high resolution 2D re-analysis is being done the results from the 3D HIRLAM re-analysis. The 2D analysis is based on the SMHI MESAN analysis. There are a number of advantages compared with the large scale 3D HIRLAM:
  • More and other observational data can be used in a 2D analysis compared with a full 3D NWP based analysis: 2m temperatures, winds over land, precipitation data, snow accumulation.
  • The first guess for these high resolution (5km) analyses is created using downscaling methods, taking into account high-resolution topography and local structures in precipitation.

    HIRLAM at 22 km and 5 km
    First guess from HIRLAM at 22 km and the down-scaled result at 5 km before any analysis. (Source: SMHI).

    The analysis uses anisotropic structure functions taking land-sea and height differences into account (see left figure below).

    The precipitation analysis requires more care in order to take into account very varying local climate, primarily due to topography and wind. Climatologic standard deviations of 24 hour precipitation amounts have been analyzed and used both for the analysis (see right below) and for a regression from topography related parameters and model wind in order to create a first guess for the analysis.

    Precipitation analysis
    Left: Local structure functions used for the mesoscale analysis (OI) with anisotropy (Source: SMHI).
    Right: Analysis of observed 24-hour precipitation with first guess from a regression, based on physiography and HIRLAM forecasts. (Source: SMHI).

    The developments have been done together with Météo-France and extensive testing of different methods have been done against high resolution data and the existing analysis systems at MF.