Geoscience Reference
In-Depth Information
dedicated observations, and to run a 500 m grid length model in a configuration comparable
with conventional nowcasts (Ludovic Auger, MeteoFrance, personal communication
WMO/WWRP Workshop on Use of NWP for Nowcasting, Boulder 2011).
DWD has a 2.8 km forecast model with a nudging assimilation scheme (Consortium for
Small-scale Modeling, COSMO) covering Germany (COSMO-DE, Stephan et al., 2008) and is
developing PP KENDA ( Priority Project "KENDA" - Km-scale Ensemble-based Data
Assimilation) for a 1 km-3 km scale Ensemble Prediction System known as LETKF (Local
Ensemble Transform Kalman Filter; Ott et al., 2004). MeteoSwiss is running a 2 km version
of COSMO. Various collaborating meteorological services are running, or are planning to
run versions of these systems.
The HIRLAM (HIgh Resolution Limited Area Model) European community run their 3D-
Var system (Gustafsson et al., 2001) with the HIRLAM model at grid lengths down to about
3.3 km and are developing a new HARMONIE system to run at about 2.5 km. They have
also run experiments comparing three hourly and hourly cycling at 11 km and are exploring
the impact of GPS and Doppler radar radial wind data (Magnus Lindskog, HIRLAM
personal communication WMO/WWRP Workshop on Use of NWP for Nowcasting,
Boulder 2011 and HIRLAM Newsletter No. 58, November 2011).
In the USA, NCEP (National Center for Enviromental Prediction) has an operational RUC
(Rapid Update Cycle) system with hourly data assimilation (Benjamin et al., 2004). As of
September 2011, this was due to be replaced by the Rapid Refresh. The RR uses a version of
the WRF model (currently v3.2+) and the Grid-point Statistical Interpolation (GSI) analysis
largely developed at NCEP/EMC (Environmental Modelling Center, NOAA), using hourly
cycling and a 13 km grid length. NCEP also run a 3 km model nested in the RR, but this has
no separate data assimilation (Steve Weygandt et al., Earth System Research Lab, Boulder,
personal communication WMO/WWRP Workshop on Use of NWP for Nowcasting,
Boulder 2011, Stensrud et al., 2009; Smith et al., 2008; Weygandt et al., 2008).
In Japan, JMA (Japan Meteorological Agency) runs a Mesoscale Model (MSM) for Japan and
its surrounding areas using a 5 km grid length and 4D-VAR with forecasts every 3 hours to
15 or 33 hours (Honda et al., 2005; Saito et al., 2006). This is a non-hydrostatic model (JMA-
NHM). Development of NWP at a higher resolution (Local Forecast Model, LFM) is also in
progress to help produce sophisticated disaster-prevention and aviation information
services.
A trial operation of a 9-hour LFM forecast run on a 2 km grid length was performed in 2010
and 2011, and operational implementation is scheduled to start in 2012. LFM also uses JMA-
NHM as a forecast model, and its initial condition is generated from a 3D-Var rapid update
cycle. The cycle uses the MSM forecast as the first guess, and runs a JMA non-hydrostatic
model-based variational data assimilation system (JNoVA) - a 3DVar (a degenerate version
of JNoVA-4DVar) analysis and 1-hour JMA-NHM forecast in turn - over 3 hours using a
5 km grid length.
The Korean Meteorological Agency (KMA) is currently running the WRF 3D-Var (Barker et
al., 2004, Xiao et al., 2008) at 10 km but is planning to use the 1.5 km, variable resolution Met
Office Unified Model (UM) system with 3D-Var in the near future. In the past they have
tested a 3.3 km version of WRF 4D-Var (Huang et al., 2009).
