Geoscience Reference
In-Depth Information
Over recent years, CAPS (Center for Analysis and Prediction of Storms, Oklahoma, USA)
has been carrying out experimental, real time forecasting including the generation of 1 km
grid length forecasts on a continental U.S. domain once a day, and the production of rapidly
updated NWP-model-based nowcasts producing two hour, 1 km forecasts every 10 minutes
(Xue et al., 2011; Kong et al., 2011; Clark et al., 2011; Brewster et al., 2010). These forecasts
assimilate US operational WSR-88D radar data and/or high-resolution experimental X-band
radar data, with and without assimilation cycles.
Comparison forecasts show systematically positive impacts of assimilating radar data on
short-range precipitation forecasting, lasting up to 12 hours on average. To address forecast
uncertainty and to provide probabilistic forecast information, storm-scale ensemble
predictions have also been carried out and the products have been evaluated at an
experimental forecasting facility. Extensive research has also been undertaken using
ensemble-based data assimilation methods for initializing storm-scale NWP models, with
very promising results.
CAPS has investigated a Mesoscale Convective System/vortex case study exploiting nested
400m /2 km grids and assimilating radar data at 5 min intervals using their Advanced
Regional Prediction System (ARPS) 3DVAR+cloud analysis (Schenkman et al., 2011a) as
well as EnKF (Snook et al., 2011). These show data impact on Collaborative Adaptive
Sensing of the Atmosphere (CASA; Schenkman et al., 2011b) and probabilistic forecast skill
with EnKF analyses (Snook et al., 2011; 2012).
Environment Canada has begun developing a convective-scale EnKF in order to examine
the assimilation of radar data (e.g. over the Montreal region; Luc Fillion, personal
communication WMO/WWRP Workshop on Use of NWP for Nowcasting, Boulder 2011).
This is based on adaptation of the Global EnKF code available at Environment Canada
(Houtekamer & Mitchell scheme) to a limited-area domain. The analysis step and the
forecast model configuration (1 km horizontal grid length) are being validated.
5.4 Development of an NWP-based nowcasting system in the UK
5.4.1 Progress to date
The Met Office has run an operational, 4 km grid length NWP model for the UK (UK4) since
December 2005. It has also run a 1.5 km UK configuration (UKV) routinely since summer
2010. Both models use three-hourly cycling 3D-Var and produce forecasts to 36 hours ahead
with 70 levels. These are based on the Met Office's Unified Model (Davies et al
.,
2005) and
variational data assimilation system (Lorenc et al., 2000; Rawlins et al
.,
2007), plus latent
heat nudging (Macpherson et al
.,
1996; Jones & Macpherson, 1997; Dixon et al., 2009). They
also include direct variational assimilation of analysed 3D cloud cover via associated
relative humidity. The UKV has a 1.5 km grid length over the UK and a 4km stretched
boundary nested in the 12 km NAE (North Atlantic and European) model. The UKV uses
3 km 3D-VAR over the whole domain. Collaborations with KMA and CAWCR (Centre for
Australian Weather and Climate Research, Australian Government Bureau of Meteorology)
are aiming to implement 1.5 km versions of the UM with 3D-Var or 4D-Var.
The Met Office's UK Post-Processing system (UKPP) incorporates a STEPS precipitation
nowcast (Bowler et al., 2006). This combines a stochastic, radar-based extrapolation nowcast
