Geoscience Reference
In-Depth Information
orographic rainfall and predictors describing mesoscale
flow and air mass stability, to
identify past events with predictors similar to those derived from real time observations.
The authors present verification results showing that NORA performs better than Eulerian
persistence for nowcasts with lead times of more than an hour.
5. NWP-based nowcasting
5.1 Introduction
In the past few years, increasing availability of high powered computers and the
implementation of non-hydrostatic models have made NWP at the convective scales (1 km-
4 km horizontal grid length) a reality for national weather services. Many centres are
already running these models operationally with update cycles of between 3 and 6 hours to
generate short-range forecasts up to about T+36 hours. Traditionally, these forecasts have
been deployed in combination with nowcasting techniques to deliver optimal guidance.
However, recently, centres have begun to explore the use of NWP-based systems for
nowcasting.
5.2 The challenges
For nowcasting purposes, the key component of NWP is the data assimilation of high
resolution observations in space and time, especially radar and geostationary satellite data.
Traditional nowcasting techniques use these observations to produce forecasts of rain, cloud
and associated weather with observation derived advection velocities, or NWP forecast
wind fields, or a combination of both. Nowcasts are also produced from analyses of other
weather elements including screen temperature, visibility, 10 m wind and wind gusts.
However, these systems do not use the observations in an optimal manner and may not use
all available observation types.
Data assimilation into NWP models potentially offers the ability to use all observations in a
consistent and synergistic manner to provide the best estimate of the state of the atmosphere
from which to produce a nowcast. At this time, nudging, variational data assimilation
(3D-Var and 4D-Var) and ensemble Kalman filters (EnKF; Sun, 2005b) for high resolution
data assimilation are being used in weather services or are under development in research
centres around the world. Indeed, some national weather services are already running
operational NWP models with data assimilation at grid lengths in the range 1 km-10 km.
Most of this work relies heavily on the exploitation of Doppler radar measured radial winds
and reflectivity data or derived surface rain rates.
One challenge for NWP-based nowcasting is to match the skill of traditional methods in the
first two hours. Traditional nowcasts closely fit the observations because they employ
extrapolation techniques and so use the observations themselves (i.e. radar derived surface
rain rate) at analysis time. This is challenging for NWP because unresolved scales are
excluded from the model state, data assimilation systems are designed, not to match
observations, but to achieve a good and balanced forecast over a longer period of time, and
the T+0 fields from the NWP system are essentially a weighted fit to both the NWP forecast
and the observations.
