Geoscience Reference
In-Depth Information
3.9 NWP-based nowcasting
The 1990s saw the first attempts to run convection resolving NWP model forecasts
assimilating radar data (Lin et al., 1993). These early experiments were focused on
predicting convective storms.
Some success was demonstrated in cases involving convection strongly forced by the large
scale environment. However, other studies showed that convective initiation in a weakly
forced environment is difficult to predict because the location and timing of initiation are
very sensitive to variations in low level temperature and moisture. It seems likely that the
ability of convection resolving NWP models to predict convection is a function of the
predominant scale of the associated forcing. In the UK, much of the convection is forced by
small-scale orography, as demonstrated by the Convective Storms Initiation Project
(Morcrette et al., 2007; Lean et al., 2008).
Despite these challenges, Lean et al. (2008) found that convection resolving models
performed better in terms of convective initiation than a 12 km grid length model with
parameterized convection, and a number of national weather services are now running
operational, convection resolving NWP models. Indeed, some are trialling configurations
with hourly or sub-hourly assimilation of radar data. NWP nowcast experiments in the UK
show some improvements in NWP forecast skill in the nowcast time frame. Prospects for
NWP nowcasting will be discussed in more detail later in this Chapter.
4. Conventional nowcasting techniques
4.1 Deterministic techniques
4.1.1 Cell tracking
Cell trackers or object-based nowcasting schemes are typically developed in areas where
severe convective storms are a significant hazard, and are best suited to the generation of
qualitative warnings of severe convective weather. In general, object-based algorithms are
used to predict the location of a (convective) object in the future and thereby assign the
properties of the object to that location. For example, a storm might be deemed to contain
large hail, and therefore a warning of large hail will be issued for the locations on the
forecast storm track.
The basic elements of cell tracking are:
1.
devise a set of rules that will be used to identify the bounds of an object in either two or
three dimensions;
2.
analyse current data to identify objects and assign attributes to them (heavy rain,
damaging wind, large hail etc);
3.
link the objects to existing tracks and estimate the advection velocity;
4.
predict the location of objects in the future.
Most cell tracking algorithms define an object, either as a set of contiguous points that
exceed some threshold in radar reflectivity, typically 35, 40 or 45 dBz (e.g. Dixon & Weiner,
1993; Han et al., 2009), or as a small region of increased reflectivity (Crane, 1979), or both
(e.g. Handwerker, 2002). Defining the object in three dimensions allows one to compute the
volume and height of the cell. This adds value when assigning the elements of severe
