Geoscience Reference
In-Depth Information
This latter work led to the operational implementation of an algorithm based upon global
cross correlation at McGill University in the mid-1970s. Bellon and Austin (1978) reviewed
the operational performance of this scheme, known as SHARP (Short-Term Automated
Radar Prediction), on two years' worth of data. The experience gained allowed subsequent
enhancement of their cross correlation method to enable independent tracking of different
echoes (Austin & Bellon, 1982) using a nine vector motion field. Rinehart (1981) describes a
similar, multi-vector, cross correlation approach to determine and extrapolate the motion of
individual storms within a multi-storm system.
3.3 Cell tracking
Algorithms founded on the tracking of radar echo centroids evolved in parallel with field-
based pattern matching techniques. These were developed specifically for nowcasting
thunderstorms, initially in North America. Amongst the earliest of these echo centroid
trackers were those described by Wilk and Gray (1970) and Zittel (1976). The extrapolation
vectors were diagnosed using a linear least squares fit through successive positions of the
echo centroids. Duda and Blackmer (1972) and Blackmer et al. (1973) formulated clustering
techniques to resolve difficulties in cases involving the merging and splitting of echoes.
Refinements to these early techniques were subsequently developed and implemented
within operational tools during the following decades. Several good examples are the Storm
Cell Identification and Tracking (SCIT) algorithm (Witt & Johnson, 1993) and the
Thunderstorm Identification, Tracking, Analysis and Nowcasting (TITAN) system (Dixon &
Wiener, 1993).
3.4 Steady state versus growth and decay
The proto-type, operational nowcasting algorithms developed during the 1970s were
generally reliant on the steady state assumption. Tsonis and Austin (1981) explored echo
size and intensity trending with a view to improving the prediction of long lived convective
cells. They found negligible improvement in skill, even using sophisticated non-linear time
trending schemes. Wilson et al. (1998) drew similar conclusions in a study involving the use
of the TITAN system (Dixon and Wiener, 1993). These results are consistent with the
findings of theoretical and NWP modelling experiments showing that the evolution of
convective scale features in the atmosphere is non-linear and, to a degree, chaotic (Tsonis,
1989).
3.5 Fractal properties of precipitation
During the 1980s and 1990s, an improved understanding of the chaotic influence of
atmospheric processes such as turbulence on the predictability of precipitation was reflected
in a growing number of publications exploring the so called
scaling
or
multi-fractal
attributes
of meteorological fields, including those of radar derived precipitation fields. Scaling
behaviour or self-similarity implies that similar features can be observed in the atmosphere
over a wide range of space and time scales, and that the relationship between certain
statistical attributes of a precipitation field measured at different scales can be described by
equations which incorporate a scaling factor. The formative papers in this area include those
by Lovejoy and Schertzer (1985, 1986).
