Geoscience Reference
In-Depth Information
In the following section, we describe some of the key milestones in radar-based precipitation
nowcasting and review these in the context of parallel advances in relevant areas of science
and technology.
3. Radar-based nowcasting - A brief history
3.1 Origins of weather radar, and early research
Operational weather radar has its origins in the development of military radar during
World War Two. The invention of the resonant cavity magnetron by John Randall and
Harry Boot at the University of Birmingham in England in 1940 allowed the construction of
high powered, centimeter-band radars, suitable for detecting precipitation. The sharing of
this technology with American scientists early in the 1940s facilitated its subsequent
development for meteorological applications.
Important early papers include those on rain drop size distributions (Marshall & Palmer,
1948) and shapes (Browne & Robinson, 1952; Hunter, 1954; Newell et al., 1955), the
measurement of precipitation (Ryde, 1946; Byers, 1948; Bowen, 1951; Twomey, 1953; Battan,
1953; Stout & Neill, 1953), its vertical structure (Langille & Gunn, 1948) and associated
estimation errors (Hitschfeld & Bordan, 1954), and those on thunderstorm identification,
behaviour and dynamics (Wexler & Swingle, 1947; Byers & Braham, 1949; Wexler, 1951;
Ligda, 1951; Battan, 1953).
3.2 Extrapolation techniques
The concept of extrapolating radar echoes for the short term prediction of precipitation was
first proposed by Ligda (1953). The earliest demonstration of the application of objective
extrapolation to radar echoes is described by Hilst and Russo (1960), whilst Noel and
Fleischer (1960) were amongst the first radar meteorologists to explore the predictability of
precipitation echoes using this approach. Further noteworthy papers are those published by
Russo and Bowne (1962) and Kessler and Russo (1963). Kessler (1966) and Wilson (1966)
explored the use of cross correlation statistics to diagnose a best estimate of echo pattern
average motion. Wilson (1966) used the maximum value of the cross correlation coefficient
as an indicator of pattern development.
Two important conclusions were drawn from these early studies. The first of these was the
positive correlation between the predictability of precipitation features and their size: large
features tend to be longer lived than small ones. The second conclusion is an adjunct to the first,
namely that small scale features are generally short lived - typically a few tens of minutes.
These findings are consistent with early investigations into the multi-scaling properties of the
atmosphere and associated limits on atmospheric predictability (Lorenz, 1963; 1973).
The 1970s saw the further development of cross correlation-based nowcasting algorithms
and their automation. Zawadzki (1973) developed an optical device for measuring the
space-time statistical properties of radar inferred precipitation fields. Austin and Bellon
(1974) evaluated an automated, computerized pattern matching programme for nowcasting
precipitation up to 3 hours ahead. They concluded that the useful range of these nowcasts
varied with the nature and extent of the precipitation. Nonetheless, this approach was
shown to be consistently skilful up to one hour ahead over a wide range of events.
