On the Forecasting of Frontal Rain Using a Weather Radar Network

Abstract

This paper is concerned with the quantitative forecasting of hourly rainfall for the period 0–6 h ahead using linear extrapolation techniques. It deals with results obtained as part of the Meteorological Office Short Period Forecasting Pilot Project. The primary data used in this study are composite maps of rainfall echo distribution generated automatically and in real time using digital data received from a network of four weather radars covering parts of England and Wales. Forecasts have been derived during a total of 29 frontal rainfall events between November 1979 and June 1980. The forecasts wore derived both subjectively in real time and objectively using a computerized echo centroid tracking technique. The objective procedure, which was used to derive forecasts on a grid of 32 × 32, 20 km squares, is a practical way of quickly producing detailed forecasts for a large, number of target areas but its accuracy suffers from a number of factors. The subjective procedure, which was applied to a single target zone, was used to investigate some of the sources of error and their impact on forecasts. It is shown that radar rainfall measurement errors accounted for as much as half of the errors in the forecasts, and it is suggested that the biggest improvements in forecast accuracy are likely to accrue from improved analysis of the radar data prior to input into the forecast procedure. The radar measurement errors are due more to the variability of echo intensity with height than to straightforward radar calibration difficulties. Subtle procedures are required to identify these errors based on an analysis of the meteorological situation in which the radar data are viewed in the context of other kinds of meteorological information. Factors such as the development and decay of rainfall systems, which lead to the breakdown of the basic assumption underlying the linear extrapolation approach, accounted for about a quarter of the errors in the forecasts.