April 1991 to April 1994
The ARC system comprises two main stages of processing
Adaptive Calibration
It is widely beleived that a point rain gauge rainfall measurement is a much more accurate estimate of the true surface rainfall intensity than the corresponding radar rainfall estimate. Thus at each calibration rain gauge site the comparison between raingauge and radar rainfall intensity estimates are used to produce a calibration parameter. The calibration parameter values are estimated using a recursive algorithm which allows the parameter values at each point to change smoothly through time. The calibration parameter value at any point in time is thus a function of all the previous radar and rain gauge rainfall data at that point, with the most recent data given greatest weighting. These calibration parameters are used together to generate a calibration surface for the entire radar domain with smoothly changing values in space. The whole radar image is then recalibrated using this calibration surface.
Calibration Performance Results
The calibration is performed using data from twelve tipping bucket raingauges (TBR's) within the radar area (about 18,000 km2). Data from about sixty aditional raingauges, sited throughout this radar area, are used to perform the assessment.
The ARC calibrated radar data are shown to estimate the storm total rainfall to within a few percent of that measured by the independent raingauges, and the magnitude of the squared errors between the radar and raingauge time series are reduced by one half. The System is able to perform adequate calibration of images displaying Bright Band.
This system is currently being implemented on-line for the NRA (North West).
See also Jim Mann's radar research page, for other recent radar work in the group.