Hydrological application of S band radar reflectivities in South Africa.

The South African National Precipitation Research Project (NPRP) operates a Russian built MRL-5 dual wavelenth (S and X band) radar. This radar was purchased by the Water Research Commission (WRC) primarily to measure rainfall over the Wilge River catchment. This catchment is the major source of water for the industrial heartland of South Africa. The WRC also support the University of Pretoria (UP) to conduct research on the hydrological applications of this radar set.

During the summer seasons October to March 1995/96, 1996/97 and 1997/98 meteorologists from UP with NPRP support operated a small rainfall network some 50 km east of the radar site. This network comprised of 16 automatic rain gauges deployed in a 16 sq. km area. Each rainfall gauge recorded the time and date of each 0.2 mm rainfall on an event datalogger. Apart from maintaining and calibrating the MRL-5 radar the NPRP staff also archived S band reflectivity on Constant Altitude Plan Position Indicator (CAPPI) format as well as in volume scan mode.

Analysis of this rainfall network and radar reflectivity data, using the Marshall Palmer drop size distribution, showed that S band data can produce area rainfall estimates far superior to those possible with the current rainfall gauge network in the NPRP research area. This is a very pleasing result considering that the NPRP rainfall network, comprising approximately 400 thousand sq. km, is probably the only such network in Africa.

Inter comparison between the radar data and gauge data also showed that the radar estimation of rainfall remains good even at individual rainfall pixel resolution. The radar does very well under conditions of general rainfall or when large well organized cloud clusters abound. Under these conditions analysis indicated that the radar overestimates by approximately 10% when the rainfall accumulation over the entire season is considered. Individual case studies during one or two day events produce similar results. During events when a bright band is present results suffer. However data incorporating radar derived storm structure may resolve this problem.