ANALYSIS OF DARWIN RAINFALL DATA - IMPLICATIONS ON SAMPLING STRATEGY

Rainfall data collected by radar in the vicinity of Darwin, Australia, have been analyzed in terms of their mean, variance, autocorrelation of area-averaged rain rate, and diurnal variation. It is found that, w hen compared with the well-studied GATE (Global Atmospheric Research P rogram Atlantic Tropical Experiment) data, Darwin rainfall has larger coefficient of variation (CV), faster reduction of CV with increasing area size, weaker temporal correlation, and a strong diurnal cycle and intermittence. The coefficient of variation for Darwin rainfall has l arger magnitude and exhibits larger spatial variability over the sea p ortion than over the land portion within the area of radar coverage. S tationary and nonstationary models have been used to study the samplin g errors associated with space-based rainfall measurement. The nonstat ionary model shows that the sampling error is sensitive to the startin g sampling time for some sampling frequencies, due to the diurnal cycl e of rain, but not for others. Sampling experiments using,data also sh ow such sensitivity. When the errors are averaged over starting time, the results of the experiments and the stationary and nonstationary mo dels match each other very closely. In the small areas for which data are available for both Darwin and GATE, the sampling error is expected to be larger for Darwin due to its larger CV.