Sampling errors for raingauge-derived mean areal daily and monthly rainfall.

Abstract Radar data from two geographical locations are used to simulate the mean standard error in using a sparse raingauge network to estimate daily and monthly mean areal convective rainfall over areas ranging from 45 000 to 180 000 km 2 . It was found that a network with a regular configuration gave somewhat less-variable errors than the random raingauge network, although the mean errors were very similar. The difference became more pronounced for the very sparse networks. The mean standard error for a particular network and rainfield was found to be a function of the number of gauges in the network, the raining fraction of the area and the ratio of the standard deviation over the mean of the non-zero portion of the rainfield. A simple three-parameter relationship was proposed to relate the mean standard error, expressed as percentage of the mean areal rainfall, to these variables. It was found that a single relationship was able to explain 63% of the variability in the estimated mean standard estimation error, combining data from both regions. Finally, the domain over which the relationship is able to make reasonable predictions is discussed, the principal constraint being that the raining fraction of the area should not exceed 0.5 for networks with > 200 raingauges.