Content area
Estimation of the sampling error in the rainfall measurement is an important issue because the accuracy of these measurements can influence the accuracy of the results from global circulation models (GCMs). This study addressed the issue of the sampling errors in the rainfall measurements from space using the statistical analyses of the rainfall data. The rainfall data collected during 1988, in the vicinity of Darwin, Australia, were analyzed in this study. The statistical analyses were conducted in one, two, and three dimensions. One dimensional analyses were performed on area averaged time series of land, ocean, and combined precipitation of Darwin I and Darwin II subsets. A strong diurnal signal was detected from periodograms and correlograms. Periodograms and correlograms also indicated the presence of the semidiurnal cycle. Simulated sampling error studies conducted using area averaged precipitation time series for Darwin I and II, indicate that the sampling errors range from 3 to 20% of the mean for sampling intervals from 5 to 12 hr. Removal of the semidiurnal cycle from the data reduced the errors by about 40 to 50%. Sampling errors were as high as 60% for sampling interval of 24 hr. In this case, the removal of diurnal cycle from the data, reduced the sampling errors by about 30 to 40%. Two dimensional rainfall fields were obtained by averaging the data along West-East, North-South, and along Time axes. Two dimensional periodograms estimated for these fields show the diurnal and semidiurnal cycle very clearly. The variations in the data are primarily in time. The rainfall fields were found to be almost isotropic in space. In three dimensional analyses, periodograms were obtained using a three dimensional Fourier transform which were used to obtain the sampling errors using North-Nakamoto method. The sampling errors range from 5 to 30% for sampling intervals from 5 to 13 hr. Significant increase in the sampling errors can be noticed for sampling interval of 12 hr. Sampling errors for sampling interval of 24 hr were 45% for Darwin II and about 64% for Darwin I. These results can be used in the satellite mission planning activities.