Sampling-Rate Effects on Radar-Derived Rainfall Estimates.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH
Pagination or Media Count:
This study investigates the errors due solely to sampling intervals that occur with radar-derived total rainfall estimates. The study was limited to nine cold-front passages over eastern Texas, in the Fall of 1984. Digitized, 10.3 cm wavelength radar observations were recorded using a one minute sampling-rate. Total rainfall estimates, for 10 km by 10 km areas, based on these data were considered ground truth totals. Sample-rates, ranging from 5 to 60 minutes, were applied to the recorded data to calculate total rain estimates for each sample rate. These derived rain totals were compared to the ground truth totals, with the differences referred to as errors. These errors were plotted against the sampling-rate. They ranged from over 100 for sample intervals greater than 50 minutes, to less than 25 for intervals less than 15 minutes. The errors were also plotted against the number of samples taken. There was no significant increase in estimate accuracy when greater than seven samples were taken per 80 minute period. Other variables, the mean rain rate, total rain, sequential variability, storm width, and storm speed of movement, were found to have very low correlations with the errors.