Comprehensive Statistical Analysis of Raindrop Size Distribution Data
Project Summary
This research is directed toward
developing improved statistical techniques for analyzing observed samples of
raindrop (or cloud droplets) sizes to obtain mathematical descriptions of the
underlying population drop size distributions (DSDs). The objective is to obtain more accurate descriptions
of the population DSDs, especially with small sample sizes, than are provided
by the widely-used moment methods (which are biased and subject to substantial
errors). Methods considered include the
maximum likelihood and L-moment approaches; modifications of the standard
versions of these techniques will be made to deal with the practical problems
of (a) observations from surface-sampling instruments (most disdrometers) and
(b) observations lacking data on very small drops, because of instrument
shortcomings. Means of assessing the goodness
of fit of the resulting DSD functions to the sample data and to the underlying
population distributions will be established; this includes means for
evaluating the fit to the size distributions of important physical quantities
like the mass concentration (as contrasted to the usual measures of agreement
with the number distribution). The
methods will be tested first on computer-simulated data (for which the true population
distributions are known) and then on actual samples of surface and aircraft drop
size data.