The Design of a Multivariate Mesoscale Field Experiment.

Radar reflectivities and rain gage data can be combined in many ways to estimate convective-storm surface rainfall.However, optimum (interpolation-produced) esti mates used to evaluate modification experiments or to imple ment sampling techniques require that statistical properties of such estimates be knovm.Such properties derive from the storm structure, implied by the observed data, and are deduced by using space-time covariance and cross-covariance functions.A four dimensional Gaussian-damped function can reflect relevant characteristics (physical and statistical) of Southeastern Montana convective systems.Functional parameter values relate to system features of size, raotionspeed and preferred storm track.An average Southeast Montana system compares and contrasts with the features of an Oklahoma counterpart.The optimum interpolation methodology is enhanced to account for multivariate means and variances.Bivariate analyses that use radar/rain gage data sets are shown supe rior to the best univariate results.The analyses reflect patterns derived from radar rainfall estimates and scaled to rain gage magnitudes.The influence of a Z-R relation ship on analysis accuracy is minimal and the model's signal recoverability qualities are shown.Consequences of filter ing data set observations improperly are discussed.The development of an experimental-design evalua tion function is completed through modelling the parameter means and variances.Predictand-related sensors are shown essential to network design.Trade offs in multivariate sensor deployments (spatial and temporal) are explained.Deployment along and across a preferred storm track is related to covariance anisotropy, gage density, temporal sampling intervals, the availability of radar data, and the interrelationships among the multivariate predictor data sets.Moreover, optimal sensor orientation to sample a moving convective system is found best to observe the system's accumulated rainfall pattern.ACKNOWLEDGEMENTS This dissertation represents a culmination of efforts by many friends and associates.Surely the list of support would be long.However, significant financial support by the National Oceanic and Atmospheric Administration and the National Weather Service created an opportunity to accelerate the technique applications reported herein.I am also grate ful to Dr. B e m i e Silverman and his staff at the Bureau of Reclamation for enthusiastically providing a data base and computer resources.Amos Eddy, through his creative genius, represents a fundamental force behind my work.Though philosophically we differ, I admire his many positive traits.Through the years he has patiently prodded my intellect with generous doses of his keen practical knowledge.Dr. Pat Brady provided a foundation that insured my work would not take forever to complete.He never failed to answer and reanswer my many questions.I have had the continual cooperation and interest of Dr.