Optimal Sampling and Analysis Using Two Variables and Modeled Cross-Covariance Functions

View More View Less
  • 1 Department of Aviation, University of North Dakota, Grand Forks 58202
© Get Permissions Rent on DeepDyve
Restricted access

Abstract

An objective analysis technique of the Eddy-Gandin type is discussed which permits the analysis and investigation of multivariate as well as univariate data sets. A multivariate data configuration consisting of raingage/radar precipitation measurements is analyzed. This includes determination of spatial-temporal correlation and cross-correlation structure functions from the observations, univariate and multivariate analysis of the surface precipitation field, and an exploration of the relative worth of different Z-R relationships in a multivariate environment. Investigative results indicate that the structure functions were quite dependent on the spatial-temporal form of the precipitation system, that the multivariate analyses were consistently better than the univariate analyses, and that the Z-R relationship did not normally produce noticeable differences when used in a multivariate environment.

The incorporation of this analysis technique with a nonlinear programming algorithm for use as an experimental design tool is also discussed. The potential of this design methodology is presented using raingage/radar structure functions. Optimal spatial sensor configurations are determined for one sensor type, and trade-offs between different instrument types using that configuration are explored.

Abstract

An objective analysis technique of the Eddy-Gandin type is discussed which permits the analysis and investigation of multivariate as well as univariate data sets. A multivariate data configuration consisting of raingage/radar precipitation measurements is analyzed. This includes determination of spatial-temporal correlation and cross-correlation structure functions from the observations, univariate and multivariate analysis of the surface precipitation field, and an exploration of the relative worth of different Z-R relationships in a multivariate environment. Investigative results indicate that the structure functions were quite dependent on the spatial-temporal form of the precipitation system, that the multivariate analyses were consistently better than the univariate analyses, and that the Z-R relationship did not normally produce noticeable differences when used in a multivariate environment.

The incorporation of this analysis technique with a nonlinear programming algorithm for use as an experimental design tool is also discussed. The potential of this design methodology is presented using raingage/radar structure functions. Optimal spatial sensor configurations are determined for one sensor type, and trade-offs between different instrument types using that configuration are explored.

Save