An experimental design model has been developed to evaluate trade-offs involved in the optimal sampling of surface rainfall. The model reveals that deployment of statistically adequate sensor arrays depends critically on the storm physics, the sensor engineering aspects, and the procedure chosen for analysis of information from the deployed network. The model is used to examine the effect of sensor density reduction on the ability of a sampling system to detect signal variation. Predictand-related sensors are shown to be essential to network design. Options in multivariate sensor deployments (spatial and temporal) are explained. Deployment along and across a preferred storm track is related to convective system anisotropy, rain gage density, temporal sampling intervals, availability of radar data, and interrelationships among the multivariate predictor data sets. Any proposed sensor array is shown to represent a balance between the demands of economy and climatology.