Various statistical methods were applied to spatially discrete data from 14 intensively sampled small estuarine systems in the mid-Atlantic U.S. The number of sites per system ranged from 6 to 37. The surface area of the systems ranged from 1.9 to 193.4 km2. Parameters examined were depth, bottom temperature, bottom salinity, surface chlorophyll a, bottom dissolved oxygen, lead concentration in sediments, silt-clay content of sediments, and number of infaunal benthic species. Statistical methods included means, standard deviations, coefficients of variation, empirical cumulative distribution functions, and contours determined by bivariate interpolation and interpolation by kriging. All of these methods were found to be appropriate depending upon the purpose of the characterization. Contouring was applied only to those systems with at least 23 discrete sample sites (7 systems). Cross-validation and randomization techniques were used to compare the two interpolation methods. Kriging was advantageous over bivariate interpolation when moderate to strong spatial correlation existed in the residuals (that is, after removal of the spatial trend with a nonparametric regression model). When kriging was conducted, the removal of the trend was necessary if the stationarity assumption was to be valid. The Delaware/Maryland coastal bays are shallow, well-mixed (horizontally and vertically) systems that exhibit little or no spatial correlation for the parameters examined. The South and Severn Rivers, subsystems of the Chesapeake Bay, exhibited moderate to strong spatial dependence for some parameters. Randomization techniques were used to evaluate the effect of decreasing the number of sites in kriged parameters. Based upon these randomizations, it was found that 23 discrete sites could be used for kriging in estuaries with characteristics similar to those in the mid-Atlantic and if the samples were collected with a comparable design.