Annual loads and flow-adjusted concentration trends were estimated by use of water-quality and streamflow data collected from 1990 through 1999 at monitoring stations on two tributaries to Chesapeake Bay in Virginia—James River at Cartersville, Va., and Rappahannock River near Fredericksburg, Va. The effects of storm-sampling frequency on the accuracy and precision of load and trend estimates were determined by use of data sets containing 0, 20, 40, 60, 80, and 100 percent of all storm samples collected in these two basins of different size, relief, and land use. Data sets included a range of dissolved and particulate constituents for the 10-year period from 1990 to 1999 and the 5-year period from 1995 to 1999.

Loads of dissolved constituents were estimated with greater accuracy and precision with fewer storm samples than loads of particulate constituents in both basins and for both time periods. All constituent loads were estimated with greater precision with fewer storm samples in the James River than in the Rappahannock River for both periods. The high relief and smaller drainage area of the Rappahannock River Basin caused quicker and more variable stream response than in the James River Basin, which led to less precise load estimates of all constituents, regardless of how many storm samples were included. For the James River, the magnitudes of the load estimates in the 5-year period were close to the estimates from the same years for the 10-year period for the dissolved constituents, but were smaller for the particulate constituents. Load estimates were more variable for the Rappahannock River than for the James River during the shorter period. In both basins, all estimates in the 5-year period had higher prediction errors than those in the 10-year period. Overall, loads of dissolved constituents were estimated with greater accuracy and precision with fewer storm samples than loads of particulate constituents; loads of all constituents were estimated with greater accuracy and precision over the longer time period; and load estimates of all constituents were more precise and required fewer storm samples in the larger and less flashy James River Basin than in the Rappahannock River Basin.

As with load estimates, estimates of flow-adjusted concentration trends were sensitive to the length of the monitoring period and the size of the basin; however, trend estimates generally were less sensitive than load estimates to the number of storm samples in the data set. Trends in flow-adjusted concentrations were estimated reasonably well with fewer storm samples for both dissolved and particulate constituents in the James River for the 10-year period, with the exception of total suspended solids. Data sets containing more storm samples were needed to obtain reasonable trend estimates for the 5-year period in this river. For the 10-year period in the Rappahannock River, more storm samples were necessary than in the James River to obtain reasonable estimates of trends for all constituents. No significant trends were observed for the 5-year period in this river, so the effect of storm-sampling frequency could not be determined. Because of the small number of significant trends throughout these data sets, it was not possible to determine whether fewer storm samples were required for estimating trends of dissolved constituents than particulate constituents. The results indicate that more storm samples were necessary for accurate estimation of trends during the shorter time period and in the smaller and flashier Rappahannock River Basin.