Although regressions of mean concentration and discharge indicated a decrease of daily mean concentration at similar daily mean discharges over the study period, a different method can be used to examine for the same effect. Instantaneous sample concentrations and associated stream discharges are used in this method, which is independent of estimation procedures for calculating daily mean values.
Sediment concentrations from samples collected during the study period were examined for significant changes with time when collected at similar discharges. Sediment concentrations were taken from cross-section samples and single-vertical samples that had been analyzed for sand division. Plots of all sediment concentrations versus time indicated that the range of sediment concentration shifted downward in most affected streams during the study period. To test whether this apparent shift was possibly independent of stream discharge, sediment concentrations from samples were separated into narrow ranges of stream discharge. The change in sediment concentration with time was then evaluated for each discharge range. The null hypothesis of no significant slope with time was tested for discharge and concentration with the non-parametric Kendall tau analysis.
Ranges of stream discharge were based on the percentage distribution of discharges sampled. Six discharge ranges were selected for each gaging station, including three 20-percentile ranges, two 10-percentile ranges, and an 8-percentile range that excluded the top 2 percent of sampled discharges ( table 14) . (For Clearwater Creek, the number of samples in the "30-20" percentile range was only 9, so a wider range of "30-10" was used for that station.) Table 14 shows that at least 70 percent of the sampled discharges at each gaging station were greater than the 50-percent exceedance discharges listed in table 11. Simply put, samples were collected more often at higher discharges.
No annual or seasonal separations were made of the sediment concentrations. The concentration data were further divided into sand concentrations within a discharge range using the sand-division percentage. Because sand transport is more dependent on hydraulic conditions (Allen, 1985), limiting the analyses to sand concentration is more likely to reveal changes at similar discharges.
Plots of sand concentration by time are given in figures 62 , 63 , 64 , 65 , 66 , 67 . The non-parametric Kendall tau analysis was performed on both the discharge and sand concentration data. Where the analyses indicated that a trend of discharge with time was not significant at the 90-percent confidence level, the tau and probability values for concentration trends with time are given in table 14. Only the Muddy River below Clear Creek showed no significant trends of concentration with time in any discharge range at the 95-percent confidence level. All other stations showed at least one negative correlation of concentration with time at similar discharges at the 95-percent confidence level.
