The fractal dimension of an individual ¯oc is a measure of the complexity of its external shape. Fractal dimensions can also be used to characterize ¯oc populations, in which case the fractal dimension indicates how the shape of the smaller ¯ocs relates to that of the larger ¯ocs. The objective of this study is to compare the fractal dimensions of ¯oc populations with those of individual ¯ocs, and to evaluate how well both indicate contributions of sediment sources and re¯ect the nature and extent of ¯occulation in streams.
Suspended solids were collected prior to and during snowmelt at upstream and downstream sites in two southern Ontario streams with contrasting riparian zones. An image analysis system was used to determine area, longest axis and perimeter of ¯ocs. The area±perimeter relationship was used to calculate the fractal dimension, D, that characterizes the ¯oc population. For each sample, the fractal dimension, D i , of the 28 to 30 largest individual ¯ocs was determined from the perimeter±step-length relationship. Prior to snowmelt, the mean value of D i ranged from 1 . 19 (Cedar Creek, downstream) to 1 . 22 (Strawberry Creek, upstream and downstream). A comparison of the means using t-tests indicates that most samples on this day had comparable mean values of D i . During snowmelt, there was no signi®cant change in the mean value of D i at the Cedar Creek sites. In contrast, for Strawberry Creek the mean value of D i at both sites increased signi®cantly, from 1 . 22 prior to snowmelt to 1 . 34 during snowmelt. This increase re¯ects the contribution of sediment-laden overland ¯ow to the sediment load. At three of the sampling sites, the increase in fractal dimensions was accompanied by a decreases in eective particle size, which can be explained by an increase in bed shear stress. A comparison of fractal dimensions of individual ¯ocs in a sample with the fractal dimensions of the ¯oc populations indicates that both fractal dimensions provide similar information about the temporal changes in sediment source contributions, about the contrasting eectiveness of the riparian buer zones in the two basins, and about the hydraulic conditions in the streams. Nevertheless, determining the individual fractal dimensions of a set of large ¯ocs in a sample is very time consuming. Using fractal dimensions of ¯oc populations is therefore the preferred method to characterize suspended matter.