Fine sediments or particles are important determinants of water quality and ecosystem health. Various contaminants such as nutrients and heavy metals are transported by fine particles and the deposition of fine particles at the surface of stream beds often causes serious impairments of benthic ecosystems. For these reasons understanding fine particle transport within watersheds and interaction with the stream bed are important for assessing impairment of water quality and aquatic ecosystems. Besides a qualitative understanding, restoration and remediation efforts would benefit from quantitative models that predict fine particle dynamics. This thesis adopted an approach that first explored the patterns of fine particle transport within California watersheds, from those patterns processes were investigated that dominated particle transport, and then finally, developing a quantitative model with the goal being able to predict particle dynamics that replicated observed patterns and represented dominant processes.
The rich data base of US Geological Survey stream monitoring data within California provided opportunities for recognizing common patterns in fine sediment loading rates as a function of flow rate. The majority of 38 minimally developed watersheds with extensive flow and particle transport data illustrated a common dependence of particle loading rate on flow rate. Physical surveys of watersheds, collection of stage-discharge data from historical gauging records, and sediment bed analysis revealed that gravel-bedded streams underwent a transition to accelerated rates of fine particle transport above a flow rate sufficient to initiate mobilization of bed sediments. Additionally, continuous flow and turbidity data had hysteresis loops when fine particle concentration is plotted against flow rate that demonstrated stream bed release of fine particles and a limited supply of those particles. These patterns were in qualitative agreement with observations by others over the last 75 years.
The transition from pattern recognition to process analysis required incorporation of the dominant processes controlling fine particle dynamics within gravel-bedded streams into a model. The process analysis was guided by the use of continuous flow and turbidity data at two locations on the Russian River in California to test process descriptions and then calibrate a quantitative model to represent those processes. The resulting process model coupled fine particle retention within the sediment bed by filtration and sedimentation with the release of accumulated fine particles in response to flood events. Model parameters such as a critical flow rate required to initiate sediment bed fluidization, the maximum fine particle storage capacity within the watershed, and background particle concentration for the watershed were identified directly from monitoring data. Model calibration consisted of optimizing the filtration parameter and the sediment bed fluidization parameter over two or three years of data. Overall the modeled fine particle release was within 5% of what was measured during flood events.
The successful process modeling for two sites formed the basis for partially validating the model for data not used in calibration within the Russian River of California and testing its applicability to other watersheds. The calibrated model parameters combined with over a year of 15-minute flow data was able to replicate within 35% the observed fine particle release by flood events. Six other watersheds were utilized in testing the calibration of the model and providing a preliminary analysis of the sensitivity of the two model parameters representing filtration and fluidization. The model could be successfully calibrated to these watersheds, and there was a limited range observed for the fluidization parameter and a possible watershed area dependency on the filtration parameter. As a further test of the model, particle loading rates were generated from measured flow data and these loading rates were similar to those observed. This agreement provided a demonstration that the model was able to quantitatively replicate the shape and scatter of particle loading observations, both under current and historical conditions.
