Date
5-2015
Document Type
Dissertation
Degree Name
Ph.D.
Institution
Oregon Health & Science University
Abstract
Streams and rivers provide many important functions and services for society, of which many depend not only on the quantity of water available, but also the stream water quality. While quantity (i.e., streamflow) is routinely forecast for many U.S. streams several days into the future, the forecasting of stream water quality remains relatively limited. This dissertation utilizes the extensive existing networks of observed and forecast streamflow, precipitation and air temperature as the basis for practical forecasting of stream turbidity for three days into the future. To accomplish this, an improved power-law based model for turbidity during hydrologic events was developed and applied to ~6000 events from 110 U.S. gages. The resulting event model parameters were examined in the context of catchment characteristics and event characteristics (e.g., hydrologic, meteorologic and antecedent moisture conditions) in order to understand the variability of turbidity response between streams and for different events within a particular stream. The results indicated that gage median parameter values were mainly correlated with catchment land cover and baseflow index, while the individual-event parameter values for a particular stream were largely correlated with the antecedent moisture conditions preceding the event. These analyses were, in part, facilitated by the use of a power-law parameter decorrelation methodology, which clarified the relationships between the power-law coefficient and the catchment and event characteristics. The information gained from these analyses was used to develop regression equations to forecast turbidity model parameters based on the event characteristics. Using archived streamflow and meteorologic forecast inputs and gage-specific regression equations, turbidity forecasts were made for events from two mid-sized streams. The turbidity forecast errors were examined, and the results indicated that the forecasts were “useful” compared to a persistence reference. The uncertainty in the turbidity forecasts due to uncertainty in the streamflow forecasts was also explored. Overall, the results indicated that practical and useful turbidity forecasts can be produced from currently-available observed and forecast inputs. The widespread availability of these inputs for streams across the U.S., and the value of turbidity as a surrogate for other water quality constituents, suggests that the forecasting demonstrated here could be implemented for many other streams and for additional water quality constituents.
Identifier
doi:10.6083/M42J69M1
Division
Division of Environmental and Biomolecular Systems
School
School of Medicine
Recommended Citation
Mather, Amanda L., "Stream turbidity modeling : a foundation for water quality forecasting" (2015). Scholar Archive. 3648.
https://digitalcommons.ohsu.edu/etd/3648