Collaborative Approach to Calibration of a Riverine Water Quality Model

Abstract

Stakeholder involvement in watershed management decisions remains challenging, particularly when one is relying upon modeling efforts as one of the primary means to understand and communicate the complicated processes controlling a natural riverine system. To determine how to meet the instream water quality standards as required by the total maximum daily load (TMDL) program, a QUAL2Kw water quality model was developed and calibrated for the Jordan River, Utah. To improve transparency, increase stakeholder engagement, and reach agreement more effectively, a collaborative approach was taken to calibrate the model that included Utah Division of Water Quality (UDWQ) staff, consultants, key stakeholder representatives, and independent experts. The study reach of the Jordan River flows 83 km from Utah Lake to the Great Salt Lake through the urbanized Salt Lake Valley in Utah. Several segments of the river have been listed by UDWQ as impaired for a number of constituents, but the present effort focuses on the dissolved oxygen (DO) impairment. In cooperation with stakeholder groups, four seasonal synoptic sampling surveys and additional research studies were conducted for use in calibrating and validating the model. The collaborative calibration process involved many meetings and workshops where model performance was iteratively evaluated in order to reach consensus on model inputs and parameterization. The end result of the collaborative calibration of the QUAL2Kw model was the development of a scientifically defensible tool used to establish a consensus that decomposition of organic matter and poor reaeration are the primary causes of the DO impairment in the Jordan River. Without the collaborative approach to calibration, agreement regarding the causes of water quality impairments would not likely have occurred. While clearly improving stakeholder believability in the model, a comparison of the collaborative calibration model performance to an autocalibrated version demonstrated that both approaches resulted in similar predictive accuracy.