Sensitivity and Uncertainty Analysis of the QUAL-TX Model to Aid In the Management of the Lower Colorado River, Texas

Environmental models, even relatively simple ones, have uncertainty. A practical method for assessing the uncertainty of a given model is the use of Monte Carlo Simulations (MCS), which provide an understanding of the distribution of potential model outputs given a distribution of different model inputs. This presentation will describe the analysis and present specific results for a sensitivity and uncertainty analysis of the QUAL-TX model, which is similar to the nationally more widely used QUAL2E model. The QUAL-TX model, which is a steady-state BOD/DO/nutrient model, was developed and calibrated to a river in central Texas. The primary constituent of concern for the modeling effort was dissolved oxygen. The objective of this effort was twofold: 1) to determine the sensitive parameters that most affected the dissolved oxygen results of the model and 2) to estimate the uncertainty in the model output using a MCS. In particular, the concern was to obtain an understanding of the predicted remaining assimilative capacity of the river for dissolved oxygen (i. e., the amount of dissolved oxygen (in mg O2 /L) above the state criteria remaining). For the sensitivity analysis, a base case of a standard state "permit evaluation" run was established using the calibrated model at critical summer low-flow conditions and all permitted dischargers at their fully permitted limits. The sensitivity of the model's predicted dissolved oxygen concentration was determined by performing a one-at-a-time parameter value change for 63 parameters within the model. For each of the 63 parameters, the value was set at its maximum and minimum to determine the model response to the extremes, resulting in 126 model runs. The parameters chosen to vary included reaeration, sediment oxygen demand, and BOD decay rates, as well as uncertain model input, such as wasteload concentration and headwater (i. e., upstream boundary) flow rate. Spider plots showing the slope of the change in assimilative capacity of the river vs. percent change in parameter value were plotted in order to determine the parameters with the greatest impact. Once the most sensitive parameters were selected from the sensitivity analysis, a MCS was performed to obtain a frequency of dissolved oxygen criteria violation, given the parameter uncertainty. The MCS considered the chemical or biological parameters that are uncertain and sensitive. The intent was to understand the effect of parameter uncertainty for those model parameters that are "beyond management control" without additional field data to constrain their ranges. From the 63 parameters tested, 15 were chosen as sensitive to the dissolved oxygen results. Of particular interest were reaeration and sediment oxygen demand, which showed the most impact on the model results. Results of the MCS indicated that under the future permitting conditions, flow regimes will be such that the dissolved oxygen standard will be not be met a majority of the time. As a result, although there is uncertainty in the modeling result, the future water quality of the river needs to be considered when managing future growth and water needs. This result aided the decision makers in developing a water management plan that would protect the river, while trying to address the needs of the surrounding communities. The sensitivity analysis also aided in understanding the model parameters that would be most important to refining the model calibration.