Modelling retention time in a clearwell

Clearwells are large water reservoirs often used at the end of the water treatment process as chlorine contact chambers. Contact time required for microbe destruction is provided by residence time within the clearwell. The residence time distribution can be determined from tracer tests and is the one of the key factors in assessing the hydraulic behaviour and efficiency of these reservoirs. This work provides an evaluation of whether the two-dimensional, depth-averaged, finite element model, River2DMix can adequately simulate the flow pattern and residence time distribution in clearwells. One question in carrying out this modelling is whether or not the structural columns in the reservoir need to be included, as inclusion of the columns increases the computational effort required. In this project, the residence time distribution predicted by River2DMix was compared to results of tracer tests in a scale model of the Calgary Glenmore water treatment plant northeast clearwell. Results from tracer tests in this clearwell were available. The clearwell has a serpentine baffle system and 122 square structural columns distributed throughout the flow. A comparison of the flow patterns in the hydraulic and computational models was also made. The hydraulic model tests were carried out with and without columns in the clearwell. The 1:19 scale hydraulic model was developed on the basis of Froude number similarity and the maintenance of minimum Reynolds numbers in the flow through the serpentine system and the baffle wall at the entrance to the clearwell. Fluorescent tracer slug injection tests were used to measure the residence time distribution in the clearwell. Measurements of tracer concentration were taken at the clearwell outlet using a continuous flow-through fluorometer system. Flow visualization was also carried out using dye to identify and assess the dead zones in the flow. It was found that it was necessary to ensure the flow in the scale model was fully developed before starting the tracer tests, and determining the required flow development time to ensure steady state results from the tracer tests became an additional objective of the work. Tests were carried out at scale model flows of 0.85, 2.06, and 2.87 L/s to reproduce the 115, 280, and 390 ML/day flows seen in the prototype tracer tests. Scale model results of the residence time distribution matched the prototype tracer test data well. However, approximately 10.5 hours was required for flow development at the lowest flow rate tested (0.85 L/s) before steady state conditions were reached and baffle factor results matched prototype values. At the intermediate flow, baffle factor results between the scale model and prototype matched well after only 1 h of flow development time, with improvements only in the Morril dispersion index towards prototype values with increased flow development time (at 5 h). Similar results were seen at the highest flow tested. For fully developed flow, there was little change in the baffle factor and dispersion index results in the scale model with varied flow rate. With the addition of columns to the scale model, there was no significant change in the baffle factor compared to the case compared to without the columns, but up to a 13.9 % increase in dispersion index as compared to the tests in the scale model without columns for fully developed flow. Further, the residence time distribution results from the scale model tests without columns matched the entire residence time distribution found in the prototype tests. However, for the model with columns, the residence time distribution matched the prototype curve well at early times, but departed significantly from it at times later in the tests. It appears the major effect of the addition of columns within a model clearwell is to increase the dispersion index and increase the proportion of the clearwell which operates as a mixed reactor. The results also showed there was good agreement between the physical model tests and River2DMix simulations of the scale model tests for both the flow pattern and residence time distributions. This indicates that a two-dimensional depth-averaged computer model such as River2DMix can provide representative simulation results in the case where the inlet flow is expected to be quickly mixed throughout the depth of flow in the clearwell