Modelling shallow and narrow urban salt-wedge estuaries: Evaluation of model performance and sensitivity to optimise input data collection

Complex three-dimensional estuarine hydrodynamic models require large quantities of high-resolution data for model forcing and initialisation. The data are often expensive and difficult to collect with high accuracy (e.g. bathymetry data, riverine flows, water depths, etc.). It may be possible to reduce input data requirements, whilst maintaining predictive capabilities. This is the first study that assesses the sensitivity of a three-dimensional hydrodynamic model of a shallow and narrow urban salt-wedge estuary to input data used for model forcing and initialisation. The model was built using the TUFLOW FV modelling platform and its performance was tested against high-resolution water level, flow velocity, vertical salinity and temperature distribution data. A number of scenarios were used in which data used for model forcing and initialisation, including flow rates, salinity and temperature, wind, bed roughness, bathymetry and vertical mesh discretisation were systematically varied. To assess the sensitivity of model outputs, model predictions were compared to the optimised model predictions for ten periods covering different hydrologic and hydrodynamic conditions. The analysis showed that all model outputs (i.e. water level, velocity, temperature and salinity) were influenced by large and localised water inputs. Due to limited wind fetch of narrow water bodies, wind inputs are expected to have limited impact on hydrodynamic model outputs. However, in this study, flow velocity, salinity and temperature outputs were all influenced by wind inputs. Whilst, accurate bathymetry data are considered essential for developing three-dimensional hydrodynamic models of shallow regions, in this study, uncertainty in the bathymetry data had limited influence on model outputs. Removal of stormwater inputs (i.e. 208 stormwater drains), setting constant salinity for fresh water inputs, weekly averaging of temperature and errors in bathymetry all had minimal impact on model outputs. The results of this case-study can help inform future modelling exercises of narrow and shallow salt-wedge estuaries by focussing efforts on the most important input data. This would potentially lead to substantial reductions in cost and time needed to set up the model.