HySalt: A metamodel for estimating water level and spatial-temporal distribution of salinity in estuarine environments

Estuarine salinity results from the interaction between tidal forcing, sea level variability and freshwater discharge, and is commonly simulated using three-dimensional hydrodynamic models. Although physically robust, these models are computationally expensive and limit their use for rapid assessments and large scenario ensembles. This paper presents HySalt, a hybrid metamodel developed to efficiently estimate water level and the spatio-temporal distribution of salinity along the full water column at predefined stations within an estuary over a tidal cycle. HySalt combines a reduced set of high-resolution three-dimensional hydrodynamic simulations with statistical learning techniques. A synthetic ensemble of hydrodynamic conditions, defined by eight forcing parameters, seven of which vary across the simulation ensemble, is generated and dynamically downscaled using a calibrated Delft3D-FLOW model. Principal Component Analysis is applied to reduce the dimensionality of water level and salinity outputs, and Radial Basis Functions are used to interpolate the dominant modes, allowing rapid reconstruction of the corresponding fields for any combination of forcing parameters. HySalt is validated against the calibrated Delft3D-FLOW simulations, accurately reproducing the spatio-temporal variability of water level and salinity. In addition, a direct comparison with field observations is provided to assess the performance of the approach under real-world conditions. Results demonstrate the applicability of the metamodel to obtain fast and accurate estimates of water level and salinity fields for a wide range of coastal and riverine forcing conditions, offering a practical tool for the assessment of salt intrusion and the support of estuarine water management under present and future scenarios. A web-based dashboard enables interactive exploration of scenarios and salinity thresholds for decision support.