ABSTRACT:

The MHIA model combines a soil water balance with two response branches (fast and slow), each routed using a Gamma-type instantaneous unit hydrograph (IUH). The objective is to represent, in a conceptually consistent and computationally efficient manner, the processes of storage, release, and flow propagation within a catchment. The formulation is designed to support gradient-based optimization, enabling its use within differentiable architectures or hybrid neural networks, while also allowing global optimization through Particle Swarm Optimization (PSO).

ABSTRACT:

This data represents outcomes from experiments focused on optimizing a distributed hydrological model via a Surrogate-Assisted Evolutionary Algorithm (SAEA). Two scripts perform data analysis, visualization, and strategy comparison. The first script imports pertinent datasets, visualizes the model's performance, and applies the Generalized Likelihood Uncertainty Estimation (GLUE) methodology to create a confidence band for predictions. The second script manages data for multiple generations and compares Monte Carlo (MC) and Evolutionary Algorithm with Surrogate Modeling (EA-SM) strategies, visualized in a scatter plot. Together, these scripts provide comprehensive insights into the model's performance, efficiency, and potential enhancements.

ABSTRACT:

MHIA is a continuous hydrological model that computes a balance of the volume of water in the soil, taking into account the following processes: precipitation, infiltration, percolation, evapotranspiration and exfiltration. From these variables, the model evaluates surface and subsurface runoff, generating a hydrograph at the outlet of the modelled catchment. As input data, the model needs to be fed with time series of precipitation and temperature, with whatever time resolution. The model has 14 parameters that must be defined by the user or calibrated from observed discharge time series.