Abstract

This work presents and illustrates the application of hydroPSO, a novel multi-OS and model-independent R package used for model calibration. hydroPSO allows the modeller to perform a standard modelling work flow including, sensitivity analysis, parameter calibration, and assessment of the calibration results, using a single piece of software. hydroPSO implements several state-of-the-art enhancements and fine-tuning options to the Particle Swarm Optimisation (PSO) algorithm to meet specific user needs. hydroPSO easily interfaces the calibration engine to different model codes through simple ASCII files and/or R wrapper functions for exchanging information on the calibration parameters. Then, optimises a user-defined goodness-of-fit measure until a maximum number of iterations or a convergence criterion are met. Finally, advanced plotting functionalities facilitate the interpretation and assessment of the calibration results. The current hydroPSO version allows easy parallelization and works with single-objective functions, with multi-objective functionalities being the subject of ongoing development. We compare hydroPSO against standard algorithms (SCE_UA, DE, DREAM, SPSO-2011, and GML) using a series of benchmark functions. We further illustrate the application of hydroPSO in two real-world case studies: we calibrate, first, a hydrological model for the Ega River Basin (Spain) and, second, a groundwater flow model for the Pampa del Tamarugal Aquifer (Chile). Results from the comparison exercise indicate that hydroPSO is: i) effective and efficient compared to commonly used optimisation algorithms, ii) “scalable”, i.e. maintains a high performance for increased problem dimensionality, and iii) versatile to adapt to different response surfaces of the objective function. Case study results highlight the functionality and ease of use of hydroPSO to handle several issues that are commonly faced by the modelling community such as: working on different operating systems, single or batch model execution, transient- or steady-state modelling conditions, and the use of alternative goodness-of-fit measures to drive parameter optimisation. Although we limit the application of hydroPSO to hydrological models, flexibility of the package suggests it can be implemented in a wider range of models requiring some form of parameter optimisation.