hydroMOPSO: A Versatile Multi-Objective Optimization R Package for Calibration of Environmental and Hydrological Models
Abstract
Several real-world engineering applications face problems where competing objectives need to be optimized at the same time, and where the improvement in one objective can only be achieved at the expense of worsening one or more of the remaining objectives. Among the many nature-inspired families of heuristic algorithms, several multi-objective particle swarm optimizers (MOPSOs) have been developed in the last decades to find the Pareto-optimal front for multi-objective problems.
In this work we introduce hydroMOPSO, a novel multi-objective R package that combines two search mechanisms to ensure population diversity and accelerate its convergence towards the Pareto-optimal front. hydroMOPSO is model-independent, i.e., its calibration engine can be linked to any model code, including models available in R (e.g., TUWmodel, airGR, topmodel), but also any other model that needs to be run from the system console (e.g. SWAT+, Raven). In addition, hydroMOPSO is platform-independent (e.g., GNU/Linux, Mac OSX, Windows), and it can take advantage of multi-core machines and network clusters. To assess the accuracy and diversity of solutions provided by hydroMOPSO, we compare its results against another recent multi-objective optimization R package (caRamel), using three well-known DTLZ benchmark optimization problems. Additionally, we test its performance in the calibration of two real-world hydrological models (TUWmodel and GR4J) to obtain a good representation of both high and low streamflows. Our results show that hydroMOPSO is more effective in solving the three DTLZ benchmark problems and the two hydrological model calibrations, delivering more accurate and diverse Pareto fronts. Beyond the results obtained for the previous study cases, we believe hydroMOPSO can be included within existing modeling frameworks requiring some form of multi-objective or multi-variable parameter estimation.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.H15M0961M