Conditional interval reduction method: A possible new direction for the optimization of process based models. (December 2022)

• Record Type: Journal Article
• Authors: Hollós, R.; Fodor, N.; Merganičová, K.; Hidy, D.; Árendás, T.; Grünwald, T.; Barcza, Z.
• Is Part Of: Environmental modelling & software. Volume 158(2022)
• Publication Date: 2022-12
• Subjects: Model optimization -- Bayesian calibration -- Parameter constraints -- Decision tree; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2022.105556
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Application of process-based models at different spatial scales requires their proper parameterization. This task is typically executed using trial-and-error parameter adjustment or a probabilistic method. Practical application of the probabilistic methods is hampered by methodological complexity and lack of interpretability. Here we present a novel approach for the parameterization of process-based models that we call as conditional interval refinement method (CIRM). The method can be best described as the combination of a probabilistic approach and the advantages of the expert-based parameter adjustment. CIRM was demonstrated by optimizing the Biome-BGCMuSo biogeochemical model using maize yield observations. The proposed approach uses the General Likelihood Uncertainty Estimation (GLUE) method with additional expert knowledge, supplemented by the construction and interpretation of decision trees. It was demonstrated that the iterative, fully automatic method successfully constrained the parameter intervals meanwhile our confidence on the parameters increased. The algorithm can easily be implemented with other process-based models. Highlights: A novel approach (CIRM) is presented for the parameterization of process based models. Output data conditioning is combined with a probabilistic method using likelihood function. CIRM performance is demonstrated in a low-data environment using the Biome-BGCMuSo model. CIRM reduces prior parameter intervals and eliminates the equifinality issue.