Optimization of LuKARS

Applicant

Prof. Dr. habil. Gabriele Chiogna
Professor for Applied Geology and Modeling of Environmental Systems
Friedrich-Alexander-Universität Erlangen-Nürnberg

Project Overview

In a previous research project, we developed a numerical model (LuKARS, Land use change modeling in KARSt systems) to predict discharge in karst aquifer systems. We developed versions of LuKARS using R and Python (NumPy and Pandas libraries are mainly used). LuKARS is a semi- distributed lumped parameter model that was developed to simulate the hydrological impacts of land use changes in a karst system.

The model consists of different buckets that represent the dominant hydrotopes in a considered recharge area. The water balance for each bucket is solved using an explicit temporal integration scheme. Each hydrotope is characterized by a different response and contributes to the spring discharge with individual quick flow and infiltration.

The parameter space for LuKARS increases depending on the number of hydrotopes. Therefore, to perform parameter optimization, we take advantage of Bayesian inversion using the active subspace method. Neither the model LuKARS nor the application of the active subspace method are optimally structured and we would like to parallelize and optimize the codes. Moreover, we would like to investigate also alternative algorithms to develop surrogate models, perform uncertainty quantification and parameter estimation in a computationally efficient way.