Context and General Objective

Over the decades, expensive field experimentation has been replaced by numerical simulation to understand physical phenomena and human impacts on the environment. However, numerical simulation has become complex due to the modeling of physical phenomena. To simplify, the problem’s dimension is reduced by identifying non-influential input parameters. A statistical model, called a metamodel, is then created from selected simulations to replace the computational code. This metamodel is used to prioritize the influence of input parameters and aid in decision-making in uncertain scenarios. The thesis aims to adapt these approaches for correlated inputs and proposes a sequential, computation-efficient method for building accurate metamodels, applied to pesticide transfer hydrological models to facilitate their use by non-modelers.

More Specifically...

The methodology will first be tested on toy models and then on an operational model developed within Riverly, the BUVARD-MES grass strip sizing model, which couples water, sediment, and pesticide transfer within a grass strip (Figure 2, Lauvernet and Helbert, 2020; Carluer et al., 2017). A first phase of the thesis has focused on developing a sampling method that accounts for dependencies between input parameters of the BUVARD-MES model (Lambert et al.). The subsequent steps will explore the second objective described in Figure 1.

Funding

PhD Scholarship: École Centrale de Lyon, working environment: Water4All’s AQUIGROW project (Horizon Europe Program Grant 101060874), and the academic-business consortium CIROQUO.

References

• N. Carluer, C. Lauvernet, D. Noll, and R. Munoz-Carpena. Defining context-specific scenarios to design vegetated buffer zones that limit pesticide transfer via surface runoff. Science of The Total Environment, 575:701–712, 2017. ISSN 0048-9697. doi: https://doi.org/10.1016/j.scitotenv.2016.09.105 . URL https://www.sciencedirect.com/science/article/pii/S0048969716320332.
• G. Lambert, C. Helbert, and C. Lauvernet. Quantization-based latin hypercube sampling for dependent inputs with an application to sensitivity analysis of environmental models. Applied Stochastic Models in Business and Industry, n/a(n/a). doi: https://doi.org/10.1002/asmb.2899.  URL https://onlinelibrary.wiley.com/doi/abs/10.1002/asmb.2899.
• C. Lauvernet and C. Helbert. Metamodeling methods that incorporate qualitative variables for improved design of vegetative filter strips. Reliability Engineering & System Safety, 204:107083, 2020. ISSN 0951-8320. doi: https://doi.org/10.1016/j.ress.2020.107083.  URL http://www.sciencedirect.com/science/article/pii/S0951832020305846.