TY - JOUR

T1 - A fuzzy-weighted finite-volume flow model of flooding on the river Thames in a fuzzy possiblistic framework

AU - Hankin, B. G.

AU - Beven, K. J.

PY - 2000/12/1

Y1 - 2000/12/1

N2 - The propagation of large scale floodwaters through complex environmental systems cannot be uniquely modelled using deterministic physically-based models in real time owing to the large degree of fuzziness in the boundary conditions, including topographic detail, distributed roughness and hydrological inputs. A way around this problem is to accept a degree of fuzziness in model predictions, and examine the relative performance of a large number of model structures within a Generalised Likelihood Uncertainty Estimation (GLUE) framework. This is applied to the parameter space of a new and efficient fuzzy-weighted finite-volume (FWFV) flow model to produce fuzzy possibilistic maps for a flood event on the river Thames, UK. The finite volume approach to modelling the St Venant flow equations produces a set of linear weighted equations for the solution variable at each node in terms of its values at adjacent nodes. These weights comprise components which relate to the connectivity of the nodal value of the solution variable to its surrounding nodal values, generally in terms of the local diffusive conductance and convective mass flux per unit area. These physical transport properties are affected by the local boundary roughness, which often cannot be specified exactly. Furthermore, the diffusive term is dependent on the time averaged turbulent properties of the flow field, for which there is no analytical model. The FWFV model uses fuzzy inference systems (FIS) to estimate these connectivity-weights, based on training information from measurements in complex flows, and implicitly reflects the uncertainties in the distributed boundary conditions and flow properties.

AB - The propagation of large scale floodwaters through complex environmental systems cannot be uniquely modelled using deterministic physically-based models in real time owing to the large degree of fuzziness in the boundary conditions, including topographic detail, distributed roughness and hydrological inputs. A way around this problem is to accept a degree of fuzziness in model predictions, and examine the relative performance of a large number of model structures within a Generalised Likelihood Uncertainty Estimation (GLUE) framework. This is applied to the parameter space of a new and efficient fuzzy-weighted finite-volume (FWFV) flow model to produce fuzzy possibilistic maps for a flood event on the river Thames, UK. The finite volume approach to modelling the St Venant flow equations produces a set of linear weighted equations for the solution variable at each node in terms of its values at adjacent nodes. These weights comprise components which relate to the connectivity of the nodal value of the solution variable to its surrounding nodal values, generally in terms of the local diffusive conductance and convective mass flux per unit area. These physical transport properties are affected by the local boundary roughness, which often cannot be specified exactly. Furthermore, the diffusive term is dependent on the time averaged turbulent properties of the flow field, for which there is no analytical model. The FWFV model uses fuzzy inference systems (FIS) to estimate these connectivity-weights, based on training information from measurements in complex flows, and implicitly reflects the uncertainties in the distributed boundary conditions and flow properties.

M3 - Journal article

AN - SCOPUS:0034451655

VL - 65

SP - 532

EP - 533

JO - PIK Report

JF - PIK Report

SN - 1436-0179

IS - 2

ER -