Date of Award

9-1-2022

Publication Type

Thesis

Degree Name

M.A.Sc.

Department

Civil and Environmental Engineering

Keywords

Hydrological models;parameter uncertainty;Polynomial Chaos Expansion;surrogate model;SWAT

Supervisor

Tirupati Bolisetti

Abstract

Hydrological models are powerful tools that simulate the natural hydrological cycle and natural processes like surface runoff, groundwater flow, and evapotranspiration, which are needed to be understood and quantified for a wide range of applications like water resource management, climate change impact assessment, flood studies and water quality assessment. Errors and uncertainties are bound to creep into the modelling process because of various reasons like incomplete understanding and representation of the natural phenomena, model errors, approximation errors, and parameter uncertainties. This study aims to efficiently quantify the parameter uncertainty by making use of surrogate modeling techniques. There is an inherent trade-off between model complexity and the parameter uncertainty i.e., parameter uncertainty usually increases if complex hydrological model with high model accuracy is employed, but the required computational effort would increase significantly. Considering this tradeoff, one lumped, conceptual model (HYMOD) and one semi-distributed, process-based model (SWAT) for a small (179 km2) and mid-sized (2318 km2) watershed are considered for this study. Hydrological modelling processes are frequently hampered by computationally costly simulations. Consequently, modellers can opt a surrogate model which is a machine learning model that approximates another model but requires less computational effort. This thesis uses Polynomial Chaos Expansion (PCE) method of surrogacy which represents an accurate approximation of the model as the sum of carefully selected polynomials, each separately weighted. The aim of this study is to use Polynomial Chaos Expansion to 1) Improve the calibration and uncertainty assessment procedure for HYMOD, and 2) Quantify parameter uncertainties in SWAT

Download

Share

COinS