Demystifying hydrological monsters: can flexibility in model structure help explain monster catchments?

Esse, Wouter R. van (2012) Demystifying hydrological monsters: can flexibility in model structure help explain monster catchments?

Abstract:Rainfall-runoff hydrological models are commonly used to investigate and simulate catchment behaviour and predict discharges. The simulation of the discharge is never perfect and in some cases a hydrological monster is created, a combination of a model and a catchment that together result in a poor simulation. This study compares two reservoir based modelling approaches to investigate the role of model structure on model performance and ultimately explain the hydrological monsters. 237 French catchments are modelled using the fixed GR4H and flexible SUPERFLEX approach. GR4H is a single model structure that is calibrated using four parameters and generally shows good average performance on a wide range of catchments. In the SUPERFLEX approach, model components and functions can be combined in any way to a create specific model for each catchment. Twelve SUPERFLEX structures with varying complexity are used to analyse the influence of model structure. All models are calibrated using a split sample test, ten year time series where split in two periods. Calibration took place on the first and second period and validation on the second and first respectively. Inconsistency between parameter sets or model structures (SUPERFLEX) is considered as a failure of the approach. This study found that relatively simple model structures with some key components can lead to a good simulation of the discharge. The analysis of the thirteen individual model structures (GR4H + 12 SUPERFLEX structures) showed that: - The use of a power function to describe reservoir outflow significantly increases mean model performance on the catchment set, - Independently calibrated parallel reservoirs increase model performance in permeable catchments with dominant base flow, - A lag-function between reservoirs in SUPERFLEX structures leads to no significant increase in mean model performance on the catchment set, and - Increasing model complexity beyond a certain point does not lead to higher average performance. However, complex structures do show a smaller range in performance, meaning there are less catchments for which they perform very poor and that fewer monsters are created. On the whole catchment set, the flexible modelling approach does not provide better results than the fixed modelling approach on average. However, it manages to provide consistent results between test periods on a larger number of catchments. On 69 catchments, both modelling approaches performed poorly or inconsistently. These catchments were selected as monsters in this study and were classified into three groups: - Catchments where severe climatic differences between calibration and validation periods are too large for the models to correctly simulate discharge across these periods, - Catchments where models are unable to simulate the extreme flashy behaviour, and - Catchments where small scale disturbances in flow or measurement errors hinder good simulation. Generally, selecting an individual model structure for each catchment helps to rehabilitate some hydrological monsters but adding complexity is no guarantee for better results.
Item Type:Essay (Master)
Irstea: Hydrosystems and Bioprocesses Research Unit
Centre de Recherche Public - Gabriel Lippmann
Faculty:ET: Engineering Technology
Subject:56 civil engineering
Programme:Civil Engineering and Management MSc (60026)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page