Investigation of the model complexity required in runoff simulation at different time scales / Etude de la complexité de modélisation requise pour la simulation d'écoulement à différentes échelles temporelles

Abstract

The "optimal" model complexity is defined as the minimum watershed model structure required for realistic representation of runoff processes. This paper examines the effects of model complexity at different time scales, daily and hourly. Two watershed models with different levels of complexity were constructed and their capability to simulate runoff from a watershed was evaluated. Both models were tested on the same watershed using identical meteorological input, thereby assuring that any difference between model outputs is due only to their model structure. It is demonstrated that, at a daily time scale, a simple model gives good results. For the mountain situation, in which snowmelt is a dominant influence, the nonlinearity of the runoff processes is moderate, and therefore a simple model works well. The model produced good results over a period of 28 years of continuous simulation. However, this simpler model was inadequate when tested on an hourly time scale due to greater nonlinear effects, especially when modelling high-intensity rainfall events. Therefore, the hourly simulation benefited from the more complex model structure. These model results show that optimal watershed model complexity depends on temporal resolution, namely the simulation period and the computational time step. It was shown that certain process representations and model parameters that appeared unimportant during the long-term simulation had significant effects on the short-term extreme event model simulation.Key words watershed model; model complexity; runoff simulation; temporal resolution; UBC watershed model Etude de la complexité de modélisation requise pour la simulation d'écoulement à différentes échelles temporelles Résumé La complexité "optimale" d'une modélisation est définie comme étant la structure minimale du modèle de bassin versant requise pour représenter de manière réaliste les processus d'écoulement. Cet article examine les effets de la complexité de modélisation à différentes échelles temporelles, journalière et horaire. Deux modèles de bassin versant de niveaux de complexité différents ont été construits et leurs aptitudes à simuler l'écoulement d'un bassin versant ont été évaluées. Les deux modèles ont été testés pour le même bassin versant à partir des mêmes données météorologiques en entrée, ce qui assure que toute différence entre les sorties ne provient que de leurs structures. Il est mis en évidence que, au pas de temps journalier, un modèle simple donne de bon résultats. En situation montagneuse, où la fonte nivale a une influence dominante, la non-linéarité des processus d'écoulement est modérée, et un modèle simple donne par conséquent de bons résultats. Le modèle donne de bons résultats pour une simulation continue sur une période de 28 ans. Cependant, ce modèle plus simple n'est plus adéquat lorsqu'il est testé au pas de temps horaire, en raison d'effets non-linéaires plus accentués, en particulier lors de la modélisation d'événements de forte intensité pluvieuse. Par conséquent, la simulation ho...

Quantification of uncertainty in the assessment of future streamflow under changing climate conditions

“…This is a continuous hydrological model and only need Pr, Tmax, and Tmin to simulate flow. As the UBCWM was designed from minimum meteorological parameters, it is very useful in the mountainous watershed, for example, Campbell River watershed where meteorological and flows data are often spare (Micovic & Quick, ). Because the hydrologic response of a mountainous watershed depends on elevation, UBCWM adapted the “area‐elevation band” concept.…”

Seasonal flows of international British Columbia-Alaska rivers: The nonlinear influence of ocean-atmosphere circulation patterns

“…This high seasonal resolution, which is becoming widely used in hydroclimatological analyses (e.g., [13,19,28,29,78]), facilitates the study of seasonally transient physical hydrologic processes (such as the onset of the spring snowmelt freshet, or late-summer peak glacier meltwater production following exhaustion of the seasonal snowpack), which may last only a few weeks and can be easily obscured in monthly or seasonal averages or totals. It also expedites the identification of nonlinear effects which can similarly be lost to such temporal aggregation [13,34,53,62,66]. Previous work has empirically demonstrated that the specific data processing and analysis methods used here enable detection of statistically and physically significant hydrologic teleconnections that cannot be identified using more temporally coarse-grained datasets [19].…”

The Ensemble Framework For Flash Flood Forecasting (EF5) v1.2: Description and Case Study

“…The argument about model structural error suggests that future modeling systems should be able to account for these uncertainties with different model structures. Micovic and Quick (2009) look at the complexity of model representation needed as the temporal resolution of the hydrologic model decreases. So as simulations move from long-term climate simulations at a daily time step to simulations for individual days with extreme flood events is there a need for more hydrologic model complexity?…”