Seasonal forecasting of spring floods in snow-covered basins is
challenging due to the ambiguity in the driving processes, uncertain
estimations of antecedent catchment conditions and the choice of
predictor variables. In this study we attempt to improve the prediction
of spring flow peaks in southern Quebec, Canada, by studying the
preconditioning mechanisms of runoff generation and their impact on
inter-annual variations in the timing and magnitude of spring peak flow.
Historical observations and simulated data from a hydrological and
snowmelt model were used to study the antecedent conditions that control
flood characteristics in twelve snow-dominated catchments. Maximum snow
accumulation (peak SWE), snowmelt and rainfall volume, snowmelt and
rainfall intensity, and soil moisture were estimated during the
pre-flood period. Stepwise multivariate linear regression analysis was
used to identify the most relevant predictors and assess their relative
contribution to the interannual variability of flood characteristics.
Results show that interannual variations in spring peak flow are
controlled differently between basins. Overall, interannual variations
in peak flow were mainly governed, in order of importance, by snowmelt
intensity, rainfall intensity, snowmelt volume, rainfall volume, peak
SWE, and soil moisture. Variations in the timing of peak flow were
controlled in most basins by rainfall volume and rainfall and snowmelt
intensity. In the northernmost, snow-dominated basins, pre-flood
rainfall amount and intensity mostly controlled peak flow variability,
whereas in the southern, rainier basins snowpack conditions and melt
dynamics controlled this variability. Snowpack interannual variations
were found to be less important than variations in rainfall in forested
basins, where snowmelt is more gradual. Conversely, peak flow was more
sensitive to snowpack conditions in agricultural basins where snowmelt
occurs faster. These results highlight the impact of land cover and use
on spring flood generation mechanism, and the limited predictability
potential of spring floods using simple methods and antecedent
hydrological factors.