TEMPORAL SCALES AND HYDROLOGICAL REGIMES - IMPLICATIONS FOR FLOOD FREQUENCY SCALING

This paper investigates the interactions among storm, within-storm, be tween-storm, and seasonal variabilities of rainfall and catchment resp onse time and how these interactions can help identify different hydro logical regimes. Using the theoretical framework provided by a simple linear rainfall-runoff model and intuitive reasoning, five hydrologica l regimes, ranging from very fast to very slow, are identified on the basis of two dimensionless ratios of the constituent timescales. For e ach regime, the paper addresses the question of how the interactions b etween the timescales in rainfall and in the runoff response manifest in the flood peak response and in the shape of the flood frequency cur ve, and the resulting implications for the Scaling of the flood freque ncy curve between catchments. It is found that, for example, in fast r egimes, within-storm patterns are the important determinant of flood p eaks, while in slow regimes multiple storms and seasonality are critic al. The latter result is verified by means of simulations with a deriv ed flood frequency model far an actual catchment, Salmon Creek, in Wes tern Australia. Slow regimes are characterized by low values of the co efficient of variation, CV[Q(p)], of the flood peaks, while fast regim es are characterized by higher values of CV[Q(p)]. Analysis of the sim ulation results on the Salmon catchment, which belongs to the slow reg ime and exhibits a nonlinear rainfall-runoff response, also shows that nonlinearity in catchment response substantially increases CV[Q(p)] a nd may even dominate the scaling of CV[Q(p)] with catchment size. The scaling exponent in the relationship between mean annual flood and cat chment size, for a linear runoff response, is higher for slow catchmen ts and lower for fast catchments, and in both cases it remains constan t with catchment area. A major conclusion of the paper is that the com bined effects of within-storm patterns, multiple storms; and seasonali ty have an important control on the observed scaling behavior in empir ical flood frequency curves, each being dominant over a different rang e of catchment sizes.