Modeling channel-floodplain hydrologic connectivity under non-stationary conditions

Abstract:

Traditional floodplain inundation models typically rely on the assumption of stationary flood frequency distributions and static channel geometries. However, changes in climate, land cover, or water management have been shown to systematically shift the mean and variance of flood flows. Further, changes in flood magnitudes have been shown to cause systematic changes in channel widths and depths. Therefore, some amount of change in the flood regime will be absorbed by changes in channel conveyance, but the subsequent changes to the frequency and magnitude of floodplain inundation are not obvious. This work presents a numerical model of channel-floodplain hydrologic connectivity under non-stationary conditions. Specifically, the model predicts the width of floodplain inundation in response to a time series of synthetically generated floods. Flood time series are generated via generalized extreme value probability density functions (PDFs) with specified mean, variance, and skew parameters. To simulate non-stationary conditions, we modify the mean, variance, and/or skew systematically during each run. Throughout each model run, the geometry of the alluvial channel changes in response to flood flows according to simple hydraulic geometry relations. Results suggest that a river's inundation regime is more sensitive to changes in the variance parameter of a governing PDF than to changes in the mean parameter. Simple changes in the mean of the governing PDF result in changes to channel geometry (e.g. channel widening or narrowing), but the frequency and magnitude of inundation may adjust with time after the onset of non-stationary hydrology towards a state of dynamic equilibrium similar to the previous inundation regime. However, changes in the PDF's variance parameter can induce greater variation in channel geometry, often resulting in less frequent, but greater magnitudes of floodplain inundation.