Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy

This work addresses the geomorphic response of mountain rivers to extreme floods, exploring the relationships between morphological changes and controlling factors. The research was conducted on six tributaries of the Magra River (northern Apennines, Italy) whose catchments were affected by an extreme flood (estimated recurrence interval > 100 years in most of the basins) on 25 October 2011. An integrated approach was deployed to study this flood, including (i) analysis of channel width changes by comparing aerial photographs taken before and after the flood, (ii) estimate of peak discharges in ungauged streams, (iii) detailed mapping of landslides and analysis of their connectivity with the channel network. Channel widening occurred in 35 reaches out of 39. In reaches with channel slope < 4% (here defined as nonsteep reaches), average and maximum ratios of post-flood and pre-flood channel width were 5.2 and 19.7 (i.e., channel widened from 4 to 82 m), respectively. In steep reaches (slope ≥ 4%), widening was slightly less intense (i.e., average width ratio = 3.4, maximum width ratio = 9.6). The relationships between the degree of channel widening and seven controlling factors were explored at subreach scale by using multiple regression models. In the steep subreaches characterized by higher confinement, the degree of channel widening (i.e., width ratio) showed relatively strong relationships with cross-sectional stream power, unit stream power (calculated based on pre-flood channel width), and lateral confinement, with coefficients of multiple determination (R2) ranging between 0.43 and 0.67. The models for the nonsteep subreaches provided a lower explanation of widening variability, with R2 ranging from 0.30 to 0.38; in these reaches a significant although weak relation was found between the degree of channel widening and the hillslope area supplying sediment to the channels. Results indicate that hydraulic variables alone are not sufficient to satisfactorily explain the channel response to extreme floods, and inclusion of other factors such as lateral confinement is needed to increase explanatory capability of regression models. Concerning hydraulic variables, this study showed that the degree of channel widening is more strongly related to unit stream power calculated based on pre-flood channel width than to cross-sectional stream power and to unit stream power calculated with post-flood channel width. This could suggest that most width changes occurred after the flood peak. Finally, in terms of hazard, it is crucial to document the type and magnitude of channel changes, to identify controlling factors, and most importantly, to develop tools enabling us to predict where major geomorphic changes occur during an extreme flood.