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

GEOMOR-05503; No of Pages 14 Geomorphology xxx (2016) xxx–xxx Contents lists available at ScienceDirect Geomorphology journal homepage: www.elsevier.com/locate/geomorph Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy Nicola Surian a,⁎, Margherita Righini a, Ana Lucía b, Laura Nardi c, William Amponsah d,e, Marco Benvenuti c, Marco Borga e, Marco Cavalli d, Francesco Comiti b, Lorenzo Marchi d, Massimo Rinaldi c, Alessia Viero d a Department of Geosciences, University of Padova, Italy Faculty of Science and Technology, Free University of Bozen-Bolzano, Italy Department of Earth Sciences, University of Florence, Italy d CNR IRPI, Padova, Italy e Department of Land, Environment, Agriculture and Forestry, University of Padova, Italy b c a r t i c l e i n f o Article history: Received 11 May 2015 Received in revised form 30 December 2015 Accepted 2 February 2016 Available online xxxx Keywords: Channel widening Unit stream power Lateral confinement Sediment sources a b s t r a c t 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 N 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 b 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 crosssectional 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. © 2016 Elsevier B.V. All rights reserved. 1. Introduction Geomorphic effectiveness of large floods has been long studied and debated (e.g., Wolman and Miller, 1960; Gupta and Fox, 1974; Wolman and Gerson, 1978; Magilligan, 1992; Costa and O'Connor, 1995; Phillips, 2002; Kale and Hire, 2004; Thompson and Croke, 2013; Magilligan et al., 2015). A major issue has been the role of large floods ⁎ Corresponding author. E-mail address: nicola.surian@unipd.it (N. Surian). in comparison to more frequent floods with lower magnitude. Several studies have contributed to developing the concept of effective and formative discharge proposed originally by Wolman and Miller (1960), pointing out that (i) it may be more appropriate to consider a range of discharges rather than a single formative discharge (Pickup and Rieger, 1979; Surian et al., 2009) and (ii) large floods may play a major role in certain fluvial systems such as steep channels (Johnson and Warburton, 2002; Lenzi et al., 2006), in ephemeral streams in arid and semiarid areas (Harvey, 1984; Reid et al., 1998; Hooke and Mant, 2000), and in bedrock channels (Jansen, 2006). http://dx.doi.org/10.1016/j.geomorph.2016.02.002 0169-555X/© 2016 Elsevier B.V. All rights reserved. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 2 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx Another major research question concerns the factors controlling channel response to a large flood event. Most works have focused mainly on hydraulic variables (e.g., unit stream power, flow duration above a critical threshold; see Magilligan, 1992; Cenderelli and Wohl, 2003; Kale, 2007; Krapesch et al., 2011; Magilligan et al., 2015) but, as suggested by Costa and O'Connor (1995), understanding and prediction of channel and floodplain response to a large flood should incorporate additional factors. Some works have confirmed that hydraulic forces may not be sufficient to explain geomorphic effects (e.g., Heritage et al., 2004; Nardi and Rinaldi, 2015), and consequently, attempts have been made to include other factors. For instance, human interventions and structures have been considered by Langhammer (2010); bedload supply and pre-flood channel planform by Dean and Schmidt (2013); lateral confinement by Thompson and Croke (2013); a bend stress parameter by Buraas et al. (2014). This work deals with an extreme flood that occurred in the Magra River catchment (northern Apennines, Italy) on 25 October 2011. Channel widening, the dominant geomorphic effect of this event along the channel network, was analyzed in six subcatchments by comparing aerial photographs taken before and after the flood. The working hypothesis was that explanation of geomorphic effects requires models that include other variables (e.g., lateral confinement, sediment supply) besides hydraulic-related variables (cross-sectional or unit stream power). The main aim was thus to explore the relationship between channel widening and a range of controlling factors. Other specific questions addressed were (i) which channel width (i.e., pre- or post-flood width) should be considered to calculate unit stream power in order to have a better explanation of channel response?; and (ii) is sediment supply from hillslopes (i.e., landslides) a key factor driving channel changes in mountain environments? We were able to address such questions in relatively small catchments (drainage areas between 8.5 and 38.8 km2) because an integrated approach was deployed to study this flood event (Rinaldi et al., 2016). Besides the analysis of morphological changes, the approach includes field measurements coupled to a rainfall-runoff model to estimate peak discharges in the ungauged streams, detailed mapping of landslides and analysis of sediment connectivity, as well as information concerning other fundamental aspects of the event (e.g., sedimentological characterization of flood deposits, dynamics of large wood transport; Lucía et al., 2015). 2. Study area 2.1. General setting The Magra River catchment is located in the northern Apennines (northwestern Italy) and covers an area of 1717 km2, ranging from a maximum elevation of 1901 m asl to sea level (Ligurian Sea) (Fig. 1). The catchment is characterized by ridges with a NW-SE direction, associated to thrust faults, which define two main subcatchments: the Magra (1146 km2) and the Vara (571 km2) subcatchments. The catchment is mainly composed of sedimentary rocks (predominantly sandstones and mudstones), with some outcrops of magmatic (ophiolites) and metamorphic rocks. The climate is temperate, with dry summers and most precipitation occurring in autumn. The mean annual precipitation is 1707 mm, reaching maximum values of about 3000 mm in the upper part of the catchment. The Magra catchment is predominantly forested (about 66% of the whole catchment), while urban areas are relatively small and mostly located at low elevations. 2.2. The extreme event on 25 October 2011: rainfall distribution and intensity Rainfall maps for the study event were obtained based on data collected by the Monte Settepani meteorological radar placed at 1386 m asl on the Apennines, at the border between the Piemonte and Liguria regions. The radar data were processed for a number of error sources (Marra et al., 2014) and were merged with rain-gauge Fig. 1. Location map of the Magra River catchment, the six study catchments, and the study reaches. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx data by using the procedure described by Martens et al. (2013). Because of the large differences in sampling area, the merging was carried out at the event temporal scale and then was scaled down to the temporal resolution used for the flood event analysis and the hydrological modelling. The obtained rainfall estimates show that maximum hourly rates were up to 149 mm/h, whereas 3-hour maximum and eventaccumulation maxima were up to 326 mm and 500 mm, respectively. Recurrence intervals up to 300 years were estimated for this event based on precipitation records. Fig. 2 reports the spatial distribution of rainfall maxima corresponding to 3-hour rainfall duration over the Magra River catchment. We selected the 3-hour duration because this is the duration that best corresponds to the response time of the catchments selected for the analysis, which are characterized by sizes ranging between 8.5 and 38.8 km2 (Table 1). Basin average event-accumulation rainfall amounts over the study basins are also reported in Table 1, showing that all the basins were impacted by high rainfall totals ranging from 267 (Geriola) to 380 mm (Gravegnola). However, the spatial pattern of the 3-hour rain maxima shows huge differences among the basins, with a maximum over the Pogliaschina and a decreasing gradient toward the other basins (Fig. 2). This is reflected in the higher value of the unit peak discharge over the Pogliaschina (Table 1) with respect to the other basins. 2.3. General characteristics of the six study streams Six catchments where rainfall was very intense were selected to analyze channel response (Table 1). Teglia, Mangiola, Geriola, and Osca rivers are tributaries of the Magra River, while Gravegnola and Pogliaschina rivers of the Vara River (Fig. 1). Unit peak discharge in these rivers was estimated between 12.8 (Osca) and 23.7 m3 s−1 km−2 (Pogliaschina) (Table 1). The recurrence interval of the peak discharge has been estimated by comparing the values of peak discharge (Table 1) with the results of regional equations relating 3 peak discharges to catchment area (Autorità di Bacino interregionale del Fiume Magra, 2006). In the Teglia and Mangiola rivers the peak discharge is associated to a recurrence interval slightly lower than 200 years. The highest recurrence intervals have been obtained for the Gravegnola (between 200 and 500 years) and Pogliaschina (N500 years). The relatively low recurrence interval of the peak discharge in Geriola and Osca (between 30 and 100 years) is consistent with lower rainfall amounts in these catchments (Table 1). Streams within the six catchments are characterized by average channel slope ranging from 4.1% (Osca) to 8.8% (Geriola), coarse sediments (mainly gravels and cobbles), and a wide range of conditions in terms of lateral confinement, i.e., from highly to poorly confined reaches. The study reaches correspond to the middle and lower portions of these rivers, which are characterized by partly confined or unconfined conditions (Fig. 1). In addition to the main stems, some tributaries of the Gravegnola and Pogliaschina rivers were also analyzed. According to field observations, intense bedload was widespread in the study streams during the 25 October 2011 flood, and sediment transport also occurred as debris flood (Hungr et al., 2001) in some reaches. 3. Methods An integrated approach was adopted to investigate the geomorphic effects of the 25 October 2011 flood in the Magra catchment. Analyses of rainfall, peak discharge, channel changes, depositional features and sediment structures, wood dynamics (Lucía et al., 2015), and sediment sources were carried out by field surveys, remote sensing, and numerical modelling at different spatial scales (i.e., from catchment to cross section scale). The whole methodological framework is described in detail by Rinaldi et al. (2016), while in this work we focus on delineation of spatial units, channel width changes, and potential controlling factors of channel changes. The estimation of peak discharges — used to calculate cross-sectional stream power and unit stream power — and analysis of Fig. 2. The 25 October 2011 event in the Magra River catchment: spatial distribution of rainfall maxima corresponding to three-hour rainfall duration. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 4 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx Table 1 General characteristics of the six study streams and rainfall and discharge characteristics of the 25 October 2011 flood event (D50 - median diameter of bed sediment; Qpk - estimated peak discharge; n.a. - not available). Stream Drainage area (km2) Basin relief (m) Stream length (km) Channel slope (%) D50 (mm) Total rainfall (mm) 3-hour maximum rainfall (mm) Runoff ratio Qpk (m3 s−1) Unit Qpk (m3 s−1 km−2) Teglia Mangiola Geriola Osca Gravegnola Pogliaschina 38.8 26.2 8.5 21.8 34.6 25.1 1035 1012 884 962 1106 625 14.8 12.9 7.2 9.9 12.8 9.1 4.9 6.6 8.8 4.1 7.0 5.6 47–69 41–95 n.a. 44–65 33–79 24–36 335 376 267 243 380 350 116 148 116 125 176 209 0.53 0.57 0.51 0.52 0.62 0.61 538 406 121 279 523 595 13.9 15.5 14.2 12.8 15.1 23.7 sediment sources and of their connectivity with study reaches are also described. The last part of the methodological section deals with statistical analysis carried out to explain channel response to the flood event, by exploring the relationships between changes in channel width and controlling factors. 3.1. Morphological characteristics and delineation of spatial units The first step of this study included (i) analysis of morphological characteristics and (ii) delineation of spatial units. This step is crucial for the subsequent analyses, specifically for a sound interpretation of channel changes. The material used in this step included orthophotos with a spatial resolution of 50 cm, topographic maps at 1:5000 scale, and a digital elevation model (DEM) with a spatial resolution of 10 m. The analysis was carried out by geographic information system (GIS) software (ArcGIS 10.2). The analyzed morphological characteristics were pre-flood channel areas, alluvial plain areas, lateral confinement, and channel slope. Identification of the pre-flood channel area was straightforward (where channel width was N 5–6 m) while more difficult for smaller channels or at locations with dense riparian vegetation. In the latter cases, photo interpretation was supported by use of the topographic maps to reduce the degree of uncertainty. Similarly, definition of the alluvial plain, which includes present floodplain and low terraces (i.e., surfaces that can be some meters higher than the floodplain and can be infrequently flooded), was not always straightforward (e.g., in narrow valley bottoms with dense vegetation cover). Alluvial plain was mainly identified using the DEM and topographic maps, while aerial photographs turned out to be useful where sharp changes in land cover could be clearly associated to elevation changes (e.g., a change from agricultural land to forest). As for lateral confinement, three valley settings were differentiated (Brierley and Fryirs, 2005): confined, partly confined, and laterally unconfined reaches. Lateral confinement was defined by combining two aspects: the degree of confinement, which is the percentage of channel banks directly in contact with hillslopes or ancient terraces (Brierley and Fryirs, 2005), and the confinement index, which is defined by the ratio between the alluvial plain width and the channel width (Rinaldi et al., 2013). The DEM was employed to estimate channel slope, as the difference in elevation divided by the planimetric distance relative to each reach. Delineation of spatial units was carried out according to the approach proposed by Rinaldi et al. (2013), which is a modification of the approach by Brierley and Fryirs (2005). According to that approach, stream sectors were defined as macroreaches having similar characteristics in terms of lateral confinement, while reaches are homogeneous also in terms of channel morphology (channel pattern, width, slope) and hydrology. We used the reach scale (reach length was commonly from 1 to 3 km) for an overall assessment of magnitude of channel changes and for a preliminary investigation of controlling factors. For a more accurate analysis of the relation between channel changes and controlling factors, reaches were divided into subreaches having a constant slope. To identify the proper length of the subreaches (i.e., the minimum distance with constant slope), the method proposed by Vocal Ferencevic and Ashmore (2012) was applied. Because DEM resolution was rather low (10 m), the length of subreaches turned out to be on the order of 300–500 m. 3.2. Morphological changes: analysis of channel widening Morphological changes induced by the 2011 flood were assessed by field surveys and interpretation of aerial photographs. The dominant process observed in the study reaches was channel widening, which was analyzed in detail by comparing aerial photographs taken before and after the flood. Pre-flood orthophotos have a spatial resolution of 50 cm and were taken in 2006 (Gravegnola and Pogliaschina) and 2010 (Teglia, Mangiola, Osca, and Geriola). Because the photos for Gravegnola and Pogliaschina catchments were taken 5 years before the flood, we verified them to be representative of the pre-event situation by the inspection of images taken in July 2011 and available through Google Earth©. Four sets of orthophotos were used to analyze the post-flood situation. The Gravegnola and Pogliaschina channels were analyzed with photos taken on 28 October 2011 by the Liguria region (resolution 10 cm) and 28 November 2011 by the Civil Protection of the Friuli Venezia Giulia (resolution 15 cm). For the other four catchments, we used photos taken on November 2011 by the Toscana region (resolution 20 cm) and on December 2012 (resolution 30 cm). The latter photos were taken by an ad hoc flight contracted to have full coverage of the study catchments. To assess changes in channel width, channel banks, and islands (i.e., in-channel surfaces covered by woody vegetation), these features were digitized on pre- and post-flood orthophotos. In this work, the term channel refers to an active channel, which includes low-flow channels and unvegetated or sparsely vegetated bars (i.e., exposed sediments). Channel width was calculated dividing channel area by the length of the reach or subreach, and changes in channel width were expressed as a width ratio, i.e., channel width after/channel width before the flood (Krapesch et al., 2011). Estimate of channel width, and consequently of width ratio, is affected by errors, in particular related to photo interpretation and digitization. Although rigorous error assessment was not carried out, we judged that, overall, errors are relatively small in this analysis because (i) images with high spatial resolution were used and (ii) the magnitude of changes (i.e., width ratio) was very high in most of the reaches. As mentioned above, larger errors in channel width estimate could concern the smallest channels or some reaches where there was a dense vegetation cover. 3.3. Hydraulic analysis: estimates of peak discharge The flash flood of 25 October 2011 in the Magra River and its tributaries was studied following an approach that encompasses analysis of rain-gauge data and weather radar observations, post-event surveys aimed at estimating peak discharge and time evolution of the flood in ungauged catchments, and model-based consistency check of rainfall and discharge (Borga et al., 2008). Post-flood assessment of peak discharge was based on the survey of high water marks and crosssectional geometry and computation of flow velocity using the Manning-Strickler equation under the assumption of uniform flow; for each cross-section a central (more probable) discharge value was Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx assessed, and upper and lower bounds of the estimate were computed. More details on the procedures applied in post-flood surveys and their assumptions can be found in Gaume and Borga (2008) and Marchi et al. (2009). The consistency of rainfall and discharge data was verified by applying a distributed rainfall-runoff model (Borga et al., 2007). The hydrological model applied in this study uses a mixed Curve Number– Green Ampt method (Grimaldi et al., 2013) for rainfall excess modelling. The procedure consists of applying the SCS-CN approach (Ponce and Hawkins, 1996) to quantify the storm net rainfall total amount and using this value to estimate the effective saturated hydraulic conductivity of the Green-Ampt method. The model simulates the runoff propagation by considering distinct hillslope and channel pathways. A full de Saint Venant model is used to reconstruct the propagation of the flood wave along the main channels of the Magra and Vara rivers. 5 The rainfall-runoff model was first calibrated at three stream-gauge stations located on the Vara and Magra rivers; calibration parameters were then transposed to the ungauged catchments where peak discharges had been estimated by means of post-flood field observations. This enabled a check of the consistency of rainfall and discharge data at the scale of subcatchments and provided the basis for peak discharge computation at channel reach and subreach scale, as required for the analysis of morphological changes of the study streams. 3.4. Sediment sources, delivery, and connectivity Sediment supplied from hillslopes to channel network during the 2011 flood derived essentially from landslides. The analysis of sediment sources was thus carried out by analyzing GIS-based landslide Table 2 Morphological characteristics, channel width changes, and controlling factors at reach scale. Wafter (m) Wratio AS (%) Qpk (m3 s−1) Ω (W m−1) ωbefore (W m−2) ωafter (W m−2) 8.9 11.9 11.8 17.7 35.2 60.1 42.7 48.3 4.0 5.1 3.6 2.7 0 0 20 50 457 495 529 538 136,895 116,478 102,567 110,833 15,364 9788 8691 6267 3887 1937 2402 2294 2884 2630 0 0 6.5 8.4 9.7 11.1 6.5 22.2 39.1 34.6 77.3 3.5 5.3 3.5 10 0 0 255 360 402 92,888 173,897 98,493 8353 26,590 4447 2375 5030 1274 10,768 76,650 4958 28 65 107 – 5.5 7.8 10.6 u.c. 6.7 8.2 10.1 18.5 18.7 35.7 49.4 51.1 4.0 4.3 4.9 2.8 0 0 0 0 47 79 106 114 62,952 59,213 49,405 39,425 9356 7180 4901 2128 3371 1658 1001 772 1681 1363 0 0 Steep Nonsteep Nonsteep Nonsteep 39 52 143 135 13.0 16.5 20.6 4.2 3.0 3.1 6.9 32.4 28.8 24.8 25.8 54.4 9.6 7.9 3.7 1.7 0 0 4 0 106 217 268 278 67,084 70,989 66,526 42,924 22,361 22,570 9629 1324 2328 2858 2582 789 7055 4226 312 0 0.121 0.069 0.045 0.172 0.134 0.034 0.023 0.019 Steep Steep Steep Steep Steep Nonsteep Nonsteep Nonsteep 20 30 208 24 62 42 96 177 2.4 3.6 21.1 2.5 5.8 6.3 7.2 5.0 8.6 8.3 9.5 9.4 10.7 6.6 13.4 35.5 21.2 27.4 88.0 19.6 59.2 29.5 60.0 105.4 2.5 3.3 9.2 2.1 5.5 4.5 4.5 3.0 0 0 0 0 0 0 10 30 56 84 159 53 72 293 449 519 66,855 58,850 70,762 89,299 92,029 95,275 106,600 90,636 7784 7051 7429 9494 8578 14,541 7946 2550 3147 2150 804 4565 1553 3230 1776 860 7496 17,906 764 21,828 4477 27,226 13,718 232 0.026 0.015 0.020 0.004 0.041 0.034 0.068 0.166 0.053 0.026 0.029 0.088 0.044 0.025 0.089 0.068 Nonsteep Nonsteep Nonsteep Nonsteep Steep Nonsteep Steep Steep Steep Nonsteep Nonsteep Steep Steep Nonsteep Steep Steep 35 58 90 137 20 48 18 16 19 39 62 27 26 42 17 29 11.1 15.8 21.8 26.6 6.0 12.5 3.6 2.3 3.2 10.0 20.3 4.0 4.1 5.7 3.0 5.7 3.1 3.7 4.1 5.1 3.3 3.8 5.0 6.8 6.0 3.9 3.0 6.6 6.4 7.3 5.8 7.3 6.9 28.5 81.7 32.3 4.1 7.2 5.0 9.1 14.8 39.0 26.4 7.1 6.4 13.3 5.8 7.3 2.2 7.7 19.7 6.3 1.2 1.9 1.0 1.3 2.5 10.0 8.7 1.1 1.0 1.8 1.0 1.0 10 30 70 40 0 0 0 0 0 0 0 0 0 0 0 0 31 146 188 581 54 61 6 4 118 170 206 34 54 126 11 34 7712 21,732 8952 25,569 29,056 15,675 3752 5538 61,378 43,919 68,978 29,901 25,843 26,126 9837 21,048 2465 5881 2158 4965 8705 4106 757 817 10,306 11,241 22,617 4501 4027 3574 1686 2903 1120 763 110 793 7112 2180 750 607 4134 1125 2611 4200 4027 1959 1686 2883 5085 40,685 1628 0 113 8661 1419 25,354 24,664 18,239 4824 3200 22,449 8827 3843 18,675 L (m) S (m m−1) Reach typology Wpl (m) Ci Wbefore (m) 2816 1380 1975 374 0.031 0.024 0.020 0.021 Nonsteep Nonsteep Nonsteep Nonsteep 43 97 132 – 4.9 8.2 11.2 u.c. Mangiola M2.2 1994 M2.3 3616 M3.1 3600 0.037 0.049 0.025 Nonsteep Steep Nonsteep 72 55 215 Geriola GE1.2 GE1.3 GE2.1 GE3.1 1804 1582 1555 968 0.140 0.077 0.048 0.035 Steep Steep Steep Nonsteep Osca O1.2 O1.3 O2.1 O2.2 2227 3296 2543 891 0.064 0.033 0.025 0.016 Gravegnola CR1.1 1486 CR1.2 1685 CR1.3 1338 SU1.1 1168 SU1.2 646 V1.1 1772 GR1.1 2817 GR2.1 1917 Pogliaschina P1.1 2008 P1.2 2686 P1.3 461 P2.1 499 PO1.1 492 PO1.2 844 RG1.1 557 SO1.1 742 CN1.1 2567 CN1.2 1188 CN1.4 503 RN1.1 700 RN1.2 1207 RN1.3 1450 B1.1 264 GI1.1 1722 Code Teglia T2.3 T3.1 T3.2 T4.1 SS (m2) Code - reach identification code (T - Teglia; M - Mangiola; GE - Geriola; O - Osca; CR - Casserola; SU - Suvero; V - Veppo; GR - Gravegnola; P - Pogliaschina; PO - Pogliasca; RG - Redovego; SO - Sottano; CN - Cassana; RN - Redarena; B - Benoia; GI - Ginepro); L - reach length; S - channel slope; reach typology - steep, reach with slope ≥ 4%; nonsteep, reach with slope b 4%;Wpl width of the alluvial plain; Ci - confinement index (Wpl/Wbefore); Wbefore - channel width before the flood; Wafter - channel width after the flood; Wratio - channel width after the flood/ channel width before the flood; AS - percentage of reach length with artificial structures (e.g., walls, ripraps) preventing lateral mobility; Qpk - peak discharge; Ω - cross-sectional stream power at the peak discharge; ωbefore - unit stream power calculated based on channel width before the flood (Ω/Wbefore);ωafter - unit stream power calculated based on channel width after the flood (Ω/Wafter); SS – sediment supply area expressed in terms of landslide area connected with the stream reach. Note. u.c.: unconfined. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 6 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx inventories that were prepared through visual interpretation of digital aerial photographs taken in different periods: 3 days, 33 days, and 14 months (December 2012) after the event, and of a high-resolution image (0.5 m) taken by the WorldView II satellite four days after the event. The satellite image and the post-event aerial photographs cover almost the entire surface of the Pogliaschina and the Gravegnola basins, while the complete coverage of the remaining four basins was guaranteed by the aerial photographs collected 14 months after the event. We visually compared the pre-event orthophotos acquired in 2006 with a satellite image acquired on 20 July 2011, 3 months before the event, available through Google Earth©. The visual comparison of the two pre-event images allowed verification that no significant landslide had occurred in the area between 2006 and the date of the satellite image (20 July 2011). Anecdotal information confirmed that landslides did not occur between 20 July 2011 and the October 2011 event. The orthophotos flown in December 2012 were taken with less favorable lighting conditions with respect to the other images, making the detection and mapping of the landslides more difficult and locally less accurate. Furthermore, snow cover, natural soil erosion from landslide scars, and artificial sediment removal occurred between the rainfall event and the date of the orthophotos, obliterating partially or completely some of the soil slips and the landslide deposits. Because of the different quality of images, the photointerpretation was focused on three types of shallow landslides recognizable in all available images: (i) translational slides, (ii) earth flows (including also some debris flow phenomena), and (iii) rotational slides (Mondini et al., 2014). Field surveys on landslide sites allowed us to validate the obtained inventories. In order to evaluate which sediment source areas were effectively coupled to the studied reaches of the channel network (i.e., areas responsible for sediment supply), a geomorphometric analysis of sediment connectivity was carried out. In each catchment, a map of sediment connectivity was derived by calculating the sediment connectivity index (IC, Cavalli et al., 2013) using the SedInConnect tool (Crema et al., 2015). The IC, originally developed by Borselli et al. (2008), is a distributed GIS-based index mainly focused on the influence of topography on sediment connectivity and aiming at representing the linkage between different parts of the catchment (i.e., hillslopes and features of interest such as catchment outlet, main channel network or a given cross section along the channel). The IC is defined by the logarithm of the ratio between an upslope and a downslope component expressing, respectively, the potential for downward routing of the sediment produced upslope and the sediment flux path length to the nearest target or sink. A weighting factor appears in both components of IC to model the impedance to runoff and sediment fluxes. More details can be found in Cavalli et al. (2013). In this study, IC was applied to evaluate the potential connection between hillslopes and the studied reaches. In the IC calculation, Manning's n roughness coefficient, assigned according to different land use types, was used as a weighting factor. The resulting IC maps were used as a support for the removal of decoupled sediment source areas from the total inventory. This operation was then checked by visually inspecting post-event orthophotos. Finally, the variable expressing the sediment supply from the hillslopes to the studied subreaches was computed by summing all the areas of the sediment sources connected to each subreach in the six basins. 3.5. Analysis of controlling factors Analysis of controlling factors was carried out at the subreach scale, taking into account four geomorphic and three hydraulic factors. The geomorphic factors were channel slope as a proxy of stream morphology (see e.g., Montgomery and Buffington, 1997); confinement index, which represents the width of the alluvial plain and a natural constraint to channel widening; artificial structures that may hinder channel 7 lateral mobility; sediment-supply area, in terms of landslide areas effectively coupled to the main channel network. Stream energy was analyzed taking into account three hydraulic variables closely related to flood power: cross-sectional stream power (W m−1) defined as Ω = γQS, where γ is the specific weight of water (N m−3), Q is the discharge (m3 s−1), and S is channel slope; unit stream power (W m−2) obtained by dividing cross-sectional stream power by channel width measured before and after the flood (ωbefore = Ω / Wbefore and ωafter = Ω / Wafter). Hence statistical analysis was performed considering seven geomorphic and hydraulic variables but taking into account separately independent variables related to stream energy (i.e., channel slope, cross-sectional stream power, unit stream power). Least squares multiple regression analysis was used to investigate which set of variables gave the best explanation of channel response (i.e., channel widening). Software STATGRAPHICS centurion XVI (version 16.2.4) was used for all statistical analyses. 4. Results 4.1. Morphological changes at the reach scale In this section the morphological characteristics of the study streams and channel changes that took place during the 25 October 2011 flood are illustrated at the reach scale. The aim of this section is to show the magnitude of changes and analyze some of the possible factors that could play a role in the geomorphic response of stream channels. A more accurate analysis of controlling factors was carried out at subreach scale, as described further in the paper. Following the delineation procedure described above, stream sectors and reaches were defined in the six catchments. Then, only the partly confined and unconfined reaches were considered for the following morphological analysis (i.e., analysis of changes in channel width) (see study reaches in Fig. 1). The minimum, average, and maximum length of the 39 study reaches is 264, 1573, and 3616 m, respectively (Table 2). All these reaches display typical characteristics of mountain streams and cover relatively wide ranges in terms of channel slope, channel width, and lateral confinement. Channel slope varies between 0.4% and 17.2%, with 5.3% being the average of the 39 reaches; channel width ranges from 3 to 36 m, being 9 m on average; confinement index ranges from 2.3 to 26.6 (Table 2). Considering such a range of morphological characteristics, and specifically the variability in channel slope, the whole data set was analyzed considering two subsets: the first including reaches with a slope b 4% (hereafter called nonsteep reaches), and the second with reaches having slope ≥ 4% (steep reaches). The two subsets consist of 21 and 18 reaches, respectively. The selection of such threshold stems from a widely accepted definition of steep channels, characterized as sediment supply-limited and with stepped morphology, as those having slopes higher than ~3–5% (Montgomery and Buffington, 1997; Comiti and Mao, 2012). Channel widening occurred in 35 reaches, while no significant change in channel width was detected only in 4 reaches (see width ratios in Table 2). In the subset of nonsteep reaches, the minimum, average, and maximum width ratio was 1.7, 5.2, and 19.7 respectively. Most intense changes occurred along the Pogliaschina River (reach P1.3) where the channel widened from 4.1 up to 81.7 m (Table 2, Fig. 3). Although there were several artificial structures along this reach (i.e., along 70% of the reach), in several sites the channel took up the whole alluvial plain, and locally, widening of the alluvial plain occurred by the erosion of valley slopes (Fig. 3D). Fig. 3 shows two reaches of the Teglia River (the downstream part of reach T3.2 and the short, unconfined reach T4.1) where widening was less intense (width ratio was 3.6 and 2.7 in T3.2 and T4.1, respectively). Artificial structures (along 20% and 50% of T3.2 and T4.1, respectively) had some effect in these Fig. 3. Pre- and post-flood aerial photographs showing morphological changes along nonsteep reaches of the Teglia River (A and B), reaches T3.2 and T4.1, and Pogliaschina River (C and D), reach P1.3. Refer to Table 2 for more information about morphological characteristics and channel width changes in these reaches. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 8 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx cases as shown by sharp changes in the degree of widening along the two reaches. In the 18 reaches characterized by steep slopes, the minimum, average, and maximum width ratio was 1.0, 3.4, and 9.6, respectively. There are no artificial structures along these reaches, but the presence of narrow alluvial plains was likely a limiting factor for channel widening. Some reaches, e.g., GE1.2, CR1.2, SU1.1, were characterized by relatively moderate or low width ratios, 4.0, 3.3, and 2.1, respectively; but such ratios approximate their confinement index values, meaning that channel widening took up most of the alluvial plain (Table 2). Two examples of widening along steep channels are shown in Fig. 4. The first example refers to the Suvero River (tributary of the Gravegnola River): in reach SU1.2 (slope = 13%) the channel widened from 10.7 up to 59.2 m (width ratio = 5.5), took up almost completely the alluvial plain and, locally, eroded significant portions of the valley slopes (Fig. 4D). The other example refers to the Geriola River where in reach GE1.3 (slope = 7.7%) widening was slightly less intense than in the Suvero (i.e., from 8.2 to 35.7 m; width ratio = 4.3) but still with local erosion of valley slopes (Fig. 4B). 4.2. Estimate of peak discharge Post-flood field estimates of peak discharge were carried out in five out of the six studied catchments: one cross section was surveyed in the Mangiola and Teglia rivers (in both catchments close to basin outlet), two in the Osca and Gravegnola, and six in the Pogliaschina. Only in Geriola were no cross sections found suitable for recognition of high water marks and topographic survey because of major channel changes caused by the flood. The agreement of model-computed peak discharges with post-flood estimates can be deemed rather satisfactory, with 9 out of 12 cross sections lying within the range of discharges resulting from the computation based on field surveys. Modelcomputed discharges were then used in the analysis of the factors controlling morphological changes in the studied channels. This was done by applying the rainfall runoff model at multiple cross sections located at the end of each investigated channel reach and subreach (Table 2). In the Teglia River, ~75% of the basin area lies upstream of a dam for hydroelectric power production; peak discharge in the investigated reaches, which are located downstream of the dam, was assessed by summing the outflow from the dam to the model-computed discharge of the interbasin downstream of the dam. Estimates of peak discharge, cross-sectional stream power, and unit stream power at the reach scale are reported in Table 2. Peak discharge ranged from 4 m3 s−1 (Sottano River, a small tributary of Pogliaschina River) to 595 m3 s− 1 (Pogliaschina River). Cross-sectional stream power varied between 7712 and 136,895 W m− 1 in the nonsteep reaches and between 3752 and 173,897 W m−1 in the steep reaches. Unit stream power calculated using channel width before the flood ranged between 757 and 26,590 W m− 2, while that calculated using post-flood channel width ranged between 110 and 7112 W m−2. Notably, in most of the reaches unit stream power largely exceeded the threshold value of 300 W m−2 that Magilligan (1992) and later other researchers referred to for differentiating reaches where major geomorphic changes occurred from those reaches where changes were more limited. 4.3. Sediment sources, delivery, and connectivity The sediment source inventory compiled through photointerpretation in the six catchments featured 1196 landslides for a total surface of about 917,000 m2, which represents 0.6% of the study area (Table 3). Pogliaschina and Gravegnola catchments show the highest number of mapped landslides. This result is attributable mainly to the 9 Table 3 Summary results of the inventory of all landslides in catchments and landslides coupled to the study reaches according to sediment-connectivity analysis. Catchment Teglia Mangiola Geriola Osca Gravegnola Pogliaschina a All landslidesa Coupled landslides No. Surface (m2) No. Surface (m2) 186 157 33 125 257 438 133,070 151,322 16,054 67,589 206,232 342,585 10 42 7 21 100 189 7345 92,377 5508 12,593 99,189 213,384 Percentage of coupled landslides 5.5 61 34.3 18.6 48.1 62.3 Initial inventory not considering the soil slip class. extreme intensity of the rainfall (Fig. 2) and the resultant flood (Table 1) in these two basins, but also to the higher quality of the available orthophotos in terms of spatial resolution and date of acquisition than to data sets available for the other four catchments. Among the analyzed landslide types, earth flow is the most represented class, with percentages ranging from 66% to 86% of the individual landslide inventories. The remaining sediment sources are mainly translational landslides, while the rotational landslide class represents b2% of the cases. The total sediment source inventory was filtered to limit the inventory to the landslides effectively contributing to sediment supply in the study reaches (Table 3) by assessing the sediment-connectivity pattern using the index of connectivity (Cavalli et al., 2013). Fig. 5 illustrates the adopted criteria for selecting the landslides coupled to the study reaches. Notably, even if all the mapped slope instabilities (red polygons in Fig. 5A) are located very close to the analyzed channel network, most of these features in Fig. 5B are characterized by low values of IC (i.e., low connectivity) due to the gentle local slope and to the land use type that favor sediment storage. Conversely, the two landslides classified as coupled are well-linked to the stream network as can be observed on the orthophoto (Fig. 5A) and are characterized by medium values of IC, very close to high values of IC (hot colors in Fig. 5B). The connectivity analysis allowed us to reduce the initial inventory by about 62%, with the highest reduction (by around 94.5%) for the Teglia basin where the presence of a dam in the catchment significantly reduced the former mapped landslides (Table 3). Excluding all the sediment sources upstream of the dam, only 10 landslides with a total area of about 7345 m2 were coupled to the stream reaches. Among all the analyzed stream reaches, SO1.1, RN1.2 in the Pogliaschina, M2.3 in the Mangiola, and SU 1.1 in the Gravegnola (Table 2) feature the highest value of sediment supply per unit channel length (34.2, 18.6, 21.2, and 18.7 m2 m− 1, respectively), and several reaches in the Pogliaschina and in the Gravegnola exceed the value of 10 m2 m−1. We also note that about 60% of the total sediment sources appear to be connected to the studied reaches in the Mangiola and Pogliaschina catchments (Table 3). 4.4. Analysis of controlling factors The relationships between the degree of channel widening and possible controlling factors were explored using multiple regression analysis. The analysis was carried out for the widening (i.e., width ratio) at subreach scale. The whole data set includes 157 subreaches, with a minimum, average, and maximum length of 157, 392, and 630 m, respectively. Seven controlling variables were considered (i.e., confinement index, percentage of reach length with artificial structures, sedimentsupply area, channel slope, cross-sectional stream power, and unit stream power calculated using pre-flood and post-flood channel width), but each regression model incorporated only three to four variables. Each model included only one of the variables expressing Fig. 4. Pre- and post-flood aerial photographs showing morphological changes along steep reaches of the Geriola River (A and B), reach GE1.3, and Suvero River (C and D), reach SU1.2. Refer to Table 2 for more information about morphological characteristics and channel width changes in these reaches. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 10 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx Fig. 5. Example of the sediment connectivity analysis carried out to limit the initial sediment source inventory to landslides effectively coupled to the study reaches. (A) The initial inventory superimposed on an orthophoto. (B) Coupled and decoupled landslides according to the analysis of the sediment connectivity map. The IC is the sediment connectivity index (see the text for more details). (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.) potential or flood flow energy (i.e., channel slope, cross-sectional stream power, unit stream power). The first analysis was carried out on the whole data set. All four multiple regression models turned out to be significant (p b 0.001) and gave moderate coefficients of multiple determination (R2 and adjusted R2 ranged between 0.36 and 0.51 and between 0.35 and 0.50, respectively) (Table 4). The best model was the one including unit stream power calculated based on pre-flood channel width and confinement index as explanatory variables (model 3 in Table 4). To achieve a better understanding of controlling factors, the data set was split into two subsets using the same criteria as adopted for the analysis at reach scale (i.e., nonsteep and steep subreaches, based on the 4% threshold). Therefore, multiple regressions were carried out on one subset including 89 nonsteep subreaches and a second one including 68 subreaches with steep slope. Percentage of reach length with artificial structures was not taken into account for the steep subreaches because only 1 subreach out of 68 has some structures. All four multiple regression models for the nonsteep subreaches turned out to be significant (p b 0.001) and gave moderate coefficients of multiple determination (R2 and adjusted R2 ranged between 0.30 and 0.38 and between 0.27 and 0.36, respectively) (Table 5). The best model including unit stream power calculated based on pre-flood channel width (R2 = 0.38 and adjusted R2 = 0.36) has the following equation: Wratio ¼ −0:719 þ 0:174  ωbefore þ 0:292  Ci þ 0:275  AS þ 0:026  SS ð1Þ where Wratio is the ratio of channel width after the flood/channel width before the flood, ωbefore is the unit stream power calculated based on pre-flood channel width (W m−2), Ci is the confinement index, AS is the percentage of reach length with artificial structures, and SS is the sediment supply area (m2). The variable being the best predictor of width ratio was confinement index (R2 = 0.23; significant in all four regression models). Significant relationships, although with very low R2 values, were also found for the rate of sediment supply (R2 = 0.07) and for unit stream power calculated in the two ways (R2 = 0.14 and R2 = 0.12, using pre-flood and Table 4 Multiple regression models between width ratio and controlling factors for the whole data set (157 subreaches). Model 1 Model 2 2 R = 0.36 R2adj = 0.35 p-value b 0.001 Ci Percentage of reach length with artificial structures Sediment supply area Channel slope Ω (W m−1) ωbefore (W m−2) ωafter (W m−2) Model 3 2 Model 4 2 R = 0.46 R2adj = 0.45 p-value b 0.001 R2 = 0.37 R2adj = 0.35 p-value b 0.001 R = 0.51 R2adj = 0.50 p-value b 0.001 R2 p-value R2 p-value R2 p-value R2 p-value 0.33 0.04 8.07E−05 0.03 – – – b0.001 0.696 0.008 0.696 – – – 0.33 0.04 8.07E−05 – 0.07 – – b0.001 0.195 b0.001 – b0.001 – – 0.33 0.04 8.07E−05 – – 0.23 – b0.001 0.051 0.016 – – b0.001 – 0.33 0.04 8.07E−05 – – – 0.09 b0.001 0.748 0.031 – – – 0.024 Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx 11 Table 5 Multiple regression models for the relationships between width ratio and controlling factors for the nonsteep subreaches. Ci Percentage of reach length with artificial structures Sediment supply area Channel slope Ω (W m−1) ωbefore (W m−2) ωafter (W m−2) Model 1 Model 2 Model 3 Model 4 R2 = 0.30 R2adj = 0.27 p-value b 0.001 R2 = 0.34 R2adj = 0.30 p-value b 0.001 R2 = 0.38 R2adj = 0.36 p-value b 0.001 R2 = 0.37 R2adj = 0.33 p-value b 0.001 R2 p-value R2 p-value R2 p-value R2 p-value 0.23 0.05 0.07 0.01 – – – b0.001 0.481 0.006 0.756 – – – 0.23 0.05 0.07 – 1.33E−06 – – b0.001 0.203 0.001 – 0.042 – – 0.23 0.05 0.07 – – 0.14 – b0.001 0.085 0.013 – – 0.001 – 0.23 0.05 0.07 – – – 0.12 b0.001 0.997 0.006 – – – 0.005 post-flood channel width, respectively). The relationships with percentage of reach length with artificial structures, channel slope, and crosssectional stream power were weak or very weak and were statistically not significant (p N 0.05). The four multiple regression models for the subset of steep subreaches were significant (p b 0.001) and gave higher coefficients of multiple determination than those obtained for the nonsteep subreaches (R2 and adjusted R2 ranged between 0.43 and 0.67 and between 0.41 and 0.65, respectively) (Table 6). The best model including unit stream power calculated based on pre-flood channel width (R2 = 0.67 and adjusted R2 = 0.65) has the following equation: Wratio ¼ −2:118 þ 0:317  ωbefore þ 0:366  Ci þ 0:004  SS increased following a non-linear function (Fig. 6E). In nonsteep subreaches, unit stream power also was the most important predictor among hydraulic variables, but its relation with the degree of channel widening was much weaker (R2 = 0.23) (Fig. 6B). Notably, the most intense channel widening (width ratio up to 20) that occurred in the Pogliaschina and Gravegnola catchments is not associated to the highest values of unit stream power (Fig. 6B). Hence, most interesting, the plots of Fig. 6 showed that the relationships between width ratio and controlling factors are not linear. This helps to explain why multiple regression models, which rely on linear relations, did not have very high coefficients of multiple determination (adjusted R2 were 0.36 and 0.65, respectively). ð2Þ 5. Discussion and conclusions where Wratio is the ratio of channel width after the flood/channel width before the flood, ωbefore is the unit stream power calculated based on pre-flood channel width (W m−2), Ci is the confinement index, and SS is the sediment supply area (m2). The width ratio showed clear relationships with three variables: confinement index (R2 = 0.43), cross-sectional stream power (R2 = 0.44), and unit stream power calculated based on pre-flood channel width (R2 = 0.50). In turn, the relationships with the three other variables (sediment-supply area, channel slope, and unit stream power calculated based on post-flood channel width) were weak and statistically not significant (p N 0.05). To explore further results of the multiple regression models, we examined the controlling factors of the two best models (i.e., model 3 for the nonsteep and steep subreaches, see Tables 5 and 6) by estimating simple regression models (Fig. 6). Simple regression analyses confirmed that confinement index and unit stream power (using pre-flood channel width) are the best explanatory variables, for the nonsteep and steep channels. Width ratio was more clearly related to confinement index in steep subreaches (R2 = 0.53) (Fig. 6D) than in nonsteep subreaches (R2 = 0.32) (Fig. 6A). In steep subreaches, unit stream power — being the most important predictor among hydraulic variables — explained 52% of the degree of channel widening, which 5.1. Regression models and controlling factors Results confirmed the main hypothesis of this work that hydraulic variables alone are not sufficient to explain channel response to an extreme flood event. The inclusion of other factors, specifically lateral confinement, channel slope, hillslope sediment supply, and percentage of reach length with artificial structures, led to satisfactory models explaining the observed variability in the degree of channel widening. Differences between the two sets of subreaches (with nonsteep and steep slopes) appeared in terms of overall explanatory capability of the models and most significant explanatory variables. In the steep (≥ 4%) subreaches, which were characterized also by higher confinement, channel widening occurred mainly through lateral erosion and, as confirmed also by field observations, depositional processes were less significant. In these subreaches, cross-sectional stream power, unit stream power (calculated based on pre-flood channel width), and lateral confinement showed good relationships with the degree of channel widening (i.e., width ratio) and significant statistical models were obtained (Table 6, Fig. 6). These results suggest that the widening process is essentially controlled by two factors: flood power and valley confinement. Notably, flood duration above a critical Table 6 Multiple regression models for the relationships between width ratio and controlling factors for the steep subreaches. Model 1 Model 2 2 R = 0.44 R2adj = 0.42 p-value b 0.001 Ci Sediment supply area Channel slope Ω (W m−1) ωbefore (W m−2) ωafter (W m−2) Model 3 2 Model 4 2 R = 0.65 R2adj = 0.64 p-value b 0.001 R2 = 0.43 R2adj = 0.41 p-value b 0.001 R = 0.67 R2adj = 0.65 p-value b 0.001 R2 p-value R2 p-value R2 p-value R2 p-value 0.43 0.06 0.02 – – – b0.001 0.925 0.295 – – – 0.43 0.06 – 0.44 – – b0.001 0.65 – b0.001 – – 0.43 0.06 – – 0.50 – b0.001 0.751 – – b0.001 – 0.43 0.06 – – – 0.02 b0.001 0.813 – – – 0.704 Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 12 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx threshold (e.g., related to bedload transport) was not included in our analysis, but it is a variable that very likely would increase the robustness of regression models in these subreaches (Costa and O'Connor, 1995; Magilligan et al., 2015). Unfortunately, data on thresholds for bedload transport are not available for the study streams, neither are grain size distributions for all the analyzed reaches. The models obtained for the nonsteep subreaches, compared to the steep subreaches, are less satisfactory as they provide a lower explanation of widening variability. Besides, the degree of channel widening showed relatively weak relationships with all the significant explanatory variables (i.e., lateral confinement, sediment supply area, unit stream power; Table 5, Fig. 6). These results suggest that widening at the lower slopes and with less confined channels is a more complex process and that additional factors should be considered to better understand geomorphic response in these reaches. Likely, widening in these cases results from a combination of bar formation and lateral Fig. 6. Simple regression models between width ratio and controlling factors. Only significant factors of the two best models obtained by multiple regression analysis (see Tables 5 and 6) are shown: (A), (B), and (C) refer to nonsteep subreaches; (D) and (E) to steep subreaches. In (A) and (D) the dashed lines represent the 1:1 relationship. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx erosion, with sediment (volumes and size) supplied from landslides and upstream reaches becoming significant. Notable bar formation and channel aggradation were observed in several of these reaches (Rinaldi et al., 2016), and repeated avulsion processes might have occurred during the event in these aggrading subreaches. As to additional factors, large riparian trees coupled to wood jams could have played a role by occasionally reinforcing banks and, therefore, hampering channel widening. 5.2. Unit stream power and geomorphic response The analysis carried out in the six subcatchments of the Magra River basin showed that unit stream power calculated based on pre-flood channel width has stronger relations with channel widening in comparison to unit stream power calculated based on post-flood channel width and to cross-sectional stream power. Because peak discharge was used for stream power calculation, we are aware that neither pre-flood nor post-flood channel width is actually appropriate for the estimation of unit stream power, as the most appropriate would be the (unknown) width at the flood-peak time. The fact that using the pre-flood width gives better relations with the degree of channel widening (i.e., width ratio) could suggest that most width changes occurred after the flood peak. This hypothesis is supported by a video that allowed flood reconstruction along one reach of the Mangiola River (the video documents channel changes during the whole flood event; see Rinaldi et al., 2016, for more details). Statistical analyses and the video on the Mangiola River are not sufficient to validate the hypothesis that most channel widening took place after the flood peak, but the results of this work show clearly that using unit stream power based on post-flood width is not effective to explain geomorphic response, at least in streams that underwent intense widening in terms of width ratio. 5.3. The role of sediment supply in channel response Some previous works (e.g., Harvey, 2001; Sloan et al., 2001) pointed out that sediment supply from hillslopes was a major driving factor of channel response during extreme flood events. On the other hand, examples also show that landslides, although coupled with channel network, may have minor effects on channel processes (e.g., Milan, 2012). The detailed landslide inventory and the analysis of sediment connectivity allowed us to assess the extent to which sediment supply from hillslopes influenced channel response in the six study catchments. It is worth clarifying that we only assessed sediment source areas, without an estimation of the involved volume that would require a rather uncertain estimation of landslide thickness. Moreover, once a landslide was considered to be coupled to the study reaches, its entire area was assumed to contribute to sediment supply without taking into account that part of the landslide material, especially in channelized earth flows, could be blocked by large wood (Lucía et al., 2015). In the steep reaches, although notable sediment sources were coupled with some reaches (see Table 2), no significant relation was found between the degree of channel widening and the hillslope area supplying sediment to the channels. Likely, freshly-eroded colluvium was transferred to downstream reaches given the typical supply-limited conditions of high-energy channels, or remained stored in the smallest tributaries before entering the main channels. In the nonsteep reaches, a significant although weak relation was found (Table 5 and Fig. 6C). This would suggest that during the flood a redistribution of the material stored in the alluvial plains was likely the dominant process, whereas the contribution of material from hillslopes was relevant only at specific sites and was thus not reflected in the analysis of the whole data set, which included 89 nonsteep subreaches. Notably, remarkable channel widening (i.e., width ratio up to 6) also took place along several subreaches that received no sediment input from landslides (Fig. 6C). 13 5.4. Practical implication for river management and risk mitigation Although the six study catchments are not densely populated, the 25 October 2011 flood caused severe damages to infrastructures, specifically to roads and bridges, and loss of lives. Such catastrophic effects of the flood mainly reflected channel dynamics (i.e., bank erosion, bed aggradation, channel avulsion, intense transport of large wood) or inundation processes caused or enhanced by bridge clogging due to large wood. Therefore, in terms of hazard, documenting the type and magnitude of channel response is crucial in identifying controlling factors of such response and in developing tools to enable channel dynamic predictions. Recently, Buraas et al. (2014) stated that there is still a general lack in the capability to predict where major geomorphic changes take place during an extreme flood event. In this respect, the regression models and other outcomes of this study can be used to predict the degree of channel widening and to support planning along river corridors through hazard mapping. For instance, the 2011 flood showed that in steep, relatively confined mountain reaches the whole, relatively narrow, alluvial plain may undergo major geomorphic changes, which are effectively explained by regression models that include lateral confinement and unit stream power. This study pointed out that in nonsteep reaches predicting the degree of channel widening is more uncertain and that risk mitigation may not rely only on protection structures that can be partly effective during extreme floods in mountain environments. Acknowledgements The Basin Authority of the Magra River, the Hydrological Service of Regione Toscana, and the Environmental Agency of Regione Liguria are acknowledged for providing hydrological and other essential data. Fondazione CARIPARO and CNR IRPI are gratefully acknowledged for funding M. Righini and W. Amponsah PhD fellowships, respectively. Project TRUMPS, funded by Autonomous Province of Bolzano, supported part of this research. We thank B. Wyżga and the two anonymous reviewers for their comments and very helpful suggestions. References Autorità di Bacino interregionale del Fiume Magra, 2006. Piano stralcio “Assetto Idrogeologico” del bacino del Fiume Magra e del Torrente Parmignola – Relazione Generale. Unpublished report, 209 p. (http://www.adbmagra.it/), Sarzana (Italy) (in Italian). Borga, M., Boscolo, P., Zanon, F., Sangati, M., 2007. Hydrometeorological analysis of the August 29, 2003 flash flood in the eastern Italian Alps. J. Hydrometeorol. 8 (5), 1049–1067. Borga, M., Gaume, E., Creutin, J.D., Marchi, L., 2008. Surveying flash floods: gauging the ungauged extremes. Hydrol. Process. 22 (18), 3883–3885. Borselli, L., Cassi, P., Torri, D., 2008. Prolegomena to sediment and flow connectivity in the landscape: a GIS and field numerical assessment. Catena 75 (3), 268–277. Brierley, G.J., Fryirs, K.A., 2005. Geomorphology and River Management: Applications of the River Style Framework. Blackwell, Oxford, p. 398. Buraas, E.M., Renshaw, C.E., Magilligan, F.J., Dade, W.·.B., 2014. Impact of reach geometry on stream channel sensitivity to extreme floods. Earth Surf. Process. Landf. 39, 1778–1789. Cavalli, M., Trevisani, S., Comiti, F., Marchi, L., 2013. Geomorphometric assessment of spatial sediment connectivity in small Alpine catchments. Geomorphology 188, 31–41. Cenderelli, D.A., Wohl, E.E., 2003. Flow hydraulics and geomorphic effects of glacial-lake outburst floods in the Mount Everest region, Nepal. Earth Surf. Process. Landf. 28, 385–407. Comiti, F., Mao, L., 2012. Recent advances in the dynamics of steep channels. In: Church, M., Biron, P.M., Roy, A.G. (Eds.), Gravel-bed Rivers: Processes, Tools. Wiley, Environments. Chichester, pp. 353–377. Costa, J.E., O'Connor, J.E., 1995. Geomorphically effective floods. In: Costa, J.E., Miller, A.J., Potter, K.·.W., Wilcock, P. (Eds.), Natural and Anthropogenic Influences in Fluvial GeomorphologyMonograph 89. American Geophysical Union, Washington, D.C., pp. 45–56. Crema, S., Schenato, L., Goldin, B., Marchi, L., Cavalli, M., 2015. Toward the development of a stand-alone application for the assessment of sediment connectivity. Rend. Online Soc. Geol. Ital. 34, 58–61. Dean, D.J., Schmidt, J.C., 2013. The geomorphic effectiveness of a large flood on the Rio Grande in the Big Bend region: insights on geomorphic controls and post-flood geomorphic response. Geomorphology 201, 183–198. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002 14 N. Surian et al. / Geomorphology xxx (2016) xxx–xxx Gaume, E., Borga, M., 2008. Post-flood field investigations in upland catchments after major flash floods: proposal of a methodology and illustrations. J. Flood Risk Manag. 1 (4), 175–189. Grimaldi, S., Petroselli, A., Romano, N., 2013. Green-Ampt Curve-Number mixed procedure as an empirical tool for rainfall-runoff modelling in small and ungauged basins. Hydrol. Process. 27 (8), 253–264. Gupta, A., Fox, H., 1974. Effects of high magnitude floods on channel form: a case study in the Maryland Piedmont. Water Resour. Res. 10, 499–509. Harvey, A.M., 1984. Geomorphological response to an extreme flood: a case from Southern Spain. Earth Surf. Process. Landf. 9, 267–279. Harvey, A.M., 2001. Coupling between hillslopes and channels in upland fluvial systems: implications for landscape sensitivity, illustrated from the Howgill Fells, northwest England. Catena 42, 225–250. Heritage, G.L., Large, A.R.G., Moon, B.·.P., Jewitt, G., 2004. Channel hydraulics and geomorphic effects of an extreme flood event on the Sabie River, South Africa. Catena 58, 151–181. Hooke, J.M., Mant, J.M., 2000. Geomorphological impacts of a flood event on ephemeral channels in SE Spain. Geomorphology 34 (3–4), 163–180. Hungr, O., Evans, S.G., Bovis, M.J., Hutchinson, J.N., 2001. A review of the classification of landslides of the flow type. Environ. Eng. Geosci. 7, 221–238. Jansen, J.D., 2006. Flood magnitude–frequency and lithologic control on bedrock river incision in post-orogenic terrain. Geomorphology 82, 39–57. Johnson, R.M., Warburton, J., 2002. Flooding and geomorphic impacts in a mountain torrent: Raise Beck, central Lake District, England. Earth Surf. Process. Landf. 27, 945–969. Kale, V.·.S., 2007. Geomorphic effectiveness of extraordinary floods on three large rivers of the Indian Peninsula. Geomorphology 85, 306–316. Kale, V.·.S., Hire, P.·.S., 2004. Effectiveness of monsoon floods on the Tapi River, India: role of channel geometry and hydrologic regime. Geomorphology 57, 275–291. Krapesch, G., Hauer, C., Habersack, H., 2011. Scale orientated analysis of river width changes due to extreme flood hazard. Nat. Hazards Earth Syst. Sci. 11, 2137–2147. Langhammer, L., 2010. Analysis of the relationship between the stream regulations and the geomorphologic effects of floods. Nat. Hazards 54, 121–139. Lenzi, M.A., Mao, L., Comiti, F., 2006. Effective discharge for sediment transport in a mountain river: computational approaches and geomorphic effectiveness. J. Hydrol. 326, 257–276. Lucía, A., Comiti, F., Borga, M., Cavalli, M., Marchi, L., 2015. Dynamics of large wood during a flash flood in two mountain catchments. Nat. Hazards Earth Syst. Sci. 15 (8), 1741–1755. Magilligan, F.J., 1992. Thresholds and the spatial variability of flood power during extreme floods. Geomorphology 5, 373–390. Magilligan, F.J., Buraas, E.M., Renshaw, C.E., 2015. The efficacy of stream power and flow duration on geomorphic responses to catastrophic flooding. Geomorphology 228, 175–188. Marchi, L., Borga, M., Preciso, E., Sangati, M., Gaume, E., Bain, V., Delrieu, G., Bonnifait, L., Pogačnik, N., 2009. Comprehensive post-event survey of a flash flood in Western Slovenia: observation strategy and lessons learned. Hydrol. Process. 23 (26), 3761–3770. Marra, F., Nikolopoulos, E.I., Creutin, J.D., Borga, M., 2014. Radar rainfall estimation for the identification of debris-flow occurrence thresholds. J. Hydrol. 519, 1607–1619. Martens, B., Cabus, P., De Jongh, I., Verhoest, N.E.C., 2013. Merging weather radar observations with ground-based measurements of rainfall using an adaptive multiquadric surface fitting algorithm. J. Hydrol. 500, 84–96. Milan, D.J., 2012. Geomorphic impact and system recovery following an extreme flood in an upland stream: Thinhope Burn, northern England, UK. Geomorphology 138, 319–328. Mondini, A.C., Viero, A., Cavalli, M., Marchi, L., Herrera, G., Guzzetti, F., 2014. Comparison of event landslide inventories: the Pogliaschina catchment test case, Italy. Nat. Hazards Earth Syst. Sci. 14, 1749–1759. Montgomery, R.D., Buffington, J.M., 1997. Channel-reach morphology in mountain drainage basin. Geol. Soc. Am. Bull. 109 (5), 596–611. Nardi, L., Rinaldi, M., 2015. Spatio-temporal patterns of channel changes in response to a major flood event: the case of the Magra River (central-northern Italy). Earth Surf. Process. Landf. 40, 326–339. Phillips, J.D., 2002. Geomorphic impacts of flash flooding in a forested headwater basin. J. Hydrol. 269, 236–250. Pickup, G., Rieger, W.A., 1979. A conceptual model of the relationship between channel characteristics and discharge. Earth Surf. Proc. 4, 37–42. Ponce, V.M., Hawkins, E.R.H., 1996. Runoff curve number: has it reached maturity? J. Hydrol. Eng. 1 (1), 11–19. Reid, I., Laronne, J.B., Powell, D.M., 1998. Flash-flood and bedload dynamics of desert gravel-bed streams. Hydrol. Process. 12, 543–557. Rinaldi, M., Surian, N., Comiti, F., Bussettini, M., 2013. A method for the assessment and analysis of the hydromorphological condition of Italian streams: the Morphological Quality Index (MQI). Geomorphology 180-181, 96–108. Rinaldi, M., Amponsah, W., Benvenuti, M., Borga, M., Comiti, F., Lucìa, A., Marchi, L., Nardi, L., Righini, M., Surian, N., 2016. An Integrated Approach for Investigating Geomorphic Response to Extreme Events: Methodological Framework and Application to the October 2011 Flood in the Magra River Catchment. Italy, Earth Surface Processes and Landforms http://dx.doi.org/10.1002/esp.3902 (in press). Sloan, J., Miller, J.R., Lancaster, N., 2001. Response and recovery of the Eel River, California, and its tributaries to floods in 1955, 1964, and 1997. Geomorphology 36, 129–154. Surian, N., Mao, L., Giacomin, M., Ziliani, L., 2009. Morphological effects of different channel forming discharges in a gravel-bed river. Earth Surf. Process. Landf. 34, 1093–1107. Thompson, C., Croke, J., 2013. Geomorphic effects, flood power, and channel competence of a catastrophic flood in confined and unconfined reaches of the upper Lockyer valley, southeast Queensland, Australia. Geomorphology 197, 156–169. Vocal Ferencevic, M.V., Ashmore, P., 2012. Creating and evaluating digital elevation model-based stream-power map as a stream assessment tool. River Res. Appl. 28, 1394–1416. Wolman, M.G., Gerson, R., 1978. Relative scales of time and effectiveness of climate in watershed geomorphology. Earth Surf. Proc. 3, 189–208. Wolman, M.G., Miller, J.P., 1960. Magnitude and frequency of forces in geomorphic processes. J. Geol. 68, 54–74. Please cite this article as: Surian, N., et al., Channel response to extreme floods: Insights on controlling factors from six mountain rivers in northern Apennines, Italy, Geomorphology (2016), http://dx.doi.org/10.1016/j.geomorph.2016.02.002