Census and typology of braided rivers in the French Alps

Abstract

Following the implementation of the European Water Framework Directive (WFD) and the need to reach a “good ecological status”
for rivers, key-questions are being raised about braided rivers. Before any environmental policy can be drawn up, these rivers
need to be located, long term changes must be evaluated, and the regional diversity of such systems must be understood, as
their inner complexity has not yet been well studied. Therefore, the aim of this work is to carry out a census of the braided
channels of the French Alps and to establish a typology based on basic geomorphic indicators. A minimum estimate of the cumulative
length of braided rivers prior to major infrastructure construction amounted to 1214km. Around 53% of these rivers have disappeared
during the last two centuries in relation to embankment or channelization, but a loss of 17% is still unexplained. The range
in catchment size, mean slope and active channel width has been determined for the Western Alpine braided channels as well
as the range in changes due to narrowing, widening and shifting. Seven types of braided rivers have been distinguished based
on geographical settings (climate conditions and geology) and differences in terms of adjustment to human pressure on peak
flow and sediment delivery. The percentage area of islands in the active channel and the relative length of banks also show
a regional difference. Maximum and minimum thresholds of braided activity have been established taking into account the active
channel width and the catchment area. The position of the studied reaches between these two thresholds are discussed in relation
to position of rivers known in the literature, considering both long-term trends and short-term fluctuations in channel width.

Full text

Research Article
Census and typology of braided rivers in the French Alps
Herv Pigay1,*, Adrien Alber1, Louise Slater1and Laurent Bourdin2
1University of Lyon, CNRS-UMR 5600, Site Ens-lsh, 15 Parvis R. Descartes, F-69342 Lyon, France
2Agence de lEau Rhne-Mditerrane-Corse, Lyon, France
Received: 7 January 2009; revised manuscript accepted: 26 August 2009
Abstract. Following the implementation of the Euro-
pean Water Framework Directive (WFD) and the
need to reach a “good ecological status” for rivers,
key-questions are being raised about braided rivers.
Before any environmental policy can be drawn up,
these rivers need to be located, long term changes
must be evaluated, and the regional diversity of such
systems must be understood, as their inner complexity
has not yet been well studied. Therefore, the aim of
this work is to carry out a census of the braided
channels of the French Alps and to establish a
typology based on basic geomorphic indicators. A
minimum estimate of the cumulative length of braided
rivers prior to major infrastructure construction
amounted to 1214 km. Around 53% of these rivers
have disappeared during the last two centuries in
relation to embankment or channelization, but a loss
of 17% is still unexplained. The range in catchment
size, mean slope and active channel width has been
determined for the Western Alpine braided channels
as well as the range in changes due to narrowing,
widening and shifting. Seven types of braided rivers
have been distinguished based on geographical
settings (climate conditions and geology) and differ-
ences in terms of adjustment to human pressure on
peak flow and sediment delivery. The percentage
area of islands in the active channel and the relative
length of banks also show a regional difference.
Maximum and minimum thresholds of braided
activity have been established taking into account
the active channel width and the catchment area. The
position of the studied reaches between these two
thresholds are discussed in relation to position of
rivers known in the literature, considering both long-
term trends and short-term fluctuations in channel
width.
Key words. Channel narrowing; riparian zone; classification; good ecological status; channel shifting; channel
change; human pressure.
Introduction
Braided rivers are characterized by several low flow
channels, which diverge and converge, interspersed
with sparsely vegetated gravel bars (Rust, 1978). All
these features form a large straight band of fluvial
activity, so called the active channel, with a high width/
depth ratio. The bars are topographically low, and the
low flow channels differ in slope, depth and relative
elevation over a given cross-section within this active
channel. Both reach and catchment conditions influ-
ence braiding. Braided rivers are characterized by an
abundant bed load, erodible river banks, an extensive
and rapid variability in discharge and relatively steep
slopes (Leopold and Wolman, 1957; Ferguson, 1987;
Gurnell et al., 2009).
Braided rivers are considered by several authors to
be very interesting ecological systems, notably when
considering riparian communities and ecological
* Corresponding author e-mail: herve.piegay@ens-lsh.fr
Published Online First: October 8, 2009
Aquat. Sci. 71 (2009) 371 – 388
1015-1621/09/030371-18
DOI 10.1007/s00027-009-9220-4
 Birkhuser Verlag, Basel, 2009
Aquatic Sciences

functions (Ward et al., 1999a b, 2002; Rempel et al.,
2000; Nakamura and Shin, 2001). Nevertheless, be-
cause aquatic habitats in braided systems evolve
rapidly, they are not always as interesting as the ones
located in more stable reaches in terms of biodiversity
because of a severe reduction in the number of
permanent aquatic bodies and because of trophic
effects from frequent movement of the channel floor
(Pont et al., 2009). Braided rivers are characterized by
a high sensitivity to planform changes in response to
extreme floods (Burkham, 1972) and over longer time
periods (Bravard, 1989; Brewer and Lewin, 1998), and
can then be plotted on a trajectory from highly active
to senescent stages (Pigay et al., 2006).
Collective contributions on braided-river process-
es, forms and ecology were published in 1993 and
again in 2006, targeting challenging issues for the
scientific community (see Best and Bristow, 1993;
Sambrook Smith et al., 2006). The most recent
publication notably stressed the need to quantify the
range in feedbacks operating between the geomor-
phological and ecological components of braided
reaches and also to develop long-term monitoring of
channel evolution and to increase the number of
studies across a range of braided river environments,
scales and grain sizes, in order to assess correctly the
degree to which ecological, geomorphological or
sedimentological models can be applied (Sambrook
Smith et al., 2006).
Because of the implementation of the European
Water Framework Directive (WFD) and the need to
reach a “good ecological status” for rivers, key-
questions are being raised about braided rivers. What
is a braided river with good ecological status? Can we
expect a progressive disappearance of braided rivers
following the first census carried out by Bravard and
Peiry (1993) and the narrowing dynamics observed
on many of them (Libault and Pigay, 2002)? What
kind of restoration measures can be promoted? Is
artificial sediment reintroduction a self-maintaining
strategy (Libault et al., 2008)? The French Alps are
one of the last regions of the Alps where braided
channels are still well extended, notably in the
southern part (Durance catchment, left side tribu-
taries of the middle Rhne) compared to Austria,
Switzerland, Italy and Bavaria (Habersack and
Pigay, 2008).
Before suggesting any environmental policy for
braided rivers, there is a need to locate them, to
evaluate long-term changes, and to understand the
regional diversity of such systems. Since the pioneer-
ing work of Leopold and Wolman (1957), distinguish-
ing geomorphic patterns (straight, meandering and
braided channels), and of Schumm (1968) who
suggested a distinction between anastomosing and
braided rivers, a great amount of work has been done
to define channel pattern types (Kellerhals and
Church, 1989; Church, 1992) with detailed focus on
anabranching rivers (Nanson and Knighton, 1996).
Nevertheless, the braided pattern has been interpret-
ed as a single type whose intrinsic mechanisms have
not yet been well studied. In 1983, Ferguson and
Werrity suggested a definition for wandering gravel-
bed rivers, as a transitory stage between a braided
river and a meandering river with a main sinuous low-
flow channel within a straight active channel. Com-
pared to a braided river, the number of flow branches
is low and the active channel is narrower. New
contributions have also been made recently on
island-dominated patterns (Osterkamp, 1998; Gurnell
and Petts, 2002; Francis et al., 2009) and the inter-
actions between the active channel and vegetation as a
control of braided pattern characteristics, defining
whether they are truly braided or partly hybrid with
island development (then called island-braided chan-
nels) (Gurnell et al., 2009; Francis et al., 2009).
Because incision and narrowing characterized Alpine
braided rivers (Bravard et al., 1997; Libault and
Pigay, 2002), the question then arises of the effect of
vegetation and decrease in sediment delivery on
braided river planform evolution.
The aim of this work is to carry out a census of the
braided channels of the French Alps and to establish a
typology based on basic geomorphic indicators to
assess the potential diversity of braided patterns. The
aims were firstly to determine the historical evolution
of braided reaches over the past centuries in South-
East France; secondly, to establish specific types of
braided channels based on key-geomorphological
parameters; and finally, to improve our understanding
of how these braided channels evolve by quantifying
the lateral mobility and decadal adjustment for a
selection of reaches.
Materials and methods
A census of braided river reaches from aerial
photographs and historical maps
The aims were to establish where the current braided
reaches are situated; to find out where they were
before the main period of dam and dike building in the
19 and 20 centuries; and finally to determine where
braids have disappeared and where they still exist (Fig.
1; Table 1).
Braided reaches were evaluated in the Rhne-
Mediterranean district over three periods, the late 18
century, the end of the 19 century, and the end of the 20
century, to determine the historical evolution of
braided reaches.
372 H. Pigay et al. Census and typology of braided rivers in the French Alps

Figure 1. Location (A) of the Rhne-Mediterranean district in France (excepting the lowland and western parts of the Sane basin), (B)of
the braided reaches in the district prior to the main embankment and damming period (18 to mid-19 century.) and (C) at the early 21
century. (D) Location of the 49 studied braided reaches and indication of the river names used in the text.
Aquat. Sci. Vol.71, 2009 Research Article 373

Several difficulties exist in defining a braided river
using aerial photographs and maps because the width
and shape of the river depend on its discharge at the
time when it is observed, which is highly variable from
one period to another. Unfortunately, airborne sur-
veys are often taken at different periods of the year.
Therefore, to consider a reach as braided, and as
previously discussed by Egozy and Ashmore (2008), it
is not accurate to use a threshold of the braided index,
which is based only on the number of low flow
channels. Most of the pictures used were collected
during summer when braided channels in the north-
ern French Alps undergo the highest discharge,
whereas rivers in the southern French Alps under a
Mediterranean climate exhibit the lowest flow, often
with a single wet channel. We define a braided
channel as a corridor of gravel with evidence of a
network of low flow or dry beds and a total active
width (the active channel) of at least 6 to 8 times the
summed low-flow channel widths. The main low-flow
channel flows sinuously, without touching the two
sides of the gravel corridors, showing a clear underfit-
sream pattern, which distinguishes it at low flow from
a wandering channel.
The material used to localise existing braided
reaches was SCAN 100 maps from the IGN (French
Institut Gographique National) to have a prelimi-
nary scanning of the whole network at 1:100,000
scale; as well as the BD Ortho(Ortho-rectified
airborne photographs at ca 1:25,000 scale). The dates
of these documents are variable, late 20 century
(mainly 1990s) for the SCAN 100 maps but around
2000 for the orthophotographs. The BD Carthage,
which is a GIS vectorial layer of the entire network of
rivers, and ranges from Strahler orders 1 to 8, was also
used to locate reaches. To determine the historical
state before the main period of river regulation, we
combined data provided by several maps. A geo-
metric map of France called the Carte de Cassini or
Carte de lAcadmie was used. The Carte de Cassini
was created by Cassini de Thury using a geodesic
triangulation. It was started in 1750 and finished in
1815. The scale is approximately 1:86,400. Another
mapofFrancecalledtheCarte dtat-Major,at
1:80,000 scale, created by military staff officers (the
Corps dEtat-Major) between 1818 and 1881, and
supervised by geographic engineers, was equally
used to define historically braided reaches. We
considered braided reaches when the channel was
drawn with more than 2 low flow channels. Because
of scale, we did not classify the upper narrow-braided
channels but we identified the lower ones that have
been the most affected by human pressures.
Planimetric characterization of contemporary active
braided channels
Within the braided river continuum identified above,
several reaches were selected to study different spatial
and temporal frameworks (Table 1).
Forty-nine reaches were firstly selected and char-
acterised to provide a synchronic analysis of spatial
heterogeneity in the study area (see Fig. 1D for their
location and names). These reaches were selected at
some distance from cities to minimize effects of local
infrastructure and other human pressure on their
geometry. Sampling was done so that the regional area
is covered from upstream to downstream as well as the
main river segments (river reaches between two main
confluences). The reaches were downloaded for
characterization from the BD Ortho of the IGN
using ArcGis 9.2 and at the same scale (1:3,000). They
were defined so that their length was 20 times the
average width of the active channel, which is an
acceptable compromise to get a significant braided
reach, homogeneous in terms of pattern (not affected
by any confluence, valley constraining or other human
pressures). The channel reach lengths vary between
725 and 7000 m.
Table 1. Summary of historical maps and IGN aerial photographs used for this study.
200 years historical censing
(entire Rhne catchment) Regional organization
(n=49 sampling
reaches)
Channel shifting
assessment at decadal
scale (n=29)
Riverscape
characterization
(n=16)
Riverscape
decadal changes
(n=6)
Carte de Cassini
1750 X
Carte dEtat
major X
SCAN 100 maps X
Orthophotos
early 21th c. XXXXX
Archived air
photographs of
1940s
X
Archived air
photographs of
1970s
XX
374 H. Pigay et al. Census and typology of braided rivers in the French Alps

Several variables were measured from GIS vec-
torial layers : the mean width of the active channel
(bare gravel bars and low flow channels), the mean
width of riparian forests, and the width of the
available valley floor. The term “available” is used
because valley floors are constrained by alluvial fans,
terraces and valley sides in mountain areas, and in
lowland valleys by dikes, which are usually located
further away from the active channel. Here, the
valley floor was defined as the largest area that could
possibly be occupied by the river (Slater, 2007). To
this, we added the mean slope and mean elevation of
the reach, as well as the catchment area provided by
Pont and Rogers (2004). Simple bivariate analyses
showed that the mean active channel width is linked
to catchment size (Fig. 2A) as well as slope (Fig. 2B).
In a similar way that the hydraulic geometry ap-
proach considers that the flow width results from
catchment size, and is an indirect indicator of flow
regime in a fairly homogeneous regional area, the
active channel width has been rated to catchment size
to provide a normalized value, the width of gravel
bars increasing with basin size as long as the sediment
delivery area increases. These two variables were
rated to catchment area to remove the size effect
(e.g., respectively named W*, S*).
The set was also compared to existing systems that
are well-known in the literature on braided rivers
(Tagliamento, Brenta and Piave, Italy, at various
dates, from Surian, 2006; and Val Roseg, Switzerland,
from Ward and Uehlinger, 2003) and to the largest
braided systems at their activity peak (Rhine in 1828,
from Kleinas, 2003; and Durance prior to the Serre-
PonÅon dam construction, from Miramont et al.,
1998). Secondly, a sub-set of 29 braided reaches was
selected to quantify the decadal evolution of their
active channel at two dates, separated by an average
of 32 years (D25:22years;D
75: 36 years) between
around 1975 and 2000 (Orthophotographs available).
The most recent aerial photographs were extracted
from the BD Ortho, whereas the oldest were selected
within the archived aerial photographs of the IGN.
The archived photographs were scanned at a reso-
lution of 600 to 1200 DPI depending on their scale to
provide a final resolution of 0.5 m, similar to that of
the orthophotographs. They were then georectified
using a first order polynomial transformation with an
average RMS error of 2 to 3 m for each photograph.
The boundaries of the active channels were then
digitized at both dates. The mean width and sinuosity
of the active channels were characterized for each
state. Then, the mean eroded and constructed areas
of the floodplain were measured from overlays of
channel position at two different dates, providing a
minimum estimation of the intensity in lateral
processes.
A remote sensing analysis of 16 of these reaches
was performed to determine the extent and geo-
metrical characteristics of riparian vegetation as well
as the average size of islands, six of them being
observed at three dates (ca. 1940s, 1970s and early 21
century). For each aerial photograph, three classes of
land-use were defined within the active channel:
gravel, water, and vegetated island. A ratio of islands
to the entire surface was calculated, and the number
of islands per reach counted using eCognition soft-
ware (Definiens) with an object-based approach,
which is well adapted for treating imagery with low
radiometric resolution by taking into account the
radiometric values of the three canals of the images
(red, green, blue) as well as the shape, size, texture,
and topological environment of each object. The
approach is based on the idea that visual perception
uses contextual and topological information to
understand an image and to identify objects that it
contains.
Figure 2. Statistical relationships between catchment area and (A) active channel width, and (B) active channel slope.
Aquat. Sci. Vol.71, 2009 Research Article 375

Results
Change in cumulated braided reaches length over the
last two centuries (19 to 20 centuries)
Six hundred and thirty kilometres of braided reaches
were made out from the Cassini Atlas, and 684 km
from the Etat-Major maps (Fig. 1). The Cassini atlas
was not precise enough for small rivers, so this
underestimated the length of cumulated braided
reaches, but it was useful to locate large braided
reaches that had been destroyed in the late 19 century
and not visible on the more recent Etat-major maps.
As a consequence, the combination of the two surveys
provides a minimum estimate of the whole cumulative
length of braided rivers prior to major infrastructure
construction of about 1049 km.
For the 20 century, braided reaches were deter-
mined from the Orthophotographs database
(1:25,000). These showed that there were still 650
km of braided reaches in the early 21 century. When
combining the two sets of maps, 559 km of braided
reaches had disappeared but 165 km had equally
appeared. In fact, the 165 km of new reaches
correspond to small rivers that were not detectable
on large-scale old maps. As a consequence, we should
consider that braided reaches were present on 1214
km, prior to major human pressures of the 19 century,
as a reasonable minimum estimate.
If we refer to the 1049 km of braided rivers
identified above, 53 % of them disappeared in 200
years, so that the occurrence in braiding is much less
homogeneous in space now than it was in the 18 and 19
centuries. Rivers where braiding disappeared most
were the largest rivers: Isre, Rhne, Durance, Arve,
Verdon, and Var (Fig. 1). Change in pattern was most
important in the Northern Alps. Braiding underwent
fewer changes in the Southern Pre-Alps. Nevertheless,
although the planform pattern is still visible on the
documents, this does not mean that it has not changed
through time, notably in terms of intensity.
The disappearance of braided rivers was due to
various causes (Fig. 3). Before the major regulation
works of the 19 century, a latitudinal gradient existed
within the French Alps, with braiding frequency
decreasing from south to north. The existence of
braided reaches was widespread in the Durance
network and in the southern Isre catchment, whereas
further north, the existence of braided reaches was less
frequent and sparse.
This first regional distribution of braiding under-
lined the fact that embankment and braiding decline
occurred in the northern Alps during the Piedmont-
Sardinia kingdom period, when embankments were
developed, and before the Cassini map was published.
A specific study of the Sarde Map would help to define
this (see Peiry, 1987 for the Arve). The 1049 km
estimate is therefore an underestimation of braided
river conditions prior to Alpine region development.
With the industrial revolution in the French Alps and
the use of “white gold” (hydroelectricity), braided
rivers strongly declined during the 19 century, mainly
in the north.
The specific study of reaches where a braided
pattern disappeared during the last century showed
that 21% of these reaches were clearly embanked (ca
117 km) to protect agricultural land or urban areas,
whereas 48% (ca 269 km) were channelized. There is
no evidence of dike construction but evidence of
straightening and bank stabilisation does exist.
Around 5.1% of the reaches were replaced by
reservoirs (28 km) when the branch was dammed.
The rest of the disappearance needs additional study
to be fully understood.
Around 6.5% of the concerned reaches (ca 36 km)
are downstream of major reservoirs, explaining evi-
dent planform adjustment associated with upstream
controls. As shown along Italian rivers (Surian and
Rinaldi, 2003), in-channel gravel mining occurred in
some areas and could explain at least 2.1% of the
planform changes, if using the district Atlas of 1997 in
which in-channel mining activity was established
(SDAGE RMC, 1997). But 17% of the reaches
where changes were observed are still unexplained
without further investigation at an individual level for
each specific case.
Figure 3. Potential causes of the disappearance of the braided
pattern within the network of the Rhne-Mditerranean French
hydrographic district (excepting the lowland and western parts of
the Sane basin).
376 H. Pigay et al. Census and typology of braided rivers in the French Alps

Geographical characteristics of braided reaches in the
French Alps
General characteristics of braided reaches and com-
parisons with known cases. The catchment size of
sampled braided reaches in the study area varies
between 25 km2and 14,000 km2(Fig. 4). This can be
compared to the Rhine in the Alsacian plain, which
has a catchment size of 35,000 km2, where it used to be
braided in 1828 (Carbiener, 1974; Kleinas, 2003), and
was the most important braided system in this area
during the last two centuries. The slope and mean
elevation of the study reaches range, respectively,
from 2.3 m km1to 52 m km1(average of 15 m.km1)
and from 75 m to 1787 m. So the persistent braided
reaches occur in various contexts but most are
localised in the intra-mountainous plains and the
upper part of valleys, which are sometimes directly
connected with glaciers (e.g. Val Roseg in Switzerland,
which is located slightly higher than the French set at
1980-2050 m, or the Durance and Vnon rivers in the
Oisans Massif). Some reaches are also outside the
Alpine area in the piedmont zone and are character-
ized by lower slope and a larger catchment size.
The average active channel width of the study
rivers is 118 m with a maximum of 351 m, and can be
compared with the middle Durance, Rhine and
Tagliamento at their maximum: 1,200, 1,250 and
1,630 m. The “available valley floor” is equal to 243
m on average, with a maximum of 740 m (Fig. 4). It
covers almost the same area as the natural corridor
that is defined as the area covered by the active
channel and its riparian forest. The mean width of the
natural corridor is equal to the sum of the mean width
of the active channel and the mean width of the
riparian forest. Therefore, the width of the riparian
corridor, which is mainly occupied by spontaneous
tree-dominant units, ranges from 11 m to 402 m
(mean of 104 m) and displays a distribution that is
similar to that of the active channel width. In this
context, most of the study reaches evolved in a
relatively wide natural corridor and are not directly
constrained by the valley sides.
Regional distribution of braided channel types. A
classification of braided reaches was established to
identify contemporary functional braided types and to
assess any spatial pattern in their distribution at a
regional level (Figs. 5 and 6).
The 1st,2
nd and 3rd types are characterized by low
values of normalised slope and channel width. They
are mainly located in the lower part of the network.
The three types can be distinguished by their normal-
ised width and relative forest cover:
(i) Reaches corresponding to the 1st type have
particularly narrow active channels for a given catch-
ment size, but a wide riparian forest. They are
concentrated in the western part of the catchment,
the southern pre-Alps, mainly the lower tributaries of
the Rhne and Bech, which is a right side tributary of
the Durance draining a similar regional area.
(ii) Reaches of the 3rd type have a wider normalised
active channel width and exhibit a significantly
narrower riparian forest compared to types 1 and 2.
They do not show a clear geographical distribution.
(iii) The 2nd type is an intermediate type between
the 1st and 3rd types, in terms of normalised active
channel width and relative forest cover, with a steeper
normalised slope. These reaches are mainly to the east
of the previous types, in the upstream sections.
The 4th and 5th types have both a steep normalised
slope and are in the upper elevated part of the
catchments. They are also strongly different in terms
Figure 4. Statistical distributions of the six general indicators of the 49 braided reaches observed within the network of the Rhne
Mediterranean hydrographic district.
Aquat. Sci. Vol.71, 2009 Research Article 377

of geometry, as type 4 shows a large normalised active
channel width and a very narrow riparian forest
compared to type 5. Type 4 is only characterized by
two reaches located in the Durance catchment, drain-
ing very small areas, notably the Saint Pierre Torrent
in the Oisans Massif, which has glaciers and low
human density and pressure. Type 5 also characterizes
small upper catchments of eastern basins (Drac,
Romanche, Arc, Guil).
The 6th and 7th types are characterized by a fairly
large normalised active channel but an intermediate
slope and riparian forest width. Type 7 is situated at a
higher elevation with a percentage of forest cover
that is significantly larger than type 6. They are both
located in the eastern Rhne catchment. Type 6
strongly characterizes the tributaries of the middle
Durance (e.g., Asse, Blone, Jabron, Sasse) but also
the Var and Vnon in the Drac catchment.
Planform evolution and channel shifting at decadal
scale
A temporal analysis was performed to test if channel
shifting has a regional organisation, and for a better
understanding of the sensitivity of riverscape pattern
to changes.
Twenty-nine reaches provide additional informa-
tion on the decadal evolution of the planform and the
potential in channel shifting of braided rivers between
ca. 1975 and 2000. The average active channel width
was 108 m in 2000 against 102 m in the first year. The
average change in channel width over the period was
+/-20%, but it can reach a maximum of -77% and -
66% on the Bech and Petit Bech, respectively, and a
widening of 35% on the Mouge. Fifty percent of the
reaches in the set range from -18% to +18%, with a
median of 5%.
Two main trends were clear between the two dates:
a narrowing-widening process and a shortening-length-
Figure 5. Summary results following the cluster analysis performed on the 49 study braided rivers: (A) dendrogram highlighting the
hierarchy of groups, (B) box plots showing the distribution of the different parameters for each of the seven selected groups : W*, S*, mean
elevation above sea level (ASL), percent of forest area in natural corridor. Box plots provide the 10th,25
th,50
th,75
th and 90th centiles and the
mean (in gray).
378 H. Pigay et al. Census and typology of braided rivers in the French Alps

ening process. When the channel becomes wider, it also
becomes straighter; and when it narrows, it becomes
more sinuous. A statistical linear relationship was
established between the difference in the active
channel widths and the difference in the sinuosity
rates of the axis of the active channel observed at the
two dates (R2=0.32; n=29; p<0.0001). Considering the
position of the reaches according to their catchment
size, the large braided rivers are mainly plotted above
the regression line, whereas the smaller catchments are
below. The distribution of the residuals for this
regression shows a clear link with catchment size.
Figure 6. Distribution of the seven braided river types within the south-eastern France basedon W*, S*, mean elevation, and the percent of
forest area in the natural corridor. Names of the reaches are indicated in figure 1.
Aquat. Sci. Vol.71, 2009 Research Article 379

The mean bank retreat is 1.1 m yr1against 1.02 m
yr1for active channel encroachment by vegetation
with maximums of 4.6 m yr1and 3.8 m yr1,
respectively. There was a clear statistical relationship
between bank retreat or vegetation encroachment and
active channel width (R2=0.51 and =0.49, respective-
ly; n=29) as it has been shown in previous works
(Pigay et al., 1996). The wider the active channel, the
higher was the bank retreat or active channel en-
croachment by vegetation. Moreover, the relative
erosion rate (% of channel width per year and per
meter of river length) was linked to the sinuosity rate
(R2=0.35; n=29)(Fig. 7) measured during the first
year of the survey period, showing that the erosion was
highest not in the truly braided systems but in the
transitory systems from braided to wandering.
Additional work was done on 16 braided reaches
observed on the most recent orthophotographs, and 6
of them have also been analysed at two prior dates
(mid 1940s and mid 1970s) to analyse the riverscape
changing rate, and to determine which landscape
criteria influenced the interactions between the
braided channel and riparian area (Fig. 8). We
focused on the comparison of types 1 and 6 following
the regional setting established in figures 5 to 6,
considering two parameters, the percentage of is-
lands in the active channel, which varies from 1.2 to
22.6%, and the sinuosity of the river banks, which
ranges from 1.03 to 1.51. There was a statistical
difference between river types 1 and river types 6
when considering the first parameter (Fig. 5): type 1,
which was defined primarily by large natural corri-
dors compared to the width of the active channels,
has the highest percentage of islands in the active
channel compared to the eastern type (see Fig. 8A).
The sinuosity of the river banks is highly variable
whatever the braided types. If there is no difference
between the two groups in the medium term, a clear
difference is shown in the variation term; the western
reaches being much more variable than the eastern
ones. This shows that a large natural corridor will
tend to create river banks with a large range of
sinuosity values (Fig. 8A). The analysis of the six
river reaches for which historical trend is available
showed two significant patterns: i) a peak of island-
development in the 1970s, which is significant for the
Drme, Eygues and Asse, and ii) a decrease in bank
sinuosity through time, notably after the 1970s, once
the islands are attached to external riparian margins.
These data show that regional differences exist, at
least between two groups of braided patterns, and
that the present pattern is significantly evolving with
time.
Discussion
A major impact of human infrastructures on the
braided rivers
The historical census of braided reaches over the last
two centuries showed a strong reduction of this
geomorphic river-pattern. Embankment and channe-
lisation strongly affected the northern Alps, the
Rhne corridor and the main Mediterranean systems
(lower Durance, Var), whereas reservoir construction
and its downstream effect on peak flow mainly
affected the Durance (Fig. 3). Obviously, we presume
that the changes are more complex than what one can
detect on maps. The impact of mining activity, which is
supposed to affect at least 2.1% of the cumulated
disparaged braided length, was underestimated at the
regional scale (Landon, 1999; Rinaldi et al., 2005).
Figure 7. Relationships between the % of floodplain eroded area during the surveyed period rated to the channel area at year 1, and the
sinuosity rate of the active channel at year 1.
380 H. Pigay et al. Census and typology of braided rivers in the French Alps

Figure 8. A) Statistical distribution of the percentage of active channel area occupied by islands and of the sinuosity of the riverbank (active
channel perimeter minus two mean widths, divided by two and by the reach length) for 14 reaches characterising the Rhne and Durance
braided river types (see Fig. 5), and B) Temporal evolution of the two indicators for 6 of the river reaches between ca 1940 and 2000.
Figure 9. (A) Distribution of the 49 studied braided reaches and other known reaches from the literature on a scatter plot linking channel
width and catchment size. Data from Miramont et al. (1998) for the Durance; from Surian (2006) for the Tagliamento, Brenta and Piave;
from Kondolf et al. (2007) for the Eygues, from Wardand Uehlinger (2003) for Val Roseg; communication by J. Toone for the upper Drme
and by C. Kleinas (2003) for the Rhine. (B) Same scatter with the positionof the different braided reach types (see Fig. 6). The grey crosses
are the natural corridor widths. Type 1: y =6.24 x0C94;R
2=0.74; Type 2: y =4.29 x1C05 ;R
2=0.70; Type 3: y =19.6 x0C67;R
2=0.70; Type 5: y =
7.73 x0C97;R
2=0.80; Type 6: y =33.5 x0C60 ;R
2=0.61; Type 7: y =18.8 x0C87;R
2=0.43. (C) Same scatter showing the sinuosity rate of the 29
reaches studied between ca 1975 and 2000 at year 1 and year 2. The circles are proportional to the sinuosity rates of the active channel. (D)
Same scatter showing the narrowing – widening rate of the 29 reaches studied in ca 1975 and 2000. The black circles correspond to reaches
characterized by channel widening and the white circles by channel narrowing (in % of the channel width at year 1). In (C) and (D), the
black crosses are the natural corridor widths and the grey the contemporary active channel widths of the 49 sampled reaches.
Aquat. Sci. Vol.71, 2009 Research Article 381

Previous case-study analyses have shown that the
braided planform evolution is associated to a set of
factors that interact in a complex manner (Landon et
al. 1998; Libault and Pigay, 2001; Surian and
Rinaldi, 2003; Kondolf et al., 2007). The active
construction of infrastructure took place during a
period when sediment delivery was decreasing be-
cause of depopulation and decline in grazing pressure
in mountain areas, thus reducing the sensitivity of
slope to erosion. However, it may be possible to
distinguish a variety of factors influencing each reach
individually. Since we did not carry out a detailed
reconstruction of channel evolution but just studied a
few dates, this creates some uncertainty in determin-
ing the full causes of change. Nevertheless, it demon-
strates how infrastructure development affects braid-
ed geometry -in terms of cumulative length- more
importantly than all other human-controlled factors
acting only on channel adjustment conditions and
pattern metamorphosis.
The change in braided planform, which is not yet
fully explained, is more widespread, but mainly
concerns upstream sections. This change could be
explained by a disconnection between sediment
Fig. 9 (continued)
382 H. Pigay et al. Census and typology of braided rivers in the French Alps

sources and the channel network. We hypothesise that
they are associated with a reduction in sediment
supply, due to hillslope afforestation and stabilisation,
and also to local channel adjustment, with vegetation
encroachment following changes in floodplain land-
use, where grazing was replaced by the spontaneous
development of forests (Libault and Pigay, 2002;
Kondolf et al., 2007).
We also showed that 50% of braided rivers from
the 18/19 centuries are still present in the French Alps,
which is significant compared to Austria, for example,
which evolved from 28% of the hydrographic network
(>500 km2) to 1% (Muhar et al., 2008). It would be
helpful to have other exhaustive censuses in the rest of
the European Alps, notably in Italy where significant
braided reaches are located to have a good idea of
changes and existing natural heritage.
A typology driven by a regional diversity of controlled
factors or by differences in adjustment stages?
The typology showed that the braided types 4, 6 and 7
are characterized by a significantly larger normalised
channel than the other types. In the case of types 4 and
7, they are further upstream in the catchment, strongly
dependent on sediment sources. Type 6 is slightly
lower than the two previous types, but in a region that
should deliver more abundant sediment than that
where the other types are located.
While it seems clear that braided activity is
controlled by regional settings (east versus west,
upstream versus downstream), typology also showed
that when the active channel width is significantly
large for a given catchment area, the riparian forest is
narrow in the valley corridor (e.g., type 4, partly types
3 and 6), and in reverse, when the active channel width
is significantly narrow for a given catchment area, the
riparian forest is large (e.g., type 1, type 5). How can
this be explained? Three interpretations can be
considered : i) The active channels are undergoing a
change, a narrowing by vegetation encroachment due
to bedload delivery decrease, which is variable in
space and time, so that when the active channel is
narrowing (e.g., less and less active), the forest is
encroaching the bars and vice versa. ii) The presence
of riparian forest along a channel could also be
controlled by human pressure on its external margins,
promoting deforestation for agricultural or other land
use purposes, which may also vary in space at a
Figure 10. (A) Temporal evolution of the normalised active channel width of five well-known river reaches from early 19 to early 21
centuries. (B) Same evolution of the 29 reaches observed between ca 1975 and 2000. (C) Distribution of the difference between the
normalised widths measured in ca 1975 and 2000.
Aquat. Sci. Vol.71, 2009 Research Article 383

regional level according to the conditions of develop-
ment. iii) The presence of riparian forest can also be
explained by historical lateral shifting of active
channels in the floodplains. The renewal rate seems
to be highly variable in space, and we presume that
some braided reaches entirely renewed their natural
corridor at a pluri-centennal scale, as estimated for the
Waimakari river (New Zealand) by Reinfelds and
Nanson (1993), when others would be more stable. In
this way, the natural corridor could be an indicator of
the total eroded area during the last two centuries,
rather than a historical active channel width at its
maximum.
Following our earlier works, the detailed analysis
performed on the Eygues River (Kondolf et al., 2007)
but also on a large set of rivers in the Drme, Roubion
and Eygues catchments (Libault and Pigay, 2002), it
is now well known that the riparian forest area
observed along the French Alpine rivers primarily
results from bar encroachment due to vegetation
development throughout the 20 century, following
grazing abandonment and decreased bedload supply.
In this context, we assumed that the first interpreta-
tion is the strongest, so that the natural corridor should
be a good estimate of the active channel width at its
maximum (ca late 19 to early 20 century.). We then
tested this assumption by plotting the mean natural
corridor width against catchment size in the figure 9A
(black crosses). We plotted the most active braided
reaches known in the European Alps during the
maximum pressure on land in the middle 19 century.
The black crosses indicate the maximum normalised
active channel width potentially observed in the
French Alps within the range of study catchment
areas, and the grey crosses the position of the 49 study
braided channels. The Val Roseg and Rhine highlight
the maximum range of braided rivers in the European
Alps in terms of catchment size (from 66 km2to 35,000
km2) and slope (from 58 to 0.91 m km1).
Following these results, it seems that an empirically
derived maximum law of braiding could be defined for
the study area, assuming that the natural corridor is
representative of the size of the river during the 19
century. The Val Roseg as well as the middle Durance
and Drme in the middle of the 20 century, but also the
Brenta and Piave, fit the maximum activity law well.
Inversely, the Tagliamento at its known maximum was
slightly above this maximum and appears as a real
outlier. It is interesting to see the position of the Rhine
and lower Durance at their maximum, in relation to
the maximum law, as they are the largest braided
systems in terms of catchment size. They are signifi-
cantly much less braided than what is expected.
Therefore, it is also possible to interpret the distance
from the maximum braided law to the black crosses
partly as the effect of the local human pressure on land
in the valleys (see the arrow on Fig. 9A). This is
coherent with what has been observed on the Roubion
and Eygues, where part of the riparian forests were
deforested in the 1970s for agricultural purposes,
gravel and dump storage; but it is only a minor factor
explaining the evolution of the riparian forest width
through time.
When we fit the partial relationships between the
active channel width and catchment size for each of
the observed types (Fig. 9B), it is possible to highlight
interesting trends. Figure 9 could then be used to tell
where the braided rivers are situated on the meta-
morphosis trend and how braided channels in the
Alpine and Mediterranean areas have narrowed over
the past century, if indeed the current riparian forest
was formerly occupied by active channels. Types 4, 6
and partly 7 follow parallel trends compared to types
1, 2 and 3, underlining clear differences in terms of
adjustment. The first ones are still much more active
than the second ones, which followed a similar trend a
century ago. This result suggests that types 1 to 3 but
also type 5 have the largest forest area, which would
mean that they have been the most affected by the
narrowing of the active channel since the end of the 19
century. In upland areas, type 4 would not have been
affected by a reduction in sediment delivery as it is not
affected by forest encroachment, whereas type 5
reaches, which are also closer to sediment sources,
would have undergone significant historical adjust-
ment.
The difference observed between types 1 to 3 and
types 4/6/7 in figure 9B agrees with the hypothesis
according to which the western prealpine braided
reaches would be more advanced in terms of deficit in
bedload delivery compared to the eastern ones, island
formation and extension being an indicator for
channel metamorphosis. Because a peak in island-
extension was observed in the 1970s, it should confirm
that the island-pattern in this context may be more a
transitory pattern appearing during the narrowing
process than a truly hydraulic controlled pattern. The
chronology of grazing decline and spontaneous affor-
estation may be slightly different between the two
types, as shown on the Drme compared to the
Eygues, which explains that metamorphosis is still
going on in the inner part of the Southern Alps
compared to the middle Rhne valley. Differences
may also be associated with geological and climate
variations controlling flow conditions and sediment
delivery. Compared to the western French braids, the
eastern French catchments are more sensitive to
erosion and they exhibit more intense Mediterranean
rainfall, explaining higher sediment delivery and a
lesser narrowing.
384 H. Pigay et al. Census and typology of braided rivers in the French Alps

Whatever the answers, additional historical anal-
ysis of riverside land-use is needed to validate the
hypotheses. At this stage, it is clear that it is a
challenging issue to distinguish what it is linked to a
regional complexity (higher sediment delivery be-
cause of climate, geologic or orographic conditions)
with the temporal complexity (position of the reach
along its adjustment trajectory due to secular sedi-
ment decrease associated to changes in human
pressures, and possibly with climate change). It is
important to understand the temporal range of
variation in channel width for each of the cases and
to understand the regional organisation, which is a
challenging issue for research because of the time
investment needed to analyse so many historical
documents. It is possible that the eastern Alps,
notably the Dolomites, may deliver more sediment
than the western Alps, explaining higher braided
width per square kilometre, as shown in the case of
the Tagliamento (Surian, 2006). It is also necessary to
consider the complexity of downstream sediment
transfer because of discontinuities along the hydro-
graphic network (tributaries, confinement).
Longitudinal structure of natural corridors could
be analysed in order (i) to identify a disruption of the
corridor that potentially affects local dynamics; and
(ii) to understand the spatial complexity of historical
adjustment and the associated controlling factors. As
a consequence, this maximum law should evolve in
the future with a better understanding of the regional
complexity at the Alpine scale. The cases of the lower
Durance and Rhine at their maximum were surpris-
ingly lower than expected. This also enables a
discussion about the furthest downstream distance
of sediment propagation, from the sediment sources
down to the bottom of the Alpine massif. Are
downstream sections narrower than upstream ones
because they have not yet adjusted to the sediment
propagation or because the downstream sections are
progressively influenced by low sediment transport
tributaries, so flow becomes dominant compared to
sediment delivery, progressively creating a new
pattern downstream?
It is therefore possible to derive another relation-
ship corresponding to the minimum of braiding
activity (Fig. 9A) below which we should consider
that a river is not truly braided, but wandering. The
width of the Durance after damming is not at its
narrowest in braided activity, as it has been artifi-
cially maintained for flood risk management and
would have been much narrower if it had been
completely adjusted. In this context, a reference is
provided by the Drac downstream from Saint-
George de Commiers dam built in 1962 where the
channel, which was fully braided in 1956, moved from
an average width of 145 m in 1956 to 41 m in 1987
(Peiry and Vivian, 1994), with a narrowing of 72%
due to peak flow reduction and intense gravel
mining. This reach was no longer braided in 1987,
and is situated a fair stretch down from the lower
boundary of the braided area (Fig. 9A). Another
good example is the upper Drme near Luc, where
the channel width fluctuations classify as braided
after floods, but near wandering when vegetation
encroaches bars over the decade following the flood.
The observation of aerial photographs and field
reconnaissance provide confident visual ground
truth. The minimum law of braided activity could
be used to detect, in the narrowing process, which
braided stretches are near to metamorphosis and
which ones are still fairly active, so as to improve the
understanding of regional factors controlling braid-
ing (Gurnell and Petts, 2006).
Short-term fluctuations versus long-term adjustments
and consequences for ecological potential
The observations here also question the role of short-
term fluctuations in channel width evolution com-
pared to long-term changes. Is the long-term narrow-
ing, which exists both on French and Italian rivers,
reversible in short time scale, as is possibly suggested
with the recent channel rewidening following the
large floods of the 1990s (Rollet, 2007; Surian et al.,
2009; Rinaldi et al., 2009)? The long-term narrowing
(e.g. 50 to 60%) is not as high as the short-term
fluctuations (ca. 30%) along the Drme. From the
period of maximum intensity, it has been possible to
see how the Tagliamento, Eygues and Drme have
behaved through time (Fig. 9A) and to understand
the comparison between short-term fluctuations and
long-term adjustments better. Whereas the narrowing
of the Eygues and Tagliamento reached, respectively,
60% and 53% over the last two centuries, the range in
change for the upper Drme (e.g., -33%) is associated
only with flood periodicity, since the channel width
varies across a stationary trend, widening significantly
after the floods of 1951, 1978, 1994 and 2003, but
narrowing in the periods after.
The Durance in 1986, almost three decades after
peak flow decrease resulting from the Serre-PonÅon
dam, narrowed from 1200 m to 400 m (-66%). In fact,
understanding the size effect provides arguments to
consider changes according to normalised width
values. When considering not the relative narrowing
(in % of the initial active channel width) but the
absolute normalised width W*, then the range in
fluctuation is significantly different between those
associated with flood periodicity and the long-term
changes shown for the Tagliamento, Eygues, Piave
and Brenta at a secular scale. So that a fluctuation of
Aquat. Sci. Vol.71, 2009 Research Article 385

+/- 5 times in normalized channel width, observed on
a specific reach, seems to be associated with flood
periodicity, whereas values of 10 to 20 underlined a
real long term adjustment (Fig. 10 A, B).
Concerning the hypothesis according to which the
natural corridor width may be controlled by channel
shifting (e.g., higher the channel shifting, higher the
natural corridor), the results have shown that the
shifting increases or decreases in relation with plan-
form fluctuations at a short time scale (1 to 2
decades). When they become narrower and more
sinuous, they shift more than when they become
straighter and wider (Fig. 9 C, D). These trends,
which have both been observed between ca 1975 and
2000 on various rivers, are therefore not driven by
geographical controls (Fig. 9 C, D). When the reach
goes down to the lower limits of activity, it is
characterized by a higher sinuosity (Fig. 9C). This
also shows that the rate of changes between 1975 and
2000, in terms of normalised active channel width,
ranges from -2 and 3 with a maximum at -6 (Fig. 10 B,
C), slightly lower than 10 to 20, which was considered
previously as a critical value for detecting a truly
secular adjustment (Fig. 10A). Moreover, the most
sinuous and narrowest channels underwent signifi-
cant floodplain erosion (black circles, Fig. 9D),
whereas the less sinuous and widest channels under-
went floodplain encroachment during the survey
period (ca 1975 – 2000) (white circles, Fig. 9D).
As a consequence, we consider that shifting does
not vary regionally but temporally on the different
reaches, according to their state on the dynamic
equilibrium, depending on when the observation
was done, in relation to the flood distribution
through time. When the reaches suffer a large flood,
such as on the Bech, reaches widen tremendously,
whereas in other cases, following a long period of low
floods, reaches undergo a more significant narrowing.
It is therefore important to know when the measure-
ments were made, with respect to the occurrence of
the floods, to be able to compare the braided river
geometry and riparian corridor characteristics. Thus,
braided patterns evolve significantly with time, indi-
cating that additional knowledge must be sought to
understand how peak flow chronology and channel
recovery due to catchment and riparian afforestation
affect channel geometry and its interactions with the
riparian area.
The distinction existing between the wandering,
island-braided and truly braided patterns should also
be discussed in terms of controlling factors. The
minimum law of braided activity is a threshold that
should be explored, introducing an additional variable
that is the sinuosity of the braided active channel, as
there is a link between the reduction in length and the
increase in sinuosity when moving from a truly
braided to a wandering channel. A braided reach
can be also a wandering one, depending on where the
observed situation is positioned in its flood periodicity.
This is independent of any adjustment conditions, and
shows the temporal complexity of such systems. It
opens the way to a new reflection on the effects of the
changing rate of geomorphic patterns on ecological
assemblages and richness. When can we consider a
braided island as a permanent planform or as a
transitional one? Is it then a spatial ecotone or a
temporal ecotone in the sense given by Naiman et al.
(1990)? Following Zanoni et al. (2008), islands are
very ephemeral features, which increase in frequency
during the narrowing process but decrease during
widening, showing an interesting temporal pattern
similarly to what we observed here in relation to flood
history and channel shifting controlled by fluctuating
planform pattern.
These findings open the question as to whether
the longitudinal organisation of the planform pattern
observed, for example, on the Tagliamento by
Gurnell and Petts (2006) with truly braided straight
sections and more island-braided sections, is associ-
ated with a geographical setting or with complex
temporal fluctuations that are dissociated from one
reach to another. The geometrical characteristics of
the remnant braided reaches and their evolution
through time are complex. While it is possible to
highlight the general trend in narrowing and decrease
of braided activity, thus distinguishing several geo-
graphical areas, additional elements are needed to
understand the process of narrowing and lateral
rejuvenation, and to determine which elements influ-
ence the development of islands in space and time.
Does the presence of islands indicate a specific
hydraulic condition in the hydrographic network, as
suggested by Gurnell and Petts (2006), or do they
appear specifically during the phase of encroachment
and disappear when a new active flood period
occurs? Are they then a transitory pattern of braided
river existence in time and space? It is also expected
that the control and response of vegetation may
differ from one reach to another, according to its
position in the hydrographic network. Summer con-
ditions (low flow, hot temperature) affect vegetation
growth in lowland Mediterranean reaches, whereas
higher flows and mild temperatures are observed
along the northern braids during the growing season.
A temperature gradient, limiting tree establishment
as far as the reach is located in the upstream reaches,
is also to be expected. All this amounts to explain a
complex pattern of island-braided distribution in the
network, independently of the stage of adjustment of
the individual braided reaches.
386 H. Pigay et al. Census and typology of braided rivers in the French Alps

Conclusions
The spatial and temporal analysis of braided rivers in
the French Alps has clearly shown their progressive
disappearance due to various types of human pres-
sure. A comparative analysis of a set of braided
reaches sampled in the different areas of the Rhne
catchment also showed that the existing braided
reaches underwent complex adjustment processes,
mainly a narrowing, variable in intensity, in relation to
the decrease in bedload delivery, the intensity of which
also differs from west to east and from upstream to
downstream (depending on the distance to sediment
sources). Braided planforms can also undergo short-
term fluctuations (in terms of island frequency,
sinuosity and shifting) in relation to flood periodicity
with clear consequences for ecological assemblages.
This preliminary contribution provided a series of
hypotheses that are to be validated by additional
research. A detailed historical analysis of the corridor
mosaics should now be performed in relation to the
series of peak flows in order to go further in the
understanding of channel changes over the last
century and its consequences in terms of ecology.
Acknowledgments
The authors kindly thank the Water Agency that is
supporting the research and the ISIG technical plat-
form of the ENS-lsh. Adrien Alber got a PhD grant
from the French Ministry of Agriculture. The research
is conducted in the LTER group of the ZABR (Zone
Atelier Bassin du Rhne) and the IWRnet FOR-
CASTER project (2008-2010). Damien Saulnier and
Jules Toone are also thanked for their contributions in
GIS work.
References
Best, J. L. and C. S. Bristow, 1993. Braided Rivers. Special
Publication 75, Geological Society Publishing House, Bath.
Brewer, P. A. and J. Lewin, 1998. Planform cyclicity in an unstable
reach: complex fluvial response to environmental change.
Earth Surface Processes and Landforms 23: 989–1008.
Bravard, J. P. 1989. La mtamorphose des rivires des Alpes
franÅaises  la fin du moyen-ge et  lpoque moderne. Bull.
de la Soc. Gog. de Lige 25: 145 – 157.
Bravard, J. P. and J. L. Peiry, 1993. La disparition du tressage fluvial
dans les Alpes franÅaises sous leffet de lamnagement des
cours deau (19-20me sicle). In: I. Douglas and J. Hagedorm
(eds), Geomorphology and Geoecology, Suppl. Berlin: Zeits-
chrift fr Geomorphologie. pp. 67 – 79
Bravard, J.P., C. Amoros, G. Pautou, G. Bornette, M. Bournaud, M.
Creuz des Chatelliers, J. Gibert, J. L. Peiry, J. F. Perrin J. F. and
H. Tachet, 1997. River incision in South-east France: morpho-
logical phenomena and ecological effects. Regulated rivers:
research and management 13: 1–16.
Burkham, D. E. 1972. Channel changes of the Gila River in Safford
Valley, Arizona, 1846-1970: U.S. Geological Survey Professio-
nal Paper 655-G, 24 p.
Carbiener, R., 1974. Le Rhin et lAlsace, histoire de lvolution des
rapports entre lhomme et un grand fleuve. Bull. Soc. Industr.
Mulhouse 757: 61– 69.
Church, M., 1992. Channel morphology and typology. In : P. Calow
and G.E. Petts (eds), The Rivers Handbook. Hydrological and
Ecological Principles. Oxford, Blackwell scientific publica-
tions. 1: 126– 143.
Dufour, S. and H. Pigay, 2009. The myth of the lost paradise to
target river restoration: forget natural reference, focus on
human benefits. River Research and Applications 25(5): 568–
581
Egozi, R. and P. Ashmore, 2008. Defining and measuring braiding
intensity. Earth Surf. Process. Landforms 33: 2121–2138.
Ferguson, R. I., 1987. Hydraulic and sedimentary controls on
channel pattern. In : K. Richards (ed.), River Channels :
Environment and Process. Blackwell Ldt., U.K., pp. 129–158.
Ferguson, R. I. and A. Werrity, 1983.Bar development and channel
changes in the gravely river Feshie, Scotland. In: J. D.Collinson
and J. Lewin (eds), Modern and Ancient Fluvial Systems.
Intern. Assoc. of sedimentology sp. Publ. 6: 181– 193.
Francis,R. A. , D.Corenblit and P.J. Edwards, 2009. Perspectives on
biogeomorphology, ecosystem engineering and self-organisa-
tion in island-braided fluvial ecosystems. Aquatic Sciences
DOI 10.1007/s00027-009-9182-6
Gurnell, A. M. and G. E. Petts, 2002. Island-dominated landscapes
of large floodplain rivers, a European perspective. Freshwater
Biology 47: 581– 600.
Gurnell, A. M. and G. Petts, 2006. Trees as riparian engineers: The
Tagliamento River, Italy. Earth Surf. Process. Landforms 31
(12): 1558–1574.
Gurnell, A., N. Surian and L. Zanoni, 2009. Multi-thread river
channels: A perspective on changing European alpine river
systems. Aquatic Sciences DOI 10.1007/s00027-009-9186-2
Habersack, H. and H. Pigay, 2008. River restoration in the Alps
and their surroundings: past experience and future challenges.
In H. Habersack, H. Pigay and M. Rinaldi (eds), Gravel-bed
River 6: From Process Understanding to River Restoration,
Elsevier, Amsterdam. pp. 703– 737.
Kellerhals, R. and M. Church, 1989. The morphology of large rivers
: characterization and management. International Large River
Symposium, Can. Spec. Publ. Fish. Aquat. Sci. 106: 31–48.
Kleinas, C., 2003. Lvolution du style fluvial du Rhin Suprieur de
Ble  Lauterbourg. Unpublished Masterthesis, University of
Lyon 2 and University of Strasbourg 1, 54 pp.
Kondolf, M. G., H. Pigay and N. Landon, 2007. Changes in the
riparian zone of the lower Eygues river, France, since 1830.
Landscape Ecology 22: 367 – 384
Landon, N., 1999. Lvolution contemporaine du profil en long des
affluents du Rhne moyen, constat rgional et analyse dun
hydrosystme complexe, la Drme. PhD thesis, Universit
Paris IV–Sorbonne, Paris.
Landon, N., H. Pigay and J. P. Bravard, 1998. The Drme River
incision (France): from assessment to management. Landscape
and Urban Planning 43: 119 –131.
Leopold, L. B. and M. G. Wolman, 1957. River channel patterns-
braided, meandering and straight, U.S. Geol. Surv. Prof. Paper
282 B: 39 –85.
Libault, F. and H. Pigay, 2002. Causes of 20th century channel
narrowing in mountain and piedmont rivers and streams of
Southeastern France. Earth Surf. Process. Landforms 27: 425–
444.
Libault, F., H. Pigay, P. Frey and N. Landon, 2008. Chapter 12
Tributaries and the management of main-stem geomorphol-
ogy. In: S. Rice, A. Roy and B. L. Rhoads (eds), River
Confluences and the Fluvial Network, J. Wiley and Sons,
Chichester pp. 243– 270.
Miramont, C., M. Jorda and G. Pichard, 1998. Evolution historique
de la morphogense et de la dynamique fluviale dune rivire
mditerranenne: lexemple de la moyenne Durance (France
Aquat. Sci. Vol.71, 2009 Research Article 387

du Sud-Est). Gographie Physique et Quaternaire 52 (3): 1–
13.
Muhar, S, M. Jungwirth, G. Unfer, C. Wiesner, M. Poppe, S.
Schmutz, S. Hohensinner and H. Habersack, 2008. Restoring
riverine lanscapes at the Drau River: successes and deficits in
the context of ecological integrity. In: H. Habersack, H. Pigay
and M. Rinaldi (eds), Gravel-bed River 6: From Process
Understanding to River Restoration, Elsevier, Amsterdam.
pp. 779– 807.
Naiman, R. J., H. Dcamps and 19 authors, 1990. The Ecology and
Management of Aquatic-Terrestrial Ecotones. Man and the
Biosphere series, Unesco, 4, 316 pp. Paris.
Nakamura, F. and N. Shin, 2001. The downstream effects of dams
on the regeneration of riparian tree species in northern Japan.
In: J. M. Dorava, D. R. Montgomery, B. B. Palcsak and F. A.
Fitzpatrick (eds), Geomorphic Processes and Riverine Hab-
itat. Am. Geophys. Union Wat. Sci. Appl. 4: 173–181.
Nanson, G. C. and A. D. Knighton, 1996. Anabranching rivers:
their causes, character and classification, Earth Surf. Process.
Landforms 21 (3): 217–239.
Osterkamp, W. R. , 1998. Processes of fluvial island formation, with
examples from Plum Creek, Colorado and Snake River, Idaho.
Wetlands 18: 530–545.
Peiry, J. L., 1987. Lutilisation du cadastre sarde de 1730 pour
ltude des rivires savoyardes: lexemple de la valle de lArve
(Haute-Savoie). Revue de Gographie de Lyon 64 (4): 197–
203.
Peiry, J. L. and H. Vivian, 1994. Dynamique des crues et rduction
des capacits dcoulement du chenal conscutives  la
construction dun barrage hydrolectrique: lexemple du
Drac infrieur en amont de Grenoble. Conference on ”Crues
et inondations”, 23mes Journes de lHydraulique, Socit
Hydrotechnique de France, Nmes, 14-15 et 16 sept. 1994, pp
321–329.
Pigay, H., O. Barge and N. Landon, 1996. Streamway concept
applied to river mobility / human use conflict management.
Proceedings of the International Water Resources Associa-
tion: ”Rivertech96 : new/emerging concept for rivers”, Chica-
go, USA, pp. 681– 688.
Pigay, H., G. Grant, F. Nakamura and N. Trustrum, 2006. Braided
river management: from assessment of river behaviour to
improved sustainable development. In: G. H. Sambrook-
Smith, J. L. Best, C. S. Bristow and G. E. Petts (eds), Braided
Rivers: Process, Deposits, Ecology and Management, Special
publication 36 of the International Association of Sedimentol-
ogists, Blackwell Publishing, pp. 257 –275.
Pont, D. and C. Rogers, 2004. Modlisation des distributions
piscicoles  lchelle du rseau franÅais. Comparaison avec les
Contextes. Unpublished Final Report, Conseil Suprieur de la
PÞche, 116 pp.
Pont, D., H. Pigay, A. Farinetti, S. Allain, N. Landon, F. Libault,
B. Dumont and A. Mazet, 2009. Conceptual framework and
interdisciplinary approach for the sustainable management of
gravel-bed rivers: the case of the Drme River basin (SE
France). Aquatic Sciences DOI 10.1007/s00027-009-9201-7
Reinfelds, I. and G. Nanson, 1993. Formation of braided river
floodplains, Waimakariri River, New Zealand. Sedimentology
40: 1113–1127.
Rempel, L. L., M. Healey and J. R. Richardson, 2000. Macro-
invertebrate community structure along gradients of hydraulic
and sedimentary conditions in a large gravel-bed river.
Freshwat. Biol.45: 57 – 73.
Rinaldi, M., C. Simoncini and H. Pigay, 2009. Scientific strategy
design for promoting a sustainable sediment management: the
case of the Magra River (Central – Northern Italy). River
Research and Applications 25: 607 – 625.
Rinaldi, M., B. Wyzga, and S. Surian, 2005. Sediment mining in
alluvial channels: physical effects and management perspec-
tive. River Research and Applications 21: 1– 24.
Rollet, A. J., 2007. Etude et gestion de la dynamique sdimentaire
dun tronÅon fluvial  laval dun barrage: le cas de la basse
valle de lAin. Unpublished PhD thesis, University of Lyon 3,
Lyon, France, 305 pp.
Rust, B.R. , 1978. A classification of alluvial channel systems. In: A.
D. Miall (ed.), Fluvial Sedimentology, Canadian Society of
Petroleum Geologists, Calgary, Alberta, pp. 187–198.
Sambrook Smith, G. H., J. L. Best, C. S. Bristow and G. E. Petts
(eds), 2006. Braided Rivers: Process, Deposits, Ecology and
Management, International Association of Sedimentologists,
Blackwell Publishing, 368 pp.
Schumm, S. A., 1968. River adjustment to altered hydrologic
regimen; Murrumbidgee River and paleochannels, Australia.
USGS prof. paper 598, 65pp.
SDAGE RMC, 1997. Schma Directeur dAmnagement et de
Gestion des Eaux du basin Rhne, Mditerrane, Corse.
Prfecture de bassin et Comit de Bassin, Lyon. 3 volumes.
http://sierm.eaurmc.fr/sdage/sdage.php
Slater, L., 2007. Caractrisation des rivires en tresses franÅaises.
Unpublished Masters thesis, University of Lyon / ENS-lsh, 53
pp.
Surian, N., 2006. Effects of human impact on braided river
morphology: examples from Northern Italy. In: G. H. Sam-
brook-Smith, J. L. Best, C. S. Bristow and G. E. Petts (eds),
Braided Rivers: Process, Deposits, Ecology and Management,
Special publication 36 of the International Association of
Sedimentologists, Blackwell Publishing, pp. 327–338.
Surian, N. and M. Rinaldi, 2003. Morphological response to river
engineering and management in alluvial channels in Italy.
Geomorphology 50: 307 – 326.
Surian, N., M. Rinaldi, L. Pellegrini, C. Audisio, F. Maraga, L.
Teruggi, O. Turitto, and L. Ziliani, 2009. Channel adjustments
in northern and central Italy over the last 200 years. In: L. A.
James, S. L. Rathburn and G. R. Whittecar (eds.), Management
and Restoration of Fluvial Systems with Broad Historical
Changes and Human Impacts, Geological Society of America
Special Paper 451, in press.
Ward, J. V. and U. Uehlinger, (eds), 2003. Ecology of a Glacial
Flood Plain, Kluwer Academic Publishers, the Netherlands,
Amsterdam, 306 pp.
Ward, J. V., K. Tockner and F. Schiemer, 1999a. Biodiversity of
floodplain river ecosystems: ecotones and connectivity. Regul.
River. Res. Manag. 15: 125–139.
Ward, J. V., K. Tockner, P. J. Edwards, J. Kollmann, G. Bretschko,
A. M. Gurnell, G. E. Pettsand B. Rossaro, 1999b. A reference
system for the Alps: the Fiume Tagliamento. Regul. River.
Res. Manag 15: 63– 75.
Ward, J. V., K. Tockner, D. B. Arscott and C. Claret, 2002. Riverine
landscape diversity. Freshwat. Biol. 47: 517–539.
Zanoni, L., A. Gurnell, N. Drake and N. Surian, 2008. Island
dynamics in a braided river from analysis of historical maps and
air photographs. River Research and Applications 24 (8):
1141–1159.
To access this journal online:
http://www.birkhauser.ch/AS
388 H. Pigay et al. Census and typology of braided rivers in the French Alps