Earth Surface Processes and Landforms
Earth Surf. Process. Landforms 27, 425­444 (2002)
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/esp.328
CAUSES OF 20th CENTURY CHANNEL NARROWING IN MOUNTAIN
AND PIEDMONT RIVERS OF SOUTHEASTERN FRANCE
F. LIÉBAULTa,b* AND H. PIÉGAYa,b
a UMR 5600 - CNRS Environnement-Ville-Société, 18 rue Chevreul, 69 362 Lyon cedex 07, France, and Laboratoire Rhodanien de
b Géographie de ľEnvironnement, Université Lumi`ere Lyon 2, 5 avenue Pierre Mend`es-France, 69 676 Bron cedex, France
Received 23 April 2001; Revised 22 August 2001; Accepted 22 August 2001
ABSTRACT
Extensive channel narrowing in southeastern France provides an illustration of geomorphic response to land-use changes.
The study region comprises a range of environments, from large piedmont and intramountain gravel-bed rivers, to
small mountain streams. Field measurements and analysis of historical data demonstrate two distinct periods of channel
change. From 1850 to 1950, channel narrowing is interpreted to be the result of a recovery process in response to
widespread channel destabilization induced by major floods during the second half of the 19th century. At the time, the
largely deforested basins were highly responsive to flooding, whereas the recovery process was accelerated by floodplain
and basin-scale land use changes (afforestation) and torrential control works, which in turn reduced sediment delivery
and enabled vegetation development in channels. From 1950 to 1970, channel narrowing accelerated in most of the
studied rivers. This recent phenomenon is considered as a human-induced fluvial adjustment, directly related to forest
development on river margins and human abandonment of intensive floodplain land uses. At the same time, long-profile
degradation occurred as a result of long-term bedload supply decrease. In small mountain streams, channel narrowing is
mainly explained by channel incision which seems to progress from upstream where sediment sources are progressively
stabilized by afforestation. Copyright  2002 John Wiley & Sons, Ltd.
KEY WORDS: channel width adjustment; active channel narrowing; riparian vegetation; incision; land-use change; hydrological
change; mountain stream; gravel-bed river; France
INTRODUCTION
Rivers have been viewed as systems at least since Schumm's (1977) exposition of the complex links between
control variables of bedload sediment supply, Qs, and the discharge Q, and dependent variables of channel
geometry. Research conducted on hydraulic geometry following Leopold and Maddock's (1953) pioneering
work established power relationships between the discharge and channel pattern, depth, width, and meander
wavelength. For example, empirical relationships have been established on several rivers to predict the channel
width (W in m) from Q, the bankfull discharge (in m3
s 1
:
W D a Qb
1
The importance of bedload sediment supply notwithstanding, channel geometry can be strongly influenced
by local controls such as slope imposed by the geological setting, the vegetation cover of banks, and the
grain size of the valley floor. Dense vegetation can increase roughness and, via binding effects of the roots,
increase bank resistance to erosion. As a consequence, the coefficient a of the relationship between discharge
and channel width (Equation 1) decreases when the vegetation cover is greater and more arboreous (Andrews,
1984; Hey and Thorne, 1986; Huang and Nanson, 1997). Herbaceous cover was found to be more effective
than arboreous cover in maintaining narrow channels on small streams (Zimmermann et al., 1967; Bergeron
and Roy, 1985; Clifton, 1989; Trimble, 1997). The contrasting results demonstrate that vegetation plays a
* Correspondence to: F. Liébault, UMR 5600 - CNRS Environnement-Ville-Société, 18 rue Chevreul, 69 362 Lyon cedex 07, France.
E-mail: frederic.liebault@free.fr
Copyright  2002 John Wiley & Sons, Ltd.
426 F. LIÉBAULT AND H. PIÉGAY
Qs--
Channel narrowing
due to vegetation
encroachment
Floodplain
afforestation
a
d
Modification of
agricultural
practices
End of Little Ice Age
(1850-1880)
Torrent control works
Channel degradation
due to dredging and mining
operations
c
Particular
hydrological period
during the 20th century
Planned or
spontaneous
afforestation
b
Climate change
River
damming
Land use change
a : Qs decrease > Q decrease
i) dyschronism between Qs decrease and channel narrowing due to system response time to
change (the response time is mainly related to the time of sediment transfer through the basin)
ii) progressive channel narrowing from the sediment sources
b : Q decrease > Qs decrease
i) synchronism between floodplain afforestation and channel narrowing
c : Hydraulic deconnection of former active surfaces by channel degradation
i) channel incision prior to vegetation encroachment in active channels
ii) progressive channel narrowing from locations where the bed is lowered (upstream and
downstream progression of degradation from disturbed reaches)
d : Channel adjustment to floodplain afforestation
i) synchronism between floodplain afforestation and channel narrowing
Q--
Figure 1. Conceptual model of factors controlling channel narrowing: working hypothesis (a to d)
variable role depending on other parameters, such as bank height or sediment characteristics (Gregory and
Gurnell, 1988).
Just as channel geometry can adjust over time in response to changes that affect bedload and discharge, it can
also adjust to changes in bank and floodplain vegetation. Reservoir-induced peak flow reduction, bed sediment
trapping, and resulting channel narrowing have been described by Eschner et al. (1983), Collier et al. (1996)
and Peiry (1997). Examples of channel narrowing associated with decreased discharges include Schumm and
Lichty (1963), and Friedman et al. (1996). Documented effects of vegetation change on channel form include
channel widening after bank vegetation loss (Kondolf and Curry, 1986) and channel narrowing associated
with vegetation development on channel margins (Hadley, 1961; Nevins, 1969; Graf, 1978). Cross-section
adjustment to vegetation change is generally explained by the modification of bank roughness and resistance
to erosion (Millar, 2000) or by the modification of in-channel sediment trapping efficiency which can induce
floodplain construction in former active channels (Schumm and Lichty, 1963; Osterkamp and Costa, 1987).
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 427
European rivers have undergone complex adjustments during the last two centuries because human activities
have influenced discharge, bedload supply and transport at different levels within the basins and in different
periods. As a consequence, channel narrowing at different times and places has been interpreted in different
ways. Channel shrinkage has been associated with effects of reservoirs and discharge modifications (Peiry and
Vivian, 1994; Surian, 1999), floodplain land-use changes (Piégay et al., 1994, in press), hydrological fluctuations
(Miramont and Guilbert, 1997) or bedload supply decrease (Bravard and Peiry, 1993; Garcia-Ruiz et al.,
1997; Liébault and Piégay, 2001). If the first cause is easy to establish because of the strong chronological
links existing between the dates of reservoir construction and consequent vegetation encroachment, the two
others are more difficult to prove. They often occurred in the same basins and their respective influences on
vegetation encroachment have still not been established.
The aim of this study is to review the different observations made on wide gravel-bed rivers and small
streams in the mountains and piedmont of southeastern France to determine the respective influence of basin
and valley-floor scale controls on the channel narrowing process (Figure 1). Based on the chronological and
spatial patterns of channel and environmental changes in multiple basins in the region, the relative influence
of natural and human factors on narrowing is assessed.
MATERIAL AND METHODS
A contrasted sample from catchment size and region concerned
A large set of rivers and streams located in southeastern France (Figure 2 and Table I) has been selected.
These basins range in catchment area from 63 to 17 600 km2
and span a wide range of climates (oceanic,
continental and Mediterranean) and geomorphological environments (intramountain plains covering lacustrine
or periglacial deposits, piedmonts). In the first instance, a comparison analysis on the large rivers in the
data set (i.e. >450 km2
) is performed. Secondly, a detailed analysis on the Eygues, Roubion and Drôme
basins is documented. These basins adjoin one another in the southern Prealps, a homogeneous geomorphic
region of low-elevation limestone mountains. Field and archival data were collected from 51 tributaries to
test chronological links between channel narrowing and basin land-use changes and/or channel regulation.
These tributaries range from 10 to 150 km2
in drainage area, from 450 to 1390 m in mean elevation, and
from 0008 and 003 m m 1
in slope. They all have floodplains that are sufficiently wide to allow lateral
adjustment in their downstream reaches. These streams represent small fluvial systems with a strong coupling
between hillslopes and channels in their upstream reaches, suggesting a high sensitivity to environmental
changes.
Assessment of channel narrowing
Historical data (old maps and air photos) and dendrochronology were used to characterize and date contemporary
changes in channel width and vegetation encroachment on formerly active alluvial surfaces. The
widths of active channels (low-flow channels and unvegetated gravel bars) were measured on aerial photographs
from the mid-1940s to the 1990s (scales range from 1 : 17 000 to 1 : 25 000) and on detailed maps
and land surveys from the 19th century. On wide gravel-bed rivers, it was possible to measure active channel
width for regularly spaced segments (generally 500 m in length). Investigated reaches ranged from 21 to
93 km in length and were generally selected on downstream parts of the basins. On small streams, active
channel areas were measured in downstream reaches (all in alluvial portions of the streams) and mean channel
width was calculated by dividing the active area by the length of the reach. Comparing measurements of
active channel width from aerial photos and the field on two small tributaries of the Eygues showed that the
air photos underestimated channel width because part of the channel was obscured by bank vegetation. The
portion of channel obscured by bank vegetation was up to 10 m, with most values between 0 and 4 m, and a
mean of 309 m (Liébault et al., 2001). Thus, analysis of sequential air photos should be capable of detecting
width changes of more than 3 m. A Mann­Whitney U-test was performed to characterize differences of
channel width (non-normal distribution) between dates. An ANOVA test was performed to detect significant
differences between rates of channel narrowing (normal distribution) obtained on tributaries of the Drôme,
Eygues and Roubion rivers.
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
428 F. LIÉBAULT AND H. PIÉGAY
Ain
Doubs
Giffre
Drôme
Roubion
Eygues Ouvze
Ardche
Ubaye
Buëch
Bléone
Fiume Secu
Loire
Allier
Rhône
Rhine
CORSICA
a)
main French rivers
studied rivers
studied reaches
main towns
Gaging stations :
1
2
3
4
5
6
Basel (Rhine River)
Besançon (Doubs River)
Luc-en-Diois (Drôme River)
Sauze-St-Martin (Ardche River)
Serres (Buëch River)
Barcelonnette (Ubaye River)
0 200 km100
1
2 3
4
5 6
7
8
9
10
1 1
12
13
14 15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37 38
39
40
41
42 43
44
45
46
47
48
49
50
51
Roubion River
Eygues River
Drôme River
Luc-en-Diois
Nyons
Die
10 km0
main trunks
main tributaries
watershed limits
sub-basins
main villages
b) Prealpine tributaries :
1 : Gervanne 27 : Vbre
2 : Charsac 28 : Eyzarette
3 : Riousset 29 : Bine
4 : Sure 30 : Soubrion
5 : Marignac 31 : Upper-Roubion
6 : Comane 32 : Rieu Sec
7 : Meyrosse 33 : Sauve
8 : Valcroissant 34 : Bordette
9 : Archiane 35 : Cougoir
10 : Gats 36 : Bentrix
11 : Boulc 37 : Rieu Montaulieu
12 : Mians 38 : Ennuye
13 : Contcle 39 : Merderie
14 : Colombe 40 : Arnayon
15 : Esconavette 41 : Cénas
16 : Barnavette 42 : Aiguebelle
17 : Aucelon 43 : Establet
18 : Courance 44 : Upper-Oule
19 : Upper-Roanne 45 : Pommerol
20 : Brette 46 : Baudon
21 : Béoux 47 : Marcijaye
22 : Rif Miscon 48 : Lidane
23: Maravel 49 : Armalause
24 : Charens 50 : Esclate
25 : Upper-Drôme 51 : Upper-Eygues
26 : Rimandoule
1
2
3
4
5
6
Paris
Lyon
Marseille
Figure 2. The study site: (a) location of large gravel-bed rivers studied; (b) detailed map of the selected southern Prealps small mountain
streams
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 429
Table I. Main characteristics of studied rivers
Rivers Drainage
basin
(km2
Max.
elevation
(m)
Geomorphological
characters
Dam regulating
floods (year of
construction)
Mean
annual
discharge
(m3
s 1
)
Reference
Ain 3630 1300 Free-meandering/
piedmont
1968 122 Piégay (1995)
Ard`eche 2240 1500 Single-bed sinuous
river/piedmont
­ 26 Piégay (1995)
Bléone 980 2819 Braided/inner
southern Alps
­ 92 Saulnier (1999)
Bu¨ech 974 2709 Braided/southern
Prealps
­ 15 Gautier (1992)
Doubs 7700 1460 Anastomosing and
meandering/
piedmont
­ 177
Drôme 1642 2051 Braided/piedmont ­ 19 Landon (1999)
Eygues 1150 1757 Braided/piedmont/M ­ 6 Landon Piégay
(unpublished
work)
Fiume Seccu 63 2029 Single-bed sinuous
river/M
­ Gaillot and
Piégay (1998)
Giffre 459 3100 Braided/northern
Alps
­ 19 Piégay (1995)
Loire (upstream
of Nevers)
17 570 1800 Wandering/piedmont 1984 185 Gautier et al.,
(2000)
Allier 13 000 1800 Wandering/piedmont ­ 160 Gautier et al.
(2000)
Ouv`eze 1818 1900 Braided/piedmont/M ­ 5 Piégay (1995)
Roubion 600 1606 Single-bed sinuous
stream/piedmont
­ 2 Liébault and
Piégay (2001)
Southern
Prealps
tributaries
10­150 2051 Single-bed sinuous
streams/southern
Prealps
­ <2 Liébault et al.
(2001)
Ubaye 970 3000 Braided/inner
southern Alps
­ 11 Piégay (1995)
M, Mediterranean.
Trees were cored and rings counted for a total of 405 riparian trees established on the first vegetated surfaces
adjacent to the active channels on both (i) wide gravel-bed rivers draining more than 900 km2
(Drôme, Ain
and Ubaye), and (ii) small mountain streams draining areas between 10 and 100 km2
(Archiane, Béoux, Bine,
Fiume Secu, Gats, Rif Miscon, Upper-Roubion). The oldest species present (Alnus incana, Populus nigra,
Pinus sylvestris) were sampled to obtain a minimum date for establishment of the riparian forest.
Assessment of environmental changes
Modifications of environmental controls on channel morphology were examined at both the basin and reach
scales to assess their respective influence. In addition, gauging records were analysed whenever long records
were available.
Basin-scale controls included land-use changes on hillslopes and, in some areas, torrent control works conducted
between 1860 and 1920 to stabilize sites of active erosion. Land-use modifications since the beginning
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
430 F. LIÉBAULT AND H. PIÉGAY
of the 19th century have been documented in the Ouv`eze, Ard`eche, Ubaye, Eygues, Drôme and Roubion
river basins. Resources utilized included Napoleonic land surveys for the period 1822­1840, agricultural surveys
of 1929, 1954 and 1988, and the National Forest Inventory of 1991. These various sources all indicate
areas occupied by different types of land use at the administrative scale of the municipality. An exhaustive
inventory of torrent control works conducted in the southern Prealps was completed, including reafforestation
of hillslopes (mostly planting of Pinus nigra), and channel stabilization in headwaters (check-dams, fascines
and wattlings, brush gully checks). The archives of the National Forest Office recorded works done annually
in each municipality, allowing determination of the chronology and spatial distribution of regulation works
on torrents. Other written archives and historical statistics (population growth, grazing pressure, industrial
activity) provided further information.
At the reach scale, floodplain land-use changes were documented using the Napoleonic land surveys
(1820­1840) and more recent sources, such as aerial photos and recent land survey maps on the middle
Ard`eche, Ouv`eze, Eygues, Drôme and Ain rivers (Piégay, 1995). Long-profiles surveyed in the 1920­1930s
were compared with equivalent surveys from the 1980­1990s to evaluate vertical channel changes in most
of the study streams (see details in Landon, 1999). Information on long-profile changes was also available on
other basins, such as some small mountain streams of the southern French Prealps. Cross-sections were surveyed
to determine the difference in elevation between the present active channel and remnants of previously
active alluvial surfaces, notably those still visible on 1948 air photos.
Finally, flood records were analysed for those rivers where narrowing had been determined and which had
sufficiently long hydrological records to test for possible decreases in flood magnitude or frequency during
the period of active channel narrowing. The annual peak flow and the annual number of days for which the
discharge is above the one-year recurrence-interval flood was plotted for three rivers ­ the Ubaye, the Drôme
and the Rhine ­ for which available records begin during the 19th or early 20th century. The Rhine River
did not narrow during the historical period because it was channelized in 1868­1870, but it has a gauging
record back to 1809 at Basel and it is interesting to study its trend because the catchment underwent the same
land-use changes as the Rhône. For the period 1900­1995, daily flood frequency records are compared for
three distinct subperiods: 1900­1945, 1945­1970 and 1970­1995. The intermediate period is characterized
by high rates of channel narrowing in the studied basins, whereas the two other periods show a relative
stability or a minor decrease in channel width.
CONTEMPORARY CHRONOLOGY OF CHANNEL NARROWING
A general trend
In most of the studied basins, a general trend of channel narrowing can be established. A significant channel
width decrease over the last 50 years can be observed not only on large rivers, but also on small mountain
streams of the southern Prealps (Figure 3). Most of the streams underwent more rapid channel narrowing from
1950 to 1970 than from 1970 to 1990. It is notable that width changes were similar on the large piedmont
rivers (Drôme, Eygues, Roubion, Ouv`eze, Ard`eche, Ain, Loire, Allier) and on the mountainous tributaries of
the Prealps. Thus, a common pattern of channel narrowing is evident in space and time, regardless of position
in the stream network. It is notable that the Drôme and the Roubion underwent strong narrowing not only
between 1945 and 1970 but also from 1970 to 1990 (Figure 3).
The chronology of narrowing on 51 tributaries of the three Prealpine piedmont rivers also indicates an acceleration
between 1950 and 1970. The mean rate of channel narrowing between 1948 and 1991­1996, calculated
for the 51 tributaries, was 55 per cent. ANOVA performed to test the differences of this variable between the
three basins reveals that mean rates were similar (p-values of F test are >005 for each couple). Mean values
of narrowing are 50, 58 and 62 per cent for the Drôme, the Eygues and the Roubion tributaries respectively.
On some basins, it was possible to characterize the pattern of active channel change over the last two
centuries (Figure 4). A generalized trend toward channel narrowing was noted from the 19th century, with a
strong acceleration during the 1950s and 1960s in most of the studied streams. On the Ubaye River, in the
inner Alps, a 1895 photograph of its confluence with the Gimette torrent shows the valley floor completely
unforested, whereas the forest corridor was well established by the air photo of 1948 (and still occupied this
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 431
1
10
100
1000
1
10
100
1000
1
10
100
1000
1
10
100
1000
1940 1969 1957 1993 1957 1993 1948 1956 1972 19911996
DOUBS RIVER
1
10
100
1000
1946 1970 1991
DRÔME RIVER
UBAYE RIVER
1
10
100
1000
OUVEZE RIVER
LOIRE RIVER
1
10
100
1000
1
10
100
1000
1
10
100
1000
1
10
100
1000
1947 1969 1989 1945 1980 1996 1948 1956 1973 1982 1992 1948 1956 1972 1991
ARDECHE RIVER
ALLIER RIVER
1
10
100
1000
1
10
100
1000
1948 1972 1996 1956 1971 1991 1947 1973 1991
EYGUES RIVER
AIN RIVER
UPPER-ROUBION RIVER
activechannelwidth(m)activechannelwidth(m)activechannelwidth(m)
Small mountain streams
of the southern French Prealps
UPPER-DRÔME RIVER
n = 188 n = 67 n = 19 n = 125
W46 = 110 m
W70 = 83 m
W91 = 63 m
- 25 %
p = 0.0004
S
- 18 %
p < 0.0001
S
+ 6 %
p = 0.1233
NS
- 24 %
p < 0.0001
S
- 21 %
p = 0.0005
S
- 16 %
p = 0.9657
NS
- 32 %
p < 0.0001
S
- 24 %
p = 0.0002
S
- 39 %
p = 0.0015
S
- 24 %
p = 0.0081
S
+ 3 %
p = 0.3678
NS
+ 7 %
p = 0.0967
NS
- 7 %
p = 0.3050
NS
- 50 %
p = 0.0023
S
- 31 %
p < 0.0001
S
+ 22 %
p = 0.0084
S
+ 0 %
p = 0.1843
NS
+ 33 %
p < 0.0001
S
- 16 %
p = 0.1146
NSW48 = 200 m
W72 = 152 m
W96 = 143 m
W48 = 21 m
W71 = 14 m
W91 = 7 m
W47 = 83 m
W73 = 57 m
W91 = 48 m
W47 = 86 m
W69 = 70 m
W89 = 72 m
n = 187 n = 144 n = 103 n = 85
W45 = 88 m
W80 = 60 m
W96 = 64 m
W48 = 81 m
W56 = 84 m
W73 = 85 m
W82 = 73 m
W92 = 68 m
W48 = 22 m
W56 = 27 m
W72 = 18 m
W91 = 18 m
W40 = 141 m
W69 = 150 m
W96 = 125 m
n = 107 n = 56 n = 87
W57 = 223 m
W93 = 177 m
W57 = 188 m
W93 = 142 m
W48 = 11.47 m
W56 = 9.92 m
W72 = 6.01 m
W91 = 4.65 m
n = 51
- 1 %
p = 0.2304
NS - 39 %
p = 0.0031
S - 23 %
p = 0.1862
NS
p = 0.08
NS
p = 0.41
NS
p = 0.31
NS
p = 0.27
NS
Figure 3. Active channel width changes that affected the studied rivers during the last 50 years (1950­2000); boxes represent inner and
outer quartiles; vertical lines represent inner and outer tenths; open circles are extreme values; n, number of width measurements; W,
mean values of active channel width for each date; values in per cent represents the rate of channel narrowing between dates; results
of a non-parametric statistical test (Mann-Whitney U-test) are presented; S indicates significant differences of channel width between
dates; NS, not significant
area in 1994; Figure 5). Dendrochronological analysis conducted in this forest showed that the trees were
established between 1912 and 1935, with the average date of establishment being 1921. The old floodplain
units in the study reaches of the Ubaye River were vegetated earlier than those of other rivers (Figure 6).
Dendrochronology on riparian forests established on former active channels indicates that forests because
established mainly between 1950 and 1970, regardless of the size of the channel (Figure 6). Riparian forests
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
432 F. LIÉBAULT AND H. PIÉGAY
activechannelwidth(m)
10
15
20
25
30
35
40
2000
0
50
100
150
200
250
300
1800 1850 1900 1950
EYGUES (48 km)
activechannelwidth(m)
0
10
20
30
40
50
1800 1850 1900 1950 2000
SURE (5395 m)
0
5
1800 1850 1900 1950 2000
GERVANNE (1950 m)
0
10
20
30
40
50
60
70
1850 1900 1950 2000
BEOUX (323 m)
activechannelwidth(m)
0
10
20
30
40
50
60
70
80
1800 1850 1900 1950 2000
BENTRIX (730 m)
10
15
20
25
2000
0
5
1800 1850 1900 1950
BINE (9322 m)
2000
0
2
4
6
8
10
12
14
16
1800 1850 1900 1950
SOUBRION (6373 m)
Figure 4. Pattern of active channel width evolution since the middle of the 19th century on one large gravel-bed river (Eygues River)
and six small mountain streams of the southern Prealps; lengths of studied reaches are indicated in brackets
encroached at the same time on the channel of large piedmont rivers and on small mountain tributaries. These
results indicate a remarkable temporal and spatial homogeneity in vegetation encroachment, with no evidence
of lag time in responses of large and small rivers.
Some exceptions
The Giffre and Ubaye rivers, which are located in the inner Alps (an area with strong connection to hillslope
sediment sources), did not undergo narrowing between 1950 and 1970 (respectively C2 per cent and C5 per
cent). The Doubs River was also unusual in that it widened slightly between 1945 and 1970 (C6 per cent)
(Figure 3). Its trend is all the more unusual in that the Doubs River is located in the piedmont zone of the
Jura Mountains, within a basin similar to the Ain River.
Like the Roubion and the Drôme, the Ain River underwent a constant narrowing between 1947 and
1983, without inflection around 1970. Moreover, between 1980 and 1996, the Ain River widened by an
average of 6 m year 1
, its surface area increasing from 435 ha to 457 ha over a reach of 35 km (see
Figure 3). The case of the Soubrion stream (Figure 4) is also characterized by a more regular decreasing
trend at the 20th century scale, without specific evolution between 1950 and 1970. By 1950, the
Soubrion stream had already narrowed to 4 m wide. It did not experience accelerated vegetation encroachment
after 1950, because narrowing between 1820 and 1950 was sufficient to stabilize all its previously active
surfaces.
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 433
a) 1895
b) 1994
Figure 5. The Ubaye at its confluence with the Gimette torrent: (A) in 1895, the valley floor is occupied by discontinuous herbaceous
plants and gravel bars; (B) in 1994, the valley floor is forested and the channel becomes narrower and deeper
REGIONAL CAUSES OF CHANNEL NARROWING
Channel narrowing: a complex process­response of the fluvial system
Results from this study illustrate the complexity of process­responses involved in the evolution of active
channel width during the last 150 years. Channel narrowing occurred both on large rivers draining more than
1000 km2
and small mountain streams with a high sensitivity to environmental change, and at about the same
time. The long-term narrowing trend accelerated from 1950 onwards in rivers with channels wide enough for
vegetation to establish. Historical information on channel widths from 1850 to 1950 is inadequate to quantify
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
434 F. LIÉBAULT AND H. PIÉGAY
FiumeSeccu
Upper-Drôme
Ain
dateoftreeestablishment
Bine
Upper-Roubion
n = 27
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
1900
1890
1880
1870
1860
Archiane
Béoux
RifMiscon
Barnavette
Esconavette
Ubaye
small mountain streamslarge gravel-bed rivers
11517863 3415 456710 12 20
Figure 6. Distributions of the dates of tree establishment on first adjacent alluvial surfaces of active channels in different rivers based
on dendrochronological measurements; boxes represent inner and outer quartiles; vertical lines represent inner and outer tenths; open
circles are extreme values; n, number of sampled trees
variations in narrowing rates within this period. However, historical photographs of c. 1900 attest that most
of the studied rivers maintained their braided pattern through this time, consistent with results of the historical
analysis showing that post-1950 channel narrowing was a major fluvial change at the century scale.
Channel narrowing was commonly associated with pattern change from braiding to wandering, or from
wandering to straight or meandering. On large piedmont rivers, as channel width decreased, sinuosity increased
and braiding index decreased (Piégay, 1995; Landon, 1999). On small mountain streams most former active
gravel bars stabilized, and channel morphology changed from wandering riffle­pool to straight confined
patterns without well-developed open gravel bars.
Confronted with these fluvial changes, some authors have argued that the braided patterns could be considered
as relict forms in most of the European mountains, in equilibrium with past climatic and erosive
conditions, which resulted in high bedload supply and flood discharges (Bravard, 1991; Gautier, 1992; Miramont
and Guilbert, 1997). Climatic changes following the end of the Little Ice Age (Le Roy Ladurie, 1983;
Grove, 1988) and basin reafforestation induced by torrent control works and rural depopulation should have
decreased bedload supply and peak flows, in turn progressively shrinking channels (see working hypothesis
on Figure 1). This hypothesis can explain a long-term, gradual decrease in active channel width since the end
of the 19th century but not the abrupt acceleration of the phenomenon after 1950. Moreover, in the southern
Prealps, a similar intensity and pattern of narrowing occurred on the studied basins after 1950, whereas
conditions of bedload supply decrease, notably torrent control works, differed among them.
1830­1950: channel narrowing associated with changes in bedload transport and discharge
Because of the long gap between the detailed maps and land-use descriptions of the Napoleonic cadastre
(c. 1830) and the first aerial photographs (c. 1950), the analysis of narrowing is not easy for this period. Did
narrowing occur slowly and consistently between 1830 and 1950 or more abruptly during a shorter period?
Torrent control works. On the Ubaye River, Piégay and Salvador (1997) previously showed from dendrochronological
evidence that vegetation encroachment began abruptly around 1920. This major change
has been strongly associated with torrent regulation realized from 1880 to 1900 15 km upstream. If average
annual bedload movement is about 500 m, as observed on the Drôme basin where distances of transport are
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 435
Table II. Inventory of torrent control works carried on the Drôme, the Eygues and
the Roubion basins between 1860 and 1978
Drôme Eygues Roubion
Number of restoration perimetersŁ
53 23 1
Total area of restoration perimeters (ha) 27 428 6840 150
Turfing operations (t)
406 41 0
Reafforested areas (ha) 13 217 2393 137
Number of wattlings and fascines
91590 1134 36
Number of check-dams 13 544 2387 0
Brush gully checks (km)§ 599 158 02
Ł Area purchased by the French Forest Administration for the restoration of degraded lands.
 Use of grass to aid in revegetating a bare surface.
 Small dams made of twigs or flexible saplings woven between upright stakes.
§ Use of brush mulching or fascines in gullies to aid in revegetation.
determined after each flow event since 1997 on three mountain streams (Liébault et al., 2001), the bedload
retention upstream would affect the study reach of the Ubaye about 30 years after the works (consistent with
the observations).
A comparison has been made of the chronology of torrent control works on a set of 51 mountain streams in
the southern Prealps with the timing of active channel evolution downstream. The torrent control works from
1860 to 1920 were not uniformly distributed in space (Table II). Most were conducted in the Drôme basin,
with very few structures (and little reafforestation) in the Eygues and the Roubion basins. The different density
of torrent control works is a possible factor explaining the narrower widths of tributaries in the Drôme versus
the Eygues basin in 1948 (mean active channel widths in 1948 are respectively, 10 and 14 m, significantly
different at the p-level 003, Mann­Whitney U-test). Another factor is climate, with the Eygues basin being
more Mediterranean, and thus more prone to high hillslope erosion rates due to the thinner vegetation cover
and more intense rainfalls (cf. Wolman and Gerson, 1978; Descroix, 1994).
Results obtained on the Ubaye River and the southern Prealps suggest that the main period of fluvial
adjustment to torrent control works was the first half of the 20th century. As such, they cannot explain post1950
narrowing, even if the lag time for effects of mountain torrent control on piedmont rivers downstream
is considered. Moreover, they cannot be the key factor explaining narrowing because shrinkage occurred on
rivers with almost no torrent control works, such as most of the Eygues and Roubion tributaries.
Bedload supply decrease versus peak flow decrease: land-use change or climate change? The industrial
revolution in the second half of the 19th century led to massive migration from rural areas to towns and
cities. As a result, extensive ploughed lands were replaced by open shrub land and meadows in the first
decades of the 20th century (Taillefumier and Piégay, in press). Comparison of historical land surveys in the
southern Prealps indicates that area of shrub and meadow increased from 34 per cent in 1830 to 51 per cent
in 1954, while arable lands declined during the same period from 32 to 6 per cent. These changes could
have induced a progressive stabilization of hillslopes and a slow decrease of sediment supply to the stream
network, leading to a progressive stabilization of gravel bars.
It is difficult to assess potential hydrological changes in the 19th and 20th centuries because few data exist.
On the Ubaye River, gauging records back to 1904 reveal no change in annual peak flow that could explain
post-1920 channel change (Figure 7a). On the Drôme, the gauging record at Luc-en-Diois back to 1907,
augmented by historical records of large floods back to 1850, suggests that the 1850­1900 period had higher
floods (Figure 7b). On the Rhine River, large floods occurred during the last part of the 19th century but the
one-year recurrence-interval discharge series, at least since 1869, has been essentially stationary (Figure 7c).
The response of the Ubaye River to an estimated 1000-year flood in 1957 (Figure 7a) provides a test of
the role of flood magnitude in determining channel width: the channel widened during the flood but narrowed
within a few years after. The Upper Drôme River (Figure 3) illustrates a similar pattern of widening/narrowing
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
436 F. LIÉBAULT AND H. PIÉGAY
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
0
50
100
150
200
0
50
100
150
200
Annual daily peak flow (m3 s-1)
a)
0
100
200
annual occurence of 10 yr-recurrence
interval daily discharge (Landon, 1999)
Q10 not reached
Q10 reached
1850 19501900 2000
b)
maximum instantaneous
peakflows of the Drôme River
at Luc-en-Diois gauging station,
in m3
. s-1
(Landon, 1999)
1810 1830 1850 1870 1890 1910 1930 1950 1970 1990
Annual daily peak flow (m3
s-1
)c)
20
40
60
80
0
Annual number of days with discharge > 1-yr-RI discharge
Annual number of days with discharge > 1-yr-RI discharge
1000
2000
3000
4000
5000
6000
1870 1890 1910 1930 1950 1970 1990
Figure 7. Hydrological fluctuations observed during the last two centuries: (a) the Ubaye record at Barcelonnette gauging station (drainage
basin: 549 km2, period of measurement: 1904­1999); (b) the Drôme record at Luc-en-Diois gauging station (194 km2, 1907­2000);
(c) the Rhin record at Basel gauging station (36 000 km2, 1809­1999)
associated with a small sequence of moderate floods: four Q10 events occurred during 1953, resulting in a
wider channel in 1956, but a few years later the channel returned to its pre-flood width. These observations
suggest a sort of lateral periodic variation of the gravel-bed channel at time scales of one to ten years, which
can be considered as simple fluctuations above and below a mean width (the dynamic equilibrium). These
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 437
kinds of channel adjustments are independent of long-term changes such as those observed between 1830
and 1950.
All these numerous examples of channel adjustment following floods displayed recovery times of less
than ten years. However, these observations of post-flood channel recovery occurred in a general context of
decreased bedload supply, and were associated with chronologically isolated events, and thus would be less
geomorphically effective than the sequence of large 19th century floods because the context was different.
The question of channel response to hydrological changes during the 1850­1950 period can be posed
in terms of relative effectiveness of floods of different magnitude on fluvial geomorphology (Wolman and
Gerson, 1978). Given that the large 19th century floods occurred in basins that were highly sensitive due
to the degradation of the vegetation cover, their morphologic effectiveness may have been great enough to
generate real modifications of the fluvial landscape. The morphological effects of these floods may have been
enhanced by the fact that they occurred over a short period.
Floodplain land-use changes: replacement of ploughed lands by grazing. The Napoleonic land surveys of
the first half of the 19th century clearly show that agricultural land use extended up to the edge of the active
channel on most French rivers. Different floodplain areas studied on the Ouv`eze River in 1830­1837 were
mainly occupied by meadows or ploughed land (Figure 8a and b). Floodplain forest was very infrequent.
There were stands of willows used by traditional farmers and artisans, precisely located on maps, but these
b)
03570
willow
meadows
forest
shrub
ploughed
land
garden
35 70
Left bank Right bank
% of each line
First line of parcels along the Ouvze
Second line of parcels
0
orchards
?
?
?
Pont de Viols
Vaison-la-Romaine
Landuse types within the
riparian area of the Ouvre River
Willow unit
Forest
Meadows
Vineyards
Ploughed lands
800 m
N
a)
Figure 8. Riparian landscape of the Ouv`eze River in the first half of the 19th century, based on the Napoleonic land survey (1830­1837);
(a) Detailed map of riparian land-use types on a 15 km downstream reach; (b) proportion of land-use types observed on the two first
lines of land parcels adjacent to the active channel
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
438 F. LIÉBAULT AND H. PIÉGAY
did not always form a buffer between ploughed area and the active channel. In the commune of Mollanssur-Ouv`eze,
from 20 to 35 per cent of the parcels of land located on the border of the Ouv`eze River were
ploughed. A detailed diachronical mapping of floodplain land use on a small portion of the Ard`eche River
illustrates changes that affected the active channel boundary between 1833 and 1993 (Piégay, 1995). In this
area, where the human pressure was greater than along the Ouv`eze River, the Napoleonic land survey of
1833 showed that arable land occupied most of the floodplain surface and was its interface with the active
channel. The 1933 land survey shows that the floodplain was dominated by shrub formations, which appear
on aerial photographs as open grazed areas with dispersed shrubs.
Today, looking at the channels and their floodplain forests, it is difficult to imagine that the local communities
used such area for cultivation, but statistics on historical population densities provide a glimpse
of the former population pressure. In the communes of the Ard`eche basin, population density was 60­100
inhabitants per km2
in 1840­1850, compared with only 10­20 inhabitants per km2
in 1990. Similar trends
were observed in the Ouv`eze basin and along the lower Ain (Piégay, 1995). Thus, the slow narrowing trend
observed between 1830 and 1950 may also be caused by a change in floodplain resistance to areal erosion.
Following rural depopulation, lower demand for corn, and changed agricultural practices, vegetation
progressively encroached upon formerly ploughed, destabilized bar and floodplain surfaces.
In summary, for the period 1850­1950 it is interpreted that post-1900 channel narrowing occurred as a
recovery process from an episode of widespread channel destabilization induced by floods in basins that
were highly responsive to change following human disturbance. The recovery process was accelerated by
floodplain and basin-scale land-use changes, and torrent control works which reduced sediment delivery and
thus permitted vegetation establishment in channels.
1950­1970: channel narrowing associated with floodplain land-use changes and channel degradation
Three hypotheses are advanced in this study to explain post-1950 channel narrowing: (i) abandonment
of intensive floodplain land uses such as grazing and riparian forest exploitation after 1950 in most French
rivers; (ii) a phase of long-profile degradation in most of the studied rivers in the second half of the 20th
century which induced the abandonment of active channels; and (iii) a possible hydrological trend toward
decreased flood discharges after 1950. Each hypothesis is supported by evidence in studied rivers (Table III),
as discussed below.
Land-use changes. Much of the historical evidence suggests that active channel narrowing since 1950 was
due to abandonment of intensive floodplain land uses. This is illustrated by two rivers with very different
Table III. General synthesis on 1950­1970 channel narrowing occurrence and potential explaining factors
Rivers Channel narrowing
between 1950 and 1970
Flood discharge decrease
after 1950
Forest development on
floodplain after 1950
Degradation occurrence
after 1950
Ain
Ardche
Buëch
Drôme
Eygues
Ouvze
Roubion
Southern Prealp tributaries ?
Doubs
Giffre
Upper-Drôme
Ubaye
well established occurrenceKey: uncertain occurrence absence ? unknown
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 439
catchment hydrology and geomorphology, the Ard`eche and Ain. The Ard`eche drains the granitic Massif
Central and flows through limestones in its piedmont reaches (Figure 2). The runoff is rainfall dominated,
with a unit runoff of 0012 m3
s 1
km 2
. At Chauzon in the piedmont, the floodplain is now densely forested
whereas 1933 land surveys and 1947 air photos show it was open meadow. The Ain River, which drains
the higher elevation Jura mountains (mostly Jurassic limestones) (Figure 2), is dominated by snowmelt (or
mixed rainfall­snowmelt), and has a unit runoff of 0034 m3
s 1
km 2
, three times greater than the Ard`eche.
Despite its different characteristics, the Ain underwent parallel changes in floodplain land use and evolution
of active channel width. Along a 70-km study reach, forested area increased from 596 to 1219 ha, and the
area of open active channel decreased by 40 per cent from 1947 to 1991.
A number of study rivers show a decrease in the rate of vegetation encroachment and reduction of active
channel width after 1970, evidently reflecting the establishment of vegetation on most available surfaces by
1970 (Piégay, 1995). The Giffre and Ubaye Rivers, large rivers within the Alps, provide a variation on this
theme. Their riparian forests were fully established by the 1940s, and cross-sectional geometry had already
adjusted to forested boundary conditions. Thus, these rivers did not display significant narrowing after 1950.
The Doubs River provides an interesting contrast to the other rivers studied. The channel width of the Doubs
was essentially stable from 1940 to 1996 (Figure 3). The explanation appears to be continuation of intense
cattle grazing on the floodplain, which inhibited establishment of vegetation in the channel. Unlike other
regions, where after the Second World War farmers generally shifted their agricultural production techniques
and no longer used the riparian zone for grazing or cultivation, the farmers along the Doubs specialized in
cattle production and continued to use the riparian zone for grazing.
Abandonment of agriculture in riparian zones is one of the major changes observed in rural areas in
France after 1945 because of demographic shifts and increased agricultural specialization (to row crops and
vineyards), instead of traditional polyculture-rearing over the entire landscape including the riparian zone. In
the context of this change, the riparian forest could rapidly expand onto the floodplain and unvegetated bars.
Moreover, the forested floodplains increased hydraulic roughness conditions, potentially reducing frequency
of bed mobilization and thereby facilitating vegetation establishment on bars. On most rivers, this adjustment
process stopped in the 1970s once a new equilibrium was attained between vegetation establishment and scour
of seedlings, thereby maintaining an active channel. As trees were established on the floodplain, seedling stock
increased, allowing for more rapid and dense colonization of the active channel.
Channel narrowing and channel degradation. Results from this study show that post-1950 channel narrowing
was accompanied by degradation on several rivers (Table III), raising the question of the chronological
relation of these two phenomena. In the large gravel-bed rivers studied, accelerated channel degradation is
mainly explained by gravel mining, which was widespread in the 1970s, especially in the lower reaches. The
mining-induced incision occurred after the vegetation encroachment in active channels, and cannot be viewed
as a cause of channel narrowing. In the Ubaye and Giffre Rivers, which had already narrowed by 1950,
between 1 and 3 m of channel degradation occurred in the 1970s, without any channel narrowing (Piégay and
Peiry, 1997; Piégay and Salvador, 1997). These examples suggest that the disconnection of alluvial surfaces
by degradation is not necessarily an important factor enhancing the development of vegetation on valley floors
when the active channel has wide unvegetated gravel bars, because degradation due to mining affects not a
small part of the active channel, but the whole width, comprising annually relocated low-flow channels and
gravel bars.
In light of the link between in-channel vegetation encroachment and floodplain afforestation observed
elsewhere (Millar, 2000), and taking into account that floodplain vegetation establishment is synchronous
with in-channel vegetation encroachment on all the large gravel-bed rivers studied, it is concluded that in
these systems, vegetation establishment is mainly the result of land-use change on the floodplain (including
riparian areas), and that vegetation encroachment has accelerated channel degradation. With the reduction
in bedload supply due to afforestation, the channels would probably have narrowed even if the floodplain
land use had not changed, but at a slower rate, corresponding to the trend observed from 1850
to 1950.
It is also interesting to consider why narrowing continued after 1970 on the Roubion and the Drôme, but not
on the Eygues or the Ouv`eze rivers. A climatic explanation is proposed: the first two rivers are located in the
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
440 F. LIÉBAULT AND H. PIÉGAY
northern part of the southern Prealps, with a less Mediterranean climate (and thus lower hillslope erosion rates)
than the others. As a consequence, the relative decrease in bedload supply was greater, contributing more to
vegetation establishment downstream. On the Eygues and Ouv`eze rivers, the vegetation encroachment slowed
down sooner, reaching an equilibrium with bedload transport, and maintaining a wide active channel. The
Upper Drôme River illustrates this mechanism. This upstream reach has not narrowed, due to the persistence
of high bedload inputs, reflected in a long-term aggradation of its long-profile (Figure 3). Thus, where bedload
supply remain high, annual mobilization of the active channel bed can scour seedlings, and thereby prevent
vegetation encroachment and channel narrowing.
Evidence of channel degradation is also observed on the small Prealpine tributaries, where confined narrow
channels are incised in low terraces corresponding to the previous active channels abandoned between
1948 and 1971 (Liébault et al., 1999). Detailed topographic surveys of several tributaries showed the mean
difference in elevation between active channel and first adjacent alluvial surfaces to be 131 š 060 m (based
on 140 cross-sections). Dendrochronological dating of tree establishment on alluvial surfaces along five
streams showed two main young alluvial surfaces on the modern valley floor (19th­20th century active
floodplain) (Figure 9a). The lower surface corresponds to what were unvegetated gravel bars on the 1948
aerial photographs, the higher surface to the 19th century active channel as showed on detailed historical
maps (Liébault et al., 2001). Vegetation encroachment of these surfaces was not synchronous, with
the higher surface mostly colonized between 1930 and 1950 and the lower one between 1950 and 1970
(Figure 9b).
The lag time between floodplain and active channel encroachment observed on mountain streams suggests
that in-channel vegetation establishment cannot be considered as a simple adjustment to floodplain afforestation.
Field evidence provides arguments to support this position: (i) the absence of surficial fine sediment
deposits on low terraces (level L1 on Figure 9a) suggests that these levels were not constructed by vertical
accretion following vegetation establishment in active channels, but were primarily generated by channel
incision; (ii) the presence of pioneer species adapted to dry conditions (Pinus sylvestris, Bruxus sempervirens,
Juniperus communis, Salix eleagnos) implies that the vegetation established on surfaces which became dry
quickly following the disconnection process.
Based on these observations, channel narrowing on small mountain streams is considered to be a consequence
of channel degradation related to the decrease in sediment supply following basin afforestation. The
progressive stabilization of sediment sources induced a downstream-progressing degradation which disconnected
margins of active channels. This is attested by a comparison of the timing of forest establishment
on upstream and downstream sites in two tributaries of the Drôme basin (Figure 9c). These observations
demonstrate that forest developed earlier on upstream reaches, which are closer to the stabilized sediment
sources. The response of the vegetation to these channel changes may have been delayed by human controls
(grazing and wood cutting) on vegetation development c. 1950 onwards, a fact that probably explains the
timing homogeneity of in-channel vegetation expansion after 1950 between large and small rivers of the
southern Prealps.
Schumm and Lichty (1963) demonstrated that active channel narrowing in the Cimarron River was associated
with floodplain construction. This mechanism is also observed on large gravel-bed rivers of southeastern
France where vegetation establishment on gravel bars is associated with fine sediment deposition. Observations
made on mountain streams demonstrated that channel narrowing is not necessarily a result of floodplain
construction, but rather a consequence of channel incision which leads to the formation of dry surfaces slowly
encroached by pioneer species adapted to deep water tables.
Channel narrowing and hydrological change. Hydrological daily series for the last 100 years were analysed
for basins with sufficiently long gauging records to compare the 1950­1970 hydrological period with others
(Figure 10). Four basins were selected: the Ard`eche, Drôme, Bu¨ech and Doubs rivers.
Results for the Ard`eche basin showed that flood discharge occurrence was higher during the period characterized
by active channel narrowing. On the Bu¨ech River, accelerated narrowing occurred during a period
of frequent high flow events. On the Doubs River, the lack of active channel narrowing after 1950 cannot be
related to a specific hydrological trend.
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 441
0 10 20 30 40 50 60 70
6
5
4
3
2
1
0
relative distance (m)
relativeelevation(m)
BEOUX Stream
AC L1L2
0 10 20 30 40
relative distance (m)
6
5
4
3
2
1
0
relativeelevation(m)
ARCHIANE Stream
ACL1L2 L1 L2
GATS Stream
0 10 20 30 40 50 60 70
relative distance (m)
6
5
4
3
2
1
0 relativeelevation(m)
AC L2L2
RIF MISCON Stream
6
5
4
3
2
1
0
relativeelevation(m)
0 10 20 30 40 50 60 70
relative distance (m)
ACL1 L2AFP
AC : active channel
AFP : active floodplain
L1 : lower level (active gravel bars on 1948 air photos)
L2 : higher level (active channel in late 19th century)
a) b)
1900
1990
1980
1970
1960
1950
1940
1930
1920
1910
L1 L2
n = 42 n = 59
datesoftreeestablishment
1920
1930
1940
1950
1960
1970
1980
1990
Barnavette streamc)
datesoftreeestablishment
n = 22 11 2014
160 1750 5250 5520
distance from the mouth (m)
1920
1930
1940
1950
1960
1970
1980
1990
Esconavette stream
datesoftreeestablishment
n = 20 10 15
distance from the mouth (m)
1000 2070 4030
Figure 9. Cross-sectional geometry and forest establishment on several southern Prealps mountain streams. (a) Valley-floor cross-sections
showing different alluvial levels; (b) distribution of the dates of tree establishment on the lower and higher levels (L1 and L2), based
on dendrochronological dating sampled on the cross-sections presented in (a); (c) distribution of the dates of tree establishment on
former active channels (level L1) in upstream and downstream reaches of two mountain streams of the Drôme basin (Esconavette and
Barnavette streams), based on dendrochronological dating; boxes represent inner and outer quartiles; vertical lines represent inner and
outer tenths; open circles are extreme values
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
442 F. LIÉBAULT AND H. PIÉGAY
Ardche River
Doubs River
Recurrence interval (year)
Buëch River
Drôme River
10
100
1000
10 000
Dailypeakflow(m3s-1)
1 10 100
1900 ­ 1945
1945 ­ 1970
1970 ­ 1995
Figure 10. Comparison of daily flood-frequency curves between the period of accelerated channel narrowing (1945­1970) and previous
and subsequent periods (1900­1945 and 1970­1995) on four large gravel-bed rivers: the Ard`eche River at Sauze gauging station
(2240 km2, 1955­1997), the Bu¨ech River at Serres (771 km2, 1906­1990), the Drôme River at Luc-en-Diois (194 km2, 1906­2000);
the Doubs River at Besan¸con (4400 km2, 1952­2000)
The Drôme River showed a different pattern, as the 1945­1970 period had smaller flood peaks than
previous and subsequent periods. As recorded at Luc-en-Diois gauge since 1907, the 1945­1970 period was
characterized by an absence of daily flow events exceeding the ten-year recurrence interval. This period
coincides with the most extensive encroachment of vegetation in active channels in most of the studied
tributaries, although vegetation encroachment continued afterwards despite higher peak flows.
Taken together, the hydrologic analyses show that channel narrowing was not restricted to periods of lower
peak flows, so hydrological variations cannot provide a general explanation for channel narrowing, though
perhaps they were a contributing factor on the Drôme. Moreover, the high rates of channel narrowing, to
typically around 50 per cent of the former active channel width, seem too large to be explained by the
relatively subtle hydrologic differences in the basins displaying a change. Hence, the magnitude of flood
discharge variability is considered to explain only short-term fluctuations of the contact between active
channel and riparian forests, as observed along the Ubaye River (Piégay and Salvador, 1997). The trend of
channel narrowing after 1950 is more persistent and extensive. Some high floods occurred in several studied
streams between 1940 and 1970, without modifying the trend toward channel width decrease. On the Roubion
River, important floods occurred in 1960 and 1993, without inducing significant active channel width increase
(Liébault and Piégay, 2001).
CONCLUSION
Channel narrowing has been observed in many different areas in the southeastern part of France over the past
two centuries, including both piedmont and intramountain large gravel-bed rivers, as well as small mountain
streams. Numerous channels have narrowed since the mid-19th century, with a marked acceleration from
1950 to 1970. Although some exceptions can be identified, on many of the studied rivers the narrowing was
remarkably homogeneous in space and time. The synchronism of narrowing longitudinally along rivers, and
among rivers regionally, is an important finding of this study, as is the chronology of forest establishment in
valley floors, mostly during the 1950s and 1960s.
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
CAUSES OF CHANNEL NARROWING IN SE FRANCE 443
Two different periods are distinguished to facilitate causal interpretation. The 1850­1950 period (in which
a gradual narrowing was observed) is difficult to interpret in terms of controlling factors. First, data are
still missing to account for a possible timing complexity of channel narrowing. Although one example of
rapid forest establishment around 1920 linked to torrent control works is documented (the Ubaye River),
it is difficult to generalize and to determine strong causal relations between channel narrowing and climate
or human controls. Long streamflow gauging records are rare, and among those that exist, some indicate
a decrease in annual peak flows, but not in dominant discharges likely to control the channel width over
the long term. Thus, the geomorphic effectiveness of a possible hydrological change is questionable, since
many human disturbances (torrent control works, land-use changes) occurred at the same time and could have
overwhelmed the effects of any changes.
Channel narrowing during the 1950­1970 period was more pronounced and was clearly related to human
controls, such as floodplain land-use changes, mostly in alluvial sections, and hillslope afforestation associated
with bedload transport decrease in mountain streams. It is suggested that vegetation encroachment in the active
channel may have contributed to degradation during a period of overall bedload supply decrease because of
hillslope stabilization. On some rivers, evidence suggests that the recent channel narrowing was not related to
a period of smaller floods, a factor commonly cited as an important cause of channel width decrease. These
lines of evidence are considered to indicate that the recent channel narrowing is primarily a human-induced
phenomenon.
Human actions induced profound changes in sediment supply even at the scale of years to decades for
reaches close to the sediment sources. Thus, even if the end of the Little Ice Age has had effects on river
geomorphology, its effect is probably so slow as to be hidden by short-term human effects. This argues
for caution in interpreting causes of channel changes, notably when linking long-term climatic effects and
short-term human impacts.
ACKNOWLEDGEMENT
The authors thank G. M. Kondolf and C. Rogers for the English revision of the manuscript and two anonymous
referees for insightful comments. They also thank J. Sanchez Perez from CNRS Toulouse and P. M. Bechon
from DIREN Rhônes-Alpes for providing hydrological data.
REFERENCES
Andrews ED. 1984. Bed-material entrainment and hydraulic geometry of gravel-bed rivers in Colorado. Bulletin of the Geological
Society of America 95: 371­378.
Bergeron N, Roy AG. 1985. Le rôle de la végétation sur un petit cours ďeau. Géographie Physique et Quaternaire 39(1): 323­326.
Bravard JP. 1991. La dynamique fluviale `a ľépreuve des changements environnementaux: quels enseignements applicables `a
ľaménagement des rivi`eres? La Houille Blanche 7­8: 515­521.
Bravard JP, Peiry JL. 1993. La disparition du tressage fluvial dans les Alpes fran¸caises sous ľeffet de ľaménagement des cours ďeau
(19­20`eme si`ecles). Zeitschrift f¨ur Geomorphologie, Supplement Band 88: 67­79.
Clifton C. 1989. Effects of vegetation and land-use on channel morphology. In Practical Approaches to Riparian Resource Management,
an Educational Workshop, Gresswell RE (ed.). US Bureau of Land Management: Billings, Montana; 121­129.
Collier M, Webb RH, Schmidt JC. 1996. Dams and Rivers. Primer on the Downstream Effects of Dams. US Geological Survey circular
1126: Tucson, Arizona.
Descroix L. 1994. Ľérosion actuelle dans la partie occidentale des Alpes du Sud. PhD Thesis Université Lumi`ere Lyon 2.
Eschner TR, Hadley RF, Crowley KD. 1983. Hydrologic and morphologic changes in channels of the Platte River basin in Colorado,
Wyoming and Nebraska: a historical perspective. United States Geological Survey Professional Paper, 1277A.
Friedman JM, Osterkamp WR, Lewis WM Jr. 1996. The role of vegetation and bed-level fluctuations in the process of channel
narrowing. Geomorphology 14: 341­351.
Gaillot S, Piégay H. 1999. Impact of gravel mining on stream channel and coastal sediment supply: example of the Calvi bay in Corsica
(France). Journal of Coastal Research 15(3): 774­788.
Garcia-Ruiz JM, White SM, Lasanta T, Marti C, Gonzalez C, Errea MP, Valero B. 1997. Assessing the effects of land-use changes on
sediment yield and channel dynamics in the central Spanish Pyrenees. In Human Impact on Erosion and Sedimentation, Walling DE,
Probst JL (eds). IAHS publication 245: Wallingford; 151­158.
Gautier E. 1992. Recherches sur la morphologie et la dynamique fluviale dans le bassin du B¨uech (Alpes du Sud). PhD Thesis, Université
Paris X-Nanterre.
Gautier E, Piégay H, Bertaina P. 2000. A methodological approach of fluvial dynamics oriented towards hydrosystemic management:
case study of the Loire and Allier rivers. Geodinamica Acta 1: 29­43.
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)
444 F. LIÉBAULT AND H. PIÉGAY
Graf WL. 1978. Fluvial adjustments to the spread of Tamarisk in the Columbia Plateau region. Geological Society of America Bulletin
89: 1491­1501.
Gregory KJ, Gurnell AM. 1988. Vegetation and river channel form and process. In Biogeomorphology, Viles H (ed.). Basil Blackwell:
Oxford; 11­42.
Grove JM. 1988. The Little Ice Age. Methuen: London.
Hadley RF. 1961. Influence of riparian vegetation on channel shape, Northeastern Arizona. United States Geological Survey Professional
Paper 424-C: 30­31.
Hey RD, Thorne CR. 1986. Stable channels with mobile gravel-bed rivers. Journal of Hydraulic Engineering 8: 671­689.
Huang HQ, Nanson GC. 1997. Vegetation and channel variation; a case study of four small streams in southeastern Australia.
Geomorphology 18: 237­249.
Kondolf GM, Curry RR. 1986. Channel erosion along the Carmel River, Monterey County, California. Earth Surface Processes and
Landforms 11: 307­319.
Landon N. 1999. Ľévolution contemporaine du profil en long des affluents du Rhône moyen, constat régional et analyse ďun hydrosyst`eme
complexe, la Drôme, PhD Thesis, Université Paris IV-Sorbonne.
Leopold LB, Maddock T. 1953. The hydraulic geometry of stream channels and some physiographic implications. United States
Geological Survey Professional Paper, 252.
Le Roy Ladurie E. 1983. Histoire du Climat Depuis ľAn Mil. Flammarion: Paris.
Liébault F, Piégay H. 2001. Assessment of channel changes due to long-term bedload supply decrease, Roubion River, France.
Geomorphology 36: 3­4, 167­186.
Liébault F, Clément P, Piégay H, Landon N. 1999. Assessment of bedload delivery from tributaries: the Drôme River case, France.
Arctic Antarctic and Alpine Research 31(1): 108­117.
Liébault F, Clément P, Piégay H. 2001. Analyse géographique de la recharge sédimentaire des bassins versants de la Drôme, de ľEygues
et du Roubion, UMR5600-CNRS report for Office National des For^ets: Lyon.
Millar RG. 2000. Influence of bank vegetation on alluvial channel patterns. Water Resources Research 36(4): 1109­1118.
Miramont C, Guilbert X. 1997. Variations historiques de la fréquence des crues et évolution de la morphogen`ese fluviale en moyenne
Durance (France du Sud-Est). Géomorphologie: Relief, processus, environnement 4: 325­338.
Nevins THF. 1969. River training ­ the single thread channel. New Zealand Engineering 24: 367­373.
Osterkamp WR, Costa JE. 1987. Changes accompanying an extraordinary flood on a sandbed stream. In Catastrophic Flooding, Mayer L,
Nash D (eds). Allen and Unwin: Boston; 201­224.
Peiry J-L. 1997. Recherches en géomorphologie fluviale dans les hydrosyst`emes fluviaux des Alpes du Nord. Thesis, Université Joseph
Fourrier, Grenoble 1.
Peiry JL, Vivian H. 1994. Dynamique des crues et réduction de la capacité du chenal consécutive `a la construction ďun barrage
hydroélectrique: ľexemple du Drac inférieur `a ľamont de Grenoble. In Société Hydrotechnique de France, 23`emes Journées de
ľHydraulique: crues et inondations, N^imes (France) 1: 321­329.
Piégay H. 1995. Dynamique et gestion de la ripisylve de cinq cours ďeau `a charge grossi`ere du bassin du Rhône (ľAin, ľArd`eche, le
Giffre, ľOuv`eze et ľUbaye), XIX`eme-XX`eme si`ecles. Université Paris IV-Sorbonne.
Piégay H, Peiry J-L. 1997. Long profile evolution of a mountain stream in relation to gravel load management: example of the middle
Giffre River (French Alps). Environmental Management 21(6): 909­919.
Piégay H, Salvador P-G. 1997. Contemporary floodplain forest evolution along the middle Ubaye River, Southern Alps, France. Global
Ecology and Biogeography Letters 6: 397­406.
Piégay H, Bravard JP, Dupont P. 1994. Les ripisylves et les crues dans la France du sud-est: de ľhistoire `a la gestion contemporaine.
In Société Hydrotechnique de France, 23`emes Journées de ľHydraulique : crues et inondations, N^imes (France), 1: 277­289.
Piégay H, Pautou G, Bravard JP. (in press). Les ripisylves du Rhône et de ses affluents: une histoire contemporaine. In Les ripisylves
dans les hydrosyst`emes fluviaux, Piégay H, Pautou G, Ruffinoni C (eds). Institut pour le Développement Forestier: Paris.
Saulnier D. 1999. Analyse comparative de ľinstabilité contemporaine de 5 rivi`eres en tresses du bassin du Rhône, éléments pour la
détermination ďun espace de liberté. Masters Thesis, University Lyon 2.
Schumm SA. 1977. The Fluvial System. J. Wiley and Sons: New York.
Schumm SA, Lichty RW. 1963. Channel widening and flood-plain construction along Cimarron River in south-western Kansas. United
States Geological Survey Professional Paper, 352D: 71­88.
Surian N. 1999. Channel changes due to river regulation: the case of the Piave River, Italy. Earth Surface Processes and Landforms
24: 1135­1151.
Taillefumier F, Piégay H (in press). Environmental factors controlling contemporary land-use change in Prealpine Mediterranean
mountains: a multivariate Gis-based approach. Catena.
Trimble SW. 1997. Stream channel erosion and change resulting from riparian forests. Geology 25(5): 467­469.
Wolman MG, Gerson R. 1978. Relative scales of time and effectiveness of climate in watershed geomorphology. Earth Surface Processes
and Landforms 3: 189­208.
Zimmerman RC, Goodlet JC, Comer GH. 1967. Influence of vegetation on channel form of small streams. In Symposium on River
Morphology. IAHS. Publication 75: Wallingford; 255­275.
Copyright  2002 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 27, 425­444 (2002)