Sei sulla pagina 1di 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283056822

"Fluvial Corridor" : a new ArcGis Toolbox Package for multiscale riverscape


exploration

Article  in  Geomorphology · August 2015

CITATIONS READS

16 1,848

5 authors, including:

Roux clément Mélanie Bertrand


Ecole normale supérieure de Lyon National Research Institute of Science and Technology for Environment and Agric…
3 PUBLICATIONS   71 CITATIONS    60 PUBLICATIONS   148 CITATIONS   

SEE PROFILE SEE PROFILE

Lise Vaudor Hervé Piégay


Ecole normale supérieure de Lyon University of Lyon
32 PUBLICATIONS   350 CITATIONS    524 PUBLICATIONS   8,602 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Environnemental Project of Kembs hydropower scheme View project

Flooding risks associated with wood flux in urban environments View project

All content following this page was uploaded by Mélanie Bertrand on 26 October 2015.

The user has requested enhancement of the downloaded file.


Geomorphology 242 (2015) 29–37

Contents lists available at ScienceDirect

Geomorphology
journal homepage: www.elsevier.com/locate/geomorph

“FluvialCorridor”: A new ArcGIS toolbox package for multiscale


riverscape exploration
Clément Roux a,⁎, Adrien Alber b, Mélanie Bertrand a,c, Lise Vaudor a, Hervé Piégay a
a
University of Lyon, CNRS-UMR 5600 Environnement - Ville - Société, ENS de Lyon, 15 Parvis René Descartes, BP 7000, 69342 Lyon Cedex 07, France
b
Direction Régionale de l'Environnement, de l'Aménagement et du Logement (Région Centre), Service Eau et Biodiversité, 5 avenue Buffon, 45064 Orléans Cedex, France
c
IRSTEA-Érosion torrentielle, neige et avalanches (ETGR)-Centre de Grenoble, National Research Institute for Environmental and Agricultural Sciences and Technologies, Domaine universitaire,
2 rue de la Papeterie-BP 76, 38402, Saint-Martin-d'Hères Cedex, France

a r t i c l e i n f o a b s t r a c t

Article history: Both for scientists and river basin managers, development of automated geographic information system (GIS)
Received 17 October 2013 tools is essential today to characterize riverscapes and explore biogeomorphologic processes over large channel
Received in revised form 26 March 2014 networks. Since the 1990s, GIS toolboxes and add-in programs have been used to characterize catchments. How-
Accepted 1 April 2014
ever, there is currently no equivalent to a planimetric and longitudinal characterization of fluvial corridor net-
Available online 24 April 2014
works at multiple scales. This paper describes FluvialCorridor, a new GIS toolbox. This package allows the user:
Keywords:
(i) to extract a large set of riverscape features such as the main components of fluvial corridors from DEM and
Fluvial geomorphology vector layers (e.g. stream network or valley bottom), and (ii) to aggregate spatial features into homogeneous seg-
Riverscape ments and metrics characterizing each of them. The methodological frameworks involved have been previously
Channel reach described by Alber and Piégay (2011), Leviandier et al. (2012) and Bertrand et al. (2013) and this contribution
GIS toolbox focuses on the GIS tools allowing the user to automatically operate them. A case study on the Drôme River
Automated multiscale procedure (France) is provided to illustrate the potential of the package both for geomorphologic understanding and target
River discontinuum management actions. FluvialCorridor has been developed for ArcGIS with the related native Python library named
ArcPy and tested on ArcGIS 10.0 and 10.1. Obviously, each component of the package can be used separately;
however, it also provides a complete workflow for fluvial corridor characterization, even as the toolbox is contin-
ually under development and revision. Case study database, FluvialCorridor package and guidelines are available
online at http://umrevs-isig.fr.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction network, (iii) vertically, focusing on exchanges between the surface wa-
ters and groundwaters and (iv) in time, to highlight influences of an-
Fluvial corridor and riverscape concepts were introduced in the thropogenic pressures or changes in flood events at 1- to 100-year
1960s by Leopold and Marchand (1968) and have been widely used scale (Swanson et al., 1982; Gregory et al., 1991). Riverscape units are
since the early 2000s (Fausch et al., 2002; Wiens, 2002; Eros et al., usually viewed as nested within each other from the network to the
2010; Bertrand et al., 2013). Carbonneau et al. (2012) define the segment reach, from the meso to micro-habitats (Allen and Starr,
riverscape as an ecological representation of rivers. This fluvial object is 1982; Frissell et al., 1986; O’Neill et al., 1986). These units therefore in-
a combination of broad scale units with energy, matter and biota trans- clude valley segments, geomorphic reaches, in-channel and floodplain
fers. In reference to the early notion of a fluvial continuum (Schumm, features (e.g. bars, riffles, pools, vegetation and islands, floodplain chan-
1977; Vannote et al., 1980), riverscapes are longitudinally organized nels) (Kondolf et al., 2003).
and can be characterized by a set of nested units emerging at different Thus riverscapes are becoming more and more studied by geo-
spatial and temporal scales and corresponding to both well-identified morphologists, freshwater ecologists and managers to characterize
areal features (e.g. river styles, erosion features, active channel) and biophysical features in terms of quality and health, and to provide
their boundaries, and more synthetic features (e.g., elementary sampling understanding of the network organization, its sensitivity to
features, linear features such as centerline). Riverscape units can be de- human pressures and its ability to adjust (Fausch et al., 2002;
scribed (i) longitudinally, considering hydraulic, geomorphologic and Wiens, 2002; Thorp et al., 2006; Le Pichon et al., 2009; Carbonneau
ecological patterns, (ii) transversally, to focus on biomorphological inter- et al., 2012; Bertrand et al., 2013). With the increasing availability
actions between channels and floodplains or hillslopes through the of network-scale data, such analyses are becoming more and more
common and automatic procedures to extract information are thus
⁎ Corresponding author. Tel.: +33 6 01 31 07 26. needed. By sharing these types of tools, our intention is to better en-
E-mail address: clement.roux@ens-lyon.fr (C. Roux). able researchers and river managers to characterize networks,

http://dx.doi.org/10.1016/j.geomorph.2014.04.018
0169-555X/© 2014 Elsevier B.V. All rights reserved.
30 C. Roux et al. / Geomorphology 242 (2015) 29–37

which is a major challenge for planning and targeting restoration 2. Required extensions and toolboxes
actions.
Previous significant technical tools for characterizing rivers and fluvial FluvialCorridor toolbox has been developed for ArcGIS 10.0 and
features have already been developed (Table 1). In the text below, the ArcGIS 10.1, running with the related ArcPy and Python 2.6 or 2.7 librar-
numerical superscripts refer to the rows in Table 1. Firstly, ArcHydroTools1 ies. Some ArcGIS extensions such as SpatialAnalyst, are required for
is available primarily as a method for basin and hydrographic character- ensuring an optimal use and allowing all functionalities. Moreover,
ization. It provides a large set of tools used for two key purposes: ensur- FluvialCorridor requires additional components (e.g. NumPy and SciPy
ing either raster, vector or attributes treatments (e.g. stream burning, libraries, ArcHydroTools). Further details about compatibility of the tool-
network generation, attribute transfers or accumulation) and attribute box and installation of additional components are given within the set
assessments (e.g. unique ID assignment, ordination, arc-length or water- of guidelines attached to the toolbox.
shed area measurements). Additional tools have also been created in
order to investigate relationships between physical, chemical and biolog- 3. General framework
ical parameters at a basin scale. Thus, STARS2, SSN2 or FLoWS3 allows for
exploring and predicting catchment-scale information through spatial FluvialCorridor has been developed based on the general workflow
autocorrelation. Other researchers have attempted to provide elements initiated by Alber and Piégay (2011). They developed a methodological
for a longitudinal characterization of stream networks. TauDEM4 provides framework for delineating and characterizing fluvial features based on
tools to extract stream networks from DEMs attributing each link with a raw data available at a regional (over 1000 km2) scale (e.g. vector hy-
hydrologic and morphologic metric like upstream contributing area, drographic network, DEM, archives aerial photos). The combination of
slope, or cumulative distance. Assessment of some planimetric metrics information extracted from different sources allows the user to extract
(e.g. steepness index and concavity) has also been implemented within geomorphic characteristics of fluvial features at different spatial levels.
the Stream Profiler5ArcGIS toolset. Multiple other examples of free, Fig. 1 illustrates an example of the general framework involved during
open-source software which carry out the morphometric characteriza- a complete use of the FluvialCorridor package.
tion of catchments exist, such as GRASS6 (Neteler and Mitasova, 2002), As the methodological framework, the FluvialCorridor toolbox is
SAGA GIS7 (Böhner et al., 2006), Multi Watershed Delineation8 tool subdivided in four main processes (Fig. 2). The first step consists in
(Chinnayakanahalli et al., 2006) or Geospatial Modeling Environment9. extracting unitary geographical objects (UGOs) that delineate the fea-
As new accurate remote sensing techniques and data have become tures of interest (e.g. the active channel, the valley bottom, etc.). Then,
more widely available (e.g. LiDAR, drone imagery), several tools have the spatial disaggregation of unitary geographical objects results in disag-
also been developed in response. River Bathymetric Toolkit10 (RBT) is gregated geographical objects (DGOs).Thirdly, the phase of characteriza-
one example among others. Especially designed for reach-scale studies, tion is processed. A set of metrics is calculated at the DGO database
it enables the user to access fluvial units (e.g. centerlines, banks, bankfull scale, according the focus of interest. Finally, the spatial aggregation of
water channel) and hydromorphic metrics, mainly related to the river the DGOs into aggregated geographical objects (AGOs) allows delineation
cross-sections. High resolution DEMs are thus essential to use RBT. Devel- of physically meaningful fluvial features. Fig. 2B shows those units from
oped by Thomas Dilts, Riparian Topography11 is another ArcGIS toolbox a theoretical point a view and Table 2 details the different tools of the
which is useful for riverscape characterization. Also based on high resolu- package.
tion DEMs processing, Riparian Topography provides a basis for exploring
and modeling plant species distribution along reaches by extracting 3.1. Extracted fluvial units and spatial disaggregation
“height above river” and “inundation area” rasters (Dilts et al., 2010).
Despite, or rather because of the proliferation of toolboxes to charac- As a first step, the toolset named “Spatial components” enables the
terize channel reaches with more specificity, there remains a need for user to extract UGOs such as the hydrographic network, the valley bot-
an automated process that allows the user to characterize riverscape tom or centerline of polygon features, from a DEM or vector layers.
units along river corridor networks in multiscale and planimetric Once linear and polygon features (UGOs) have been extracted, they
ways. According to the recent geomatics and statistical improvements can be disaggregated into elementary segments (DGOs) of a given user-
(Alber and Piégay, 2011; Leviandier et al., 2012; Bertrand et al., 2013), defined and constant length using the “Disaggregation processes”
development of such an ArcGIS toolbox, which accelerates such infor- toolset. This spatial disaggregation aims to provide fluvial units at a
mation production, is now possible. This toolkit can be considered as higher resolution for the characterization process. The spatial resolution
the basis of knowledge production at a large scale, which is a new scien- (i.e. the disaggregation step) must be set precisely to ensure that spatial
tific frontier. This article describes a new toolbox named FluvialCorridor, trends are correctly detected (i.e. by analogy with the Nyquist-Shannon
designed to extract and characterize fluvial features at a range of scales sampling theorem in information theory, the disaggregation step must
along a stream network. be finer than the scale of investigation (Nyquist, 1928)).

Table 1
List of existing tools for characterizing rivers or fluvial features and dataset used in the case study over the Drôme River (France).

Material's name Developer Web-link

ArcHydroTools ESRI www.esri.com


STARS United State National Oceanic and Atmospheric Administration and Australian http://www.fs.fed.us/rmrs
SSN Commonwealth Scientific and Industrial Research Organization
FLoWS United State Space-Time Aquatic Resources Modeling and Analysis Program http://www.nrel.colostate.edu
TauDEM David Tarboton-Utah State University http://hydrology.usu.edu/taudem
Stream Profiler Noah Synder and Kelin Whipple-US National Science Foundation and NASA http://www.geomorphtools.org
GRASS Geographic Resources Analysis Support System GRASS http://grass.osgeo.org
SAGA GIS System for Automated Geoscientific Analyses SAGA GIS http://www.saga-gis.org
Multi Watershed Delineation Kiran Chinnayakanahalli-Utah State University http://hydrology.usu.edu/mwdtool
Geospatial Modeling Environment Hawthorne L. Bayer-Spatial Ecology www.spatialecology.com/gme
River Bathymetric Toolkit ESSA Technologies and US Forest Service http://essa.com/tools/rbt
Riparian Topography Thomas Dilt http://arcscripts.esri.com
BD Carthage French National Geographical Institute IGN www.ign.fr
BD Topo
C. Roux et al. / Geomorphology 242 (2015) 29–37 31

Both for linear and polygon UGOs, the disaggregation processes are on connections between the geographical units created during a pro-
based on the segmentation of a polyline feature. A python script cess. According to the objectives of the case studies and the available
named SLEM (for Split Lines Each Meter) has been developed to split input data, such a framework can be either simplified or expanded
polyline features with a user-defined and constant length. Each line is and conducted from a local to a regional scale.
processed separately, but it must be correctly oriented to ensure a con-
sistent ordering of resulting segments within the network. Rotation of 4.1. Required input data
wrongly-oriented streams is carried out by a specific tool named
“Sequencing”. Finally, disaggregation of linear UGOs is directly conduct- The full use of the FluvialCorridor toolbox presented in Fig. 1 requires
ed by the SLEM tool, whereas polygon UGOs disaggregation is based on three raw data: a DEM and two vector layers, the hydrographic network
the segmentation of the centerline and a Thiessen polygonization. Each of the study area and its active channel. External data can be joined to
DGO is spatially referenced into the network by: (i) “Order_ID” and enrich the process with new information. Here, aerial photographs are
“Rank_UGO” fields of the UGO they belong to, and (ii) “Distance” field or- so used to illustrate potential gains provided by such an external data.
dering them between each confluence from upstream to downstream. Today, DEMs, aerial or satellite photographs and hydrographic net-
The “Order_ID” field can be viewed as a stream order. Unlike the Strahler works are widely produced over the world and active channel layers
or Shreve orders, “Order_ID” is defined increasingly from downstream can be obtained from manual digitalization or from radiometric analysis
to upstream. of aerial or satellite photographs (Marcus et al., 2003; Wiederkehr,
The “Polyline disaggregation” tool ensures another disaggregation 2012; Bertrand et al., 2013).
process applied to linear UGOs. It provides (i) inflection points of the
input polyline and (ii) arcs between each of them. These DGO-scale
4.2. Extraction of UGO-scale units
units are not segmented in the conventional sense of spatial disaggrega-
tion since the length of disaggregated segments is not a user-defined con-
4.2.1. Stream network extraction (UGO1)
stant. In that case, a DGO is defined as the arc between two successive
The stream network is extracted from the input DEM. The frame-
inflection points and they are attributed as any other DGO (i.e. “Order_ID”,
work is based on the widely used process introduced by O'Callaghan
“Rank_UGO” and “Distance” fields populate the final attribute table).
and Mark (1984) and involves the assessment of the drainage accumu-
lation raster from the original DEM. Over this accumulation raster, the
3.2. Metrics calculation and spatial aggregation
stream network is vectorized into a polyline feature according a user-
defined drainage area to uniformly initiate the final stream network.
Though the final goal is to re-aggregate the basin-scale characteris-
Setting a consistent drainage area is a crucial step to extract a relevant
tics, these disaggregated continuums must first be characterized with
network. Associated scale and resolution effects are discussed in
one or a set of metrics. The FluvialCorridor toolbox allows the user to as-
Tarboton et al. (1991). Following Saunders (1999) and others, such as
sess such metrics through the “Metrics” toolset. For example, each DGO
Lamouroux et al. (2008), a stream burning step can be added to the pro-
can be described in terms of sinuosity index, channel width or cumula-
cess to improve network delineation. Ensured by the ArcHydroTools ESRI
tive linear of contact into the fluvial corridor (i.e. so called ecotones in
package, this optional step enables a vector layer from a pre-existing
ecology, it refers to the transitional zone between two different bio-
network to be embedded into the original DEM.
physical features (Ward et al., 1999) such as gravel bars, vegetation or
water surfaces).
4.2.2. Valley bottom extraction (UGO2)
Finally, the toolset named “Statistics” enables to statistically aggre-
Within the FluvialCorridor toolbox, the framework used to extract
gate the attributed DGOs by using the Hubert algorithm (Hubert,
the valley bottom is the one previously developed by Alber and Piégay
2000; Kehagias et al., 2005). Leviandier et al. (2012) compared a set of
(2011).The “Valley bottom” tool is thus based on the extraction of two
seven existing methods for delineating homogeneous segments based
rasters (i.e. a reference altimetric plan and a relative DEM) from the
on univariate series. They concluded that the Hubert test of homogene-
input DEM and a stream network. Accordingly, this unit is defined as
ity and the Hidden Markov Models are especially successful for identify-
the set of DEM cells with an elevation value between the stream eleva-
ing longitudinal discontinuities. We chose the first one, which provides
tion (assumed to be the minimal elevation in a user-defined buffer sur-
a segmentation once the difference of the mean between two consecu-
rounding the stream network) and an empirically-defined threshold
tive segments is significant. So the Hubert test is used to investigate sub-
corresponding to the submerged floodplain for a uniform waterflow
patterns encapsulated in the fluvial continuum, locate the longitudinal
height (Williams et al., 2000).
discontinuities and aggregate DGOs into AGOs (i.e. homogeneous
reaches). The number of AGOs depends on a criterion α and any change
of its value modifies constraints for the AGOs delineation (i.e. increasing 4.2.3. Centerline extraction (UGO4)
α relaxes constraints and enables the test to identify more homoge- The “Centerline” tool, which involves a framework previously devel-
neous reaches). Leviandier et al. (2012) showed this parameter not as oped by Alber and Piégay (2011), can be used to extract a centerline for
a common confidence level but rather as a level of the risk governing any ‘long’ polygon (i.e. with one larger dimension compared to the
the segmentation size. other), including valley bottoms, active channels, or any other homoge-
Over an entire disaggregated network and for a given metric, the neous and continuous unitary geographical object. The process is based
“Hubert test” tool processes each DGO of a UGO from upstream to on a Thiessen polygonization of the segmented boundaries of the input
downstream. Hence the best segmentation is found (i.e. the one induc- polygon.
ing the most relevant difference between two consecutive segments
with respect to the limiting criterion α), DGOs are attributed with 4.3. DGO-scale database from spatial disaggregation
“Rank_AGO” and “AGO_Val” fields. The “Rank_AGO” field contains a
unique ID for each AGO and “AGO_Val” is the mean value of the given The spatial disaggregation can be conducted on any UGO covering a
metric over an AGO (Fig. 2B). reach or a network. At this stage of the toolbox use, we illustrate the
method with six DGOs layers (Fig. 1).
4. Automatic production of a multiscale geomorphic database DGO1 and DGO6 represent channel patterns passing through inflection
points. They are assessed by the “Polyline disaggregation” tool. Then
Through an archetypal and theoretical workflow (Fig. 1), this section DGO2, DGO3, DGO4 and DGO5 relate to the spatial disaggregation of linear
details the different tools of the FluvialCorridor package. It also focuses or polygon units they belong to (i.e. UGO1, UGO2, UGO3 and UGO4). Linear
32 C. Roux et al. / Geomorphology 242 (2015) 29–37

Raw data
Digitized stream Aerial
DEM Vector layers
network photographs

Connectivity Eroded areas


DEM detection
Stream network Theoretical
(UGO1) network

DGO1 DGO2 Valley bottom


Channel pattern
(UGO2)
Rank / Distance Rank / Distance
Sinuosity Watershed area
Elevation
scale database

Halh-length
Half-amplitude Slope
Arc length Flows Active channel
Stream power (UGO3)
Drainage density DGO3

Rank / Distance
Valley bottom width
Confinement index
Centerline
DGO
UGO

DGO4 (UGO4)

Rank / Distance
Active channel width
Extraction Contact length
Disaggregation Confinement index DGO6
Shifting DGO5
Channel pattern
Transfer information
Widenning
Aggregation Eroded volumes Rank / Distance Rank / Distance
Intermediate data or processes Eroded Areas Sinuosity
Halh-length
External data or processes Half-amplitude
Available metrics Arc length
Non exhaustive list of derived metrics
AGO-scale

Spatial
AGO
discontinuities

Fig. 1. Example of an AGO-scale database extraction through the FluvialCorridor package. This theoretical framework involves all the tools of the package so that three input raw data are
required. Aerial photographs are used to illustrate potential investigations provided by external data. Four UGO-scale units are created: spatial components and a reference centerline.
Then, they are disaggregated into a DGO-scale database to refine the accuracy of the morphometric, geomorphic and topologic metrics stored in the related attribute tables. An AGO-
scale database is created by merging DGO-scale units into homogeneous reaches thanks to the Hubert test.

features (e.g. UGO1 and UGO4) are first sequenced and oriented with the tool enables the user to assess a set of morphometric attributes (e.g. sin-
“Sequencing” tool to have consistent results. uosity, half-length, half-amplitude and arc length) against linear features
The resolution of the DGO-scale database is a user-defined parame- thanks to their related channel patterns. Their accuracy directly depends
ter so that the user may modify the disaggregation step to investigate on the quality of the input linear feature and they can be stored into the
different biophysical patterns within riverscapes. According to the raw channel pattern or into the linear feature. Sinuosity is thus calculated for
data precision, fine and detailed disaggregation enables the user to each arc of a UGO as the ratio of the UGO arc length over the channel pat-
identify very local trends by increasing the accuracy of metric calcula- tern segment length. Processes used to assess half-length, half-
tion. Conversely, managers can attempt to catch regional behaviors amplitude and arc-length are those described in Alber and Piégay
thanks to a coarser-scale disaggregation. Thus, for each UGO, disaggre- (2011).
gation scales must be set smaller than the scale at which forms and pro- Then, DGO2 has been chosen to store geomorphic and topologic
cesses of interest occurs but coarse enough to be appropriate to the metrics. Indeed, from the raw DEM input, FluvialCorridor package pro-
precision of the raw input data. poses to assess the watershed area, the mean elevation and the slope
for each segment of the entire network, from upstream to downstream.
4.4. Metric characterization of the riverscape The watershed area is simply extracted from the drainage accumulation
raster of the DEM. The elevation is stored into three different fields:
Fig. 1 provides a general overview of these metrics, either directly (i) upstream elevation, (ii) downstream elevation, and (iii) mean eleva-
extracted with FluvialCorridor package or derived from others or from tion over a DGO. The slope is calculated by dividing the difference be-
external data or processes (e.g. connectivity DEM, image analysis). tween the upstream and downstream elevation by the DGO length.
Firstly, channel patterns (i.e. DGO1 and DGO6) can be viewed as refer- Lastly, width values for polygon units such as the valley bottom or
ence axis for linear features such as UGO1 and UGO4. The “Morphometry” the active channel can be calculated with the “Width” tool of the
C. Roux et al. / Geomorphology 242 (2015) 29–37 33

Fig. 2. Implementation of the spatial disaggregation and aggregation procedure developed by Alber and Piégay (2011). The first row describes the general framework involved in the
FluvialCorridor package, from the raw data to the AGO-scale database extraction. The second row illustrates the different databases and the related geographical objects creating during
the process.

“Metrics” toolset. Valley bottom width is thus assessed into the DGO3 corresponds to the cumulative linear boundary shared between two ad-
using the framework developed by Alber and Piégay (2011) which re- jacent biophysical features (e.g. water/bar).
quires creating points at a constant length on the UGO2 boundaries.
The resulting width values are measured as the projection of these 4.5. Extraction of homogeneous reaches and longitudinal discontinuities
points on the centerline (i.e. UGO4). The active channel width is assessed
into the DGO4. The process is based on a Thiessen polygonization of the Once the DGO-scale database is characterized with at least one met-
centerline (i.e. UGO4) and the active channel width is assessed as the in- ric, spatial aggregation can be conducted with the “Hubert test” tool of
tersection between Thiessen polygons boundaries and UGO3. Another the “Statistics” toolset. This statistical test is univariate so that only
metric, named “Contact length”, can be calculated with the DGO4 and one numerical metric can be assessed each run. For an entire network,

Table 2
Summary of tools included in the FluvialCorridor package, sorted according to four toolsets: (i) extraction of “Spatial Components”, (ii) spatial “Disaggregation Processes”,
(iii) characterization with a set of “Metrics”, and (iv) spatial aggregation with “Statistics” tools.

Tools Description

Stream network extraction of the stream network from a DEM Spatial components
Valley bottom extraction of the valley bottom with a DEM and a stream network
Centerline extraction of the centerline of a polygon
Sequencing orientation and ordination of a linear network Disaggregation processes
Segmentation disaggregation of entities, with a user-given constant step
Polyline disaggregation extraction of the disaggregated pattern of a linear network passing through inflection points
Contact length assessment of the cumulative contact length between a set of polygons Metrics
Elevation and Slope assessment of the elevation and the slope along a linear network
Morphometry assessment of a set of morphometric units along a linear network
Discontinuities assessment of the ratio between two consecutive DGOs for a given metric
Watershed assessment of the drainage area along a network
Width assessment of the width of polygons
Hubert test identification of longitudinal discontinuities thanks to the statistical test of Hubert and re-calculate Statistics
metric value into each AGO
34 C. Roux et al. / Geomorphology 242 (2015) 29–37

each different UGO is processed successively. The final output result in- thresholds, the valley bottom was extracted and cleaned by removing
cludes two fields more than the input DGOs layer: “Rank_AGO”, a unique alluvial fans and filling residual holes. We visually validated this UGO
ID for each AGO identified and “AGO_Val”, the mean value of the given with an existing layer from the Créateurs de Drôme project. Finally, we
metric over an AGO. This attribute table format allows to keep the used the “Centerline” tool with a 5 m disaggregation step to create a
DGO-scale characterization into the AGO-scale database. 89.6 km centerline related to the valley bottom.
Longitudinal discontinuities along the network can also be extracted To distinguish confined from unconfined reaches within our two
into a multi-point feature thanks to the “Discontinuities” tool of the reaches, we had to compute a confinement index, defined here as the
“Statistics” toolset. The “AGO_Val” ratio of two consecutive AGOs is ratio between the active channel width and the valley bottom width
transferred into a point. This point is precisely located at the break be- and viewed as an indicator of the potential of the stream's lateral mobil-
tween the two AGOs. This functionality enables to represent discontinu- ity. This metric must be stored into one of the geomorphic components
ities breaks as consistent geographical objects. which are representative of the fluvial continuum: the stream network,
the valley bottom or the centerline. We chose the valley bottom because
5. Application example on the Drôme River (France) it is a physically meaningful riverscape unit and because it allows a con-
sistent visualization of results. Hence, we created a DGO-scale database
The present section illustrates how the toolbox is used for character- thanks to the “Disaggregation processes” toolset (Fig. 3B step ➁). First,
izing geomorphic features. We apply it to the Drôme River (France) con- we ran the “Sequencing” tool to ensure a good orientation of the valley
tinuum to identify distinct geomorphic reaches along a river course and bottom centerline. This sequenced centerline was used to disaggregate
to distinguish unconstrained and constrained reaches by anthropogenic the valley bottom polygon every 100 m (“Segmentation” tool).
infrastructure (e.g. embankments) or geological confinements. At the same time, we executed the “Width” tool of the “Metrics”
toolset (Fig. 3B step ➂ and ➃) to calculate needed metrics. Along the
5.1. Background centerline and each 20 m, a first set of points with the valley bottom
width information (WVB) and another one with the active channel
5.1.1. Study area width (WAC) were created. Those attributes were then transferred into
The Drôme River is a 106 km long tributary of the Rhône River the DGO-scale database thanks to a spatial join (Fig. 3B step ➄). Only
(Fig. 3A) and its catchment of 1660 km2 lies between 800 and 2000 m one DGO did not receive a valley bottom width and nine had no value
of elevation. Located in the Southern French Alps, it drains the Vercors of active channel width. This is due to either a lack of information
and Diois massifs, a steep and mountainous terrain of Mesozoic lime- from the active channel layer or inconsistent DGOs (see Section 6.1).
stone and marl (Piégay et al., 2004). Lastly, we calculated the confinement index (CI).
This continuum has been chosen for illustrating the present GIS pro- Finally, we used the “Statistics” toolset, with an α criterion of 0.05, to
cedure because its geomorphic pattern is constrained longitudinally in aggregate the attributed DGO-scale database in two sets of AGOs
terms of anthropogenic and natural confinement (Landon et al., 1998). (Fig. 3B step ➅). The “Hubert test” tool was run two times: one over
We focused our example on two reaches (C1 and C2 on Fig. 3A). The the CI in the reach C1 and another one over WAC in the reach C2.
first one is situated within the upper part of the Drôme River, from
Die to Vercheny. This 15.3 km reach is characterized by two steep gorges 5.3. Results
separated and surrounded by wide alluvial valleys. The second one is in
the downstream part of the river, between Crest and Livron-sur-Drôme. This section focuses on results presented in Fig. 3C1 and C2. They are
Along this13.8 km reach, originally highly braided, the channel has been based on the two AGO-scale databases which have been produced from
locally constrained by embankments. the 897 DGOs of the Drôme River. Among this DGO-scale database, WVB
varies between 62.7 m and 4373.6 m (W VB : 831.7 m) and WAC between
5.1.2. Available data 0.5 m and 372.6 m (W AC : 81.4 m). The confinement index CI is between
Within France, several databases are available and provide a large 2.1 × 10−4 and 0.76 (CI: 0.15).
set of geographical data. Thus, respectively from the BD Carthage®12
and the BD Topo®12, we extracted the stream network and the DEM 5.3.1. Upper part of the Drôme River: reach C1
of the Drôme River catchment. The stream network is 593 km long, in- We chose the reach C1 to illustrate how useful the FluvialCorridor
cluding the main tributaries of the Drôme River. The DEM has a 50 m toolbox can be to distinguish geologically confined reaches from alluvial
resolution. valleys, within which the active channel can freely shift. Following
Over the past two decades, numerous scientific studies and integrat- Wiederkehr (2012), a CI of 0.3 can be used to identify geological con-
ed river plans have been conducted on the Drôme River, resulting in a finements, i.e. above this threshold, we assumed that the valley has a
broad range of available data. Thus, the last input of our study case, a significant control on the channel shift. In that way, we used the AGO-
vector layer of the active channel, comes from the Créateur de Drôme scale database constructed with the confinement index. The reach C1 in-
project, started in 2008 and supported by the ZABR and GRAIE programs cludes 105 DGOs aggregated into 6 different aggregated reaches which
and the Rhône-Mediterranean basin authority. It has been extracted extend from 700 m to 3.2 km. As shown in the Fig. 3C1, and according to
thanks to image analysis of infra-red orthophotos. the 0.3 threshold of confinement index, Reaches 2, 3 and 5 can be iden-
tified as confined reaches (i.e. respectively CI = 0.56; 0.33; 0.38). Con-
5.2. Methods sidering the confinement index, the 3 others AGOs appear to be
unconfined (CI(1) = 0.16; CI(4) = 0.16; CI(6) = 0.23). This is strongly
The general workflow is presented in the Fig. 3B. confirmed visually on the orthophotos and by the contour lines of
First, we used the “Spatial components” toolset to create a UGO-scale elevation.
database over the study area (Fig. 3B, step ➀). A consistent stream net-
work of 101 km has been extracted thanks to the “Stream network” tool. 5.3.2. Lower part of the Drôme River: reach C2
The raw DEM has been burned with the agree stream of the BD Car- We chose the reach C2 to illustrate how the FluvialCorridor toolbox
thage® and a minimum drainage area of 65 km2 has been set empirical- can be used to distinguish braided reaches from embanked reaches.
ly both to reduce the drainage density and also to extract the entire The reach C2 is overlaid by 124 DGOs which have been aggregated
Drôme River. We proceeded to a quick manual cleaning of the resulting into four homogeneous reaches in terms of active channel width.
network in order to only select the Drôme River main stem. Then, we Those aggregated reaches vary from 1.1 to 5.1 km and two pairs clearly
ran the “Valley bottom” tool. After having tested several elevation emerge according to WAC.: Reaches 1 and 3 [84.4 m; 91.7 m] and
C. Roux et al. / Geomorphology 242 (2015) 29–37 35

A
B

e
an
rv
Ge
La
Livron sur
Drôme Die
C2 C1 agree active
Crest
stream DEM channel
Le Rh

Vercheny
La Drôme
ône

Le Bès

La
Ro
an
ne
Drôme basin
Drôme river
main tributaries
secondary tributaries
0 5 10 km
1
main cities

FluvialCorridor tools
Stream network UGOs
1 "Spatial components" : Valley bottom stream network
Centerline valley bottom
2 "Disaggregation processes" :
Sequencing centerline
Segmentation 0 5 1 km

3 & 4 "Metrics" : Width


6 "Statistics" : Hubert test

3 4

2 271m 53m
240m 48m
242m 55m
229m 44m
228m 43m
185m 37m
199m 43m
166m 44m
170m 40m
157m 45m
141m 149m 44m
116m 47m
135m 47m 40m
110m 123m
109m 46m 14m
110m
110m
108m 21m

0 50 100 m 0 50 100 m

Metrics
0 5 1 km valley bottom width (20m)
DGOs active channel width (20m)
disaggregated valley bottom (100m)
FID Order_ID Rank_UGO Rank_DGO Distance Width_VB Width_AC Conf_Index
12 1 1 12 1200 123.32378 23.98450 0.194484
13 1 1 13 1300 110.491 43.75596 0.396014
14 1 1 14 1400 95.71298 29.85080 0.311878
15
16
1
1
1
1
15
16
1500
1600
125.86233
173.72395
52.71579
55.65886
0.418837
0.320387 5
17 1 1 17 1700 141.80922 36.20536 0.25531
18
19
1
1
1
1
18
19
1800
1900
112.21320
111.07617
30.20107
10.87599
0.26914
0.097915
Spatial join
20 1 1 20 2000 143.0410 35.41414 0.247565 attributed DGOs
homogeneous valley segments
AGOs :
based on confinement index (α =0.05)
70
60

40 0 50 0 1
0

0.16 6
2
0.56 homogeneous valley
400

AGOs : segments based on active


4
0.16
3
0.33 channel width (α =0.05)
50
0

0
40

60
0 3
91.7
0 0 1 2 km
50
4
0 0,5 1 km 194.0
5 60 0
0.38
700 active channel (water / bar) 2
196.8
isolines 100m
1
isolines 25m 84.4
AGOs boundaries
active channel AGOs boundaries
confined reaches
6 (water / bar) embanked reaches
0.23 active channel width

C1 0.23 confinement index (CI) =


valley bottom width embankments 91.7 active channel width (m)
C2
Fig. 3. (A) Location of the Drôme River basin in France and the two reaches C1 and C2 within the main hydrographic network. (B) Main steps of the study case workflow. (C1) and (C2)
Results over the two reaches of the Drôme River: homogeneous reaches in terms of confinement index in C1 and in terms of active channel width in C2.
36 C. Roux et al. / Geomorphology 242 (2015) 29–37

Reaches 2 and 4 [196.8 m; 194.0 m]. A statistical analysis shows that the 6.2. Other potential or derived metrics
two narrow reaches are also quite uniform in width. Actually, the
interquartile range (IQR) of WAC in Reaches 1 and 3 (i.e. respectively The richness of the metrics characterization directly depends on the
22.4 m and 11.5 m) is significantly lower than in the two others AGOs set of available metrics. As shown in Fig. 1, the FluvialCorridor package
(i.e. IQR(2) = 79.2 m and IQR(4) = 51.4 m). Those results are visually val- enables the user to calculate ten geomorphic or landscape metrics. Da-
idated, both with the active channel layer and the orthophotos. It clearly tabases can also be characterized either by a combination of several
appears that Reaches 2 and 4 are braided reaches, with a high width var- metrics or by importing data from external databases or processes.
iability and free to extend or laterally shift. Conversely, an embankment The first point is illustrated in the case study by the use of the confine-
layer from the Créateurs de Drôme project confirms that Reaches 1 and 4 ment index (Fig. 3C1) and in the theoretical workflow (Fig. 1) where
are confined by human infrastructures (i.e. embankments) resulting in a we assessed the drainage density in the DGO2 as the ratio between
quite uniform distribution of WAC. the cumulative stream length and the watershed area. External data-
bases (e.g. ESTIMKART for the French territory (Lamouroux et al.,
2010), CORINE Land Cover for the European Union) can be joined to a
5.4. Conclusion
DGO-scale database to add information such as flows or channel
depth values, increasing the number of available and potential derived
The present organization is quite confusing because there are two
metrics. External databases can also be created from external processes
consecutive conclusion chapters (i.e. 6 and 7).I think the Chapter 6
and then joined to the attribute table of a DGOs layer (e.g. cumulative
"Conclusion" is not correctly located. This part of the paper refers to
eroded area from image and DEM analysis, Fig. 1). Possibilities are man-
the specific conclusions of the case study and so should be placed within
ifold and field data can also be included when available. By joining sev-
the chapter 5, as a sub-chapter 5.4."?>Actually, all the workflows de-
eral databases, one may access and investigate a very large range of
scribed above can be applied in a few hours depending on the study
metrics.
area extent and the preliminary requirements, such as the selection of
input data and the empirical settings of parameters (e.g. the elevation
6.3. Added value and further improvements
threshold for the valley bottom, the spatial disaggregation step). The
case study presented took about 20 min of computation. Also, presented
As mentioned in previous sections, the FluvialCorridor toolbox pro-
results could surely be improved by increasing the DGO-scale database
vides a complete and robust workflow and offers multiple approaches
resolution, adjusting the α criterion of the Hubert test or, certainly by
to characterizing fluvial continuums and riverscapes.
introducing other parameters (e.g. transversal slope threshold (Gallant
Compared to the first version, this toolbox provides many options to
and Dowling, 2003)).
explore the database. User-defined metrics allow a more tailored ap-
proach to the exploration of one’s dataset, depending on the data avail-
6. Discussion and conclusions able and the geographical context. For example, the “Valley bottom”
tool allows the operator to choose his/her own threshold in terms of el-
6.1. FluvialCorridor limitations evation. The “Stream network” tool allows the user to choose the drain-
age area value (in km2) used to initiate the hydrographic network and
As it is still in development, the FluvialCorridor toolbox shows some in the “Hubert test” tool, α value is also chosen by the operator. As a
shortcomings. Geomatic or geometric limits are referenced within the last example, the constant length used to the spatial disaggregation of
guidelines included with the package. One of them has already been in- UGOs is a user-defined parameter.
troduced in Section 5.2 and deals with the limits of the spatial disaggre- Further improvements are projected. Major efforts are currently
gation processes, especially for polygon inputs. Currently and as under way to include imagery analysis tools. By integrating colored or
implemented in the package, with too highly curved centerlines, infra-red orthophotographs, the user will be able to access new
Thiessen polygons used for the disaggregation do not cross the entire riverscape units (e.g. water bodies or backwaters, vegetated islands or
input polygon. Thus, resulting DGOs can lead to problems during the forest cover, bars, anthropogenic elements, etc.) according to a specific
characterization of the DGO-scale database, as mentioned in the research focus. Some others utilities are investigated such as a cluster
methods section of the case study above. Similar issues occur at conflu- analysis module for exploring features typology after the spatial aggre-
ences which are very specific areas where the metric assessment can be gation. Moreover, only one algorithm of aggregation (i.e. Hubert test) is
inconsistent (e.g. fluvial widths, ecotones). currently available. One of the future improvements will be to imple-
Moreover, methodological shortcomings about the multiscale po- ment new statistical algorithms such as the multivariate Hidden Mar-
tential of the package are also noticed. Problems can occur with some kov Models (Baum and Eagon, 1963; Mari et al., 2011). Finally, and
tools used at a regional scale. Input parameters do not handle the inher- following the pioneer contributions of Tormos (2010), we plan to intro-
ent size effect of fluvial longitudinal patterns. For example, in the case of duce new procedures to explore multiscale models, enabling the user to
the “Valley bottom” tool, valley bottom extraction is done within a buff- predict local reach-scale metrics from upstream reach, network or
er, created around the different branches of the hydrographic network. catchment characteristics.
The buffer size is user-defined and constant over the entire study length.
Problems can therefore occur for upland areas, where the valley bottom, 6.4. A geomatic workflow for answering different questions at a network scale
which is generally narrower in headwater systems, can be searched for
within surrounding valleys. In order to avoid such problems, the buffer This paper can be viewed as a part of a larger work, initiated by Alber
size, and perhaps the elevation threshold used for the valley bottom def- and Piégay (2011), with a goal of developing a procedure to investigate
inition, should be set for each stream of the network according to its and characterize networks and riverscapes at a range of scales. The pre-
catchment area. This issue can be applied to other processes such as sented FluvialCorridor toolbox package goes beyond the simple concep-
the spatial disaggregation or aggregation and must be fixed in the tual and geomatic framework by providing an automated GIS tool. It can
next version. therefore be used in other contexts to answer to other questions. Belletti
Nevertheless, and as mentioned previously, the FluvialCorridor tool- et al. (2012) and Notebaert and Piegay (2013) demonstrated that this
box is still under development and geomatic and geometric weaknesses multiscale approach is powerful and operational both in research and
remain infrequent. The package still provides robust tools and frame- management contexts. The latter used the entire workflow to show
works to enrich scientific studies or management plans and produces that the scaling effect observed on the floodplain width of the Rhône
consistent information from local to regional scales. catchment is mainly controlled by lithological variations rather than
C. Roux et al. / Geomorphology 242 (2015) 29–37 37

by downstream distance. Bertrand et al. (2013) used this GIS package Gregory, S.V., Swanson, F.J., McKee, W.A., 1991. An ecosystem perspective of riparian
and framework to assess the risks of environmental changes (i.e. in zones. Bioscience 40, 540–551.
Hubert, P., 2000. The segmentation procedure as a tool for discrete modeling of hydrome-
terms of habitat diversity and trout distribution) induced by sediment teorological regimes. Stoch. Environ. Res. Risk Assess. 14, 297–304.
replenishment on the Drôme River. Wiederkehr (2012) used the proce- Kehagias, A., Nidelkou, E., Petridis, V., 2005. A dynamic programming segmentation pro-
dure to provide a typology of channel patterns (e.g. single-bed or multi- cedure for hydrological and environmental time series. Stoch. Environ. Res. Risk As-
sess. 20, 77–94.
ple channels, braided, sinuous or straight channels) and compare this Kondolf, G.M., Montgomery, D.R., Piégay, H., Schmitt, L., 2003. Geomorphic classification
level of organization with the in-channel features (e.g. the inter-pool of rivers and streams. In: Kondolf, G.M., Piégay, H. (Eds.), Tools in Fluvial Geomor-
spacing) showing the two geomorphic organizations are not linked. A phology. Wiley, Chichester, pp. 171–204.
Lamouroux, N., Pella, H., Vanderbecq, A., Sauquet, E., Chandesris, A., Capra, H., Souchon, Y.,
running project is focused on sedimentation patterns in groyne fields. Pont, D., 2008. Rapport ESTIMKART: Impacts quantitatifs sur les peuplements
Geomatic procedures are used to provide a typology of sedimentation aquatiques de la restauration hydraulique des cours d’eau. Guide de l’application.
pattern according to the inter-groyne geometry and associated hydrau- Cemagref et Agence de l’Eau Rhône, Méditerranée et Corse, Lyon (54 pp.).
Lamouroux, N., Pella, H., Vanderbecq, A., Sauquet, E., Lejot, J., 2010. Estimkart 2.0: Une
lic conditions.
plate-forme de modèles écohydrologiques pour contribuer à la gestion des cours
The FluvialCorridor package is therefore constantly being refined and d'eau à l'échelle des bassins français. Version provisoire. Cemagref, Agence de l'Eau
only V01 is presently available. Nevertheless, this first version of the Rhône-Méditerranée-Corse, Onema, Lyon (45 pp.).
FluvialCorridor toolbox has already been shown to be powerful enough Landon, N., Piégay, H., Bravard, J.P., 1998. The Drôme River incision (France): From assess-
ment to management. Landsc. Urban Plan. 43, 119–131.
to explore riverscape features at reach to network scales from existing Le Pichon, C., Gorges, G., Baudry, J., Goread, F., Boet, P., 2009. Spatial metrics and methods
vector and raster layers. for riverscapes: Quantifying variability in riverine fish habitat patterns.
Environmetrics 20, 512–526.
Leopold, L.B., Marchand, M.O., 1968. On the quantitative inventory of the riverscape.
Acknowledgements Water Resour. Res. 4, 709–717.
Leviandier, T., Alber, A., Le Ber, F., Piégay, H., 2012. Comparison of statistical algorithms for
The authors wish to thank all the researchers who actively participat- detecting homogeneous river reaches along a longitudinal continuum. Geomorphol-
ogy 138, 130–144.
ed to the development of this research project, and notably B. Belletti, Marcus, W.A., Fonstad, M.A., Legleiter, C.J., 2003. Management applications of optical re-
L. Bourdin, S. Dufour, F. Liébault, B. Notebaert, M. Rival, E. Wiederkehr, mote sensing in the active river channel. In: Kondolf, G.M., Piégay, H. (Eds.), Tools
and the different institutions who funded the projects: RMC Water Agen- in fluvial geomorphology. Wiley, Chichester, pp. 19–41.
Mari, J.F., Le Ber, F., Lazrak, E.G., Benoît, M., Eng, C., Thibessard, A., Leblond, P., 2011. Using
cy, SedAlp project (Alpine Space European Program). The authors also Markov models to mine temporal and spatial data. In: Funatsu, K., Hasegawa, K. (Eds.),
kindly thank John C. Stella and Robin Jenkinson who reviewed the paper. New fundamental technologies in data mining. InTech, Rijeka, pp. 511–584.
NB: the GIS toolbox, its guidelines and the database of the case study Neteler, M., Mitasova, H., 2002. Open source GIS: A GRASS GIS approach, 689. Kluwer Ac-
ademic Publisher (464 pp.).
are available online at http://umrevs-isig.fr. Notebaert, B., Piegay, H., 2013. Multi-scale factors controlling the pattern of floodplain
width at a network scale: The case of the Rhône basin, France. Geomorphology
References 200, 155–171.
Nyquist, H., 1928. Certain topics in telegraph transmission theory. Trans. Am. Inst. Electr.
Alber, A., Piégay, H., 2011. Spatial disaggregation and aggregation procedures for charac- Eng. 47, 617–644.
terizing fluvial features at the network-scale: Application to the Rhône basin O'Callaghan, J.F., Mark, D.M., 1984. The extraction of drainage networks from digital ele-
(France). Geomorphology 125, 343–360. vation data. Comput. Vision Graph. Image Process. 28, 323–344.
Allen, T.F., Starr, T.B., 1982. Hierarchy: Perspective for ecological complexity. The Univer- O'Neill, R.V., Deangelis, D.L., Waide, J.B., Allen, G.E., 1986. A hierarchical concept of ecosys-
sity Chicago Press, Chicago (310 pp.). tems. Princeton University Press, Princeton (262 pp.).
Baum, L.E., Eagon, J., 1963. An inequality with applications to statistical prediction for Piégay, H., Walling, D.E., Landon, N., He, Q., Liébault, F., Petiot, R., 2004. Contemporary
functions of markov processes and to a model for ecology. Bull. Am. Math. Soc. 73, changes in sediment yield in an alpine mountain basin due to afforestation (the
360–363. upper Drôme in France). Catena 55, 183–212.
Belletti, B., Dufour, S., Piégay, H., 2012. Regional variability of aquatic pattern in braided Saunders, W., 1999. Preparation of DEMs for use in environmental modeling analysis,
reaches (example of the French Rhône basin). Hydrobiologia 712, 25–41. 1999. ESRI User Conference. ESRI Online, San Diego, CA.
Bertrand, M., Piégay, H., Pont, D., Liébault, F., Sauquet, E., 2013. Sensitivity analysis of en- Schumm, S.A., 1977. The fluvial system. Wiley, New York (338 pp.).
vironmental changes associated with riverscape evolutions following sediment rein- Swanson, F.J., Janda, R.J., Dunne, T., Swanston, D.N., 1982. Sediment budgets and routing in
troduction: Geomatic approach on the Drôme river network, France. Int. J. River Basin forested drainage basins. USDA Forest Service, general technical report PNW-141,
Manage. 11, 19–32. Portland, OR (165 pp.).
Böhner, J., McCloy, K.R., Strobl, J., 2006. SAGA - Analysis and Modelling Applications. Tarboton, D.G., Bras, R.L., Rodriguez-Iturbe, I., 1991. On the extraction of channel networks
Göttinger Geographische Abhandlungen, 115 (130 pp.). from digital elevation data. Hydrol. Process. 5, 81–100.
Carbonneau, P.E., Fonstad, M.A., Marcus, W.A., Dugdale, S.J., 2012. Making riverscapes real. Thorp, J.H., Thoms, M.C., Delong, M.D., 2006. The riverscape ecosystem synthesis:
Geomorphology 137, 74–86. Biocomplexity in river networks across space and time. River Res. Appl. 22, 123–147.
Chinnayakanahalli, K., Hill, R., Olson, J., Kroeber, C., Tarboton, D.G., Hawkins, C., 2006. The Tormos, T., 2010. Analyse à l'échelle régionale de l'impact de l'occupation du sol dans les
Multi-watershed Delineation Tool: GIS software in support of regional watershed corridors rivulaires sur l'état écologique descours d'eau. Ph.D. Thesis Institut des Sci-
analyses. Utah State University, Logan (44 pp.). ences et Industries du Vivant et de l'Environnement, Paris, France (507 pp.).
Dilts, T.E., Yang, J., Weisberg, P.J., 2010. Mapping riparian vegetation with LiDAR data: Vannote, R.L., Minshall, G.W., Cummins, K.W., Sedell, J.R., Cushing, C.E., 1980. The river
Predicting plant community distribution using height above river and flood height. continuum concept. Can. J. Fish. Aquat. Sci. 37, 130–137.
ArcUser Magazine, pp. 18–21 (Winter 2010 Issue). Ward, J.V., Tockner, K., Schiemer, F., 1999. Biodiversity of floodplain river ecosystems:
Eros, T., Schmera, D., Schick, R.S., 2010. Network thinking in riverscape conservation — A Ecotones and connectivity. Regul. Rivers Res. Manag. 15, 125–139.
graph-based approach. Biol. Conserv. 144, 184–192. Wiederkehr, E., 2012. Apport de la géomatique pour une caractérisation physique multi-
Fausch, K.D., Torgersen, C.E., Baxter, C.V., Li, H.W., 2002. Landscapes to riverscapes: Bridg- échelle des réseaux hydrographiques. Élaboration d'indicateurs appliqués au bassin
ing the gap between research and conservation of stream fishes. Bioscience 52, du Rhône. Ph.D. Thesis École Normale Supérieure de Lyon, Lyon, France (287 pp.).
483–498. Wiens, J.A., 2002. Riverine landscapes: Taking landscape ecology into the water. Freshw.
Frissell, C.A., Liss, W.J., Warren, C.E., Hurley, M.D., 1986. A hierarchical framework for Biol. 47, 501–515.
stream habitat classification — Viewing streams in a watershed context. Environ. Williams, W.A., Jensen, M.E., Winne, J.C., Redmond, R.L., 2000. An automated technique
Manag. 10, 199–214. for delineating and characterizing valley-bottom settings. Environ. Monit. Assess.
Gallant, J.C., Dowling, T.I., 2003. A multi-resolution index of valley bottom flatness for 64, 105–114.
mapping depositional areas. Water Resour. Res. 39, 1347–1360.

View publication stats

Potrebbero piacerti anche