University of Montana University of Montana
ScholarWorks at University of Montana ScholarWorks at University of Montana
Biological Sciences Faculty Publications Biological Sciences
8-2005
Using Airborne Multispectral Imagery to Evaluate Geomorphic Using Airborne Multispectral Imagery to Evaluate Geomorphic
Work Across Floodplains of Gravel-Bed Rivers Work Across Floodplains of Gravel-Bed Rivers
M. S. Lorang
D. C. Whited
Richard F. Hauer
University of Montana - Missoula
, ric.hauer@umontana.edu
J. S. Kimball
Jack Arthur Stanford
The University of Montana
, jack.stanford@umontana.edu
Follow this and additional works at: https://scholarworks.umt.edu/biosci_pubs
Part of the Biology Commons
Let us know how access to this document bene>ts you.
Recommended Citation Recommended Citation
Lorang, M. S.; Whited, D. C.; Hauer, Richard F.; Kimball, J. S.; and Stanford, Jack Arthur, "Using Airborne
Multispectral Imagery to Evaluate Geomorphic Work Across Floodplains of Gravel-Bed Rivers" (2005).
Biological Sciences Faculty Publications
. 301.
https://scholarworks.umt.edu/biosci_pubs/301
This Article is brought to you for free and open access by the Biological Sciences at ScholarWorks at University of
Montana. It has been accepted for inclusion in Biological Sciences Faculty Publications by an authorized
administrator of ScholarWorks at University of Montana. For more information, please contact
scholarworks@mso.umt.edu.
Ecological
Applications,
15(4),
2005,
pp.
1209-1222
© 2005
by
the
Ecological
Society
of
America
USING
AIRBORNE MULTISPECTRAL IMAGERY TO
EVALUATE
GEOMORPHIC
WORK
ACROSS
FLOODPLAINS OF
GRAVEL-BED RIVERS
M. S.
LORANG,'
D. C.
WHITED,
F
R.
HAUER,
J. S.
KIMBALL,
and J. A. STANFORD
Flathead
Lake
Biological
Station,
Division
of Biological
Sciences,
University
of
Montana,
311 Bio Station
Lane,
Polson,
Montana
59860-9659 USA
Abstract.
Fluvial
processes
of
cut and
fill
alluviation and
channel
abandonment
or
avulsion are essential for
maintaining
the
ecological
health of
floodplain ecosystems
char-
acteristic
of
gravel-bed
rivers.
These
dynamic
processes shape
the
floodplain
landscape,
resulting
in
a
shifting
mosaic of
habitats,
both above and below
ground.
We
present
a new
and
innovative
methodology
to
quantitatively
assess
the
geomorphic
work
potential
nec-
essary
to maintain a
shifting
habitat mosaic for
gravel-bed
river
floodplains.
This
approach
can be
used
to delineate critical habitats for
preservation
through
land
acquisition
and
conservation
easements,
often
critical
elements
of river restoration
plans
worldwide.
Spa-
tially explicit modeling
of water
depth,
flow
velocity,
shear
stress,
and stream
power
derived
from surface
hydraulic
measurements was combined with airborne
multispectral
remote
sensing
for detailed
modeling
of
floodplain
water surfaces over
tens to
hundreds
of
square
kilometers. The model results
were
then combined within a GIS
framework
to
determine
potential
nodes of channel avulsion that delineate
spatially explicit
zones
across
the flood-
plain
where
the
potential
for
geomorphic
work
is the
greatest.
Results of this
study
dem-
onstrate
the
utility
of
integrating existing
multispectral
remote
sensing
data
coupled
with
time-lagged ground-based
measures
of flow
hydraulics
to
model
fluvial
processes
at rela-
tively
fine
spatial
resolutions
but over
broad
regional
extents.
Key
words: airborne
multispectral imagery; floodplain; geomorphology; gravel-bed
rivers;
river
restoration;
stream
power.
INTRODUCTION
A
new
paradigm
has
emerged
over the last decade
regarding
the
biophysical
structure and function of riv-
ers that shows
floodplains
to
be centers of
biophysical
organization throughout
the
world's
river
systems
(Stanford
and Ward
1993,
Ward et al.
2002).
This view
is
particularly
relevant to
gravel-bed
rivers where the
ecological integrity
of
accompanying floodplains
is in-
timately
linked to the
ability
of the river
to
perform
geomorphic
work
(Ward 1997).
The
dynamics
of al-
luvial
processes, particularly
cut and fill alluviation and
channel
avulsion,
lead to a
shifting
habitat mosaic that
is
essential
for
sustaining floodplain
ecosystem
integ-
rity
(Stanford 1998).
Large
alluvial
floodplains
of
gravel-bed
rivers are
particularly complex,
both above and
belowground.
The
legacy
of
scour,
deposition,
inundation,
and
drought
results
in
an
expansive, spatially complex
and
three-dimensional
floodplain
mosaic.
Lateral,
vertical,
and
longitudinal linkages,
through
which water and ma-
terials
flow,
are
maximized on
expansive
river flood-
plains
producing
inherently
high biodiversity
and bio-
complexity
(Ward
et al.
1999).
Under natural condi-
tions,
biodiversity
and
biocomplexity
are maximized
Manuscript
received 25
September
2003;
revised 25
August
2004;
accepted
7
September
2004;
final
version
received
3 No-
vember
2004.
Corresponding
Editor: J. Baron.
I
E-mail:
mark.lorang@umontana.edu
on
gravel-bed
river
floodplains
because of the conver-
gence
of
aquatic
and terrestrial biotic
assemblages
at
multiple spatial
and
temporal
scales,
and because
con-
vergence trajectories
are
regularly
disturbed
by
flood-
ing
(Ward
et al.
1999).
The
biophysical linkages
that
characterize natural
floodplains
are critical to sustained
anadromous and
resident salmonid
populations,
as well
as
other
important components
of a
floodplain
ecosys-
tem
(Naiman
1998,
Ward
et al.
2002).
Several studies
from
across western North
America
have
revealed
progressive
declines
in
the extent
and
health of
floodplain
gallery
forests
(Bradley
and
Smith
1986,
Braatne et
al.
1996).
The
primary
causes
of these
declines have been related to
dams,
water
diversions,
and the
clearing
of
floodplain
habitats for
agricultural
use and
livestock
grazing
(Rood
and
Mahoney
1990,
Rood et al.
1995,
Mahoney
and
Rood
1998).
Globally,
floodplains
are threatened
by
degradation
from
flow
regulation,
habitat
alteration,
domicile
encroachment,
agricultural
cultivation,
invasive nonnative
species,
and
pollution.
There
is an
urgent
need for
floodplain
restoration
through
conservation easements and
pur-
chases of
critical habitat to
protect
remaining
intact
floodplains
(Tockner
and Stanford
2002).
The river
can
do much of
the
work of
restoration,
but
effectiveness
of
this river restoration
protocol
largely
depends
on
the
degree
to which
regulated
rivers
are allowed to
approach
natural
flow
conditions
and
reoccupy
historical
floodplains
(Stanford
et al.
1996,
1209
This content downloaded from 150.131.192.151 on Wed, 30 Oct 2013 19:01:48 PM
All use subject to JSTOR Terms and Conditions
M. S. LORANG
ET AL.
Ecological Applications
Vol.
15,
No. 4
W
0
25
50
__==,
km
~ *.'1%^
'
...
,
"
.
,
,
,I
N
0 0.5
1.0
2.0
I
m
kmr
FIG.
1. The
figure
shows the extent of
multispectral imagery
acquired
for Union
Gap
floodplain
reach and location
within
the Columbia and
Yakima River
drainage
basins,
northwestern United States
and Canada.
Richter and Richter
2000).
These activities
will
require
the
development
of new
and robust
methodologies
to
identify,
assess,
and monitor
floodplain
areas
with the
greatest potential
for a renewed
shifting
habitat
mosaic
to exist.
We
present
such a
methodology
below
using
a
combination
of field measurements
of water
depth
and flow
velocity coupled
to
spectral
reflectance
pat-
terns
captured
with airborne
multispectral imagery.
STUDY
AREA
We studied the Union
Gap
floodplain
(Fig. 1)
along
the sixth order Yakima
River in central
Washington,
USA
(46°33'
N,
120°27'
W).
The reach
is -8 km
in
length
and 2-3
km
wide,
with
93
ha of
water
surface
and 110
km
of river
edge
during
base flow conditions.
The
floodplain
has been
severely
altered
by gravel
min-
ing,
levees,
and urban encroachment.
In
spite
of these
disturbances,
floodplain
processes
such as cut
and
fill
alluviation
and channel
avulsions continue
to
reshape
some areas
of the
floodplain,
creating floodplain
hab-
itats such
as
spring
brooks,
isolated
ponds,
and wet-
lands. Several
federal,
state,
and local
county
agencies
are
interested
in the
preservation
of critical
floodplain
habitats within
the Union
Gap
reach
through
land
ac-
quisition
and conservation
easements.
METHODS
Image acquisition
Airborne
multispectral
(blue,
0.46-0.55
Im;
green,
0.52-0.61
pm;
red,
0.61-0.70
Im;
near-infrared,
0.78-
0.92
pm)
digital
imagery
was
acquired
on
24
August
1999
for the Union
Gap
floodplain
on the
Yakima
River.
The
digital
imagery
was
acquired
at
a
1-m2
resolution.
1210
This content downloaded from 150.131.192.151 on Wed, 30 Oct 2013 19:01:48 PM
All use subject to JSTOR Terms and Conditions
August
2005
GEOMORPHIC WORK ACROSS FLOODPLAINS
1211
Discharge
at the time of
image acquisition
was 74 cm
measured at the Terrace
Heights gauging
station located
-1
km
upstream
from our
study
reach. The remote
sensing
data were
acquired using
an ADAR
System
5500
digital
camera
(Positive
Systems,
Whitefish,
Montana,
USA)
flown on-board a
light
aircraft. A
dig-
ital
image
mosaic
(Fig.
1)
was created for the
floodplain
image using
Digital Image
Made
Easy
(DIME)
soft-
ware
(Positive
Systems,
Whitefish,
Montana,
USA),
which uses a
semiautomated textural
analysis
of
over-
lapping images
to assemble
large
amounts of
spatial
data into a
single
mosaic. The
digital image
mosaics
of the
floodplain
were then
georeferenced
to a United
States
Geological Survey
(scale
-
1:2400)
Digital
Or-
tho-Photo database.
These remote
sensing
data were
then used for
spatially
explicit
classification of flow
velocity,
water
depth,
and
vegetation
cover for the en-
tire
floodplain
reach.
Field measurements
of depth
and
flow velocity
A
Sontek
RS3000
Acoustic
Doppler velocity-Profiler
(ADP,
Sontek/YSI,
San
Diego,
California,
USA)
was
used to
acquire
detailed
water
depth
and vertical
profile
measurements of flow
velocity along
channel reaches
within the
study floodplain.
The ADP
data were ob-
tained on 23
April
2001,
when river
discharge
was
59
cm measured at
the Terrace
Heights gauging
station.
The ADP
uses three transducers to
generate
a
3
MHz
acoustic
pulse
into the water. As
the sound travels
through
the
water,
it is reflected in all
directions
by
particulate
matter
(e.g.,
sediment,
biological
matter)
being
transported
with the flow. The
transmitted
energy
is most
strongly
reflected from the bottom substrate
providing
a measure
of water
depth.
Some of the
energy
reflected
from
particles
being
carried with the flow
re-
turns to
the transducers with a
frequency change,
and
the
processing
electronics
measure that
Doppler
shift,
which
is
directly
correlated to the
velocity
of the water.
By measuring
the return
signal
at different
times fol-
lowing
the transmit
pulse,
the
ADP
measures water
velocity
at different
distances from the
transducer,
from
just
below the water
surface to the bottom. This results
in
a measured
velocity
profile
and
depth
of the water
column. The
profile
of water
velocity
is divided into a
series of individual
15
cm
deep
cells from
top
to bot-
tom,
where the
average
of the
return
signals
for 5-s
time
intervals are measured for
each cell. The
ADP
operates using
three transducers
generating
beams with
different orientations relative to the flow
of water. The
velocity
measured
by
each ADP
transducer is
along
the
axis of
its acoustic beam. These
beam velocities are
converted to XYZ
(Cartesian)
velocities
using
the rel-
ative orientation of the
acoustic
beams,
giving
a
3-D
velocity
field relative to the
orientation of the
ADP.
Since it is not
always possible
to control instrument
orientation,
the ADP
includes an internal
compass
and
tilt sensor
to
report
3-D
velocity
data in
Earth
(East-
North-Up
or
ENU)
coordinates,
independent
of instru-
Remote
sensing
(multispectral
images)
GPS
ADP
Flow
V
FIG. 2. A schematic
showing
the correlation between mul-
tispectral imagery,
over
a
square
area
of water
surface,
with
flow
velocity
(V)
and
depth
(h)
data measured with
the ADP
(Acoustic
Doppler velocity
Profiler).
Velocity
was measured
in discrete
bin
intervals
throughout
the water column and
then
averaged
to determine
V.
ment orientation.
Hence,
it is
possible
to determine
the
mean flow
velocity
in
separate
cells
through
the water
column oriented
perpendicular
to the flow field.
We
deployed
the
ADP
from the front of a small
jet-
boat with
both
velocity profile
data and
depth
data cor-
related
spatially by linking
a
GPS
(Global
Positioning
System)
receiver colocated with the
position
of the
ADP
(Fig.
2).
During
data
acquisition
the ADP was
maneuvered back and forth across the channel to obtain
data from as full an
array
of
aquatic
habitats,
depths,
and velocities as
possible.
Both the
ADP and GPS data
were recorded
simultaneously
on a field
laptop
com-
puter.
The ADP data were then
processed
to create an
integrated velocity
value
(average velocity
for an
in-
dividual
ADP
profile)
as well
as a
depth
value for each
GPS location.
Image
and
field
data collection inconsistencies
We used archived
imagery
from a
previous project
on the Union
Gap floodplain coupled
with
ADP data
to model
depth
and
velocity throughout
the
floodplain
reach.
Although
there was a
two-year gap
between the
acquisition
of
imagery
and the
subsequent
field
data,
the amount of
geomorphic
change
was
minimal and the
discharges
were similar
in
comparison
to
peak
dis-
charges
that
typically
result in
geomorphic
work
(Fig.
3).
The winter of
1996-1997,
two
years prior
to the
image acquisition
date,
was dominated
by large mag-
nitude and duration
flooding (Fig.
3).
Those flood
events
greatly
exceeded bankfull
discharge
levels. Two
This content downloaded from 150.131.192.151 on Wed, 30 Oct 2013 19:01:48 PM
All use subject to JSTOR Terms and Conditions
1212
M.
S. LORANG
ET AL.
Ecological
Applications
Vol.
15,
No. 4
600
.
1997
-
1999
500
20
i0
-
2001
Approximate
bankfull
discharge
400
1)
300
c-
S200Image
acquisition
0
0
'
I
II
1
Jan 31 Jan 1Mar 31 Mar
30Apr
30May
29Jun 29Jul
28Aug
27Sep
270Oct
26 Nov 26 Dec
FIG.
3. The
figure
shows a
comparison
between a flood of record
discharge
(1997)
and the
timing
of
image acquisition
in
1999,
along
with the
2001
ADP
ground
truth data with the
discharge
between those data collections. All
discharge
data
come from the
USGS
gauging
station number
12500450 located
just
below the Union
Gap
reach,
Washington
State,
USA.
other
near bankfull
discharge
events occurred in
May
and June
1999. These flow events exceeded broad
geo-
morphic
thresholds and
played
a
significant
role in
set-
ting
the
geophysical
template
for the
Union
Gap
reach
prior
to
image
acquisition
in
August
1999.
Conversely,
no
large
magnitudes
or
long
duration
discharge
events
occurred
in
2000 and
2001
prior
to
acquisition
of the
hydraulic
ground
truth
data that could have resulted in
significant
geomorphic changes
relative to those
that
occurred in
1996 and
1999
(Fig.
3).
Ideally imagery
and
ground
truth data should be
col-
lected within the same
time frame.
However,
flow con-
ditions were similar and no
major
geomorphic
work
occurred
between the time of
image
acquisition
and the
collection of
ground
truth
data,
so it is reasonable to
believe that measures of
depth
and
velocity
were sim-
ilar. Minor
perturbations
in
depth
and flow velocities
occurred due to
rearrangement
of
large
wood,
lateral
erosion,
and bar
deposition
over
time,
but these
chang-
es were
minimal and in
isolated areas.
To
verify
the
analysis
of the
hydrologic
record,
we
used a
high
resolution satellite
image
from
27
August
2000
(similar
discharge
to the
1999
image)
to
compare
inundation
extent to the
1999
image.
Between
1999
and
2000,
there
was
only
a 4% increase in
inundation
extent and no
channel movement.
By overlaying
the
2001 coordinates of
the main
thalweg
on the
1999
im-
age
we determined
the
position
of the main channel
had not
changed.
Both the
acquisition
of
multispectral imagery
and
ADP
ground
truth data
occurred
during
similar base
flow
conditions
(Fig.
3).
The difference in
discharge
between the
acquisition
of
imagery
and the
subsequent
field data amounted to
only
a
10-cm difference
in
stage.
This is less than the resolution at which we were able
to
classify
water
depth
and
model
topographic
eleva-
tions across the
floodplain.
Thus,
the
spatial patterns
of water
depth
and flow velocities measured
in
2001
were
within
the
range
of actual
depth
and velocities
values that occurred
during image
acquisition
in
1999.
We demonstrate this
thoroughly
and discuss
the
ram-
ifications to model results
in
the
section,
(Methods:
Assessment
of classification
error).
By carefully
ex-
amining discharge
records,
geomorphic changes,
and
collecting
field data at similar flow
conditions,
we feel
archival
imagery
can be
successfully
used to model
depth
and
velocity,
as well as other
floodplain
habitats.
Classifying flow velocity
and water
depth
Water
depths
and flow velocities were modeled based
on
correspondence
between these variables and surface
water
spectral
reflectance
patterns captured
by
the air-
borne
multispectral digital imagery.
The water surface
spectral
reflectance
patterns
were
directly
related to
surface water
roughness
and color. The
technique
re-
quired
three
steps:
(1)
classification and extraction of
surface water features from the
imagery,
(2)
separation
of
off-channel and main channel
habitats,
and
(3)
clas-
sification
of
channel
depth
and
flow
characteristics
by
statistical
relationships
established between
spectral
re-
flectance characteristics and
corresponding
ground
truth data. The near-infrared band and
a normalized
This content downloaded from 150.131.192.151 on Wed, 30 Oct 2013 19:01:48 PM
All use subject to JSTOR Terms and Conditions
