1
This is a cartoon of the standard NIR lidar. Only get 5 partial returns from each
laser shot. In effect we’re wasting much of the laser energy\project costs,
because we never record much of the energy from each pulse.
2
Contrast between full-waveform recording and discrete sampling of that
waveform. Some of the fixed-interval discrete samples may be very informative,
while others are much less so, depending on where they occur in the waveform.
3
This illustrates how water absorbs most of the energy from near-infrared lidars.
Over the forest the red dots are the third-returns from somewhere in the middle
(vertically) of a forest canopy. The small white areas are individual tree crowns.
4
The Experimental Advanced Airborne Research Lidar (EAARL) is a
relatively new research instrument with several design features that are
attractive for stream bathymetry. EAARL is a full-waveform green lidar
with the capability to penetrate water (depth and clarity dependent)
Note that new system (2010) will have much improved data density
relative to the performance described in this slide. Laser power will also
be somewhat greater in new system.
5
-Cross environment veg/topo/bathy capabilities
-Maps detailed topography in shallow non-marine/marine environments
with clear water.
-Note the “waveform digitizing” capabilities and significance
6
The very wide dynamic range of aquatic environments makes the unique
EAARL design essential with multiple Photo Multiplier Tubes (PMTs)
sampling different portions of the returned energy.
7
This simplified block diagram describes the wide dynamic range of the
EAARL sensor. The Optical receiver front end is comprised of : a
telescope, filters, apertures, lenses, etc. and is depicted by the blue box.
The dotted lines represent the photon paths through a system of
distortion free optical power dividers that splits the signal so 90% goes to
PMT-1, 9% to PMT-2, and 1% to PMT-3. The group delay, or propagation
time, through the system is carefully controlled through the entire optical
and electrical signal path. The PMTs each convert their fraction of the
arriving optical energy to electrical impulses which are each delivered to
corresponding channels within one of two waveform digitizer banks.
When the laser is triggered, a small sample is coupled via fiber optic to
PMT-3 resulting is the system transmit pulse. The transmit pulse marks
the beginning of the laser distance measurement and also gives the laser
power and pulse shape characteristics.
8
This is a cartoon of how the EAARL laser interacts in water. The
EAARL system sends out a short green wavelength (532nm) laser
pulse that has a 15-20 cm footprint when it begins to interact with the
water surface. The entire time history of the outgoing laser pulse is
digitized and stored (thus this is a “full-waveform” laser system). As
the pulse interacts with the water surface, portions of the energy are
reflected from various layers in the water column. This time history
of the reflected energy is digitized by the array of PMTs and the
relevant portions of the digitized return are recorded as a waveform.
At any particular time, the amplitude of the waveform measures the
strength of the return. Inflections in the waveform that comprise
sufficient leading and trailing edges generally define a layer in the
water column. Depending on the signal strength and the water
depth and clarity, the laser will return another peak in the backscatter
indicating that the pulse has penetrated through to the bottom.
Note also that laser velocity in water is ~75% of air velocity - - so we
need to recognize the water surface, otherwise we will under-predict
the bed elevation because we assumed air velocity throughout the
travel time, when actually it was slower for part of the travel time.
9
The elevation of the bed of a channel is normally calculated by measuring the
difference in time between the centroid of the peak from the water surface and
the centroid of a peak from the bed.
10
11
Here EAARL bathymetric data for part of a channel are combined with the digital
color-infrared camera imagery also gathered in each EAARL mission. Coregistration was only done manually and so isn’t exact in this example.
12
EAARL can also be flown in a Pilatus PC-6 aircraft that is better equipped for
mountain canyon flying.
13
The darker green spots represent the pattern of elevation measurements using
the existing EAARL system. The olive green spots represent the additional point
elevation measurements that will be taken with the improved EAARL.
14
The full-waveform data requires custom-written processing code.
15
This screen capture shows many of the important elements of the ALPS GUI
interface.
16
This is an example of the data density from 2004 in Bear Valley Cr., Idaho. The
red dots are places where the lidar recorded either the exposed bare earth
elevation or the submerged ground elevation (if inside the channel). The point
cloud data were then gridded and contoured. The data density is greater in some
areas where we happened to get 2 passes of data. We try to fly multiple flight
lines with overlapping coverage everywhere to compensate for the somewhat
lower pulse frequency of EAARL (it’s a brute force way to increase the data
density). Most of the contours outside the channel have been removed to make
the image less busy. A few “off-channel habitat” areas were left as noted.
17
This is a side-by-side comparison of a survey-grade GPS wading survey and the
bathymetric lidar. The GPS surveyors missed a small deep pool at 626890E,
4913265N, so in this case the lidar survey is more accurate. The field photo is
taken from the position of the flow arrow in the right panel - - looking
downstream.
18
This slide shows a more detailed accuracy assessment of the same stream
reach as in the previous slide. The center map is a difference image produced
by subtracting an EAARL-derived DEM from a DEM made by an intensive
ground survey. The largest lidar errors are along the banks. The bottom figure is
a long profile down the channel (along the dashed line in the central figure). In
the profile the black line is from actual ground survey points. The red and green
lines are along the same track, but taken from the ground survey and lidar DEMs
respectively.
For more accuracy information see:
McKean, J., Nagel, D., Tonina, D., Bailey, P., Wright, C.W., Bohn, C.,
Nayegandhi, A., 2009. Remote sensing of channels and riparian zones with a
narrow-beam aquatic-terrestrial lidar. Remote Sensing, 1, 1065-1096;
doi:10.3390/rs1041065
19
Comparison of field surveyed and lidar surveyed hydraulic geometry metrics.
20
Stream metrics that seem to be mappable using the EAARL sensor.
21
22
23
The next 3 figures are from a study of the effects of glaciation on the spatial
pattern of spawning habitat in the upper Middle Fork Salmon River, Idaho. This
project used a combination of floodplain and in-stream topography mapped by
the EAARL.
24
Figure shows valley post-glacial excavation that has proceeded upstream to the
step in the topography at Channel Distance 4km. Note changes in channel form
along this 10km valley.
25
Here the valley gradient has been removed (the data are “detrended). This
emphasizes the local inset erosion surfaces. Insets b and c show again the
bathymetric detail that is available anywhere in the channel.
26
Example of a basic channel metric that can be extracted from the data. It is easy
to change your definition of a pool and recalculate this. The data were collected
in 300m reaches along the channel – that can also easily be changed. Note the
uneven distribution of pools along this unconfined channel. This is an example of
the very objective measurement of habitat using digital data – as opposed to the
subjectivity of most field surveys. The more detailed lidar data are inset in a
standard 10m DEM.
27
Map of channel changes over a 3 year period. In common with all airborne
lidars, the EAARL maps elevations better than it does horizontal position (X,Y).
This means that the sensor maps bed elevation changes much better than it
does the planimetric location of a laterally-migrating stream bank.
In this example a sand bed channel moves dramatically over time and it appears
to be partially abandoning the upper flow path (in blue) and reoccupying an older
meander on the lower right portion of the figure. In contrast, the older meander
in the lower left is not reattached to the main flow.
28
Another example of change over a 3 year period. The circle and rectangle show
areas where deposition has occurred.
29
An example of using EAARL data to investigate a potentially serious
management issue using a 2D computational fluid dynamics model. Many
climate change models predict more rain during winter in mountainous areas
where historically the water year has been dominated by spring snowmelt and
mid-winter flows are very low. The threatened and endangered fall-spawning
Chinook salmon lay eggs in stream beds in late summer. Presently these eggs
incubate during the very low flow conditions over the winter and emerge just
prior to the snowmelt peak in the spring. A concern is that increased flows in
mid-winter could mobilize the bed and endanger the eggs.
30
We use the bathymetric lidar data to define the boundary conditions in a
computational fluid dynamics (CFD) flow model to analyze this question. First we
model the bed shear stress with the CFD model MDSWMS (USGS; McDonald et
al, 2005). We calibrate by water levels and discharges measured at a study site
in upper Bear Valley Cr., Middle Fork Salmon River, Idaho. We can model shear
stress as a function of discharge and the spatial resolution of our stress
predictions is about 1 sq meter.
31
Here is another example of predicted bed shear stresses at low flow and
bankfull flow conditions.
32
We compare the modeled shear stresses to a critical shear stress for initial
motion. In this example, we use the Andrews and Parker equation for sizedependent critical shear stress. The grain size grading curves are for the entire
study channel bed and for a subset of just the salmon spawning sites on riffles.
The upper right graph shows mobility curves for each grain size (in mm) as a
function of discharge. This shows the hypothetical bed areas that would be
mobile if the bed was made entirely of each particular grain size. But there is a
mix of grain sizes so we convolve the upper right mobility curves with the actual
grain size distributions for the whole bed and for the spawning site subset.
Assuming that if a given grain size is mobile, then all particles that are smaller
than that size are also mobile, you end up with the cumulative mobility curves in
the lower right figure.
Conclusion: in this low-gradient, unconfined stream with a strong glacial legacy,
the existing gravels are not very mobile, even at the highest possible flows
(bankfull). This is corroborated by field measurements of bedload movement at
90% of bankfull. So salmon are not particularly vulnerable to this dimension of
climate change in this landscape.
33
This instrument allows us to map long continuous segments of a channel
network – well beyond the spatial scope of other survey methods, albeit with
lower precision and accuracy than, for example, field surveys.
This then raises the question of if there are other useful ways to interpret these
continuous data, beyond traditional studies of channels at cross-section and
reach scales. I suggest the frequency domain may be a fruitful approach.
34
Here we see a 10km-long thalweg profile hand digitized in a raster made from
EAARL data. The subject stream is in a wide unconfined alluvial valley
throughout this length with essentially no outside contributing sources of water
or sediment. The red straight line is included for perspective. There are multiple
scales of nested topographic features. Remember that a very detailed field
survey might cover 200m of this 10km (20X the channel width). Would that be a
good representative sample? If you moved the 200m sample around, would you
get the same representation of this 10km of channel?
35
There are many ways to evaluate continuous data like a thalweg elevation profile
in the frequency domain. I use wavelets as it is a technique that quickly explores
patterns in space and scale at the same time. Above is a cartoon of how the
wavelet convolution method works.
1. A reference shape or mother wavelet is chosen – normally from a library of
such wavelets. For smoothly-varying continuous data, almost any of the
mother wavelets gives a similar answer. This reference wavelet is set at
some spatial scale and then moved along the signal being analyzed (the
thalweg profile in our case).
2. At each position the reference wavelet is convolved with the portion of the
data signal that it covers. This establishes an objective quantitative measure
of the similarity between the reference and the object signal at that scale and
location.
3. Then the spatial scale of the reference wavelet is changed (it is either
stretched or shrunk, but the shape is kept the same) and the process is
repeated. Thus one builds up a field of similarity coefficients as a function of
spatial scale and position.
4. The similarity coefficients can be squared to give a measure of “spectral
power”.
36
Here we have a fairly obvious change in stream morphology that illustrates one
use of the wavelet method. The raster (made from EAARL data) has been
detrended to remove the general valley slope. At the Morphologic Domain
Boundary in the inset figure the stream exits a broad unconfined alluvial valley
and enters a confined mountain canyon. Here it also changes from a
meandering pool-riffle morphology to more of a straight plane bed stream. The
former is very good salmon-spawning habitat while the latter is very poor habitat.
37
The upper figure is the detrended thalweg profile of the entire 22km of channel
shown in the previous slide. The lower graph shows the variation in spectral
power mapped by a 2nd order Gaussian reference wavelet that was 100m long.
There is an obvious change in power at the morphologic domain boundary. The
table presents spectral statistics of the two domains.
38
We have built an ArcGIS toolkit to automatically extract many common hydraulic
geometry measures from high resolution DEMs. The DEMs can be from any
source – not confined to a bathymetric lidar
This slide shows the basic work flow in the toolkit.
39
The RBT depends critically on an accurate definition of bankfull. There are a
variety of ways this stage can be defined.
1. The software attempts to do this morphologically using a calculation of the
ratio of wetted reach volume / wetted reach area. Starting from a dry stream
bed, this ratio initially rises with stage. If a well defined floodplain is present,
the ratio will drop when the stage is just high enough to inundate the
floodplain.
2. The user can also manually adjust the bankfull by “flooding” a detrended
raster as a horizontal water stage is raised through the topography and the
changing flooded area is shown in a map view (the lower right image in this
slide).
3. Field survey locations of bankfull can also be entered into the detrended
raster.
4. A flow model can predict bankfull
5. Bankfull can be digitized from airphotos.
40
A detrended raster made from EAARL data showing about 15km of unconfined
channel. The next image is a close-up of the area in the inset dashed green box.
41
The RBT allows the user to:
1. Use several methods to define where to place cross sections.
2. Examine the shape of an individual section and its associated hydraulic
geometry metadata.
The example graph is for the blue cross section in the map and the graph shows
the geomorphically-defined bankfull level and hydraulic geometry values
Metadata are stored in a dbf and available to compute geometry statistics from
large groups of cross sections or to export data to 1D flow models.
42
At the right is a close-up of the upper end of a channel segment in which 900
cross sections were mapped at 10m intervals down the channel. The cross
section explorer is used to look at any 1 or more sections. Here we are looking
at section 21 (the one in blue on the right side map). The explorer shows the
hydraulic geometry measures of this section – these data are also stored in an
extractable data base file. The geometry measures were taken using the water
table elevation shown in the figure – that was produced automatically, but can be
adjusted by the user to wherever you think bankfull is. The toolkit has several
ways of estimating bankfull.
43
The upper figure shows the thalweg profile extracted using all 900 cross
sections. The lower figure is a close-up view in the area where the gradient
changes.
44
Any of the cross section hydraulic geometry metrics can be quickly mapped to
the channel long profile. In this figure the variation with distance along the
channel of both gradient and sinuosity are shown. The channel lengths over
which these two variables are calculated is defined by the user (in this case it
was 200m).
45
Longitudinal variation in bankfull width (top) and width/depth (bottom).
46
Scope = spatial range over which you can measure/ spatial resolution of things
you can map
47
48
49
50