Using GIS and Aerial Photography to
Monitor Riparian Changes in the
Shannon Clemens
GIS for Water Resources
December 5, 2008
INTRODUCTION
The Virgin River, a tributary of the Colorado River,
is a desert stream in
In 2002, an original baseline study was conducted by
a UWRL graduate student (Thompson, 2003) for the
Then in 2007, two years after the flood, Institute
for Natural Systems Engineering at the UWRL delineated aquatic and riparian
habitat data at selected intensive sites in the
Figure 1. Map of
PROJECT
OBJECTIVES
The goal of this project is to address if changes in
riparian habitat distribution and/or composition due to the January 2005 flood
event can be quantified using GIS and aerial photography. The project will address if high
resolution aerial photography (with the aid of ground truthing data) can be
used to map the riparian habitat for the remainder of the Virgin River Basin
and which resolution would be best for this task along with the effectiveness
of this approach. In addition, a
long term riparian habitat mapping plan will be recommended. For all intents and purposes,
“pre” indicates the 2002 mapping study and “post”
indicates the 2007 mapping study.
METHODS
In the summer of 2002, a baseline basin wide
riparian and aquatic habitat study was performed for the Virgin River
Program. Riparian habitat data was
collected by field crews through ground based delineation and aerial
photography. Delineation of polygons
covered the entire width of the 100 year floodplain and were drawn onto
hardcopy laminated maps (D size plots).
Imagery was flown by Olympus
Aerial Surveys of St. George, UT in January 2002 with a pixel resolution of
0.18 m.
In order to prepare the data for comparison with the
2007 data, the 2002 data was clipped using a Clip tool with a heads up
digitized boundary that was created by comparing pre- and post-flood aerial
photography. The boundary was
created to encompass the area of change after the flood for this site and also
served as an equal area boundary to compare the 2007 data to. Since the 2002 data was not fully mapped
to the extent of the newly drawn boundary (originally it was mapped within the
100 year flood plain), some photo interpretation was used to assign the uncoded
areas a code. See Figure
2 for a map of the areas to be coded and Table 1 for a
list of 2002 vegetation codes. Once
the data was completely attributed, below are the mapping efforts from 2002
shown in Figure 3.
Figure 2. ArcToolbox feature
Clip demonstrated with “blank” areas from 2002 vegetation mapping.
Table 1. Vegetation codes and descriptions for
the 2002 riparian mapping study.
|
Vegetation Code |
Vegetation Description |
|
V00 |
River |
|
V01 |
Native vegetation only |
|
V02 |
Native vegetation with Tamarisk present |
|
V03 |
Native vegetation with Russian olive present |
|
V04 |
Native vegetation with Tamarisk and Russian olive
present |
|
V05 |
Tamarisk dominant or subdominant in the overstory
or understory |
|
V06 |
Russian olive dominant or subdominant in the
overstory or understory |
|
V07 |
Bareground |
|
V08 |
Russian olive Tamarisk and Locust dominant or
subdominant in the overstory or understory |
|
V09 |
Russian olive and Tamarisk dominant or
subdominant in the overstory or understory |
|
V10 |
Native vegetation dominant with other exotics
subdominant in the overstory or understory |
|
V11 |
Other exotics dominant |
Because
of the purpose and the scale of the original 2002 mapping study, vegetation
polygons were in most cases generalized, large homogeneous stands of dominant
vegetation type were outlined and a code given to the entire area.
Figure 3. 2002 riparian
delineation efforts after Clip feature applied.
In January 2007, field crews collected riparian and
aquatic habitat mapping to quantitatively assess the effects of the 2005
flood. Rather than mapping the
entire river basin as in 2002, ten intensive sites were picked as ground
truthing for the initial project.
At these intensive sites, data was collected through ground based
delineation and aerial photography.
Data was either collected on laminated hard copy maps (D size plots) or
on a tablet PC with GPS. Prior to
the 2007 mapping efforts, the Virgin River Program provided 0.3 m aerial
photography taken April 2006. The
field crews discovered the river had changed significantly by January 2007 so
this imagery was not applicable at every site, including “Below
Bloomington Gage.” Field
crews decided to download the August 2006 NAIP 1.0 m aerial imagery while in
the field to use for mapping on Below Bloomington Gage for a few other
sites. Since the pixel resolution
was coarse at 1.0 m for any close up mapping, the field crews aided their
mapping using a GPS enabled tablet PC to help them located their position in
the NAIP imagery while in the field and with delineation. Field crews would walk around the area
in question and use the GPS to delineate polygons. Still, this imagery was 5 months old,
but the latest available imagery at the time. At a zoomed out scale, the imagery seems
adequate to use and flood changes were evident, but to delineate new vegetation
growth it was not sufficient unless one happened to be standing at the site
with a GPS. See Figure 4. Figure 5 shows the
2007 riparian mapping results.
Figure 4. Left represents 0.3
m resolution imagery from April 2006 (
Figure 5. 2007 riparian
delineation efforts.
Each set of data, whether it came from hand drawn
polygons on laminated maps or from digitized polygons from a tablet PC, was
QA/QC and attributed. Once each set
of data was prepared and attributed, a “union” was applied. The union tool in ArcToolbox will
compute a geometric intersection of the 2002 and 2007 data, where its table
will contain the attributes of both data to be used for analysis. Figure 6 shows the
output from a “union” with the same shapefile attributed to show
areas of no change versus areas change of change.
Figure 6. The left polygon
shows the immediate result of a “union” while the right polygons
shows that same shapefile displayed with areas of no change versus areas of
change.
With query building within the union
shapefile’s attribute table, a general synopsis of which areas changed
and into what (river, bareground, and vegetation) can be mapped. See Figure 7 for a
spatial distribution of changed areas.
Figure 7. A spatial distribution of changed areas between 2002 and 2007.
RESULTS
Dramatic vegetation loss was calculated and
identified in the aerial imagery at the Below Bloomington Gage intensive
site. In 2002, a total of 44.1
acres of vegetation was mapped and in 2007, a total of 12.5 acres of vegetation
was mapped. This resulted in a loss
of 31.7 acres or 71.8% of the originally mapped vegetation. See Table 2.
Table 2. A calculation of the
difference in total vegetation acres mapped and a percent vegetation loss.
The change and composition in vegetation types were
calculated (Table 3). For example, in 2002 there were 39.97
acres of tamarisk (V07) mapped. In
2007, that tamarisk changed into 1.63 acres of river (V00); 0.72 acres of
native vegetation (cottonwoods and/or willows) with tamarisk present (V02);
29.30 acres of bareground (V07); and 0.23 acres of phragmites (V11). Of the original tamarisk stands (V05),
8.09 acres remained intact after the flooding. See Table 1
for a list of vegetation codes and descriptions.
Table 3. Changes in
vegetation, bareground and river from 2002 to 2007.
At first examination, it appears that new vegetation
has been added. However in 2002,
the two classes of vegetation identified were Tamarisk (dominant or
subdominant) and Tamarisk/Russian olive/Locust (dominant or subdominant). In 2007, field crews mapped four types
of vegetation: Tamarisk (dominant or subdominant), Native vegetation (willows
and/or cottonwoods), Native Vegetation (willows and/or cottonwoods) with
Tamarisk present, and Phragmites (widely distributed reeds). The two classes coded as
“native vegetation” or “native vegetation with
tamarisk” in 2007 were probably not due to the addition of new native
species (cottonwoods and/or willows).
The main reason was due to the mapping styles of 2007 and interpretation
of the vegetation (what was dominant and where boundaries lie) were different
(along with the crew members) than from 2002. The mapping style of 2002 was a very
large mesohabitat scale where as the 2007 mapping style was a bit more
detailed. This is how the
cottonwoods and/or willows as defined in the “Native vegetation”
polygons were picked up and were not new growth. Upon closer examination of the 2002 and
2007 aerial photography, it is the same vegetation stands.
The addition of phragmites may be new growth
however. This vegetation class at
this site was not mapped in 2002.
If the 2007 phragmites polygons were overlaid on 2002 imagery, large
tamarisk stands are present. It can
be assumed the phragmites did not exist in tamarisk stands in 2002 which were
away from the river.
The river and bareground calculations maybe
inaccurate due to the changes in flow (more bareground being present at low
flow and more river being present at high flow). Some allowances in river and bareground
calculations should be taken into account.
It appears interpretation of the results should be done in order to
understand fully what the calculations represent.
Changes in riparian habitat were evident in the
available aerial photography.
Available imagery ranged from 0.25 m to 1.0 m resolution and from years
2002 to 2006 for the Below Bloomington Gage site (see Figure 8). At a zoomed out scaled, the 1.0 m
imagery could be used for overall very general vegetation mapping, but however
it would be difficult to detect any new growth or composition as seen in Figure 9. In Figure 9, the left image represents a 0.25 m pixel resolution
and on the right is the same location at 1.0 m resolution. Even though the images were taken
roughly one month apart, it is evident how new growth is hard to detect. If basin wide riparian
mapping/delineation were continued through GIS and aerial photography, 0.25-0.3
m pixel imagery (or finer) would be recommended so that new growth can be
detected in addition to vegetation composition.
Figure 8. Four snapshots in
time at the Below
Figure 9. Left imager
represents 0.25 m resolution imagery compared to 1.0 m resolution imagery. Imagery was taken roughly one month
apart.
Within this particular study there were some issues
that should be addressed if a future long term riparian habitat monitoring
program should be implemented. The
first issue to address is that the mapping styles should be consistent
throughout the study and future studies.
The 2007 riparian mapping was more refined with the delineation and
interpretation than the 2002 mapping.
Another issue to address was the use of outdated aerial imagery which
made it difficult to perform true ground based delineations unless the aid of
GPS was used.
The clipper polygon served originally as a rough
boundary for the 2007 field mapping crews.
Later it was discovered that this clipper was required in creating a
cookie cutter in order to compare the 2002 and 2007 data accurately. In some cases the 2002 data did not
fully lie within the clipper which led to some photo interpretation of 2002
imagery to assign vegetation codes.
Lastly,
since the entire area inside the clipper needed to be accounted for (otherwise
false bareground “gain” would be calculated), the river needed an
area too. However, this also
depended on the imagery used and the time and flow at that date. Since the aerial imagery did not reflect
the exact area of the river at the time of mapping, some adjustment has to be
considered to consider the river and bareground calculations. Though these issues existed, with
modifications the above methodology can still be applicable.
SUMMARY
AND RECOMMENDATIONS
With a few modifications, such as a universal
mapping system/minimal mapping unit, the above procedure can be an effective approach
in quantifying riparian changes and/or composition. With a more established mapping
protocol, variations among field crews and mapping seasons will not alter
mapping results and calculations. A
minimum mapping unit should be established so that mapping styles are similar
throughout the monitoring program.
With the development of UAVs (unmanned aerial
vehicles), it is highly recommended that this form of aerial photo acquisition
be used for each mapping effort if possible. UAVs are inexpensive, produce high
resolution imagery (as fine as 3 cm), and have an immediate turn around time so
that field crews are working with the most up to date imagery (day old
imagery). Otherwise imagery no
coarser than 0.3 m should be used for aerial photo based delineation.
For basin wide vegetation mapping using aerial
photography, GIS and up to date imagery can be used with photo interpretation
to delineate homogenous vegetation stands.
Many papers can be read on vegetation mapping from aerial photography
and the steps from which to conduct approach (which are too extensive for this
report). Existing ground truthing
and any future ground truthing data can be used to QA/QC existing digitized
polygons.
REFERENCES
Application
of Aerial Photography for Detection and Quantification of Changes in the
NAIP 2006
Imagery. National Agricultural
Imagery Program.
Olympus
Aerial Surveys,
Thompson, J.
W. (2003). A methodology for assessing aquatic and riparian habitat
quality at the watershed scale. Unpublished M.S. Thesis, Department
of Civil and Environmental Engineering,
Wilkowske,
C. D.; Kenney, T. A.;