CHAPTER 2
Changes In The Area And Condition Of Samphire Marshes With Time
R. L. Glasson, H. T. Kobryn and R. D. Segal
2.1 Introduction
Aerial photography is especially suited to the study of vegetation, water resources
and shoreline mapping (A.S.O.P, 1968). Air photographs
provide a perspective of the earth’s geographical features that are generally
readily understood. Air photographs also provide a plan view that can be spatially
compared to the knowledge an individual may have about a similar area. For those
reasons aerial photographs frequently provide working documents for planners and
managers. They are however, limited in spatial
accuracy because the images suffer geometric distortions, particularly near
photograph margins or when terrain varies in height. Also, a single aerial photograph
rarely covers an entire study area. There are manual and computer assisted techniques
of joining air photographs and also eliminating the geometric distortions within
and between individual photographs. An assembly of aerial photographs is called
a photo mosaic. Photomosaics can be controlled
or uncontrolled, the former having the geometric distortions removed.
The amount of geometric distortions in an aerial photograph depends on many
factors, including the physical optics of the camera and the orientation of
the camera at the instant of exposure. Where the optical axis of the camera
is near vertical (<3° from vertical), then the photograph is accepted
as vertical. The point where the optical axis of the camera meets the earths
surface is the principal point of the photograph. Vertical photographs are the
most common type of metric aerial photograph, or one that is used to derive
information about spatial measurements of
geographical features. Geometric distortion of aerial photographs increases
outwards from the principal point towards the margin. The distortions are increased
if photography is acquired at low altitude, or with cameras using wide angle
(>70°) and super-wide angle (>100°) lenses. Gross distortions of scale
occur on individual photographs when terrain slope changes suddenly, as when
a scarp, cliff or portion of a mountain is included in the photograph (Maling,
1989). Some of the distortions in vertical air photography can be avoided
by using only central portions of each photograph.
Aerial photography is acquired in flight strips called runs. Each photograph
within a run overlaps the previous one by about 60% (endlap) to enable sufficient
imagery to be acquired that is relatively distortion free, and to enable stereoscopic
viewing of the imagery to aid in feature identification and interpretation.
Each run is also overlapped to the adjacent run (sidelap), although usually
by only 25-30% (Figure 2.1).
As an aid for interpretation, stereo pairs of photographs are viewed
under a stereoscope. Stereo pairs share conjugate principal points, each
principal point of each photograph is being visible on both photographs.
The measured distance between the principal points is known as the airbase;
which is a function of the ground speed of the aircraft and the interval
between exposures. The stereoscope enables our eyes to view photographs
from the same relative position as the airbase. This means that our eyes
have the same optical separation as the air base, and stereoscopic depth
of vision is greatly enhanced.
Interpretation of air photography is conducted in two stages. In the
first stage the elements of the image are identified by attributes such
as shape, size, spacing, shadow or silhouette, tone or colour , texture
or association. The photographic appearance of these attributes can be
affected by many factors including the amount of illumination and the reflectivity
of the surface, as well as climatic factors and physical properties of
the ground cover. These can quite often change the appearance of objects
to resemble other features. For example shallow water stained brown with
tannin, and inundating ground, can appear to be quite deep and extensive
when using black and white photography.
The second phase consists of interpreting the factors in relation to
the context for which the information is required. Interpretation may be
carried out on individual photographs, in stereo pairs, on a composite
photomosaic, or involving all of these procedures.
Physical assembly of a photomosaic by cutting and pasting is an exacting process
in which failure can be very costly, as individual colour photographs cost about
$20.00 and a mosaic may include some 50 photographs.
By scanning photographs and using computer software, digital photo mosaics
can be constructed and repeated without fear of costly failure. This investigation
used the digital equivalent of a photo mosaic
using computer software and scanned photographic imagery of the Peel-Harvey
System.
A broad study into the samphire marshes of the Peel-Harvey System was initiated
by the Peel Preservation Group to determine the extent and importance of these
areas to the ecology of the Estuary. As part
of this study a temporal perspective was required. No accurate mapping of these
features existed nor had data been previously recorded about their locations
or extent. To provide this historical perspective, existing archival air photography
was combined with modern digital techniques to map the extent of samphire and
temporal changes in the marsh.
A Peel-Harvey photo mosaic was prepared using
ERDAS Imagine
software on a Sun Sparc station IPX as part of
the wider study of the samphire marshes. Air photographs covering the period
1957 to the present were used to determine temporal changes in samphire cover.
The 1994 series of air photographs was provided by Department of Transport.
2.2 Method
The selection of images was based on approximately 10 yearly intervals, and on
availability. Details of the acquired air photograph runs are presented in Appendix
1. The photographs selected had to lie within the boundaries of the Peel Harvey
Estuary including Goegrup and Black Lakes. The 1957 photo set, however, did not
include the Harvey River delta region at the southern end of the Harvey estuary
as this was beyond the limit of existing photography. Portions of the Black Lakes
region east of the Peel Inlet were excluded from the 1994 set, as this was also
on the limit of the photography for that run.
A detailed technical description of the digital photo mosaic
process is contained in Appendix 1. The air photographs
for each date were scanned and then digitally joined, removing as much distortion
as possible and orientating the photographs to conform with mapping conventions.
The photographs were then interpreted to determine the extent of the samphire,
which was then mapped over the digital images, enabling computer determination
of the extent of samphire. As well as the completed mosaic
of the entire area, smaller images were extracted for each date to enable comparison
of special interest areas. The final mosaic
images were compiled from some 80 sub-images, shown in Appendix
2. A flow diagram of the entire process is summarised in
Appendix 3. A table showing file handling procedures is contained in Appendix
4.
Air photograph interpretation was carried out for each of the dates on the air
photographs, using the keys contained in Appendix 5. Each
photograph was interpreted independently by two interpreters, using stereoscopes
to allow magnification and observation of the imagery in three dimensions. This
enabled the use of terrain to assist in the identification of samphire marsh.
The subsets of special interest areas was focused on areas of conservation
reserves. The geographical boundaries of each area are given in Table 2.1.
Table 2.1. Special Interest Area Boundaries (UTM
Zone 50 (m)). Value of X represent eastings, whereas values of Y indicate northings.
| COORDINATE/AREA |
UPPER-LEFT X |
UPPER-LEFT Y |
LOWER-RIGHT X |
LOWER-RIGHT Y |
|
CREERY
|
378727
|
6398715
|
381879
|
6395165
|
|
LAKES
|
384272
|
6402524
|
388196
|
6395780
|
|
AUSTIN
|
381234
|
6390770
|
385091
|
6386599
|
|
ROBERTS
|
377316
|
6388819
|
381174
|
6384649
|
|
HARVEY
|
377121
|
6376654
|
381174
|
6370968
|
|
STUDY AREA
|
370900
|
6402500
|
388200
|
6370900
|
As field checking of classification accuracy is impossible for historical
photography, only the 1994 photograph set was used to test for the accuracy
of interpretation. One hundred sites were selected by the intersection
of a grid overlay on the images, with areas that could reasonably be expected
to contain samphire. Fifty sites of samphire cover and fifty sites of non
samphire cover were used to determine by field checks for errors in the
process of the interpretation.
2.3 Results
The
mosaiced images for each date, including areas of samphire are shown in Figures
2.2, 2.3, 2.4,
2.5, 2.6. The red polygons indicate
areas of samphire cover. The total area of samphire for each date and each area
of interest are shown in Table 2.2. Red rectangles outlining boundaries of the
special interest areas are shown in Figure 2.7. Temporal
samphire changes for each special interest area are shown in Figures 2.8,
2.9, 2.10,
2.11, 2.12.
Table 2.2. Areas of Samphire (ha) during the period 1957 - 1994 for areas
of interest, and the total over the entire Peel-Harvey study area.
| AREA/DATE |
1957 |
1965 |
1977 |
1986 |
1994 |
|
AUSTIN
|
35.286
|
30.308
|
23.395
|
31.851
|
37.758
|
|
CREERY
|
179.115
|
170.365
|
162.994
|
132.984
|
140.156
|
|
HARVEY
|
5.453*
|
131.099
|
87.652
|
106.995
|
144.783
|
|
LAKES
|
48.404
|
120.791
|
110.660
|
95.165
|
45.289*
|
|
ROBERTS
|
155.016
|
114.836
|
87.849
|
67.370
|
80.749
|
|
OTHER
|
252.900 •
|
432.771
|
271.783
|
190.327
|
181.405 •
|
|
TOTAL
|
676.174 •
|
1000.17
|
744.333
|
624.692
|
630.140 •
|
* - incomplete air photo coverage resulting in underestimate for area.
• - area underestimates (*) affect summed total estimates.
In attempting to describe changes that have occurred in saltmarsh coverage,
two approaches can be made. One is to describe changes in a purely quantitative
sense which concentrates on the magnitude of a particular change. The other
is to describe the changes in qualitative terms which concentrates not
only on the changes to the samphire itself but also how these changes affect
the wider environment
2.3.1 Quantitative Changes
The areas of samphire are based on the number of
pixel’s classified as samphire in accordance with the image interpretation,
in which the areas are based on pixel units of
20.25 square metres. A decline in the area of samphire marsh took place from 1965
onwards. The greatest percentage loss of saltmarsh cover between 1965 and 1994
(approx.58% or 251 ha) occurred outside reserve areas. The greatest percentage
loss from an area of interest was in the Lakes area (approx.63% or 75 ha) and
in Roberts Bay (approx.48% or 75 ha). The physical distribution of samphire cover
is dynamic and changes for all dates which is consistent with current understanding
of the dynamic nature of marsh distribution. Total samphire loss was 36.9% (370
ha) between 1965 and 1994, the greater part of this loss occurring between 1965
and 1986 (376 ha or 37% of the cover of 1965).
The general trend for the whole study area is a significant loss between
1965 and 1986 consisting of a rapid decline between 1965 and 1977 and a
slower rate of loss between 1977 and 1986. Between 1986 and 1996 the loss
was arrested, and there was a insignificant increase in overall cover (0.8%).
Roberts, Lakes, and Creery interest areas showed increased rates of loss
between 1977 and 1986. Since 1986 four of the special interest areas have
shown increases in cover from previous dates; Austin 18%; Creery 5%; Harvey
35%; and Roberts 20%. Percentage changes are based on the percentage change
from the previous period for that specific area, rather than the total
area of all marsh.
Some areas behaved differently from the general trends; for instance
the Harvey Delta and Austin Bay showed a steady increase in area of samphire
since 1977, while all other areas showed decline. The Lakes area showed
a marked decline in 1986-1994 period while all other areas of interest
showed increased cover. The Lakes area was the only special interest area
to show consistent loss since 1965 and its rate of loss is increasing.
Overall the trend is seen to be one of rapid decline from 1965 to 1986
and static to 1994.
2.3.2 Qualitative changes
In Creery wetlands the cover continuity of the samphire marsh declined since 1957.
This largest single portion of samphire in the study area showed increasing damage
from human contact. Increasing numbers of vehicle tracks divided largely continuous
samphire marsh. The 1994 image shows a highly disturbed area broken by many tracks.
The construction of Boundary Island (south of Creery wetlands) provided an area
for samphire colonisation, and samphire invasion is observed in the 1994 image
Increasing human disturbance of this wetland will lead to further decline in ecological
quality. (See Chapter 6).
The Lakes area showed an increasing rate of decline over the period
of the study. The 1994 results for this area are inconclusive due to the
lack of photography, but the trend of loss is clear from the middle reaches
of the Black lakes chain, particularly in the area of the Goegrup Lake
entrance. The cause of decline is not clear from the air photography data.
The Harvey Delta showed increased area of samphire in the south-eastern portion,
with smaller separate patches of samphire increasing and consolidating their cover.
This may be due to different tidal regimes and successional invasion onto the
increasingly exposed soil (See Chapter 6).
The Roberts Bay area is very dynamic, areas in the north eastern sector
showing more continuous cover, while areas in the southern portion showed
evidence of rapid and extensive change due to changing land use, especially
on private land.
Austin Bay samphire showed consolidation along the shoreline, with
smaller patches in the north and eastern sector growing in embayments and
increased cover.
Overall the ecological quality of the samphire marsh areas is typically
dynamic, individual areas showing evidence of ephemeral changes and succession.
2.3.3 Accuracy Assessment
The interpretation identification error results obtained from comparison
of field sites and the interpretation are included in Table 2.3.
Table 2.3 Estimation of identification errors of samphire.
| INTERPRETATION/ACTUAL |
SAMPHIRE |
NO SAMPHIRE |
| SAMPHIRE |
41 (82%) |
9 (18%) |
| NO SAMPHIRE |
12 (24%) |
38 (76%) |
Samphire was correctly identified with an accuracy of more than 80%, with
an 18% chance of overlooking its presence. Where samphire was interpreted
not to be present, but was, occurred in 24% of cases. The correct identification
of absence of samphire occurred in 76% of cases, and so overall it can
be concluded that the interpretations were conservative in estimates and
that actual presence of samphire is likely to be higher than reported here.
2.4 Discussion
Where areas of samphire have consolidated cover with time, it is assumed that
the ecological niche of the samphire is assured. Conversely where large continuous
areas are fragmented then continued presence of the samphire communities and its
dependant species may be threatened. In all areas, the distribution of samphire
is changing even though total areas may have remained similar. This is accounted
for in the ephemeral nature of some of the species occupying the marsh and seasonal
differences of biomass of perennial species
(See chapter 3 & 6).
Differing water levels between the dates of photography influence the
identification of samphire. The 1957 photography was taken at a period
of high water, and this influenced the extent of cover estimation. This
was particularly evident for the Lakes region. Some areas were clearly
flooded in 1957 and were not classified as samphire. The same areas were
exposed and identified as samphire in later photo sets. The 1965, 1977,
1986 and 1994 photography was taken during low water levels. It is probable
that high water levels tend to mask areas of samphire leading to underestimates
of their extent and the converse is also true. This may partially explain
the increase in area seen in 1965, and also the relative increase in the
rate of loss for 1986, but it is not an explanation for the overall loss
trend as this was consistent, with 1965, 1977 and 1994 all being years
of low water levels.
The Lakes area for 1957 showed high water levels, and the 1957 estimation is underestimating
the samphire for this area. Since 1965 the total samphire for this area declined,
but the subset images show that the loss was centred on the middle reach of the
Black Lakes area (Figure 2.11). At this point a stream
flows in from the north east, and it is possible that this stream has in the recent
past either provided a increased flow of fresh water, altering the salinity regime,
or is a source of excess nutrients to a competitive species, upsetting competitive
relationships causing a decline in the prevalence of samphire.
The altered tidal and water level regimes of the Estuary since the permanent breaching
of the Mandurah Channel (1977-1986) may be the reason for the rapid decrease in
the area of samphire, and it would be expected that this loss should be arrested
when vegetational succession has stabilised. This also appears to be the case
for the recent small losses, in the 1986-1994 period. Further disruption to vegetation
cover can be expected with the recent opening of the Dawesville channel, which
should lead to further successional changes in the vegetation cover as a result
of water level changes. Vegetation loss due to changing hydrology and other factors
has already observed by catchment management authorities (George
and Bradby, 1993).
Increased opportunities for saltmarsh invasion have occurred with the
construction of Boundary Island south of the Creery wetlands. This accounts
for the increase shown in the Creery area in 1986 -1994.
Another influence on the estimation of samphire is the cyclic water
level changes experienced over the study period. During 1965, 1977 and
1994 the rainfall for the preceding months prior to the air photography
was below average. For 1986 and 1957 the rainfall was well above average.
Prior to 1977, rainfall would have had an effect on the water levels of
the Estuary, as the only channel was prone to closure. This explains the
low coverage seen in the Lakes area due to the elevated water levels in
1957 and the high levels in 1965 where the water level is low. This indicates
that the levels of samphire recorded in 1965 and 1977 are maximised as
the water levels would favour the interpretation. This implies that 1977
provides a good estimate of the total samphire both due to the favoured
water levels and the infra-red photography.
Even if adjustment is made for the lack of photographic cover over the Harvey
delta, and it is assumed that the area in the Harvey delta remained constant between
1965 and 1977 then there has been a steady decline in the total area of samphire
since 1957. The overall loss is worrying but the parallel decline in quality of
remaining areas is also of great concern. This is particularly so in the Creery
wetland area which presents one of the largest contiguous areas in the whole system.
Its present accessibility (and therefore threatened position) also means it’s
future planning demands the highest priority for management consideration (See
chapter 7).
The scanning and initial photo mosaic of the
orthophotographs has shown that the spatial
accuracy of the orthophotographs is questionable. Enquires of
DOLA reveal that a spatial accuracy
of 12.5 m is obtained in the initial modelling of the orthophotographs. Measured
mismatching of the edges of the orthophotographs revealed inconsistencies in both
X and Y directions. This was initially thought to be the result of scanning errors
introduced by the physical method used for orthophoto
scanning. Consequently the images were rescanned at a higher resolution, and at
the same time a comparison was made with a scan of a calibration image supplied
to the scanning agency by DOLA.
The scan of the calibration image revealed a measured error of 4 mm over a
diagonal distance of some 1200 mm. Measured mismatch errors between scan overlap
of adjacent orthophotographs revealed distance discrepancies of up to 60 m.
This discrepancy was not distributed evenly across the length of the orthophotograph
but was randomly distributed along the adjoining edges. This error would result
from the original orthophoto image production, and so the 12.5 m spatial
accuracy for the production of orthophotographs by DOLA
can be considered the goal rather than the achieved result. The displacement
of the same feature can be measured where the orthophoto image files overlap.
The maximum measured displacement was 60 m. The maximum measured error of 60
m is not considered to be the achieved accuracy of the overall orthophoto, as
the true geographical location of displaced GCP’s
could only be measured in the field after considerable effort. It was assumed
that a spatial accuracy of one half
of this distance (by assuming each GCP is incorrect
and therefore both have to be moved to merge) or 30 m should be used as the
overall accuracy.
The scanned aerial photographs were resampled to conform with the orthophotos.
This process achieved a better than 0.6 RMS
error in pixel relocation. Scanned photos therefore
are considered to be within 1.2 m accuracy of the orthophoto image. In conclusion,
a spatial accuracy of (±)
31 m can be assumed for the photomosaic image.
Different scales of photography contributed to some problems in photo
interpretation. The 1965 set at a 1:40000 scale proved particularly difficult.
This was also due to this set being black and white photography with relatively
poor contrast. Some small areas of samphire were very difficult to resolve
at that scale. Prior to 1977 black and white photography was the standard
and hence was the only available option. Misinterpretation is more likely
to occur with the black and white photography due to the inherent difficulty
of vegetation discrimination based on colour or tone. Conversely the infra
red photography of 1977 enabled a more precise estimation of vegetation
cover and type. Comparison of samphire cover over the differing dates is
therefore constrained by the limitations of the differing photography used
in the study.
In the 1994 photography, low sun angle resulted in a large degree of shadow
within areas of vegetation, which produced greatly contrasted images. This was
to some degree a trade off for water penetration which was very good, and which
reduced the specular reflection
off water. In 1965 photography on the other hand, solar angle was high and large
areas of solar reflection are apparent (Figure 2.3).
The air photographs were taken with a 25% edge overlap and a 60% forward overlap.
The 25% overlap was insufficient for edge matching of raster
files where the runs are not north/south or east/west. During georeferencing
images are geometrically altered to conform with mapping conventions (that is
north is the top of the page). The insufficient overlap results from the image
being resampled at an angle after georeferencing,
which produces edge triangles of null data as
shown in Figure 2.13. To
obtain image files without null data values
requires the production of two sub-images for each air photograph. This was
required because of a problem in the software, which does not allow elimination
of null file values from the edge overlap reduction. The problem is being investigated
by software supplier, ESRI.
Contrast matching of the final image was attempted but resulted in an as yet
unidentified error with the software. This resulted in the corruption of the
file statistics of the 80 images of the final mosaic.
The file statistics were individually recalculated for the entire set and the
final images were produced without contrast matching. Contrast matching was
achieved to some degree during the cubic
convolution resampling process of the smaller 8
Mb files so these images are aesthetically the best. The greatest detail
resolution (about 3 m on the ground) is maintained in the approx. 200 Mb
files.
2.5 Recommendations/Conclusions
Samphire areas within the Peel-Harvey estuary are declining both in quality
and quantity. The greatest areal loss has occurred outside reserve areas.
Reserve areas and those of special interest are showing evidence of decline
in quality of samphire cover. Decline in the quality of samphire can further
accelerate loss and degradation of vegetation and animal species. The loss
of samphire from the Creery wetlands presents a urgent case for management
consideration due to the proximity to urban development and development
pressures in general and also because this area represents one of the largest
remaining contiguous areas of samphire. This should not be interpreted
that other areas are less deserving of management consideration but that
Creery wetland represents unique opportunities for preservation and conservation
and its current land tenure requires urgent management consideration. Its
conservation and preservation should be considered a high priority.
If serious spatial mapping considerations
are part of the photo mosaic process then
the Orthomax component of Imagine should be used to increase the rectification
capacity of the air photograph images. This will allow a greater portion of
each air photograph to be included in the final mosaic
resulting in decreased file space requirements and time.
Any spatial accuracy of the final
product will depend to a large degree on the spatial
accuracy of the georeferenced
database. Orthophotographs provide a reasonably accurate and cost effective
method of providing this spatial information
providing that the error associated with the individual orthophoto is known.
That is the error inherent within each orthophoto and between orthophotographs
when more than one is used. Where spatial accuracy
is to be greater than that of the available orthophotographs then other methods
should be investigated. This requirement may be negated by the procurement of
orthophotographs at a larger scale than the final scale for study area where
the study area size and file space requirements permit this and were possible
procurement of orthophotographs is in the required digital format.
The contrast of any photography used is important in the aesthetic value of
the final composite mosaic. This is most apparent
where imagery which is used was flown for specific purposes such as high water
penetration and the low incident solar angle results in hot spots or large shadow
effects in areas of vegetation. Where large areas of the image are water and
specular effects are apparent
then this can detract greatly from the final aesthetic value of the mosaic.
This may be avoided where large overlaps are available and cost is not determinant
in the procurement of photography so that all photographs may be used rather
than every second photograph.
Differing colour balance between individual photographs may occur if
all photographs of the study area are not purchased at once. Where this
is not possible and colour balance is a problem, provision of existing
photography to the film developer can enable correct matching.
