International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
anisotropic field. The anisotropy being a function of sediment
preferred orientation arising from water flow direction.
There are other problems with using a semi-variogram
approach, as indicated by Atkinson & Lewis (2000).
Homogenous regions of varying texture must be large enough to
allow computation of the semi-variogram with a reasonable
number of lags. In many cases, areas of interest are too small
relative to the spatial resolution of the imagery. Computation of
the semi-variogram is also intensive, particularly if the full
variogram surface is generated. A consequent problem is how
this surface is represented and stored to a precision sufficient
for the classification process. For each desired profile direction,
à minimum of three parameters is necessary and these results
suggest that storing both downstream and cross-stream
parameters would be essential.
Despite these drawbacks it is believed that additional texture
layers represented by either variance or semi-variograms could
provide additional information necessary for bed-classification
and the derivation of grain-scale data from aerial imagery.
5. CONCLUSION
This project has demonstrated that it is possible to derive a five-
fold bed classification to a true accuracy of 4996 using just the
three original colour bands and 1:5.000 scale photography. It
has to be recognised that this was achieved using a “per-pixel”
classification within non-homogeneous bed material. If material
had been homogeneous, accuracies would have been higher. It
was found necessary to create an additional “texture” layer
using a 3x3 -variance convolution filter. It was found that
classification accuracy was not affected greatly by varying the
photo-scale, with 1:10,000 scale imagery also yielding a valid
classification. Significantly, it was found that simple grey-scale
imagery could yield useful classifications, provided a texture
layer was generated and used. Alternative methods of deriving
a texture layer were investigated. Autocorrelation yielded only a
modest improvement to 51%. The semi-variogram could
provide the basis for useful measures of bed texture and may
improve the classification accuracies further but efficient
storage of semi-variogram data is required.
Once a classified image had been generated from the aerial
imagery it was simple to derive a percentage sand map.
Comparison between this and traditional ground-based methods
requiring intensive fieldwork, highlighted the potential for
significant savings in time and effort if aerial imagery is
acquired. More detailed examination of the optimal photo-scale
for identifying sand patches remains an area for further
development.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the financial support
provided by the Natural Science and Engineering Research
Council of Canada (NSERC) for fieldwork support and
helicopter hire.
REFERENCES
Atkinson, P.M. and Lewis, P. 2000. Geostatistical
classification for remote sensing: an introduction. Computers
and Geosciences, 26, pp. 361-371.
Chandler, J.H., Shiono, K., Rameshwaren, P. and Lane, S.N.
2001. Measuring flume surfaces for hydraulics research using a
Kodak DCS460, Photogrammetric Record, 17(97), pp. 39-61.
Chica-Olmo, M. and Abarca-Hernandez, F., 2000. Computing
geostatistical image texture for remotely sensed data
classification, Computers and Geosciences, 26, pp. 373-383.
Church, 2004. Fraser River Gravel Reach Studies
http//www.geog.ubc.ca/fraserri ver/ (accessed: 4 March, 2004)
Fripp, J. B., and Diplas, P. 1993. Surface sampling in gravel
streams. American Society of Civil Engineers, Journal of
Hydraulic Engineering, 119, pp. 473-490.
Gilvear D.J., Waters, T.M., and Milner, A.M., 1995. Image
analysis of aerial photography to quantify changes in channel
morphology and instream habitat following place mining in
interior Alaska. Freshwater Biology. 34, pp.389-398.
Gilvear DJ, 1999. Fluvial geomorphology and river
engineering: future poles utilizing a fluvial hydrosystems
framework. Geomorphology, 31 (1-4), pp- 229-245.
Gilvear DJ, 2001. Discussion of: Lane, S.N., 2001. (The
measurement of gravel-bed river morphology, in “Gravel Bed
Rivers V", Mosley, M.P. (ed). NZ Hydrology Society,
Wellington, pp. 323-324.
ISO, 1992. Liquid flow measurement in open channels -
sampling and analysis of gravel-bed material International
Organisation of Standardization, Publication 9195, pp. 9.
Lane, S.N., 2001. The measurement of gravel-bed river
morphology, in «Gravel Bed Rivers V", Mosley, M..P (ed).
NZ Hydrology Society, Wellington, pp. 323-324.
Marks K, Bates P, 2000. Integration of high-resolution
topographic data with floodplain flow models. Hydrological
Processes, 14 (11-12), pp. 2109-2122.
Rubin, DM, (2004). A simple auto-correlation algorithm for
determining grain size from digital images of sediment. Journal
of Sedimentary Research, 74 (1) pp. 160-165.
Thomson AG, Fuller RM, Eastwood JA, 1998. Supervised
versus unsupervised methods for classification of coasts and
river corridors from airborne remote sensing. [International
Journal of Remote Sensing, 19(17), pp. 3423-3431.
Wolcott, J. F., and Church, M. 1991. Strategies for sampling
spatially heterogeneous phenomena: the example of river
gravels. Journal of Sedimentary Petrology, 61, pp. 534-543.
Wolman MG, 1954. A method of sampling coarse river-bed
material. American Geophysical Union Transactions, 35, pp.
951-956.
Wright A, Marcus WA, Aspinall R, 2000. Evaluation of
multispectral, fine scale digital imagery as a tool for mapping
stream morphology. Geomorphology, 33(1-2), pp. 107-120.
1084
KI
AE
His
doi
the
we
Site
33
15€
app
diff
info
clot
pres
Uni
A :
char
com
£00(
garri
exan
high
signi
Med
To d
role |
for c
resol
very
the v
spect
gradi
Medi
prope
most
propo
of ma
Norm
poor |
Meas
by vir
* Co
