492
C. Johnston and A. Davison
not have enough suitable broad categories in which to fit observations where Phase II
level information (Hiscock, 1996) on a site was not available. Difficulties were also
encountered in relating infaunal data from grab samples to epifaunal data from videos
to fit biotope descriptions in the biotope manual.
Interpolation of Acoustic Data
The acoustic track data points obtained as described above (Fig.s 2 and 4) were
converted to a continuous polygon coverage of the acoustic properties of the sea bed by
interpolating between adjacent data points to calculate values for intermediate areas
within Surfer for Windows™. Standard geo-statistical procedures were employed for
the interpolations; a review of geo-statistics by Rossi et al. (1992) suggested that the
procedure kriging was most suited to random data points (Davies et al. 1997). Surfer
for Windows™ provides a kriging algorithm to reduce the track data to a rectangular
grid of data points for the survey area. Display of the interpolated results as a grid,
rather than smoothing the data to provide hard boundaries, which in these areas do not
usually exist, subliminally reflects the resolution of the predicted data. The grid size,
in this case of 50 m by 50 m, should be selected as appropriate for the acoustic track
spacing achieved during the survey (which was between 100 m and 500 m in these
cases). A larger grid should be selected where track spacing is greater, to indicate the
lesser resolution of the track data. The resulting map is interpreted as the topographic
and physical habitat map.
Matching of Acoustic and Ground Validation Data
The map of acoustic properties of the seabed was analysed with biotope data from
ground validation sampling within MapInfo™ GIS, to produce a predictive map of
biotope distribution. Part of this analysis was to derive an acoustic signature for each
recorded biotope or group of biotopes. The MapInfo™ buffer capability was used to
create a 50 metre buffer around each ground validation site or line. Fifty metres was
selected as an appropriate size buffer, to reflect the accuracy of position fixing during
the ground truthing. Each buffer was then coded according to the biotope recorded at
that ground validation site. Within MapInfo™, acoustic data points falling within each
buffer were then captured, and assimilated into a table. In the few cases where more
than one biotope was distinguished from one video tow, the buffer around the video
tow line was split logically into sections representing each biotope. Minimum,
maximum, mean and variance values of the acoustic variables E1, E2 and depth (for
RoxAnn™ data) or Q1, Q2, Q3 and depth (for QTC™ data) were generated within
MapInfo™ from the captured acoustic data for each buffer, to form an acoustic
signature for each biotope or group of biotopes.
Using the acoustic signatures for each biotope group to construct Boolean
queries, areas which matched each signature were selected in turn and shaded
according to the corresponding biotope. An example of a query would be “select all
areas where E1 is between 0 and 0.5, and E2 is between 0.15 and 0.75, and depth is
C. Johnston and A. Davison
not have enough suitable broad categories in which to fit observations where Phase II
level information (Hiscock, 1996) on a site was not available. Difficulties were also
encountered in relating infaunal data from grab samples to epifaunal data from videos
to fit biotope descriptions in the biotope manual.
Interpolation of Acoustic Data
The acoustic track data points obtained as described above (Fig.s 2 and 4) were
converted to a continuous polygon coverage of the acoustic properties of the sea bed by
interpolating between adjacent data points to calculate values for intermediate areas
within Surfer for Windows™. Standard geo-statistical procedures were employed for
the interpolations; a review of geo-statistics by Rossi et al. (1992) suggested that the
procedure kriging was most suited to random data points (Davies et al. 1997). Surfer
for Windows™ provides a kriging algorithm to reduce the track data to a rectangular
grid of data points for the survey area. Display of the interpolated results as a grid,
rather than smoothing the data to provide hard boundaries, which in these areas do not
usually exist, subliminally reflects the resolution of the predicted data. The grid size,
in this case of 50 m by 50 m, should be selected as appropriate for the acoustic track
spacing achieved during the survey (which was between 100 m and 500 m in these
cases). A larger grid should be selected where track spacing is greater, to indicate the
lesser resolution of the track data. The resulting map is interpreted as the topographic
and physical habitat map.
Matching of Acoustic and Ground Validation Data
The map of acoustic properties of the seabed was analysed with biotope data from
ground validation sampling within MapInfo™ GIS, to produce a predictive map of
biotope distribution. Part of this analysis was to derive an acoustic signature for each
recorded biotope or group of biotopes. The MapInfo™ buffer capability was used to
create a 50 metre buffer around each ground validation site or line. Fifty metres was
selected as an appropriate size buffer, to reflect the accuracy of position fixing during
the ground truthing. Each buffer was then coded according to the biotope recorded at
that ground validation site. Within MapInfo™, acoustic data points falling within each
buffer were then captured, and assimilated into a table. In the few cases where more
than one biotope was distinguished from one video tow, the buffer around the video
tow line was split logically into sections representing each biotope. Minimum,
maximum, mean and variance values of the acoustic variables E1, E2 and depth (for
RoxAnn™ data) or Q1, Q2, Q3 and depth (for QTC™ data) were generated within
MapInfo™ from the captured acoustic data for each buffer, to form an acoustic
signature for each biotope or group of biotopes.
Using the acoustic signatures for each biotope group to construct Boolean
queries, areas which matched each signature were selected in turn and shaded
according to the corresponding biotope. An example of a query would be “select all
areas where E1 is between 0 and 0.5, and E2 is between 0.15 and 0.75, and depth is