|
|
|
|
|
|
|
© GeoAcoustics 2008
|
Application Note 3 GeoSwath Wide Swath Bathymetry Survey of the City of Bristol Port
Introduction In February 2003 GeoAcoustics arranged a demonstration of the GeoSwath Bathymetry System in the City of Bristol Port. This involved mobilisation, system calibration and collection of Bathymetry and Side scan Sonar data over a 2 day period in and around the Floating Harbour and Feeder Canal areas in the City of Bristol, UK. Survey Objectives The Survey was carried out to test the performance of the GeoSwath Sonar in Bristol Port (inside the Cumberland Basin sea lock). The two day survey encompassed several regions within the port, the main areas being the Floating Harbour and the Feeder Canal. The GeoSwath system’s performance was to be evaluated in the following tasks:
Rendered view of part of the Floating Harbour Installation
All positions were in WGS84 UTM zone 130.
The tide gauge was deployed on the mobilisation jetty at the west end of the floating harbour, and the tide gauge zero was used as the survey zero reference. This was not reduced to the local datum, so the survey zero is arbitrary. No significant tide variation was seen since the survey area was upriver of the sea locks.
The tide record showed a constant water level at +81cm (±1cm). The water level on 11/02/03 was 6.2m above ordnance datum (AOD), from the City Docks Engineer’s records. A global offset of 5.39m could be applied to the survey data to obtain depths AOD, if required. All the images in this report have depths relative to the tide gauge zero. Note that the transducer draft was 1.20m, which is 39cm below the tide gauge zero.
Roll calibration was reliable and repeatable, with clear best fits to better than 0.1 degrees. With swath widths of up to 20m per side in this shallow survey, this gives worse case errors of ~4cm at swath edge.
Yaw and Latency:
Yaw and latency calibrations interact. With only shallow flat regions available for the calibration they are difficult to separate. As an example, a 4 knot survey within shallow water with a navigation latency of 100ms will give similar calibration results to a yaw of 1 degree, for starboard to starboard overlapping runs. However, port to port overlap would give -1 degree for the same area. The calibration can also be very dependant on the seafloor features available in poor calibration areas. The key test is to set the latency and check that the yaw found for port and starboard have the same sign and value, with similar low error values, for several pairs of lines in different areas. This was the main reason for running the calibration over 12 lines in two different areas in the Port. Due to the lack of relief some lines gave poor calibration; this could be seen in the GeoSwath software from the absence of a minimum in the MD curves.
Iteration of latency and yaw values for several good port and starboard overlaps were used to find the minimum MD. Latency was then checked using other survey lines, and the calibration was adjusted until results were consistent over all the lines, and all had consistently low MDs.
Pitch
The calibration for pitch was difficult because of the shallow water depth in the port. Maximum depth of about 6m means that the 1m bin size subtends about 10 degrees at the transducers, so pitch offsets of less than 5 degrees will not affect the binned results. On the best pitch calibration lines the values were consistently less than 5 degrees but were not repeatable from different runs to better than ±2 degrees, so a value of Pitch=0 was chosen.
Calibration Results
Notes: (1) Estimated from the shape of the minimum in the MD curve. MD using above values with various calibration lines was repeatable and consistent at about 0.038. (2) All error contributions from calibration are well within the bin size and IHO special order depth accuracy requirements. The CSI DGPS Max accuracy is 40cm, and all positioning error contributions are well within this
Comparison with previously obtained values.
Notes: (1) 0.2s latency could look like a yaw of about 2 degrees. A 0.24s latency error would result in less than 50cm along track error (half a bin), while 2 degrees in yaw would give less than one bin error in the final gridded data. (2) 0.14 degree roll error would give about 6cm error at 25m out, well within IHO Special Order surveys.
Data Processing
The GeoSwath is supplied with a complete set of software for data collection, calibration, filtering, DTM generation, data quality control and visualisation.
Raw swath data was filtered line-by-line to remove outliers and noise points using the GeoSwath Swath32 software. Amplitude filters, box filters and along track filters were used to clean the data. Some manual adjustment of the filters was required near the harbour walls, where verticals, overhangs, boat keels and anchor chains would otherwise give errors in the bin depths.
The Swath32 software also applies sound velocity corrections and georeferencing, and allows navigation data to be checked and edited. In some parts of this survey the navigation data was unreliable and line segments had to be deleted; see the notes on positioning accuracy below.
The maximum swath width used was restricted to reflect the accuracy requirements of the survey, given the error budget contribution from the motion sensor and the signal to noise level of the sonar returns. In this case the maximum swath width for bathymetry was set to 50m in about 5m water depth.
The calibration parameters were applied to the filtered swath data, giving georeferenced, calibrated data files on a line-by-line basis.
The Swath32 software was also used to inspect the side scan records and output the data for later mosaicing.
Gridding and Grid Processing
Calibrated data files were binned in a 1m grid for DTM generation. using the vGrid and GridFly software.
After gridding the data was inspected, the data density and standard deviations were checked, and xyz files were generated using the GeoSwath vGrid and GridFly software.
Survey Results
The length of the survey area, from the Cumberland Basin sea lock to the East end of the Feeder Canal, was ~4km long. The first 1.5km (in the Floating Harbour) and last 1.5 km (in the Feeder canal) was surveyed.
Overview of the Survey area, with the Floating Harbour on the left and the Feeder Canal on the Right
The survey was in three parts:
The survey regions were upriver of the Cumberland Basin sea locks, so are not tidal and are calm.
The Calibration Area
The Watershed calibration area showed more depth variation and features than the main harbour, so was slightly better for yaw and latency calibration. The images from the Watershed demonstrate the high resolution of the GeoSwath Sonar and the advantage of collecting simultaneous side scan data.
Overview of the depth data used to create the calibration area DTM, looking North. Individual soundings are shown. The guide lines shown are at 1m spacing
Three views of the data from the same area as the 3D view above. Views are (from left) colour contour plot (from 1m DTM), 0.1m contours overlaid on side scan, and side scan mosaic (0.2m grid)
The Floating Harbour
The Floating Harbour region is a dredged channel about 40m wide and between 4m and 8m deep, with steep sides. A programme of dredging over the years has left features that can be identified in the contour plots, side scan records and in, ‘sun illuminated’ views of the DTM.
The coverage obtained in the regions of interest is very high, with many ‘hits’ on features of interest. This can be seen in some of the 3D views of raw data. The ability of the GeoSwath to collect data on the vertical channel edges can also be seen in these views.
Colour contour plot and shaded relief image from the 1m DTM of the Floating Harbour
The recent dredge area is towards the west of the harbour, with dredge marks clearly visible in the contour and 3D views. The shelf to the east of the dredge region appears to have eroded a few meters in from the southern bank and a depression can be seen.
Floating Harbour
Survey Accuracy and Quality Control
The advantage of the wide swath collected by the GeoSwath can be seen here, since the channel edges could be clearly identified in all the survey lines. If the channel edges did not overlap between two lines, the navigation was investigated further. Where navigation wander was apparent the line was reprocessed with the affected section removed. The side scan trace could also be used to check the consistent positioning of features between two lines (for example channel edges or pilings) to better than 1m resolution.
Depth
The overall depth range varied down to 7m, referenced to the tide gauge datum. Tide gauges were accurate to ±1cm and the heave sensor has negligible zero offset error. The measurement from the waterline to the transducer centre was measured to +/-1cm on mobilisation. Total static error contribution was less than +/-2cm.
with a standard deviation of less than 7cm
The depth repeatability of survey was measured from the standard deviation (SD) of depths within a 5m grid bin in a flat area of seafloor (using the vGrid software). This measures the effect of all the sources of dynamic errors, including the motion sensor, the sonar and the height control. Note that it also includes the contribution from any calibration errors (for example roll calibration) as there were several different lines contributing to the bin at different slant ranges.
The SD for bins containing slopes or features is greater because of the wide range of depths within each bin. This can also be seen using vGrid to inspect the data.
The survey was to be on a 1m grid. The standard error of the mean depth is given by the SD of the depths in the bin divided by the square root of the number of samples in the bin, for a normal distribution. A confidence level of 95% for the depth, as required by IHO Special Order surveys, is given by 2 standard deviations from the mean.
As an example, a SD of 7cm measured in a flat bin containing 100 points has an error in the mean of less than ± 1.5cm, with a 95% confidence level.
In the main survey region the SD for 1m bins was mostly between 5cm and 10cm, with a few very high SD regions (20cm-30cm). All the very high SD regions were traced to features on the seafloor, especially near channel edges and edges of dredged areas. Another source of high SD around seafloor features is positioning accuracy. With a positioning system accurate to 40cm, an edge can be offset by this much from one survey line to the next, which can move the feature to the next bin. This is seen in some of the feeder canal survey line images. During processing the swath widths were limited to 25m per side to maintain a high number of points per bin. Most of the main survey area had between 20 and 500 points per bin. With a 7cm SD for one point, 20 points gives a binned survey repeatability of better than ±5cm at 95% confidence level.
Density of soundings in the Floating Harbour. Only bins with more than 20 soundings are plotted
Feature Detection
Feature detection from the DTM is limited by the bin resolution used, although features of interest can be investigated further by inspecting the individual points in vGrid and the side scan records, as seen in the dredge mark images above. Confidence of detection of features larger than the bin size is high as a result of the number of individual depth measurements within each bin when surveying with the GeoSwath Sonar.
The side scan data from the GeoSwath can also be inspected for small object detection and to aid in feature identification. This can be done on the waterfall display for detailed inspection, or on the georeferenced and mosaiced side scan maps for feature location, as shown in the images of the calibration area above. Overlays of contour plots and side scan in Surfer also enhance object detection and identification, and software is available from GeoAcoustics for bottom classification from the side scan traces.
Conclusions
Total survey depth accuracy at the 95% confidence level is better than ±5cm, with ±2cm static offset error. This is well within IHO Special Order survey requirements for navigation channels. In this survey the GeoSwath has demonstrated:
More details on GeoSwath |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
