Blue_ice.jpg

Introduction

The Alfred Wegener Institute Helmholtz Center for Polar and Marine Research (AWI) commenced research in the western Canadian Arctic with the funding of the young researcher project titled Coastal Permafrost Erosion (COPER) in 2012. The group’s focus lay on investigating the mass transfer of sediments and carbon across the whole coastal tract, i.e. include both the emergent and submerged parts of the coast. In order to study the sediment dynamics in very shallow waters around Herschel Island, Beaufort Sea, Canada, AWI obtained a GeoSwath Plus Compact from Kongsberg Geoacoustics, a shallow multibeam system providing both bathymetric and side scan data. The following paragraphs will illustrate the processing steps and personal experience I made in the course of acquiring and processing the data, particularly focusing on processing of side scan data. Side scan data was processed using Kongsberg GeoTexture, a software package for processing side scan data, producing mosaics, and classifying textures. With some supervision, the software is fairly easy to use even for a novice like me.

Study Area

Herschel Island (69.6oN, 139oW) is located in the Beaufort Sea, at the northernmost point of the Yukon Territory, and about 70 km east of the Alaskan border (Figure 1). The island is a push-moraine that formed  during the westward advance of of the Laurentide Ice Sheet (Rampton, 1982). The coast is characterized by high cliffs and numerous retrogressive thaw slumps, indicating the presence of large massive ice bodies susceptible to permafrost degradation.  The tundra covered island is located where mean annual temperatures are well below 0 oC, and rise above freezing only between June and September (Burn and Zhang, 2009). Cold temperatures affect coastal processes, as well.  Wave- and tide induced processes are limited by the presence of sea ice and landfast ice. The presence of ice, however, introduces some physical processes unique to cold environments such as ice gouging, ice rafting, ice push-up, and ice pile-up. Ice gouging refers to the grounding of ice keels; ice rafting the transport of coarse sediment offshore incorporated into the ice matrix; while ice push-up and pile-up occur at the land-sea boundary, transporting sediments on- and across shore. Coastal erosion is limited to the period of open water. Average rates of erosion are 1-2 m/yr, and may reach 10-30 m/yr (Strzelecki, 2011), due to thermal abrasion - the combined effects of thermal and physical forces.

Figure 1. Study area overview showing Herschel Island (Geoeye 2012). Blue dots indicate locations of surface samples, while the yellow area shows the extent of bathymetric and sidescan data.

Methods

We set out to characterize bottom sediments in terms of texture and carbon content in the shallow coastal waters (< 15 m). We combined the mapping capabilities of the GeoSwath Plus Compact from Kongsberg Geoacoustics, a shallow multibeam system that provides both bathymetric and side scan data, with extensive bottom sampling and consequent analysis.

Sidescan and Bathymetric Data

Acquisition

To carry out the surveys we first had to mount the GeoSwath Plus Compact on an inflatable Zodiac MK 5 HD. We built a wooden platform onto which a retractable aluminum pole was mounted (Figure 2). The wooden platform attached to the boat, while retractable aluminum pole accommodated the V-plate assembly with the 500 kHz sonar system, the Octans 3000 motion reference unit (MRU), and the Valeport mini-SVS. The pole could be deployed to a fixed horizontal position for transport, and a vertical position when deployed. A Trimble R4 differential GPS unit provided the navigational strings and heading. Calibration lines for the patch test were collected on each of the three surveying days. In the course of the 2012 field season 123 km of survey lines were acquired. We periodically obtained sound velocity profiles of the water column using a Valeport sound velocity profiler. 

Figure 2. The mounting of the GeoSwath Plus on the inflatable boat using a retractable side mount prototype.

Data Processing

Data were initially processed using the GS+ software with the focus on bathymetry and bottom tracking. The software created swf (swath) and swp (swamp) files from the raw data, carrying depth and backscatter information respectively. Swamp files were then loaded into Kongsberg GeoTexture for side scan data processing.

GeoTexture can be used to process and mosaic multiple files automatically, or individual files may be treated differently. Regardless of the approach, the processing steps in GeoTexture are as follows: apply time varied gain (TVG), seabed locating, filtering bathymetry, applying trace normalization, applying slant range correction, and mosaicking.

Time Varied Gain (TVG)

To address geometrical spreading and absorption, TVG is used to increase the return signal strength from increasing distances from the transducer using a time curve.  TVG is set at the beginning of a project and remains constant unless the acquisition parameters change. The immediate effect of applying the TVG is that by increasing  it, the raw image appears darker. The GeoSwath system automatically applies a 20log(2R) dB during acquisition, however an additional 10 dB were used in our project.

Seabed Locate

We had some issues with the shadow of the boat itself showing up in the data on the port side (Figure 3), especially in very shallow water (<2 m). GeoTexture seabed locate functions perform very well, but in this case but necessitated a lot of manual seabed corrections. A better method was to use the bottom tracking in GS+ on the starboard side which GeoTexture can use. Having this information embedded in the swamp file was very useful for sidescan processing in GeoTexture.

Figure 3. Using the bottom tracking feature within GS+, selecting only starboard. The peak in the middle of the image stems from reflections off the boat itself. With decreasing depth, the peak would eventually migrate into the data.

Filter Bathymetry

GeoTexture imports the bathymetry information from GS+. The best results resulted from rigorously filtered bathymetry which was also truncated at the far ends of the swath. The seabed is very soft in our study area, and a layer of fluid mud typically blankets the bottom. This results in a spread of the returns at the far ends of the swath (Figure 4). Sidescan data usually exceeds the range of the bathymetry, beyond which the software extrapolates the last pixel value. The jumps introduced by the soft seabed can be smoothed in GeoTexture.

 

Figure 4. All points along the swath are shown in the depth window in GS+ where filters can be applied. Red points indicate filtered data, while green indicates good data points. Note the spread of the seabed toward the ends of the swath. The blue box indicates the limits filter, which was varied 10-13x the water depth, depending on the return quality.

Trace normalization

Beyond the TVG applied at the beginning of the processing sequence, the results can be further improved by applying a trace normalization in GeoTexture. Trace normalization function (Tfn) consists of a instrument dependent master beam function (MstrBFn), and one or more back-scatter functions (B/S Fn) related to the seabed. The MstrBFn ideally necessitates data affected by >10o of roll. Data in this project had a maximum of 8o of roll, and adequate for a MstrBFn extraction. Scatter functions depend on the seabed sediments and its texture. Given the variance of the seafloor, two, and up to four, scatter functions were used to obtain the optimal result.

Slant Range Correction

The slant range correction removes the nadir and water column effects, and produces a georeferenced sidescan image. Satisfactory results depend on good bottom tracking, but also on the trace normalization function. The high intensity boat shadow that in shallow areas encroaches into the data could sometimes be removed by skipping a number of pixels from the seabed when slant range correction was applied.

Mosaicking

Once a swath was processed as described, it was added to a mosaic. When the first file is loaded, the mosaic extents must be specified, either by letting the software scan all the lines to determine extents, or manually. In either case, it is useful to use integers, or the some multiple of the wanted raster resolution in order to avoid pixel recalculation that may slow down the processing.

One advantage of the mosaicking mode is that multiple swaths can be loaded, processed using the set of parameters defined in swath processing mode. This approach was used whenever possible. A number of swaths, five to ten, were added to the mosaic and the result was then assessed. Individual swaths sometimes needed an adjustment to the processing queue (e.g. bottom tracking, filter bathymetry, trace normalization).

Results

Processing

Processing the data with GeoTexture proved very efficient after a work-in period. The biggest challenge was extracting a good master beam function for the trace normalization. Ideally the function would be extracted from a featureless seabed with roll >10o. Nevertheless, with the kind assistance of Francisco Gutierrez from Kongsberg, a master beam function was extracted (Figure 5). Up to four scatter functions were used (Figure 6).

Figure 5. The master beam function extraction.

Figure 6. Up to four scatter functions were used to process the sidescan data.

 Mosaicking

The result of sidescan image processing with GeoTexture is illustrated in Figure 7.

Figure 7. The sidescan image above was created using the Kongsberg Geoacoustics GS+ software and minimal filtering, the image below using the full capabilities of Kongsberg GeoTexture. Note the improvement at nadir, and contrast normalization across swaths achieved with GeoTexture.

Discussion

The previous sections outlined the steps for attaining satisfactory results with GeoTexture. However, data processing highlighted the importance of good acquisition, i.e. instrument calibration, acquisition settings, SVP profiles. The Geoswath Compact system was acquired three weeks before the departure to the Yukon. With three days of training and no previous experience with multibeam surveying, I was able to carry out the survey, but given the lack of experience on the part of me and my team, many subtleties and quality checks were overlooked. While we were in the field, we also had to make major changes to the mounting system. Having acquired and processed the data personally, I will share the lessons learned in the following paragraphs.

Acquisition

Instrument Calibration

The instrument geometry (pitch, yaw, roll), navigational and attitudinal latency, must be known in order for the software to make corrections to the acquired data. There are two methods described in the operation manual: the patch test, and the equal yaw method. Either of those, however, require surveying a significant change of slope in the seafloor with sufficient data coverage. This proved to be a major challenge in our field area, as the only feature with considerable slope was the submerged portion of the gravelly spit, about 50 m wide, and 2-5 m deep (Figure 8). Navigating over the gravel bar was difficult because offshore waves break over the spit, and strong currents cause vessel crabbing. In addition, the calibration requires that the lines overlap 50%, but as the water grows shallower, the swath width also decreases.  Through trial and error, and the generous help of the Kongsberg Geoacoustics team, the equal yaw method eventually led to a calibration that produced the best possible results.

Figure 8. An early version of the bathymetry (0.5 m resolution) from the area of Pauline Cove, Herschel Island, illustrates the general flatness of the seafloor, with the only exception being the submerged portion of the gravel spit in the SW corner of the picture. Visible also are some artifacts related to the boat shadows, which show up as peaks in shallow water, the excessive jiggle of contours stems from calibration, and sound velocity profile (between the 7 and 8 m countours in the SW corner).

 

Sound Velocity Profiles

As the acoustic wave generated by the echosounder propagates through the water column, it is absorbed and scattered, and if the water column is stratified, also refracted from the layer interfaces. The GS+ software corrects for these effects by using a sound velocity profile (SVP), periodically obtained by a separate measuring device. SVPs should be taken as often as time permits, and should be decided depending on water conditions (e.g. sediment load, wind and waves, time of day). There is simply no way to correct for SVP effects in the lab, and that more SVPs were required in the course of our project became apparent only after the data was processed in the lab.

Acquisition Settings

Two settings were changed at times during the survey: sidescan gain and power. Both have the same effect on the data: contrast banding of the sidescan image. This again stems from my lack of experience at the time of the survey, and a different priority. In the course of the survey, most attention was paid to the depth window in the GS+ software which shows the seabed returns as points across the swath. At times I changed the power or gain settings, but that did not change anything on the depth display. Only when the data were processed and mosaicked did either the banding in the data become apparent, or a set of lines was darker than the others. Ultimately, I was able to find a workaround for this problem, which I will describe below.

Processing

Processing sidescan data and generating a mosaic can go relatively quickly with a good master beam functions and a set of backscatter functions. I typically moved from the deeper water to the shallow, as the lines were surveyed parallel to shore. Once I decided on a set of parameters, I loaded up to five lines into GeoTexture while in mosaic mode. This is very fast. Whenever I detected banding in a particular swath, I had to process it manually. The banding stems from changing acquisition settings during the survey, as mentioned earlier. I addressed this issue by adjusting the base function coefficients (Figure 9). Because only a portion of the line needed correction, I created copies of the swamp file, processed them accordingly, and added only the needed portion of the line to the mosaic (Figure 10).

Having the seabed detected in GS+ greatly sped up the processing of lines in shallow water. In many cases, however, manual editing of the seabed was necessary. In some cases I discarded the bathymetry altogether, because it was noisy, particularly at nadir and at the far ends of the swath, as discussed earlier (Figure 3).

Summary

The previous paragraphs illustrated that even with difficult data, and with some expert help, a relative novice may be able to produce satisfactory sidescan mosaics with Kongsberg GeoTexture. The most important lesson I drew from this experience is to be careful during the acquisition, as it may not be possible to correct some errors later. My experience is an example of learning by making mistakes. Fortunately, these mistakes were not catastrophic in terms of useful data, and others were obstacles, a challenge for innovative solutions.

Figure 9. By adjusting the base function coefficients, the image contrast was changed.

Figure 10. To address backstatter intensity variations stemming from varying power and gain settings during acquisition, the beam function coefficients were adjusted until the aberrant region fit within the adjacent lines. Only the range of traces affected by the difference was added to the mosaic from a copy of the raw data file.

 

Acknowledgements

Special thanks is due to the Kongsberg Geoacoustics team (Franzisco Gutierrez, Marylou Gentilhomme, and Martin Gutowski, who helped me in the process of acquiring data to producing good results.

References

Burn, C.R., Zhang, Y., 2009. Permafrost and climate change at Herschel Island (Qikiqtaruq), Yukon Territory, Canada. J. Geophys. Res. Earth Surf. 2003–2012 114.

Rampton, V.N., 1982. Quaternary geology of the Yukon coastal plain. Geological survey of Canada.

Strzelecki, M.C., 2011. Cold shores in warming times-current state and future challenges in High Arctic coastal geomorphological studies. Quaest. Geogr. 30, 101–113.