Data handling methods and target detection results for multibeam and sidescan data collected as part of the search for SwissAir Flight 111

The crash of SwissAir Flight 111, off Nova Scotia in September 1998, triggered one of the largest seabed search surveys in Canadian history. The primary search tools used were sidescan sonars (both conventional and focussed types) and multibeam sonars. The processed search data needed to be distributed on a daily basis to other elements of the fleet for precise location of divers and other optical seabed search instruments (including laser linescan and ROV video). As a result of the glacial history of the region, many natural targets, similar in gross nature to aircraft debris were present. These included widespread linear bedrock outcrop patterns together with near ubiquitous glacial erratic boulders. Because of the severely broken-up nature of the remaining aircraft debris, sidescan imaging alone was often insufficient to unambiguously identify targets. The complementary attributes of higher resolution, but poorly located, sidescan imagery together with slightly lower resolution, but excellently navigated multibeam sonar proved to be one of critical factors in the success of the search. It proved necessary to rely heavily on the regional context of the seabed (provided by the multibeam sonar bathymetry and backscatter imagery) to separate natural geomorphic targets from anomalous anthropogenic debris. In order to confidently prove or disprove a potential target, the interpreter required simultaneous access to the full resolution sidescan data in the geographic context of the multibeam framework. Specific software tools had to be adapted or developed shipboard to provide this capability. Whilst developed specifically for this application, these survey tools can provide improved processing speed and confidence as part of more general mine hunting, hydrographic, engineering or scientific surveys. Introduction At ~10pm, September 2 1998, SwissAir Flight 111 crashed just off the Nova Scotia coastline. Initial radar tracking of the rapidly descending aircraft was sufficient to place the crash location only to within about a 10 km radius. As a result a massive search and rescue mission, which ultimately became a search and salvage mission, was undertaken. This mission (Operation Persistence) involved a collaborative effort between civilian fisherman and several branches of the government including the Navy, the Coastguard, the Mounties, the Hydrographic Service and the Geological Survey. Within a period of about 12 days a complete 300% multibeam and sidescan survey of the crash site and vicinity was acquired rendered and distributed. During that time, the growing database had to be made available to other units in the combined fleet to support simultaneous close-range optical search and salvage operations (including laser linescan, ROV and diver operations). Shallow Water Survey –99 2 Hughes Clarke et. al. SwissAir 111 03/21/01 Initial search efforts were focussed in the vicinity of the surface debris fields. Within 72 hours, however, the 30kHz pingers attached to the two flight recorder units were triangulated using the passive sonar arrays on board the submarine HMCS Okanagan. All subsequent surveying was focussed within about a 5km radius of this site and extending landward along the assumed last track of the aircraft. Survey Methodology Before and after the triangulation of the transponders by HMCS Okanagan, the acoustic search surveys were carried out from four platforms simultaneously. HMCS Anticosti deployed a Simrad MS972 towed conventional sidescan sonar. CCGS Matthew deployed a Simrad MS992 towed conventional sidescan sonar. HMCS Kingston deployed a Klein 5000 focussed sidescan sonar. And CSL Plover deployed a Simrad EM3000S multibeam sonar. Whilst the three sidescan platforms were mid to large sized survey vessels, the Plover is merely a 31 ft survey launch. This launch is normally used for daytime operations only and normally in nearshore, protected areas. Due to the urgent nature of this operation, however, and the fact that it was the only platform with this resolution the launch was used. Previous experimental trials (Brissette et al, 1997) had shown that this system had the potential to resolve small targets. The two other available multibeam instruments, the EM100 sonar on the Matthew and the EM1000 sonar on the CCGS Frederick G Creed (36 hours away) both have significantly inferior target detection capability when compared to the EM3000. For this operation, the launch was deployed around the clock with just a two man crew (coxswain and hydrographer). This was done by having crew changes at 6 hour periods (done by hoisting the launch up by its forward davit only) and refueling every 24 hours (by completely recovering the launch). Data Processing and Dissemination All search and survey data had to collated and distributed to all the vessels in the fleet taking part in the operation. This required a dedicated at-sea parallel processing effort to ensure that the multibeam data (delivered in 6 hour chunks) was processed (cleaned, tidally reduced and georegistered) to be ready for delivery to other field units on a daily basis. All shipboard processing of the EM3000 swath bathymetry and backscatter data was performed by OMG staff using the OMG/UNB SwathEd software toolkit. The data deliverables included the following: • Hard copy map sheets of the EM3000 bathymetry • EM3000 topographic and backscatter imagery converted to BSB format for electronic chart navigation on CCGS Hudson (laser line scan) and MV Anne S. Pierce (dragging). • EM3000 topographic and backscatter imagery used as underlay for field based and shore based interactive sidescan image analysis ( see description below). Shallow Water Survey –99 3 Hughes Clarke et. al. SwissAir 111 03/21/01 For the sidescan data, the majority of the early analysis was made from scrolling real time hard or soft copy images. This was all done by looking at a single corridor of sidescan imagery in isolation. In order to compare overlapping swath corridors, the data had to be referenced manually by time and, using a hard copy navigation plot, the adjacent swaths identified and then retrieved. This was a time consuming process. To try and alleviate this analysis bottleneck, the GSC sidescan mosaicking software was utilised to try and build a regional picture showing the interline relationships. This mosaicking approach is a standard procedure for all GSC scientific surveys. The Complementary Attributes of Multibeam and Sidescan Data The hull mounted multibeam sonar has the notable advantage over towed instrument packages of confident positioning. Because the sonar is rigidly attached to a surface vessel and the position and orientation of that vessel is known to within ~1m and 0.05 degrees on all axes, this confidence can be propagated to the seabed information (derived from narrow (1.5) beams steered at known vessel-relative angles). This position confidence is sufficient, for example to try to detect the introduction of new small bathymetric targets such as mines by differencing one survey with a pre-survey (Brissette and Hughes Clarke, 1999). In contrast, the sidescan towfishes employed were all at least 80m from the mother vessel. Only on CCGS Matthew was a short baseline acoustic transponder positioning system available. The MS992 on the Matthew had the added advantage of being deployed using a two-body tow geometry. The MS992 thus had the benefit of decoupling from the surface vessel motion together with a far steeper cable angle than the other towfishes (which results in increased tracking confidence). In all cases, the instrument packages on the towfishes consisted of no more than a magnetic compass. Thus neither the position, nor the exact orientation of the sidescan instruments could be guaranteed better than about 20-30m and about two degrees in azimuth. As a result, the total horizontal positioning confidence of the towfish-based systems was at least an order of magnitude worse than the data collected from the hull mounted multibeam sonars. Because, however, the hull-mounted sonar remained close to the surface, the total slant range to the seafloor and the aspect ratio of the imaging path are larger that that used by a sidescan. Further more, the sidescan systems all had beam width of less than 0.75 degrees in azimuth (compared to 1.5 for the multibeam). It is thus clear that from a backscatter based target detection capability, the sidescan systems had a greater advantage in resolution. Separating natural from anthropogenic targets. Traditionally the identification of anthropogenic targets using sidescan imagery is based on an assumption that the targets will differ significantly in character from natural seabed features. Man made features are commonly angular, and solitary. For most temperate (mid-latitude) continental shelves, the Holocene transgression has covered the shelf with at least a surface veneer of fine-grained sediments (muds and sands). Such unconsolidated materials generally maintain low seabed slopes (< 10 degrees). The only common short wavelength targets visible in sidescan images of these type of seabeds are current-driven features such as ripples, dunes, furrows and ribbons. All these features are normally quite characteristic and do not occur in isolation. Under such conditions it is reasonable to assume that any angular solitary targets might be man made. Shallow Water Survey –99 4 Hughes Clarke et. al. SwissAir 111 03/21/01 For those higher latitude continental shelves, however, that were affected by the Quaternary glaciations, , the sedimentary processes were very different. Due to the fact that much of the material was deposited at random during melting of rock impregnated glaciers, widespread anomalous targets are very common (glacial erratic boulders). Furthermore, because much of effect of glacial activity is erosive, extensive bedrock outcrop is common. Where the bedrock consists of lithified sediments with layers, sharp