NBP09-01 Cruise Report

Transcription

1 NBP09-01 Cruise Report Autosub3 Deployments in the Amundsen Sea RVIB Nathaniel B Palmer 5 January to 25 February 2009 Report compiled by Adrian Jenkins from the contributions of the Autosub science and technical teams: Pierre Dutrieux, Adrian Jenkins, Steve McPhail, Pete Stevenson, Andy Webb, and Dave White. 1

2 Table of Contents Overview of NBP Cruise participants 3 Acknowledgements 3 Scientific background and motivation 4 Brief historical background to the project 5 Chronological outline of the cruise 6 Sea ice conditions in the Amundsen Sea during NBP Autosub3 mission planning 9 Autosub3 Technical Team Report 13 Autosub3 Cruise Narrative and Summary of Missions 13 Analysis of the Navigation Drift during the 6 PIG ice shelf missions 22 Autosub3 Scientific Sensors 24 Sensor Synchronisation 24 Seabird 9+ CTD system 24 Kongsberg EM2000 Multibeam Swath System 25 Argos Location System 26 Autosub3 ADCP configuration 26 Autosub3 Mechanical and Operational Aspects 28 Preparation and Mobilisation in Punta Arenas 28 Vehicle configuration 29 AUV Operations: mechanical configurations 31 Vehicle maintenance after each mission 31 Recovery operations 32 Mission 431 damage 33 Acoustic operations 34 Gantry, hydraulic power pack and electrical power 36 Autosub3 Data Processing Report 38 CTD data 38 Introduction 38 Processing steps 38 Preliminary analyses 42 ADCP data 45 Introduction 45 Settings 45 Mechanical configuration 46 Processing 47 SWATH data 48 Introduction 48 Processing steps 49 Appendix 1. Pre-cruise Risk Assessment Procedures 51 Appendix 2. Mission Summaries 74 Appendix 3. Autosub Logged Data Formats 97 Appendix 4. Fault Log 107 2

3 Overview of NPB0901 Adrian Jenkins Cruise participants Personnel involved in the UK component of NBP0901 included a science party of two from the British Antarctic Survey (High Cross, Madingley Road, Cambridge, CB3 0ET): Adrian Jenkins Pierre Dutrieux and a technical team of four from the National Oceanography Centre, Southampton (Waterfront Campus, European Way, Southampton, SO14 3ZH): Steve McPhail James Perrett Andy Webb Dave White Acknowledgements The successes of the project would not have been possible without the assistance of many. Particular thanks are due to: Stan Jacobs for all his efforts in making the Autosub work a part of NBP0901, and providing the best possible opportunity, during a busy and shortened cruise, to take advantage of the favourable ice and weather conditions. Captain Mike Watson and the crew of Nathaniel B Palmer for their help with all aspects of the Autosub operation. Eric Hutt and his team from Raytheon Polar Services for their invaluable assistance before, during and after the cruise, and a special thank you to the MT s Jullie Jackson, Amy Schaub, Mike Lewis and Robert Zimmerman for their help getting Autosub safely back on board, and to Kathleen Gavahan for advice on the use of the MBSystem software. And finally, to all those who contributed to the Autosub Under Ice Programme, both on the technical and scientific side. Experiences gained and developments made as part of that programme were the foundations upon which the successes of NBP0901 were built. 3

4 Scientific background and motivation The Antarctic ice sheet, which represents the largest of all potential contributors to sea level rise, appears to be losing mass at a rate that has accelerated over recent decades. Ice loss is focussed in a number of key drainage basins where dynamical changes in the outlet glaciers have led to increased discharge. The most significant recent mass losses have been from the basins that drain along the Amundsen Sea coast. The synchronous response of several independent glaciers, coupled with the observation that thinning is most rapid over their floating termini, is generally taken as an indicator that the changes have been driven from the ocean, but the precise mechanism has remained speculative. Observations have revealed that the deeper parts of the Amundsen Sea shelf are flooded by almost unmodified Circumpolar Deep Water (CDW) with a temperature around 1ºC, and that this drives rapid melting of the floating ice. We need to understand what processes determine the strength of the CDW inflow and how it interacts with the floating ice in order to quantify the past, present and possible future oceanic forcing on the glaciers. Addressing these problems was the main aim of cruise NBP0901 aboard RVIB Nathaniel B Palmer. Sponsored primarily by the US National Science Foundation Office of Polar Programmes, the work included conductivity-temperature-depth (CTD) stations to study the spatial distribution of CDW inflows and meltwater outflows and bottom- and ice-tethered moorings to monitor water column properties over two years. The cruise covered the Amundsen Sea continental shelf east of 120ºW, with particular focus on the area of Pine Island Bay (PIB) the eastern part of which is covered by the floating extension of Pine Island Glacier (PIG). PIG has experienced a sustained and continuing acceleration of its flow since at least the early 1970 s, and the current contribution of it and its neighbouring glaciers in the southeast Amundsen Sea to global sea level rise is estimated at 0.25 mm/yr. The UK component of NBP0901, funded by NERC standard grant NE/G001367/1 awarded to the British Antarctic Survey, involved the deployment of NERC s autonomous underwater vehicle (Autosub3) on six missions beneath the floating extension of PIG to study how the warm CDW gets beneath the glacier and how it determines the rate at which the glacier melts. The specific aims of the project were to: map the seabed beneath the glacier; map the underside of the glacier; and determine where and how heat is transferred from the inflowing CDW to the outflowing ice-ocean boundary layer. To do this Autosub3 was equipped with the following instrumentation: Seabird CTD, with dual conductivity and temperature sensors plus a dissolved oxygen sensor and transmissometer; Simrad multi-beam echosounder; upward-looking RDI 300 khz ADCP; downward-looking RDI 150 khz ADCP. Unusually light sea ice conditions allowed the ship to access Pine Island Bay during the early stages of the cruise and provide the perfect conditions for Autosub deployment. A total of eight missions, including two test missions in open water and six science missions beneath the glacier were run during the two-week period from 17th to 30th January Total track length was 887 km (taking 167 hours) of which 510 km (taking 94 hours) were beneath the glacier. 4

5 Brief historical background to the project This project was initially proposed and funded in 2004, along with a parallel proposal to NSF for a cruise on Nathaniel B Palmer that was scheduled for Feb/Mar However, following the loss of Autosub2 beneath Fimbulisen in February 2005 and the subsequent build of Autosub3, a new risk assessment procedure was introduced by NOCS for high-risk applications of autonomous submersibles. During the summer of 2006 outline plans for the missions beneath Pine Island Glacier were submitted and assessed. The conclusion was that the track record and fault history of Autosub3 to that date made the sub-ice missions too risky. The 2007 cruise went ahead without Autosub, which was taken to Norway in March 2007 for a series of test missions designed to prove its reliability for the under ice work. Following these and two successful scientific cruises in the summer of 2007, a reassessment of the risk showed a lower probability of vehicle loss. The outline missions were approved by NOCS, provided that Autosub was run for 5 hours/25 km in open water at the start of each sub-ice mission. In the spring of 2006, a new proposal was submitted to NSF, in response to the call for the International Polar Year, for further ship time on Palmer in the Amundsen Sea. The main aims of the cruise would be mooring and CTD work. Autosub work was not proposed initially, as at the time of writing it was still assumed that that work would be part of the 2007 cruise. However, following the withdrawal of Autosub from the early 2007 cruise, the award of funding for another cruise in 2009 and the approval of the missions beneath Pine Island Glacier in the summer of 2007, NSF agreed to a slight lengthening of the 2009 cruise to accommodate the postponed Autosub project. The subsequent addition of two other projects and a reduction in the allocation of ship time, on account of rising fuel costs, completed the shape of what was to become NBP0901. Details of the risk assessment protocol, the outline missions and the results of the risk assessments are presented in Appendix 1. The Autosub3 missions that were run during NBP0901 were planned and executed against the background of these documents. 5

6 Chronological outline of cruise 5 Jan 2009: NBP departed Punta Arenas, Chile, around noon and headed west along the Straits of Magellan Jan 2009: In transit to the Amundsen Sea Jan 2009: CTD and mooring work on the continental slope and across the continental shelf to Pine Island Bay. 17 Jan 2009: Arrival in PIB. CTD and multi-beam echo-sounding near the PIG ice front. Autosub3 deployed on test mission (M427) in deep water about 30 km west of the ice front. 18 Jan 2009: CTD section along the ice front. Analysis of M427 data reveals a few problems with motor control, upward ADCP configuration and CTD sensor performance and calibrations. 19 Jan 2009: CTD section at the southern coast of PIB. Motor control and instrument problems fixed. Final preparations for first science mission. M428 commenced about 10 km west of the ice front with a test dive and return to surface to download data. Checks showed that the earlier problems were all dealt with. Older sensor calibrations and cleaning of the transmissometer gave much better results. Autosub was sent on its way. 20 Jan 2009: At about 03:00 Autosub entered the cavity on its first science mission, travelling 30 km in before turning back. NBP extended multi-beam coverage in PIB during the mission. Around 14:00 Autosub was intercepted at its penultimate waypoint, 5 km west of the ice front, and given the command to surface. Plans to leave Autosub in the water between missions were scuppered when one of the recovery lines came loose. Recovery was complete by about 19:30. All data looked good. 21 Jan 2009: Detailed CTD section across PIB. Autosub launched on its second mission (M429) into the cavity, another 30 km in and back. 22 Jan 2009: Autosub entered the cavity around 05:00. NBP deployed a mooring and extended multi-beam coverage, then rendezvoused with Autosub at its final waypoint at about 17:30. M429 had gone according to plan. Autosub was recovered for a battery change. 23 Jan 2009: NBP worked on a CTD at the northern side of PIB. Autosub was sent on the third 30 km in and back mission (M430) beneath the ice shelf, entering the cavity at around 20: Jan 2009: NBP extended multi-beam coverage in PIB, then rendezvoused with Autosub at its final waypoint at about 10:00. M430 had gone according to plan, so data were downloaded and M431 was started without recovery. M431 was the first long mission to the grounding line, heading 60 km into the cavity before turning back. Autosub entered the cavity at around 19: Jan 2009: NBP undertook a 24-hour yo-yo CTD in the southern outflow from the PIG cavity, then rendezvoused with Autosub at its final waypoint at about 23: Jan 2009: Autosub was recovered damaged, following a collision with the ice, apparently sustained when it lost track of the ice base and ascended into a crevasse. Only minor problems were apparent with the data, but work was needed to repair the damage. NBP headed west to work on fast ice near the entrance to PIB. 27 Jan 2009: NBP finished sea ice work and collected multi-beam data over the outer part of PIB. Confirmation came from NOCS that the final two planned missions were approved, provided that the extra precaution was taken of not tracking off the ice base in the inner part of the cavity, where an appropriate minimum depth that would 6

7 prevent Autosub rising into a crevasse could not be set. NBP returned to the ice front and a test mission (M432), to check that everything was in order following the repairs, was run. 28 Jan 2009: Results from the test mission looked good, so Autosub was sent on its second long mission (M433) to the grounding line. It entered the cavity at about 09:30. NBP did a traverse along the front of PIG, measuring the height and draft of the ice front, then headed west for CTD s and multi-beam mapping. 29 Jan 2009: NBP completed multi-beam mapping and headed for the M433 final waypoint to rendezvous with Autosub. Autosub surfaced at around 11:00 following another mission abort. This time there was no collision or damage, but the mission was aborted after minimum water column thickness was detected about 30 km into the cavity, and the route to the next waypoint was blocked by an ice keel. Autosub was recovered and NBP undertook multi-beam mapping while batteries were changed. 30 Jan 2009: Autosub was launched on its final mission (M434) beneath the ice, this time a T -shaped track to explore the northern and southern parts of the inner cavity, as yet unexplored. Autosub was launched at about 00:30 and entered the cavity 5 hours later. NBP set off to the west for more CTD and multi-beam mapping work. 31 Jan 2009: NBP completed work to the west and returned to the M434 final waypoint. Autosub was recovered on board at about 12:00. The mission had gone according to plan, except that one arm of the T had been cut short when Autosub encountered minimum water column thickness. NBP left PIB to work further to the west on the Amundsen Sea continental shelf. 1-7 Feb 2009: CTD and mooring work around Getz and Dotson ice fronts and deployment of Ice-Tethered Profiler and Mass Balance Buoy on fast ice near Crosson Ice Shelf Feb 2009: CTD and mooring work on the outer continental shelf and upper continental slope. 14 Feb 2009: CTD section across eastern front of Getz Ice Shelf. 15 Feb 2009: Last mooring and sea ice work near the shelf edge Feb 2009: Mooring and CTD work S66º30' W129º30' Feb 2009: In transit to Punta Arenas. 7

8 Sea ice conditions in the Amundsen Sea during NBP0901 Initially the timing of NBP0901 did not look ideal in that we would have to leave the Amundsen Sea continental shelf in mid-february, around the time of the sea ice minimum. It looked like our best chances for Autosub deployments unhindered by ice would be right at the end of the cruise work period. However, satellite imagery from December showed an early breakout of fast ice from the eastern Amundsen Sea, including PIB, with a broad polynya developing over the inner shelf. By the time the cruise started, the outer shelf was still ice covered, but easily workable, with the easiest line through lying to the east and allowing relatively direct access to an ice free PIB (Figure 1). This situation persisted throughout the period of Autosub operations, with the ice cover on the outer shelf gradually thinning. This provided perfect working conditions for the Autosub deployments, the only potential problem being that the extent of the open water on the inner shelf could permit relatively high seas. However, the weather remained near perfect, and the occasional winds of knots or less were always from the east or south-east, straight off the land, giving minimal fetch and continued calm conditions within PIB. This greatly assisted the Autosub recovery operations. Figure 1: Ice conditions on 14/01/09, showing ship track (red) to that date through the band of ice on the outer shelf towards an ice-free PIB (directly south of the ship s position). 8

9 Autosub3 mission planning Outline plans for six science missions beneath the ice shelf, including three that penetrated 30 km into the outer cavity and three that penetrated the full 60 km to the grounding line, had been approved by NOCS in August 2007 (see Appendix 1). Final planning of waypoints and profiles was guided by ice thickness data (Figure 2), from airborne radar sounding, the bulk of which was collected in early 2006, and seabed soundings from NBP9402, refined with multi-beam data collected during the course of NBP0901. We also made extensive use of the Modis image, collected in late December 2008, shown as the underlay to the thickness data in Figure 2. The light and dark shading on this visible image reflects surface topography illuminated by sunlight from the north. Where the ice is grounded, the combination of ice thickness and surface elevation from the radar observations gives the depth of the bed. We therefore had knowledge of the seabed depth along the grounding line at the inner edge of the cavity. Depths there are similar to those at the ice front (800 to 1000 m) so we assumed that the bed would be fairly flat. Figure 2: Modis image from 28 December 2008 overlain with radar sounding flightlines, colour coded with the measured thickness. The bold black line indicates the estimated position of the grounding line in 1996, while the yellow line is the NBP9402 ship track with CTD stations shown by stars. Axes are labelled with the projection coordinates of the image. The standard parallel for this projection is 70ºS, so 1 km on the image represents about km on the ground. 9

10 Planned tracks and profiles for the six missions are shown in Figures 3 and 4. The three shorter missions were intended to track the seabed at 200 m altitude on the way to the turning point, approximately 30 km from the ice front, then ascend to track the ice base at 100 m clearance on the way out. A global minimum depth of 500 m was set to prevent Autosub tracking up into hollows in the ice shelf base, while the zig-zag course on the way out was designed to sample a wider range of basal features with the multi-beam echo-sounder, which was oriented up for these missions. The original plan was then to run three further missions that extended these tracks by approximately 30 km into cavity. It was intended that the actual turning point would be determined by Autosub s detection of minimum headroom, set at approximately 200 m of water (i.e. about 100 m clearance above and below). For the inbound leg of the longer missions Autosub would track the seabed at 100 m altitude, then track the ice base at 100 m clearance, so obtaining multi-beam images of the ice base, as far as the 30 km point reached on the shorter missions. From this point out the sub would undulate between bounds set by minimum clearance of the ice base and seabed or global minimum and maximum depths, whichever were encountered first. Figure 3: Planned mission tracks beneath PIG, with inbound legs shown in blue and outbound in red. Red stars indicate the turning points for both short and long missions. The white lines indicate the four longitudinal radar profiles (Figure 2) that were the main source of ice draft estimates. The 1996 grounding line is shown in black, and the NBP9402 cruise track in yellow. 10

11 Figure 4: Planned mission profiles beneath PIG, with short missions shown in red and long missions in blue. The ice thickness profile is taken from the southernmost radar flightline of the four shown in Figure 3, starting near X=50 km, Y=1630 km. The bold, dashed line joins the two points of known seabed depth at the ice front and the grounding line. The three shorter missions (M ) were successfully completed as above. Further discussions about the longer missions led to the abandonment of the zig-zag course for the first part of the return leg, as far as the point reached on the shorter missions. It was felt unwise to have multiple waypoints programmed for the inner part of the cavity, where we were uncertain about the turning point. It was quite possible that the sub would detect minimum headroom at a point further out than one of these waypoints and then attempt for a second time to go further into the cavity to reach its next waypoint. Finding the route blocked would then trigger a mission abort. The first of the longer missions (M431) showed the dangers of tracking an irregular ice base without a minimum depth protection (the global minimum of 500 m, required for the outer part of the cavity, gave no protection on the inner part), as the sub followed a course similar to the hypothetical one in Figure 4. The triggering of multiple collision avoidance (one of which was not actually avoided) signals led to the mission being aborted, and a rethink for the final two missions. For these we planned only seabed tracking with an altitude of 100 m in the inner cavity, and since we could not image the ice shelf base, switched the orientation of the multi-beam echo-sounder so that we could gather seabed imagery. The fifth mission (M433) was also aborted when minimum headroom was detected earlier in the mission than we had anticipated and the sub turned to find its path to the next waypoint blocked by an ice keel. The sixth and final mission (M434) track was then further amended to access the inner cavity by a known safe route, then explore to the south and the north (Figure 5). Beyond the known safe area, where there would be no protection from the minimum depth constraint, we once again planned only bottom tracking, although we ran at two altitudes (100 and 200 m) to the south, where we knew that the 500 m 11

12 minimum depth would keep the sub safe from the major irregularities of the ice base. This mission went according to plan, except that the sub detected minimum headroom very early on the northern leg. From the start of the straight leg back to the ice front Autosub followed an undulating course, collecting data on the water column between the ice base and seabed. Mission tracks, as executed are shown in Figure 5. Included here are the test missions, run at the start of operations and following the repairs necessitated by the collision on M431, and the preliminary open water tracks run at the start of each mission as part of the risk mitigation strategy (Appendix 1). Figure 5: Autosub3 missions completed in PIB during NBP

13 Autosub3 Technical Team Report Steve McPhail, James Perrett, Andy Webb, Dave White. Autosub3 Cruise Narrative and Summary of Missions Eight Autosub3 missions were carried out in the vicinity of Pine Island Glacier ice shelf from 19/1/2009 to 31/1/2009. Of these, two were test missions and six took the Autosub under the ice shelf, for a total of 94 hours, and 510 km (Table 1). Table 1: Totals for all Autosub missions on NBP0901 (19 th to 30 th January 2009). Under Ice Missions 6 Total Hours Run 167 hours ( 7.0 days) Total Distance 887 km Total Time under ice 94 hrs (3.9 days) Total Distance under ice 510 km Table 2 is a narrative for the Autosub operations on N B Palmer0901. Table 3 summarises all the Autosub3 missions, with basic statistics of start time and position, hours and km run, and a description of the mission plan, with comments about the execution of the mission and any faults noted. Table 2: Autosub cruise narrative for NBP /1/ Palmer sailed from Punta Arenas AUV switched on. All looking OK. 6/1/ CTD Removed from AUV as it was not producing data. 7/1/2009 CTD system fixed. The transmissometer on the seabird was incorrectly wired see fault log. Some confusion around the calibration parameters for the AUV CTD see fault log. 8/1/2009 Fault found on ADCP up penetrator and fixed see fault log. Fault found on data logger and fixed see fault log. 9/1/2009 Discussions with Stan Jacobs and Adrian Jenkins. Test mission to be in Pine Island Bay itself, at a point with water depth as close as possible to that near the ice front. Autosub switched on to continue soak test. 16/1/2009 Based on BAS charts, the magnetic correction for the tow fish is 49 degrees east (BAS chart). 13

14 17/1/ Mission 427 started. A test mission to test systems and sensors to 850m. Multibeam pointing down for the test. Ship at S:74:55.0 W:102: Sub on surface. All systems seemed to work well Sub recovered onto the ship without damage. However, problems with the recovery were apparent. The ship s propulsion system s suction proved very hazardous for the AUV. A boat hook was used several times to prevent the AUV going under the counter. 18/1/2009 Debrief of 1 st mission launch and recovery. Several problems with the recovery noted. In future we will use a small boat in the water for recovery. The small boat taking the recovery line out to the sub, and also acting as a safety tow boat, using a line attached to the AUV aft lift point. Main propulsion motor had been intermittently switching off for periods of a few seconds during the mission. Fault traced to occasionally missed keep going messages from the Mission Control (due to occasional network message collisions). Cure was to make the messages acknowledged, and to increase the motor turn off timeout see fault log. Calibrations for CTD and transmissometer still apparently a problem. Problems with up ADCP profiling range. EB beacon receiver signals not received at all clearly. Container GPS antenna not working. All: See fault log. 19/1/2009 Mission 428. First sub PIG mission. But first a short test mission to check out the fixes for the sensors, ADCP up, and motor control In water. Ship on station at S:74:59.48 W:101: Dived Surfaced. (M428a) Downloading data for analysis. Vehicle remaining in water maintaining position at the waypoint (mission not ended). Changes made to grounding arrangements for the emergency beacon receiver (see fault log). 20/1/2009 M428b. Plan: 30 km mission into the southern part of the PIG cavity. 200 m altitude in, up tracking at 100 m off the ice shelf on the way out. Exit path has slight wiggles east and west to improve data significance Dived. Start of Mission 428b Finished monitoring at WP1, and sent continue command. Sub heading for WP2, which is 5km from the ice edge Ship at WP2. Listening on Fish and Emergency Beacon Sub turned, is now heading for ice shelf Slant Range 905 m. Emergency Beacon coming in at 37 sec past 10 minute mark. 14

15 0224 Ship declutched to reduce noise. Last believable LXT range of 4037 m at Sub entering ice cavity Stopped monitoring. No more EB detected. Hydrophone recovered. Ship leaving for other work Recovery for M Ship back at S: , W: for recovery Tracking on LXT, Tracklink, and Emergency Beacon. Sent surfacing command, but ascent very slow because surfacing power had been set to zero (for safety reasons, for the initial dive) On surface Spotted All inboard. A small boat was used for the recovery, a lot safer than previous recovery. The original plan had been to leave the AUV in the water during data download, and to program it with another mission. However, a short piece of loose line was spotted. This was the forward line which loops back to the tail end. The line was most likely secure enough (as cable tied in, and shouldn t have been able to come out any further), but to be sure, we recovered the sub. End of M428 21/12009 Mission km into PIG cavity. 200 m altitude going in, ice following coming out. Central part of ice shelf Launched Autosub: Ship at S: , W: Dived AUV heading for WP2 with the Mission Exception Script 1 running. Surface Command sent On surface. Autosub ran towards the safe waypoint because the down looking ADCP had detected a high scattering layer at 500 to 600 m depth. The effect was to make the AUV systems trigger on limited headroom, and run the exception script 1, which is to head for WP2. To prevent this happening again, the maximum range of the down ADCP was reduced from 500 m to 440m. A look at the data indicated that this would be enough to stop the ADCP triggering on this layer again, and in practice real ranges beyond 440 m were rare Restart of Mission 429 (AUV still in the water) Dived No Tracklink Telemetry. Trying to find the fault in the fish. The fault in the fish was caused by a badly made set of crimp joints in the fish cable. 15

16 2120 Tow fish fixed, and tracking started, but 2 minutes too late to stop the mission stopping (as the AUV was programmed to surface if it had not heard from us for 1 hour after it had reached depth) Ascent of sub stopped at 90 m. This was due to the minimum up altitude setting preventing surfacing. The algorithm for detecting whether the surface was clear of ice, falsely detected ice above. The cause of this is probably related to the ADCP (sound speed) and depth sensor (water density) calibrations Sent end mission command by acoustic telemetry, to force vehicle to surface On surface Set ncicethickness to 8 m. This should overcome the problem related to sound speed and surfacing. 22/1/2009 M429 restart (30 km into PIG cavity. 200 m altitude going in, ice following coming out. Central part of ice shelf) 0118 Dived Following test runs underwater, acoustic Start sent Tow fish inboard. Ship heading for WP Ship at WP2. Tracking started with LXT, Tracklink, and EB hydrophone AUV turned at WP2. Heading for ice shelf 0700 Hydrophone and tow fish inboard. Ship heading off to deploy a mooring. Recovery for Mission Ship on station at S: W: (Penultimate waypoint attempting to intercept the AUV). Tow fish and EB deployed AUV is heading towards final waypoint (we were just too late). Recovered the fish, and steamed over to WP At waypoint S: , W Tracking. Sent surface command On surface All recovered. (Small boat used). END of Mission 429 Battery Change. It took 6 hours. 23/1/2009 M430. Plan: 3rd under ice shelf mission 30 km in. 200 m alt in, and 100 m distance off the ice shelf coming out Dived. Ship at S:74:57.18 W:101: Following test at depth, start command sent acoustically. Ship moving to WP2. Fish inboard Ship at WP2. Ranging with EB hydrophone AUV being tracked. Turned at WP Hydrophone recovered. AUV now 5km under the ice shelf. 16

17 Ship free for other operations. 24/1/2009 Recovery of M On station minute repetition rate of the emergency beacon indicates that the AUV has already reached WP8. Relocated to WP Tracking. Slant Range 667 m Sent Start command to begin surfacing On surface. S: , WL: End of M430 AUV left in the water for the for next mission 24/1/2009 M431. Plan: Mission 60 km under the ice shelf (southern edge). 200 m altitude in, ice following out at 120 m, to 30 km mark, then profiling the rest of the way back AUV Dived. Ship at S:74:52.21 W:101: After tests underwater, Start sent. Fish up and ship heading for WP Ship at WP2. Tracking with EB, later with LXT and Tracklink AUV reached WP2, turning towards the ice Finished tracking with emergency beacon hydrophone. AUV is 10 km under ice shelf. Ship heading for next station. 25/1/2009 Recovery of M Ship at S: , W: LXT range 1064 m Start mission sent. AUV coming up. 26/1/ AUV on surface 0147 Autosub recovered onto ship. Damage to front side panels, wings, CTD port side. It had hit the ice 50 km in, shortly after it had turned back having reached minimum water column thickness of 200 m. Problem apparently due to AUV running into crevasse when its ice tracking phase had started, following its turn back out from beneath the shelf. However, the sensor data looked good. End of M431 27/1/2009 M432. Test mission needed before the start of the second long mission beneath the ice shelf, to test systems and sensors following the crash Dived. Ship at S:74:57.71 W:101: Sent Start + Acknowledgment received AUV doing 3 hours of back and forth runs, for risk mitigation purposes. 17

18 28/1/ AUV back on surface. Test successfully completed. Data downloading. New mission loaded over WiFi link (AUV not recovered). 28/1/2009 M433. Plan: 50 km into PIG cavity, at 100 m altitude, turn, then at 30 km point start vertical profiling Dived. Ship at S:74:57.71 W:101: Sent acoustic start. 30 minutes of constant altitude runs to follow Sent start (+ acknowledged) for main part of mission to begin. AUV heading for WP Tow fish recovered and ship heading for WP Ship at WP2. Listening on EB hydrophone EB times: 30.94, Slant Range 1843m (Tracklink) Autosub turned at WP2. Slant Range 738 m, Depth 692 m, RTG 10m EB 3667, Range 9.6 km. Hydrophone and fish retrieved Ship heading to next station. 29/1/2009 Recovery of M Ship hove to at WP8 (recovery waypoint). S:74:53.42 W:101: Slant range 897 m (Tracklink) Surface Command sent + Acknowledged AUV on the surface and Mission Ended AUV recovered onboard. Small boat assisted. (as in all recoveries except the first). Data indicated that the AUV safety systems had prevented it going into the ice shelf beyond 30 km. Failed collision avoidance flag sent the AUV back on the emergency script 1. Hence the profiling was not carried out. Other data looked good, particularly the EM2000 swathing of the seafloor at 100 m altitude. 30/1/2009 M434. Plan: 40 km under the PIG ice shelf, then 10 km to south and back, at 200m then 100m altitude, then to north 10 km and back at 100m altitude. Return profiling AUV launched. Ship at S:74:58.96 W:101: Dived Depth 824 m, SR 836 m. AUV circling. All parameters checked and OK Sent Start Mission and Acknowledged. AUV beginning 1.5 hours off back and forth constant altitude runs, as test. AUV noticed to be going a little slower than previous mission m s -1 rather 18

19 than 1.49 m s -1. Possibly due to larger ARGOS antennae and consequent more drag. However, loss of speed not thought to be a problem, and mission times updated accordingly Sent Start (and Acknowledged). Tow fish in and Ship sent to WP2. (S: , W: ) Ship at WP2. Fish and EB hydrophone deployed AUV reached WP2. Slant Range 882m, Depth 868 m. Continue to track with Tracklink, and the EB hydrophone EB hydrophone 45.07, sec. Range 5.8 km. (entering ice shelf cavity) EB 47.43, Range 8.7 km Nothing more received. Hydrophone inboard. Ship to next station. 31/1/2009 Recovery of M On station at recovery point. Tow Fish and EB hydrophone in water. S:75:0.29 W:102: Telemetry, Tracking (LXT and Track point). SR 1089 m, Depth 820 m. Horizontal range 716 m. Mission Line 174 shows that the mission ended normally Sent Start to bring vehicle up. (Acknowledged) On surface and spotted Tow fish and EB hydrophone recovered Autosub3 recovered onto ship. Data downloaded and it looks like the mission was carried out as planned. 19

20 Table 3: Summary of All Autosub3 Missions on N B Palmer # Start time + pos, duration+ km Plan Comments and Faults /1/ GMT S: 74:59.48 W: 101: hrs 38 km /1/ GMT S: 74:59.48 W: 101: hrs 101 km /1/ GMT S: W: hr 113 km /1/ S: 74:57.18 W: 101: hr 107 km Test Mission to 850 m depth. Sub Ice shelf Mission. Started with short (2 hr) test mission, followed by data retrieval and checking. Then AUV run under the ice shelf. 30 km at 200 m const altitude, then turn and 100 m up altitude. 2nd run under the PIG. 30 km in. 200 m altitude in, and 100 m distance off the ice shelf coming out. 3rd under ice shelf mission. 30 km in. 200 m altitude in, and 100 m distance off the ice shelf coming out. None critical. Motor appeared to stop and restart intermittently fixed for M428 by increasing enable timeout from 15 to 30 seconds and making motor enable tick acknowledged and repeat. T1 on CTD appeared intermittent connectors cleaned and cable replaced for M428. Hotel Ground Fault reading high investigation showed that source was related to the grounded CTD chassis but no fault found. Mission completed as planned. Good multibeam, CTD, O2, & Transmission data. An unexpected seabed ridge found at about 30 km into ice shelf. Poor ADCP up profiling range noted. First attempt at mission was unsuccessful because shortly after vehicle dived the down ADCP misinterpreted a scattering layer at approximately 500m as the seabed causing the vehicle to go to a safe waypoint due to insufficient depth. To overcome this, we reduced the maximum ADCP range to 440m. Second attempt was unsuccessful as we were unable to communicate acoustically with the AUV due to faulty crimp joint in the acoustic fish cable. Final mission was successful and completed as planned. Battery changed before mission. Mission completed as planned. 20

21 431 24/1/ GMT S: 74:52.21 W: 101: hr 183 km /1/ GMT S: 74:57.71 W: 101: hrs 18 km /1/ GMT S: 74:57.71 W: 101: hrs 160km 4th under ice shelf mission for Palmer First 60 km mission. Run 60km in at 100m altitude, turn when got to far waypoint or collision avoided and then run back at 120m up altitude EM2000 set looking downwards. 4 hour test mission. Approximately 55km under ice shelf at 100 m altitude. Turn on minimum headroom setting. Profile out the last 30 km. EM2000 looking down. Mission terminated early due to the emergency exception being called (failed collision avoidance). The vehicle had, however reached 55km under the ice shelf. Consequently, the AUV carried out no profiling on the way out. Vehicle nose section damaged by collision with ice. Port CTD plumbing damaged. Linkquest transducer bulkhead connector damaged (the telemetry system was working properly on recovery). Battery changed before mission. No problems noted. Vehicle turned around due to limited headroom as expected, but further limited headroom situations on the return caused collision avoidance behaviour leading to a run back to the safe waypoint. Hence no profiling was carried out /1/ GMT S: 75:0.29 W: 102: hrs 167 km 45km under the ice, turn south for 10km at 200m altitude then return to turn point at 100m altitude, then go north for 10km at 100m altitude, return to turn point at 200m then return to safe way point profiling from 500 to 900m depth. Battery changed before mission. Completed successfully, although the vehicle turned back early during its northern leg after detecting a deep ice keel. 21

22 Analysis of the Navigation Drift during the 6 PIG ice shelf missions The navigation drift, as measured by the difference in position between the dead reckoned position and the GPS position when the AUV surfaces at the end of a mission, had a high variability between missions, with drift rates ranging from % (m per 100 m travelled), to 1.6 %. It is expected that the navigation performance will depend upon the relative amount of time the AUV spends bottom tracking (giving best accuracy), ice tracking (giving intermediate accuracy), and water tracking (giving poor accuracy dependent upon the magnitude of the currents). Table 4: The Navigation results for the six under ice missions. The offsets are the displacements from the GPS position at the surface at the end of the mission. % drift (m/100m) Average Drift (m/hr) Hours Water Tracking Hours Ice Tracking Hours Bottom Tracking Abs Bearing deg Displacement (m) Del East (m) Del North m) Mission * *Navigation estimates based on USBL fix before AUV surfaced. The worst navigation performance was for the first mission 428 (1.6 % drift). However, for that mission the AUV had a very slow rise to the surface at the end of the mission (with zero power ascent), with a prolonged period at less than 500 m, depth, where bottom track was lost. Because of this, the positional fix when surfaced was of little value, and hence the AUV position before surfacing was obtained using the Tracklink USBL data. The USBL system is not, however, well calibrated, and the positions cannot be trusted to better than 500 m accuracy. An apparent contradiction is that for mission 434, the time spent water tracking was also large, but the drift in navigation was much less. This is perhaps because most (1.8 hours) of the water tracking occurred during the profiling part of the mission, within the ice cavity, and where the currents might be significantly lower. For this mission, the total water tracking during the ascent and decent was only 0.6 hours. Apart from errors due to loss of bottom tracking, the largest errors are expected to be due to the scale factor error of the DVL, and the misalignment between the ADCPs and the INS system. Unfortunately, for a mission where the return position is close to the start position, these errors are largely unobservable (cancel out). Previous calibrations have limited these errors to 0.2% of distance travelled for the downward looking ADCP. However these errors could be larger for the up looking ADCP, as this is not mechanically indexed to the PHINS position. Errors as much as 1 degree in misalignment are possible (equivalent to 1.8% navigation error due to misalignment), 22

23 and since the upwards tracking was not carried out symmetrically with respect to going in and out of the ice cavity, this might explain the relatively large error for mission 429 of 0.46%. 23

24 Autosub3 Scientific Sensors For the NBP0901 cruise the Autosub vehicle was fitted with the following scientific sensors: RDI 150kHz ADCP looking downwards RDI 300kHz ADCP looking upwards Seabird 911 CTD system with O2 and transmissometer Kongsberg Simrad EM2000 Multibeam Sonar. The data from these plus the navigation data, and clock synchronisation data, will be made available to the cruise PI s on a set of DVD s. These instruments are described separately in the following sections. The mechanical systems section of this report shows the exact sensor locations for the CTD system. All the electronic systems on the vehicle are connected to a single control network. The data from all sensors apart from the EM2000 multibeam sonar are recorded on the Autosub data logger. The Autosub logger uses a proprietary data format but the data is translated into standard ASCII text files using a set of Matlab processing scripts which store the full data set in Matlab format together with two second sampled data in 3 text files. The general vehicle data are stored in a file with the.ls2 extension, the ADCP down data is stored in a file called ADCPDown and the general post processed navigation file is stored in the Mxxx.bnv file. Sensor Synchronisation The time synchronisation of the various on-board systems is important, especially where data from different systems are likely to be merged at a later date (post processed navigation data for the EM2000 is one example of this). Wherever possible the network time protocol (NTP - see for more details) system is used which allows for time comparisons with a resolution of better than 1millisecond. The Autosub3-2 computer is fed a GPS data stream and a program written at NOCS called GPSTime uses these data to set the clock on the system to match the GPS time. For NBP0901 the timing reference was the NMEA ZDA string derived from the ship s Seapath GPS system. All the Autosub related shipboard systems and the Autosub Logger run the NTP software and use the Autosub3-2 computer as their time reference. The Autosub logger timestamps each data record using this time and feeds it to the EM2000. The EM2000 uses this time as its initial reference but it appears to use its own internal clock for subsequent data timestamps. The offsets between the reference computer and the other NTP systems can be found in the peerstats file in the timing directory for each mission. The offset for the EM2000 can be obtained by comparing the time contained in the NMEA position datagram with the timestamp of that datagram. Seabird 9+ CTD system Autosub is fitted with a Seabird 9+ CTD system, which includes two sets of conductivity and temperature sensors. These are mounted in a ducted system with sea water pumped through them at a precisely known rate. Depth is measured by a Digiquartz pressure sensor. In addition, a Wetlabs Transmissometer is fitted to the same duct as the primary CT sensors and a Seabird SBE43 oxygen sensor is fitted in 24

25 the same duct as the secondary CT sensors. The output from these sensors is recorded at a rate of 24Hz. Table 5: The Seabird CTD sensors used and location. Sensor Location Serial Number Primary Temperature Port Side 4458 Primary Conductivity Port Side 2937 Secondary Temperature Starboard Side 4427 Secondary Conductivity Starboard Side 2938 Oxygen Sensor Starboard Side 1034 Transmissometer Port Side CST-979DR Data from the system is continuously logged whenever Autosub is switched on but, in order to prevent excessive wear on the pump, water is only pumped through the C/T sensors once a predetermined pressure threshold has been exceeded. The data are stored on the Autosub logger in a proprietary format but are translated into a Seabird format data file (.dat) at the end of each mission. This data file, together with the necessary configuration file was then passed to the scientific party for further processing. Sensor calibration data are stored in a separate file with the.con extension. For the NBP0901 cruise the data were processed using \Palmer09\CTD Setup\Palmer0901-SeabirdCCalsV2.con file, which contained calibration data from March 2005 for the T&C sensors and June 2006 for the oxygen sensor (which was unused since then). A basic clear path/blocked path calibration was done on the Transmissometer during the cruise and this was used in conjunction with the factory supplied calibration to give the calibration coefficients in the.con file. Some of the sensor calibrations are now rather old and it would be advisable to have these sensors re-calibrated when the equipment is returned. Kongsberg EM2000 Multibeam Swath System The Kongsberg EM2000 is a multibeam swath bathymetry system, which operates at a frequency of 200kHz and can give up to 111 beams of data with an angular coverage of up to +/-60 degrees under favourable conditions. On Autosub the instrument is triggered by a controller connected to the vehicle s LONWorks network. This controls the ping rate and also allows the trigger pulse to be synchronised with other systems on the vehicle in order to control interactions between instruments. This controller also sends time, range aiding and navigation information to the instrument. A second LONWorks controller sends attitude and depth information to the instrument. For NBP0901 this system was fitted with the transmit transducer mounted in the nose of the Autosub vehicle and the receive transducer mounted in the tail section. The transducers could be mounted either facing upwards or downwards depending upon the requirements of the particular mission. The transducers were mounted behind polythene windows in the vehicle s fibreglass outer panels. (See Table 6 in the Mechanical section of this report for exact sensor locations). The beam spacing was set to be equidistant and the maximum beam angles were +/- 60 degrees. The sensor roll settings were set to zero which meant that, when the 25

26 transducers were facing upwards (during missions ), the system would place the returns from the bottom of the ice shelf below Autosub rather than above it. Further post processing will be necessary to apply the correct rotations to the data to compensate for this. During missions the sensors were looking down, so no further rotations are necessary for these missions. The Kongsberg supplied control software was not used at all as it is known to have issues when used in AUV applications. The alternative NOCS EMControl and EMListen software was used to control and monitor the system. Good data were obtained from all missions where the EM2000 was within range of the target, with the exception of mission 431 where triggering appeared less reliable after the vehicle collided with the ice shelf. This fault could not be repeated in further tests. The vehicle was normally run at a distance of 100m from the target and a swath width of over 300m (total) was obtained at this distance. Argos Location System The Autosub vehicle is fitted with two transmitters that transmit to the Argos satellite system. This system can give the location of the vehicle and a limited amount of data can be transmitted, although this data capacity is not used on Autosub. A local direction finding receiver can also monitor these transmissions and give a signal strength and bearing to the transmitter. This is useful for locating the vehicle when it first surfaces after a mission. For NBP0901 the direction finding receiver was located on the Ice Bridge with the antenna mounted above the ice bridge. The length of the antenna cables is limited and they cannot be extended without compromising the direction finding performance so the receiver must be mounted in a dry space close to the antenna. The receiver has a serial output, which was fed to a Digi Port Server TS4 serial to Ethernet bridge (labelled Server 3). This was configured to send the serial data encapsulated in UDP packets to port 2002 on certain computers on the network. These computers were running the NOCS EMListen program to monitor and log the output from the receiver. Autosub3 ADCP configuration ADCP Physical Arrangement Autosub has two RDI ADCPs: A 300 khz RDI Workhorse pointing upwards. A 150 khz RDI Workhorse pointing downwards. Both can provide velocities in bottom tracking mode (or ice tracking for the upward looking ADCP), as well as current profiling. The range information for the four beams is also used in the control of the vehicle, where it is set to keep a constant distance from the seafloor, or the ice base. The collision avoidance system also takes input from the ADCP beam ranges. Both are currently set with 8m profiling bins. This can be changed, although shorter bins will give higher noise values (particularly for the down looking 150 khz instrument). 26

27 ADCP mechanical configuration: ADCP UP (300 khz Workhorse Navigator): Looking down from above. FWD 3 1 PORT STARBOARD 2 4 AFT Orientation is set as -45 degrees (EA -4500). / / ADCP DOWN (150 khz Workhorse Navigator): Looking down from above. FWD 1 3 PORT STARBOARD 4 2 AFT Orientation is set as 45 degrees (EA +4500). / / 27

28 Autosub3 Mechanical and Operational Aspects Preparation and Mobilisation in Punta Arenas The Autosub gantry places very large vertical deck loads and required a custom made deck/gantry interface plate to reduce the vertical deck loads to an acceptable level. Autosub mobilisation began in Punta Arenas in December 08 prior to cruise NBP0901. The launch and recovery gantry was bolted to the deck matrix with an adaptor plate designed by ourselves and constructed in Punta. The container/garage assembly was secured in place and the Autosub vehicle secured within the garage container set up. We had problems with some of our wooden equipment not meeting Chilean requirements for heat treatment. Our acoustic fish bases, weights box, and railway sleepers (used for blocking up the Autosub chocks) were destroyed, the replacements being made from locally sourced wood. Following discussion with Raytheon employees the aft garage roof was lowered and container doors refitted for crossing the Drake Passage (this turned out to be essential, the aft deck of the Palmer becomes awash to a depth of 0.5 m, even in moderate seas). The Autosub power pack had been modified back in the UK for use in the Antarctic and installed into a 10 foot container that was fitted to the port aft quarter. Installation of acoustic equipment into the aft winch control room was completed. The Mission control computers were set up in the main lab and aerial systems were installed. On arrival at Pine Island Bay the aft container was reconfigured for Autosub operations. Figures 6 and 7 show views of the gantry and Autosub s purpose-made containers on the aft deck. Figure 6: Side view of the Autosub containers. 28

29 Figure 7: Plan view of Autosub containers on the N B Palmer Aft deck. Vehicle configuration Figure 8 contains two photos showing the sensor layout within Autosub. The following shows the relative positions of the sensors on board, as well as listing the serial numbers of the components. Table 6 gives the sensor positions with respect to the vehicle datum, the fwd stainless bulkhead ring. 22cm 6cm Pump Oxygen sensor 46cm Exhaust Figure 8: Sensor layout. CTD mechanical arrangements on Autosub-3 for NBP09-01 Starboard side (secondary system): The C and T sensors are contained within an insulating foam box attached to the side panel with Velcro, the outline of which can be seen in black on the photographs. The outlet (upper) and exhaust (lower) ports are inside the box, which has a number of holes fore and aft (see photographs) to enable a flow of water as the vehicle moves. Overall length of tubing to Oxygen sensor: Inside the insulating foam box: 29

30 6cm of ¼ (12mm diameter) tubing to temperature probe Connector to conductivity cell approx. 4cm ¼ tubing Conductivity cell 25cm long Outside the foam box: 22cm plus 46cm = 68cm of ½ tubing to the lower end of the oxygen sensor exhaust via siphon break and seabird pump. Oxygen sensor: Seabird SBE-43 s/n Temperature sensor: Seabird SBE-3P s/n 03P4427 Conductivity sensor: Seabird SBE-4C s/n Port side (primary system): The pump and sensor positions on the starboard side are a mirror image of the port side, with the transmissometer downstream and inside the vehicle s nose cavity. Overall length of tubing to transmissometer: Inside the foam box: 6cm of ¼ (12mm diameter) tubing to temperature probe Connector to conductivity cell approx. 4cm ¼ tubing Conductivity cell 25cm long Outside the foam box: 22cm plus 48cm = 70cm of ½ tubing to the aft end of the transmissometer Exhaust via siphon break and seabird pump. Transmissometer: Wet Labs part No s/n CST-979DR Temperature sensor: Seabird SBE-3P s/n 03P4458 Conductivity sensor: Seabird SBE-4C s/n Table 6: Positions of sensors relative to vehicle datum. Sensor CTD (port and starboard) SBE-3P temp and SBE-4C cond. sensors SBE 43 Oxygen sensor Digiquartz Depth sensor EM 2000 Transmitter EM 2000 Receiver Longitudin al centre of gravity wrt vehicle datum (m) Vertical centre of gravity wrt vehicle datum (m) Remarks Ports side, Temp1, Ser No. 03P4458. Cond1 Ser No Starboard side, Temp 2 Ser No. 03P Cond2 Ser No Plumbed in the starboard side with Temp2 and Cond 2, s/n Position of transducer head looking upwards Position of transducer head looking upwards 30

31 EM 2000 Transmitter Position of transducer head looking downwards EM 2000 Receiver Position of transducer head looking downwards 300kHz ADCP Position of transducer heads looking upwards 150kHz ADCP Position of transducer heads looking downwards Wet Labs Transmissometer Installed horizontally plumbed into the port side with Temp 1 and Cond 1 s/n CST-979DR AUV Operations: mechanical configurations The ballast and trim of Autosub has to be carefully tailored for the local conditions. It is set to be 8 to 12kg buoyant, with neutral trim in order for it to both float and control depth without difficulty. Seawater density measurements were made using data from the first CTD station. The Autosub vehicle was floated in Pine Island Bay with lines attached as a check that no significant change in weight or trim had occurred during the reconstruction of the vehicle in Punta. The water density and vehicle weight estimates indicated a buoyancy of 10kg. Table 12 shows the history of the density measurements and the changes made to the vehicles ballast and trim throughout the cruise. Autosub started the leg with fresh batteries and underwent three battery changes throughout the science time. The battery packs were from the supplier Steatite Ltd, Redditch, (Fujitsu cells). The vehicle performed well with few faults. The mechanical problems encountered were:- Penetrator lead to 150kz ADCP causing network problems (believed to have been damaged in transit). Sluggish performance of the launch and recovery gantry in icy conditions. Vehicle maintenance after each mission Battery changes Flushing of CTDs with filtered seawater Cleaning of transmissometer Recovery line packing Packing the Jack in the Box grappling line (This was removed from recovery system to simplify recovery as a small boat was always used in the operation) Taping of panel seams for streamlining. 31

32 Table 7: AUV trim and buoyancy. Date Jan 09 Pine Island 17/01/09 Prior to float test 22/01/09 Pine Island 26/01/09 Pine Island 29/01/09 Pine Island Location Water Density (kg/m 3 ) Changes made to Autosub ballast Changes made to Autosub trim Remarks 1027 No change No change Start of cruise set up kg from tail Trim 0-0 Float test seems to indicate all ok -9.9kg 1027 No change No change 1st battery change -13.9kg 1027 No change No change 2nd battery change -13.7kg 1027 No change No change 3rd battery change -14.0kg Recovery operations On the first recovery we kept with the method we had developed over the preceding years. This requires a small float, jettisoned from Autosub, to be grappled. The float is connected by a light line to heavy lift lines held within the Autosub vehicle, which are pulled aboard and coupled to the Autosub gantry. Autosub is then winched aboard. R/V Nathaniel B Palmer is a twin screw, direct drive, fixed RPM, variable pitch ice breaker. During the first recovery it was found that the cut away counter and the huge suction generated from the main propellers would try and draw the Autosub under the stern. This was only averted by the fact that the ship has a very low freeboard, which enabled the use of boat hooks figure 9. Figure 9: Autosub being fended off. 32

33 A meeting was called, and after discussion with officers, ship technicians and the Autosub team, a decision was made to use a method the Autosub team had not used in the past. The method required the use of a small boat to act as a drogue and to keep the Autosub vehicle under control while the recovery operation was under way (Figure 10). This method was refined over the next couple of recoveries, with the use of the Autosub recovery (Jack) float being abandoned in favour of a direct pick up by the small boat. It became a practical, slick and more importantly safe method of recovery. The small boat was used in winds of up to 25kts with a wave height of about 0.6m (2 feet). The use of a small boat can severely limit the operational window, but with the work site being within a bay and sea being damped by ice, sea conditions were never a problem. Figure 10: Using small boat for recovery. Mission 431 damage On mission 431 Autosub was in collision with ice while under Pine Island Glacier. Damage caused consisted of the following: Starboard fibreglass cover split and abraded Conductivity quick-fit pipe connector sheared All fwd nose metal work distorted Rear wings bent FWD Argos antenna broken off Pins to the LinkQuest transducer bent The damage was fixed over the following 8 hours, with all damaged and broken components replaced or repaired. The sub was re-batteried and ready for deployment the following morning. The following photographs (Figure 11) show the damage to the nose panel and the extent of distortion to the nose section. 33

34 Figure 11: Damage to nose. Acoustic operations Autosub carries three independent acoustic communications devices, as well as the acoustic measuring devices: the upward and downward looking ADCPs and the EM2000 swath bathymetry echo sounder. From our experience, the Nathaniel B. Palmer appears to be an acoustically noisy ship. Even with all the ship s sonars switched off, ranges on the Autosub LinkQuest rarely exceeded 3 km, or less than half that achieved on the RRS Discovery, RRS James Cook, and the MV Terschelling. The emergency beacon was clearly audible to a range of 10km, and could be detected intermittently in quiet weather (particularly with the ships propulsion system disengaged) up to 20km away. Emergency beacon The emergency beacon transmits a 4.5 khz narrow-band chirp to communicate over a long range. It was used on all missions to track the sub on its outward leg under the glacier. The transducer is deployed by hand, via a sheave mounted on the ship s rail, to a depth of about 80m (100m of cable, from the exit point of the work space). The deck unit and controlling laptop (Toshiba) were housed in the aft control cabin. The laptop was used to run the NOC beaconview program. Initial electrical noise problems were traced to a fault in the sea cable producing a ground loop fault. After repair (see fault log) good ranges were obtained, limited only by the acoustic noise from the ship. LinkQuest digital telemetry system Autosub carries a LinkQuest Tracklink modem and transponder, which is used to track and command the sub when underwater, and to receive a limited subset of digital data. The transponder tube is mounted in the lower part of the nose section, with the transducer on a bracket inside and at the top of the nose section. The bracket is designed to swivel on an impact from above, for example if the sub collides with the stern of the ship. This was illustrated by the collisions during mission M431, where the 16mm threaded rod was skewed and the pins of the flat connector (Seaconn 34

35 FAWM-3P) were bent downwards but the transducer survived and was still working at the end of the mission. The damage to the mounting is visible in the left hand photograph of the damaged nose section (Figure 11). The LinkQuest worked well at ranges of up to 2-3km, although the tracking display, which was repeated on a monitor on the bridge, was only available in head-up mode. ORE LXT acoustic transponder This transponder (also known as the dumb transponder) is used for tracking, as a backup for the LinkQuest. As yet, it is not possible to interface the LXT data to the LinkQuest or any other of our tracking displays. The transponder, which contains a battery pack sufficient for a minimum of 3 months operations, is mounted in the aft section of the sub. The transducer is mounted forward of the tail fin, and protrudes through the upper panel. The LXT worked well, tracking the sub to a similar range as the LinkQuest. Acoustic Tow Fish With the exception of the emergency beacon receiver array, all the ship born acoustic transducers are mounted on the tow fish. This is a 380 kg towed body that was deployed to a depth of 12m from the starboard A-frame (Figure 12). It was suspended using the 5/16 wire on the lower waterfall winch (SWL 3 tons), the fish s own cable being paid out by hand and made fast once the fish was at depth. Figure 12: Acoustic fish being deployed from starboard 'A' frame. The fish (Figure 13) contains the LinkQuest surface transceiver, the large diameter receiver and the transmitter. The receiver looks down through a hole in the lower fairing, (about half the area of that fairing). The transmitter is held in a clamp on the tail bar. The LXT transducer is held in a clamp of the same design. These clamps are set up to hold the transducers out on whichever side of the fish is outboard during recovery and deployment, to minimise the risk of damage. Apart from one deck connector failure, the systems on the fish worked without any incident. 35

36 Figure 13: The aperture in the lower fairing for the LinkQuest receiver and the transducer tail clamps. Gantry, hydraulic power pack and electrical power The deployment and recovery gantry for Autosub was run from our own hydraulic pump (HPU) mounted in a 10 foot container, mounted adjacent to the main containers on the port quarter. The two main containers were joined to make the garage and workshop with a fabric roller door and heating to keep the vehicle batteries warm (Figure 14). Figure 14: The garage container showing the heaters. Gantry and HPU The ship was able to provide 480VAC 60Hz three phase power at up to 100 Amps, with an earth connected to the ship s earth. Phases are anticlockwise from P1 to P3. The HPU s original soft start system was incompatible with the supply voltage, and so a star-delta system was wired in immediately prior to shipping, and commissioned on the ship in Punta Arenas. Once the unit was commissioned, it ran with no problems. 36

37 Two 2.2kW heaters were fitted to the oil tanks of the hydraulic pump for cold weather operations. One heater was left permanently on, keeping the oil at a thermostatically controlled Celsius, the other available as a backup. An isolating transformer was used to supply the contactors control circuit (to overcome power supply neutral issues). In practice the lowest air temperature experienced during operations was -11C, with the norm being closer to -4 C. Containers and other power The workshop containers were supplied with 208VAC three phase, up to 60A, with a neutral and an earth. Figure 15 shows the initial planned layout, and the maximum individual hotel loads. In practice the wall sockets rarely exceeded 100W. The phases were split into two supplies after entering the container. Two phases provided 208V to run the fan heaters. The third phase and neutral were used to drive a 5kW 120V to 240V transformer. This allowed the rest of the container to be used on 240VAC 60Hz. All equipment was checked before use, and items with suitable power requirements were run from the ship s 120V 60Hz outlets. A large number of 240V converters were brought along, 100W, 500W and 1kW in size. Figure 12 shows wiring layout used in the containers and HPU system. Figure 15: Wiring layout. 37

38 Autosub3 Data Processing Report Adrian Jenkins, Pierre Dutrieux CTD data Introduction A Seabird SBE9+ CTD system with two sets of sensors was fitted on Autosub3. The primary system on the port side comprised temperature and conductivity sensors connected via the standard Seabird TC duct, then a Wetlabs transmissometer and Seabird submersible pump. The secondary system on the starboard side had a duplicate TC setup followed by a Seabird SBE43 dissolved oxygen sensor and a second submersible pump. Although some of the sensors had been recalibrated prior to the cruise, the results were unsatisfactory, giving numbers that not only differed markedly from the ship s CTD, but also showed offsets between sensors. The original calibrations, dating mainly from 2005 gave much better results, and these were used throughout the cruise. Post-cruise recalibration of all the sensors is a high priority. Processing steps The data were processed using version 7.18b of the Seabird Seasoft-Win32: SBE Data Processing Software, following the recommendations of Povl Abrahamsen and Kate Stansfield in their Autosub ADCP data processing toolbox: Quick start guide, with a couple of minor changes. Sensor calibration coefficients were from the Palmer0901-SeabirdCCalsV2.con file, and are listed below: Date: 02/24/2009 Instrument configuration file: C:\NBP0901\Autosub\M427\CTDdata\Palmer0901-SeabirdCCalsV2.con Configuration report for SBE 911plus/917plus CTD Frequency channels suppressed : 0 Voltage words suppressed : 0 Computer interface : RS-232C Scans to average : 1 NMEA position data added : No NMEA depth data added : No NMEA time added : No Surface PAR voltage added : No Scan time added : No 1) Frequency 0, Temperature Serial number : 4458 Calibrated on : 16-Mar-05 G : e-003 H : e

39 I : e-005 J : e-006 F0 : Slope : Offset : ) Frequency 1, Conductivity Serial number : 2937 Calibrated on : 11-Mar-05 G : e+001 H : e+000 I : e-004 J : e-005 CTcor : e-006 CPcor : e-008 Slope : Offset : ) Frequency 2, Pressure, Digiquartz with TC Serial number : Calibrated on : 14-Feb-08 C1 : e+004 C2 : e+000 C3 : e-002 D1 : e-002 D2 : e+000 T1 : e+001 T2 : e-004 T3 : e-006 T4 : e-011 T5 : e+000 Slope : Offset : AD590M : e-002 AD590B : e+000 4) Frequency 3, Temperature, 2 Serial number : 4457 Calibrated on : 16-Mar-05 G : e-003 H : e-004 I : e-005 J : e-006 F0 : Slope : Offset : ) Frequency 4, Conductivity, 2 Serial number : 2938 Calibrated on : 11-Mar-05 G : e+001 H : e+000 I : e-004 J : e-005 CTcor : e-006 CPcor : e-008 Slope :

40 Offset : ) A/D voltage 0, Oxygen, SBE 43 Serial number : 1034 Calibrated on : 08-Jun-06 Equation : Owens-Millard Soc : e-001 Boc : Offset : Tcor : Pcor : 1.35e-004 Tau : 0.0 7) A/D voltage 1, Free 8) A/D voltage 2, Transmissometer, Chelsea/Seatech/Wetlab CStar Serial number : CST979DR Calibrated on : 19-Jan-2009 M : B : Path length : ) A/D voltage 3, Free 10) A/D voltage 4, Free 11) A/D voltage 5, Free 12) A/D voltage 6, Free 13) A/D voltage 7, Free The processing steps were as follows: 1) Data Conversion: Convert raw data from CTD (.dat file) to engineering units, storing the converted data in.cnv file. Ouput variables were: #1 Scan count #2 Time, Elapsed (seconds) #3 Pressure, digiquartz (db) #4 Temperature (ITS-90, deg C) #5 Conductivity (S/m) #6 Temperature, 2 (ITS-90, deg C) #7 Conductivity, 2 (S/m) #8 Oxygen Voltage, SBE43 (V) #9 Beam Transmission, Chelsea/Seatech/Wetlabs Cstar (%) #10 Beam Attenuation, Chelsea/Seatech/Wetlabs Cstar (1/m) #11 Pump status Oxygen voltage was used at this stage, since the calculation of dissolved oxygen concentrations requires both temperature and salinity as inputs. Both beam transmission and attenuation were included, as it does not seem to be possible to generate the alternative at a later stage. Pump status was included to check whether some bad readings on a couple of descents were because of a delay in the start of the 40

41 pumps. Data output in binary format reduced files size and greatly speeded up the later processing stages. 2) Align CTD: Align data relative to pressure (typically used for conductivity, temperature, and oxygen). Advance values (seconds) used were: Temperature (ITS-90, deg C) 0 Conductivity (S/m) Temperature, 2 (ITS-90, deg C) 0 Conductivity, 2 (S/m) Oxygen Voltage, SBE43 6 The values for temperature and conductivity were those recommended by Seabird. Since the TC sensors were connected by the standard duct, the relative timing of these should be correct. The 6 s delay for oxygen was as recommended by Povl and Kate. The majority of this delay comes from the slow response time of the sensor; about 5 s in 0ºC water. The additional 1 s delay comes from the transit time of the water through the pipe-work from the conductivity cell. This latter number may need finetuning following measurements made by Andy and Dave of the present setup. An additional (small) delay could be added to all sensors to allow for the water transit from the intake to the temperature sensor. 3) Cell Thermal Mass: Perform conductivity thermal mass correction. Seabird recommended values were used for both conductivity sensors: Thermal anomaly amplitude (alpha) 0.03 Thermal anomaly time constant (1/beta) 7 4) Derive: Calculate salinity, density, sound velocity, oxygen, potential temperature, dynamic height, etc. Output variables were: #1 Salinity (PSU) #2 Potential Temperature (ITS-90, deg C) #3 Salinity, 2 (PSU) #4 Potential Temperature, 2 (ITS-90, deg C) #6 Oxygen, SBE43 (ml/l) #7 Density (sigma-theta, kg m -3 ) #8 Density, 2 (sigma-theta, kg m -3 ) 5) Wild Edit: Mark a data value with badflag to eliminate wild points. Seabird default values were used: Standard deviations for pass one 2 Standard deviations for pass two 20 Scans per block 100 Keep data within this distance of the mean 0 The same processing was applied to all variables. 6) Bin Average: Average data, basing bins on pressure, depth, scan number, or time range. The following settings were used: Bin type Time, seconds Bin size 2 This produced the standard two-second binned file, directly comparable to the other Autosub data files. For convenience a ten-second binned file was also produced, to make plotting of multiple files more manageable. 41

42 7) Finally the (still binary) bin-averaged files were converted to Matlab structures using the cnv2mat_asub utility: >> cnv2mat_asub('m###ctd_#s.cnv',[],0,inf); Preliminary analyses The first test mission (M427) appeared to produce poor quality data. Not only did they disagree with data produced by the shipboard CTD at a site nearby, but there was no internal consistency either, with the two conductivity cells showing a marked offset. Reverting to earlier calibrations improved matters and brought the primary and secondary conductivity sensors in line with each other and the ship CTD. However the dissolved oxygen sensor still showed a large difference between the profile obtained when the sub was diving and that recorded while surfacing and the transmissometer data showed a large offset and excessive noise (Figure 16). Differences between down- and up-cast oxygen data are quite common, as a result of hysteresis in the cell, but the differences illustrated in Figure 16 are much larger than would normally be expected. Following some remedial action, including cleaning and recalibrating the transmissometer, a test dive was run at the start of the first science mission (M428). Once the sub had surfaced, data were downloaded and checked, before the mission was allowed to continue. Performance of all sensors looked much better, with good matches between the ship and sub profiles and very good agreement on propertyproperty plots (Figure 16). The only remaining problem was a small offset of about 0.5 ml/l in the dissolved oxygen trace, which can presumably be resolved with postcruise calibration. The small remaining pressure hysteresis appeared no worse than that on the shipboard sensors. The earlier problems could have been due to the fact that this was a new sensor that had not been in the water prior to its deployment on M427. Cleaning and recalibration of the transmissometer clearly sorted out all its problems (Figure 16). The sensors all performed well for the remainder of the cruise, with the only slight problems being those caused by the collision with the ice on mission M431. There seems to have been a jump in the dissolved oxygen sensor calibration that affected the return leg after the collision (Figure 17), but it appeared to return to its previous value on later missions. The piping on the primary sensors was also disconnected by the impact between the temperature and conductivity sensors and the transmissometer, so that the pump was flushing only the latter and this with water internal to the sub s forward instrument compartment. Beam attenuation was higher on the return leg, but the sub was higher in the water column for most of that leg (Figure 17), where the attenuation should have been stronger, and the temperature and conductivity records look good until the sub stopped at the end of the mission. The forward motion of the sub was apparently enough to keep the sensors reasonably flushed and perhaps the loose panels kept the instrument bay better flushed than it would normally be. The derived salinity contains many spikes on the return journey as a result of misalignment of the temperature and conductivity records (Figure 17). However, the sub travelled at a steady speed, so the flow of water through the sensors should have been at a fairly constant rate, and it should be possible to reduce the number and size 42

43 of spikes, if not eliminate them altogether, by realigning the temperature and conductivity data. Figure 16: Comparison between CTD data collected during Autosub3 test missions (M427 and the start of M428), and those collected near-contemporaneously using the Seabird 911 CTD on Nathaniel B Palmer. Mission 427 and Station 15 were separated by about 20 km, but CDW properties near the seabed remain a good reference point. Mission 428 and Station 32 were close, so comparison through the entire water column is more valid. Between stations 15 and 32 the primary oxygen sensor on the ship CTD was swapped, eliminating the offset apparent on the earlier station. 43

44 Figure 17: Basic and derived variables observed over the two hour period either side of the collision with the ice shelf base on M431. The collision occurred near the point of minimum pressure that occurs just before 19.8 hours. Primary sensor data are in blue, secondary sensor data in red. For comparison oxygen concentrations derived using the secondary temperature and salinity data are shown in black (only visible near the time of the impact). There is a sharp spike in oxygen voltage that appears to precede the impact. Subsequently the secondary sensors were most affected (the damage to the sub was on the port side), although the primaries look noisier for much of the period immediately following the impact. After a broad positive deviation in both primary temperature and conductivity they settle back down, but the slow flushing is evident by the salinity spiking. The drop in oxygen voltage appears coincident with the onset of spiking, although the other secondary sensors show little reaction at this point. 44

45 There was a subtle problem with the configuration of the Autosub3 CTD, in that the single dissolved oxygen sensor was plumbed in with the secondary TC sensors. However, when only one dissolved oxygen sensor is in use the Seabird software defaults to using the primary temperature and salinity to calculate the oxygen concentration. The only way to force the use of the secondary sensor data is to process these separately from the primary sensor data; a tedious procedure. In practice however the errors are small. On M428, for example, differences in values obtained by normal and separate processing are of the order 10-3 ml/l. The differences are larger on M431, where for the return leg the primary sensors were free-flushing rather than pumped, but they remain of the order 10-2 ml/l or less. ADCP data Introduction Two ADCPs are fitted on Autosub3. They are used for navigation as well as for current measurements. The first one, a 150 khz Workhorse Navigator from RDI, is directed upward and is intended to track the ice surface within a maximum range of 200 m. The second one is a 300 khz Workhorse Navigator intended to track the seabed within a maximum range of 400 m. Test mission 427 revealed that the upward looking ADCP had been improperly mounted, with an orientation of +45 degrees instead of -45 degrees. This was fixed for all other missions. In general, the instruments mounted on Autosub have been found to give relatively poor results for current measurements, with short ranges and unusually high noise levels. This could be due to the presence of the Autosub protective shell, a potentially reflective surface in front of the instruments. Analysis of the raw data as well as pre-processing using the software provided by Povl Abrahamsen (epab@bas.ac.uk) and Kate Stansfield (ks1@soc.soton.ac.uk) were used to assess the reliability of the measurements following each mission. On mission 428, the upward looking ADCP was giving such poor results that its configuration was changed from broadband to narrowband. Settings A list of parameter settings is provided in table 1. Table 8: ADCP specific RDI settings used. Variable ADCP down value ADCP up value BA BC BD BE BF BI BK 0 0 BL 0320,0640, ,0320,0480 BM

46 BP BR 2 2 BX BZ CB CF EA EB EC EH EP ER ES ET EX EZ WA WB 1 1 WC WD WE WF WI 0 0 WJ 1 1 WL 001, ,005 WN 021 WP WS WT WV TE 00:00: :00:00.00 TF **/**/**/** --- **/**/**/** --- TP 00:00: :00:00.00 TS 09/01/29 18:50: /01/29 10:44:48 -- Mechanical configuration ADCP configurations are represented in table 2. The downward looking ADCP is oriented at -45 degrees in the Autosub reference frame. The upward looking ADCP is oriented at 45 degrees. Beam tilt is 60 degrees from the Autosub horizontal plane. 46

47 ADCP down Forward ADCP up Forward Port Starboard Port Starboard Backward Backward Table 9: ADCPs configuration. Processing Pre-processing of the ADCP data followed the ADCP Quickstart tutorial (Povl Abrahamsen and Kate Stansfield). It generally gave surprisingly poor results. Similar pre-processing of data from previous campaigns, namely JR97, agreed with published results, giving relative confidence in the pre-processing software. Important modifications in the Autosub ADCP and navigation format have been made over the years since JR97. The data are now available in a suite of Matlab files, sometimes under different names and/or different units. Some work was necessary to recreate proper files and variables necessary for the pre-processing programs. Errors might therefore have arisen from inadequate data units and or qualifications. Additional work is definitely required to make sense of these observations. More specifically, all pre-processing steps were done using Matlab. For each mission they are summarized in the Matlab script named process_adcp.m in the Proc directory inside the mission directory. The first step involved rewriting of the CTD data in Matlab format. ADCP pre-processing has been done using pre-processed CTD data. Using fully processed CTD data is not expected to make significant differences, but remains to be done. The second step involved rewriting the Matlab navigation (.bnv ) and Matlab ADCP (.ls2 ) files, appended with the suffix orig, into new Matlab files suitable for the subsequent processing routines. Finally, the pre-processing was done in three separate steps involving operator choices. These choices are recorded in log files. For now, all processing has used default choices. A typical result where the ADCP data has been further averaged into 100 m length along-track bins is shown in Figure

48 Figure 18: Depth, along-track diagram of Autosub pre-processed ADCP zonal (upper panel) and meridional (lower panel) velocity from mission 434. The thin black line represents the vehicle path, the thick black line represents the bottom depth as recorded by the downward looking ADCP, and the thick grey line represents the ice draft as recorded by the upward looking ADCP. SWATH data Introduction A SIMRAD EM2000 multibeam sonar was mounted on the Autosub. For missions 427 to 431, it was mounted looking upward to map the underside of the PIG. For safety reasons, later missions were not intended to be close enough (about 120 m) to the ice for extended amounts of time. The multibeam swath was then used to map the topography of the seabed beneath the glacier. The validity of the raw data was checked using the Kongsberg proprietary software made available by the Autosub team. Further processing, including ping editing will be done, and has been done to a certain extent, using the MB system software from David W. Caress and Dale N. Chayes. A free, powerful and open source software for processing and display of multibeam data, it is used on the R/V N. B. Palmer on a daily basis. We will use it mainly to edit the data, and to convert corrected data to NETCDF format. Further rotation for correction of upward looking data, gridding and plotting are then done using Matlab. 48

49 Processing steps The basic steps for processing have been tested during the cruise, and applied to a relatively small part of the entire recorded dataset. Further work following the same steps remains to be done. Raw data are placed in the em2000/raw directory within each mission directory. Processed data are placed in the corresponding em2000/proc directory. The shell script prepare_swath has been applied to each mission to convert raw SIMRAD files into the mb57 format files necessary for editing using the MB system function mbcopy, e.g.: mbcopy -F56/57 -I005_ _ all -O005_ _ mb57 where 005_ _ all is the name of the raw file. In the em2000/proc/mb57 directory, each file can now be edited using the MB system mbedit function, e.g. using the following command line: mbedit -G -S -I0005_ _ mb57 where 0005_ _ mb57 is the file to be edited. Once a file has gone through ping editing, the edit can be applied, either using the shell script process_swath which is intended to work on entire mission directories, or on individual files using the two following steps. The first one involves the mbprocess function, e.g. with the following command line: mbprocess -I0005_ _ mb57 -O0005_ _055516p.mb57 The processed file can then be written to a NETCDF file suitable for further work in Matlab using the mblist command, and then moved to the nc directory, e.g.: mblist -C -A -MA -ON#XYztPRSHdCcpl -I0005_ _055516p.mb57 mv mblist.nc../nc/0005_ _055516p.nc When the multibeam was oriented upward, the resulting netcdf file is then opened in Matlab and a simple plane rotation is applied using the invert_swath.m function, available in the swath_mfiles directory. A sample result is shown in Figure

50 Figure 19: Example of multibeam data from mission 433, showing a topographic feature in front of the PIG. Here the data has been ping edited but remains ungridded. 50

51 Appendix 1. Pre-cruise Risk Assessment Procedures 1) from Helen Beadman, NERC, to Adrian Jenkins, BAS, on 24/05/06, outlining the risk assessment procedure to be followed: Dear Adrian In response to your and Mike's I have outlined below the process that we hope to go through to decide on the risk associated with your cruise and therefore the likelihood of it going ahead. One of the key recommendations from the Autosub Loss Report is that NERC should define acceptable risk criteria for future high risk AUV projects, and that these criteria should be capable of being translated into reliability requirements for the AUV team to meet. Gwyn Griffiths has taken this recommendation forwards by producing a paper which proposes a method to do this, and I understand that he has recently sent a copy of the paper to you. This paper will be presented at the Marine Facilities Review Group (12th June) and we hope that following discussions we will then have a proposed method for assessing risk, that we can use to assess the risk of your proposal. You are very welcome to attend the meeting, or to submit some written comments if you would prefer, and I know that Gwyn has asked for your comments as well. While not wanting to pre-judge the outcome of the discussions at the MFRG we are keen to start the process of evaluating the potential risk associated with your proposal. In this light we would like, if at all possible, to run Gwyn's risk calculation for the specifics of your cruise (Box 2 in Gwyn's paper). To do this we would need information on each mission - its proposed length (km) and whether it would be in open water/sea ice/shelf ice. If you could provide us with 'preferred' and 'acceptable' options this would also give us some scope to look for best options. Running the calculation will give us some indication of the current risk level, but it should be stressed that this would only be a preliminary step as the reliability factors for Autosub will continue to be updated, to reflect the recent trials cruise and the subsequent Autosub cruises that are programmed for this year, thereby changing the resulting risk. I appreciate that you would like an answer as soon as possible, and we will try to ensure that a decision is made as speedily as we can. However, we do want to make sure that we have a robust process in place that can be used for subsequent proposals. Please give me a call if you would like to discuss any of this further. Best wishes Helen Dr Helen Beadman Science Programmes Officer Natural Environment Research Council Polaris House North Star Avenue Swindon SN2 1EU Tel: +44 (0) Fax: +44 (0) habe@nerc.ac.uk 51

52 2) from Adrian Jenkins, BAS, to Gwyn Griffiths, NOCS, 06/07/06, outlining the proposed set of missions for Pine Island Glacier: Hi Gwyn Thanks again for taking the time to update me on the current situation with Autosub. It will be very disappointing not to go next year, but taking the long-term view, I would rather wait (knowing that NERC approval/funding is in place) for an opportunity to really push the science forward, than rush into a cruise now with the risk of the scientific gain being much less. Of course, we could run open water and sub-sea-ice missions on the outer Amundesn shelf, but if that is all we could do, it would not address the fundamental questions in the proposal. It is also impossible to guess where the sea ice will be, so really difficult to state the mission length and type for exploring parts of the continental shelf north of the ice fronts. For this reason I list only the Pine Island Glacier missions that would be required. They would be: 1) 60 km open water mission to depth of 200 m, close to the ice front. 2) 60 km open water mission to depth of 600 m, close to the ice front. 3) 60 km open water mission to depth of 1000 m, close to the ice front. 4) 3 x 60 km sub-ice-shelf missions to depth 600 m, in the outer half of the cavity. 5) 3 x 120 km sub-ice-shelf missions to depth of 1000 m, to the "minimum headroom" limit of the cavity (at least one of these is likely to be somewhat shorter than 120 km). If fast ice were in Pine Island Bay it might be necessary to add ~60 km to all of the mission lengths and they would then all be beneath (effectively) permanent ice. In this case I guess none of the missions would be achieveable at present. An alternative could be Dotson Ice Shelf. A similar set of missions to 1)-5) beneath Dotson could address similar questions. If we could only run the open water missions, I do not think we would gain enough over the ship-based work to justify the effort of shipping autosub. If we could run 3 x 60 km missions in the outer half of the cavity, two to depths of 600 m and one to a depth of 1000 m, I think we would come away with a very worthwhile dataset. Therefore the break even point comes at 4). If the risk of 3 x 60 km missions were acceptable, I think it would be worthwhile going. I suspect that this will come out well beyond the acceptable risk. If it were only just beyond, and 2 x 60 km missions was deemed OK, I would have a serious think. If just one 60 km mission were acceptable I would prefer to wait until a later date. Having thought about this at length now, I have practical question about the implementation of the RMP. Supposing that your analysis of the probability of loss for 3 x 60 km open water missions and 1 x 60 km sub-ice-shelf mission happened to be less than Ed Hilll's acceptable probability of loss, but that 3 open water plus 2 or more sub-ice-shelf missions put the probability of loss as too high. Preusmably I would be allowed to go and undertake the four approved missions. Lets suppose they went ahead without problem. What now? Does the sub now sit on deck for the rest of the cruise? Or is the process interactive? Even without the new statistics (which would improve things) one more 60 km mission beneath the ice shelf would now have an acceptable risk. So would that be approved at this stage? And if the sub came back, would one further mission be approved as acceptable under the RMP? If the new statistics were included, even a 120 km mission might become acceptabe? So it might be possible to achieve much more than was originally approved?? Similarly if autosub had problems on its open water missions, they would presumably feed into the RMP at this stage and potentially exclude the sub-ice mission? These questions are hypothetical only. I have no doubt from our discussion yesterday that the current assessment will be that the only possible option is postponement. But I would be interested to know the answers. Best wishes Adrian 52

53 3) Risk assessment document prepared by Gwyn Griffiths, NOCS, 31/07/06: FAULT ASSESSMENT AND ESTIMATED PROBABILITY OF LOSS UNDER ICE SHELF FOR AUTOSUB3 DERIVED FROM PERFORMANCE ON DISCOVERY 295T JULY 2005, TERSCHELLING MAY 2006, DISCOVERY 306 JUNE-JULY 2006 AND TERSCHELLING JULY 2006, FORMING STEPS 2 AND 3 OF THE RISK MANAGEMENT PROCESS AUV. Gwyn Griffiths National Oceanography Centre, Southampton. E: gxg@noc.soton.ac.uk T: Based on mission fault information supplied by Steven McPhail SUMMARY The fault history of Autosub3 during its 2005 and 2006 engineering trials campaigns are analysed to provide an estimate of the probability of loss if used on a campaign of under ice missions proposed by Dr A Jenkins (BAS). Dr Jenkins requirement (Annex A) forms Step 2 of the Risk Assessment Process-AUV (RMP-AUV); this analysis is Step 3. The responsible owner has yet to declare their acceptable risk (Step 1). Based on four different analysis methods, we estimate that for Dr Jenkins minimum requirements, and no fast ice present, the optimistic probability of loss to be between 35 and 53% (pessimistic estimate is 57 87%). For his full requirements the optimistic estimate lies between 57 88% (the pessimistic between 85 97%). The probability of loss is increased in the vehicle would need to traverse under fast sea ice to reach the ice shelf. NEXT STEPS Step 4 of the RMP_AUV will be a comparison of the owner s acceptable risk and the probability of loss from this analysis, for which a written statement from the responsible owner covering Step 1 their risk acceptance will be needed. With mitigation measures we show that for scenario 1 (optimistic) the probability of loss would be reduced to 8.5% and 22.2% for scenario 3. However, it is our opinion that to effect these mitigation measures will take more time than is available between now and the shipping deadline for Dr Jenkins cruise (October 2006). 1. INTRODUCTION The purpose of this paper is to assess the faults with Autosub3, recorded during trials in 2005 and 2006 and on the two research cruises in 2006, with the aim of estimating the probability of loss of the vehicle if operated under an ice shelf given the performance of the vehicle on these trials. This assessment forms Step 3 of the Risk Management Process-AUV being developed for Autosub (Griffiths, 2006), as summarised in Figure 1. There is limited data with which to work, and this results in large confidence limits for the statistical estimators. There is, as well, some debate about the approach of using distancerelated statistics for this application, although Podder et al. (2004) did use time-related data in their analysis in an analogous way to the use of distance by Griffiths et al. (2003). For this reason we also include an analysis based on the very simple statistic of mean number of missions between high impact underway faults. 53

54 No 9 Work to reduce A based on Mitigation Plan so A < L 8 Owner may increase L? 10 Decisions that may postpone or cancel the campaign Yes 7 PI may reassess requirements In parallel Reassess modelled mitigation 6 Identify key risk factors. Produce Mitigation Plan. Model effect of mitigation. 1 Responsible Owner states acceptable probability of loss (L) for the campaign 2 Principal Investigator sets out campaign requirements 3 Technical team assess probability of loss in light of this defined campaign (A) No 4 Is A < L? Yes 11 Campaign takes place 5 Demonstrate that this is so. Figure 1 A flow chart of the Risk Acceptance Process within the RMP-AUV. Assessing which faults should be classed as high impact underway is at the heart of estimating probability of loss for operation under an ice shelf. It is not an exact science. Several of the faults might or might not be considered high impact underway. The approach taken was to ask a series of questions: Could this fault lead to an abort or other condition where the vehicle would stop, lose control or otherwise be in danger of being lost? If yes, and this fault happened early in the mission, would this fault always occur at this time? That is, would the team always be able to prevent the vehicle beginning its under ice traverse when this fault happened? This requires the team to be very sure of the root cause of the fault. Accepting that for some faults, they may have occurred at the surface at the end of a mission, is there a possibility that they could have occurred when underway? This approach has led to the assessment in section 2, and the assessment that some of the faults should be marked as high impact underway, but given a probability of less than one. Carrying this forward rigorously could be done using a Bayesian approach this is currently beyond what we can do, but it will be tackled in partnership with Peter Challenor later this year. 54

55 2. TABLE OF MISSIONS, DISTANCES AND FAULTS WITH COMMENTS Table 1 Discovery 295T July 2005 Mission Distance No. Fault Comment (km) Faults HIU? Y(1) 1. Mission aborted due to network failure. (Much) later tests showed general problem with the harnesses (bad crimp joints). Harnesses repaired by manufacturer for next cruise. 2. Loop of recovery line came out from storage slot, New storage arrangement needed. GG note: the Terschelling 2006 trials showed no explicit evidence of harness-connector problems N 1. Autosub headed off in an uncontrolled way. This was due to a side effect of the removal of the upwards-looking ADCP. The SW vulnerability that caused this was corrected for later test cruise. This problem would be caught immediately after launch and before the vehicle would be committed to its mission N 1. GPS antenna failed. New design for future cruises. GG note: new GPS antenna design used, but different set of problems encountered on Terschelling N 1. Homing failed, and the vehicle headed off in an uncontrolled direction. Mission was stopped by acoustic command. Problem due to the (a) uncalibrated receiver array, and (b) a network message ( homing lost ) being lost on the network. For the later test cruise, the protocol for the homing lost message was upgraded to acknowledged with 8 repeats (it had previously been unacknowledged). GG note: Different homing node configuration problems on m389. Totals

56 Table 2 Terschelling May 2006 Mission Distance No. Fault Comment (km) Faults HIU? Y (1) 1. Aborted after 4 minutes post dive, cause recorded as network Down Y. Logger data showed large gaps in data, of up to 60 seconds. The gaps were across all data from all nodes, strongly suggesting a logger problem. Apparently, such a logger problem can cause a general network failure. Would have resulted in loss under ice shelf if it occurred later into mission once the vehicle was under ice. GG note: No reason found yet for why this occurred after 4 minutes and not sooner or later. No root cause found yet (30/5/06). 2. Depth control showed instability. +/- 1 m oscillation due to incorrect configuration setting. This is not a high impact fault N 1. & 2. Homing node configuration and mission control node problems, but these emerged immediately after dive. Would have happened in open water pre-ice Shelf, so not classed as high impact underway. 3. Deck-Pam problem was unrelated to vehicle systems N ADCP, GPS and EM2000 problems were not critical to mission continuing. ADCP problem would reduce accuracy of navigation. Mission end phase not reliant on GPS as sub-surface rendezvous would be used N Continuation of GPS fault from 391 causing poor navigation N SeaPam problem not critical to mission Y (1) Jack-in-the-box came out, wrapping its line around the propeller, jamming it, and stopping the mission. Would have resulted in loss under shelf Y as ¼ Jack-in-the-box line came out, wrapped around the propulsion motor and jammed. GG note: SDM is convinced this happened on the surface and not when underway. But, as there is a possibility it might occur underway, class as ¼ likely N Human error, no implication for normal long missions Y Main lifting lines became loose, could have jammed motor. a GG note: SDM is convinced this happened on s the surface and not when underway. But, as there is a possibility it might occur underway, class as ¼ ¼ likely. Consider this fault to be independent of the separate Jack-in-the-box problem of m N Mission plan human error, the vehicle executed what it was told Totals and 2 as ¼ 56

57 Table 3 Discovery June-July 2006 Mission Distance No. Fault Comment (km) Faults HIU? N 1. Configuration Mistake. ADCP up was configured as a downward looking ADCP causing navigation problems as the sub was tracking sea surface as the reference. This velocity data was very noisy and put the vehicle navigation out by a factor of Damaged on recovery, moderately serious to sternplane. Discounted by PST as cause of sternplane actuator failure on m Y(2) and possibly 3 1. Stern Plane stuck up during attempt to dive, 2d 20h into mission. Stern plane actuator had flooded. HIU. 2. Aborted due to netydown. Abort release could not communicate with the Depth control node for period of 403 seconds. Possibly side-effect of actuator or motor problems. HIU. 3. Motor windings had resistance of 330ohm to case. Propeller speed dropping off gradually during a dive Possibly HIU. 4. Only one position fix from tail mounted ARGOS transmitter. 5. GPS antenna damaged on recovery Y as 1/4 1. Recovery light line was wrapped around the propeller on surface. Flaps covering the main recovery lines (and where the light line was towed) were open. Class a ¼ likely HIU subsurface. 2. Took over 1 hour to get GPS fix at final waypoint. GG comment: was this at all related to damage to GPS antenna on m402? 3. Propeller speed showed same problem as before. Subsequent testing of motor with Megger showed resistances of a few kohm between windings Possible (1) Totals but possibly 4¼ 1. Pre-launch. Abort weight could not be successfully loaded due to distorted keeper. If not spotted, could have dropped out during mission Considered low probability of distortion and not checked. 2. Pre-launch. Potential short circuit in motor controller that could stop motor. 3. Propeller speed showed same problem as before, Possible HIU. 4. CTD drop-out of 1 hour (shorter drop-outs noted in previous missions). 5. M404 recovery was complicated when lifting lines and streaming line became trapped on the rudder (probably stuck on the Bolen where the two were attached). Recovery from the situation required the trapped lifting lines grappled astern of the ship, attached to the gantry lines, and the caught end cut. 6. The forward sternplane was lost due to lifting line trapping between the fin and its flap. 7. The SeaPam nose transducer was damaged due to collision with the ship. 57

58 Table 4 Terschelling July 2006 Mission Distance (km) No. Faults Fault HIU? Comment N 1. LXT tracking transducer had leaked water replaced. Fault found pre-launch. 2. Stbd lower rudder and sternplane loose. Fault found pre-launch N 1. AUV ran slower than expected and speed dropped off during mission. 2. Current spikes of 3A and voltage drops in first part of mission. 3. Propulsion motor failed 500V Megger on recovery on windings to case. 4. One battery pack showed intermittent connection. 5. SeaPam gave no replies. 6. On surfacing first GPS fix was 1.2km out. 7. Spikes in indicated motor rpm N 1. SeaPam D0059 gave no replies at all no tracking, no telemetry. 2. Noise spikes on both channels of turbulence probe data N 1. Prop motor felt rough when turned by hand bearings replaced. 2. Aborted at 50m due to overdepth as no depth mode commanded. As this would occur immediately on dive, not counted as HIU. 3. No telemetry from SeaPam. 4. Difficulty stopping Autosub on surface via radio command. Separate problems with the two WiFi access points. 5. Still spikes on motor rpm that need investigating. Totals Note: the distances are those travelled over ground these missions were in an area with strong tidal currents; the through water distances for missions were ~ 180, 281 and 288 km. 58

59 3. STATISTICAL ANALYSIS FROM THE FOUR CRUISES 3.1 Simple statistics Discovery 295T A total of 69.9km distance run indicates ~12 hours running time in the water on this shared cruise of 8 days. With one high impact underway (HIU) fault recorded, the mean distance per HIU fault was 69.9/1 or ~70km. Alternatively, on a mission basis, irrespective of distance, the HIU fault rate (hazard rate ) would be 1/4 or 0.25 per mission. Terschelling May 2006 A total of 119.5km indicates ~20 hours running time in the water on a dedicated 10 day cruise. With two faults classed as HIU, the mean distance per HIU fault was 119.5/2 or 60km. If we count both 1/4 possibilities (m395 and m397) then this becomes 119.5/2.5 or ~48km. Alternatively, on a mission basis, irrespective of distance, the HIU fault rate (hazard rate ) would be 2/13 or 0.15 per mission counting only the two HIU faults, or 2.5/13 or 0.19 counting the two missions assessed as having 1 / 4 probability of the fault occurring underway. Discovery 306 With 2 and possibly 4¼ HIU faults recorded over 4 missions and km, the mean distance per HIU was 248 km (optimistic) and 117 km (pessimistic). Alternatively, on a mission basis, irrespective of distance, the HIU fault rate (hazard rate ) would be 2/4 or 0.5 per mission counting only the two HIU faults, or 4.25/4 or 1.06 counting the two possible and the one assessed as ¼. [Note: The two HIU faults occurred on one mission]. Terschelling July 2006 with no HIU faults on this cruise we cannot compute a mean distance between HIU events or a hazard rate of HIU faults per mission. We do, however, include the missions run on this cruise in the overall analysis as censored data. Four missions were completed and involved the vehicle travelling a total of 613 km over ground. Table 5 Overall simple statistics for the four cruises combined. HIU Mean km per HIU Mean HIU per mission Optimistic Pessimistic Total Total Optimistic Pessimistic Optimistic Pessimistic km missions Kaplan Meier analysis For the Kaplan Meier and Weibull analyses it only makes sense to examine the combined trials. The Kaplan Meier analysis aims to estimate a population survival curve from a sample (see Annex C for how the curve is produced). It is widely used in medical and life expectancy research, where information from a small sample is applied to a wider population. For our purposes, the use is rather different. We only have one vehicle. We assume that on any particular campaign or sequence of campaigns (including trials) the vehicle exhibits a sample from a population of faults that are actually present. That is, we assume there are undiscovered faults within the vehicle, not emerging on the missions or campaigns analysed, and that their statistics are represented by those of the sample that do emerge on the missions or campaigns analysed. Also, faults may be noted for which no satisfactory root cause could be found, or enduring cure could be effected. The predicted probability of survival is for this hypothetical vehicle that follows these assumptions, and not for the real vehicle, which may behave differently. Figure 2 shows the Kaplan Meier plots for the four combined trials and research cruises, without (left) and with (right) the possible HIU faults. The rapid decrease in probability of survival over the first few km is indicative of faults occurring shortly after launch, and is probably also a function of the short missions undertaken during the two trials, in keeping with their prime objective of testing systems and behaviours rather than seeking endurance as such. 59

60 Surviving Distance(km) Surviving Distance(km) Figure 2 KaplanMeier (left) optimistic assessment with HUI faults on four missions, and (right) pessimistic assessment with HIU faults on eight missions. 3.3 Weibull analysis Griffiths et al. (2003) showed that, over a sample of 240 Autosub missions, the fault history could be represented by a Weibull distribution. If that is also the case over these fewer missions, we can use this method to estimate probability of loss based on mission length as well as the number of missions. For the combined trials, Figure 3 shows the optimistic assessment (left) and the pessimistic assessment with the possible HIU faults included. Note that the Weibull distribution estimate is, in each case, to the right of the lines showing the actual HIU fault history. This is because of the censored data, that is, those missions that did not suffer a HIU fault. Their inclusion tends to increase the probability of survival over any set distance. Unfortunately, due to the very small sample, there are large confidence limits on the parameters of the Weibull distribution (alpha and beta), Table log(-log(surv)) log(-log(surv)) Distance(km) Distance(km) Figure 3 Weibull estimators (left) optimistic assessment with HUI faults on four missions, and (right) pessimistic assessment with HIU faults on eight missions. Table 6 Weibull distribution estimators alpha and beta, with their confidence limits (top) optimistic assessment with HUI faults on four missions, and (bottom) pessimistic assessment with HIU faults on eight missions. Weibull Parameter Estimates same as Extreme-Value with Alpha=exp(Lambda), Beta=1/Delta Alpha Beta L95 Alpha U95 Alpha L95 Beta U95 Beta N Failed 4 Weibull Parameter Estimates same as Extreme-Value with Alpha=exp(Lambda), Beta=1/Delta Alpha Beta L95 Alpha U95 Alpha L95 Beta U95 Beta N Failed 8 60

61 4. APPLYING THE ANALYSES TO DR JENKINS PINE ISLAND BAY PLANS 4.1 The campaign requirements In Annex A is the text of a communication from Dr Adrian Jenkins (BAS) to the author setting out his requirements for Autosub missions on a cruise to the Pine Island Bay glacier that would involve open water, under shelf ice and possibly under fast sea ice missions. He sets out a desirable set of missions and a minimum set. He also points out that there may be fast sea ice present, and hence we have four possible scenarios for which we calculate the probability of loss: Scenario 1 Minimum set with no fast ice o Three 60 km open water missions o Three 60 km missions under outer half of the ice shelf cavity Scenario 2 Minimum set with fast ice o Three 120 km under fast ice missions o Three 120 km missions: 60 km under fast ice and 60 km under outer half of the ice shelf cavity Scenario 3 Desirable set with no sea ice o Three 60 km open water missions o Three 60 km missions under outer half of the ice shelf cavity o Three 120 km missions under ice shelf cavity Scenario 4 Desirable set with fast ice o Three 120 km under fast ice missions o Three 120 km missions: 60 km under fast ice and 60 km under outer half of the ice shelf cavity o Three 180 km missions: 60 km under fast ice and 120 km under ice shelf cavity 4.2 Estimates of the probability of loss Full details of the calculations are provided in Annex B (simple per mission and per km statistics), Annex C (Kaplan Meier statistics) and Annex D (Weibull statistics) and the results are summarised in Table 7. For Dr Jenkins minimum requirements the probability of loss is expected to be between 35 and 53% at best (Scenario 1) and between 57 and 88% for his desirable requirements (scenario 3). Table 7 Probability (as %) of losing Autosub3 based on four statistical analyses for each scenario. The optimistic assessment is the first number in each cell, the pessimistic the second. Analysis model / Scenario Scenario 1 Scenario 2 Scenario 3 Scenario 4 Simple per mission Simple per km Kaplan Meier Weibull Mean estimate

62 5. MITIGATION AND NEXT STEPS As it is likely that the probability of loss estimated in Section 5 exceeds the risk acceptance by the responsible owner we outline what mitigation measures might be taken in order to reduce the risk: 1. Work by the technical team to identify the root causes of the faults found and to effect their rectification. To include, among others: ensuring the propulsion motor runs properly for extended periods at depth without loss of performance or increase in electrical leakage; seeking to understand if the netydown problems could have been caused by motor/actuator problems; to understand the sternplane actuator failure etc, as fully described by McPhail et al. in the cruise reports. 2. Testing in the laboratory, Empress Dock and deep water location (e.g. a Norwegian fjord), as needed, to demonstrate that the problems identified have been corrected. It is unlikely that this work could be completed in time for an October shipping date to meet the RV Palmer schedule for Dr Jenkins cruise of February In addition to the technical measures, it is suggested that the current procedure of checking all systems work correctly prior to commitment to the mission be extended so that the AUV is confined to operating in a low hazard zone (from which a recovery could be made if there was a fault) for a distance (e.g. ~25km) before it is committed to traversing a zone of higher hazard. We model this option below. 5.1 Modelling the effect of mitigation, further trials and monitoring, that is, can the probability of loss be brought down to an acceptable value? In the following, we model Dr Jenkins scenarios with: The D306 faults on mission 402 removed, which models an Autosub where these faults have been understood and completely rectified. The other faults have been left in. A series of six short 5km missions, with no faults, modelling successful lab or dock trials added to the list of missions. A series of missions of 10, 100, 200, 300 km with no faults (essentially duplicating the Terschelling Simpson cruise) added to the list of missions. Conditional probability calculated to simulate the effect on the probability of loss of being able to recover the vehicle if a fault is found within the first 25 km. The calculations in Annex E show that for scenario 1 (optimistic) the probability of loss would be reduced to 8.5% and 22.2% for scenario 3. While these are indicative, they do show that the probability of loss can be brought down to the 20% region, which this author would consider acceptable for high risk high pay-off polar marine science. REFERENCES Griffiths, G., Millard, N. W, McPhail, S. D., Stevenson, P. and Challenor, P. G., 2003a. On the Reliability of the Autosub Autonomous Underwater Vehicle. Underwater Technology 25(4): Griffiths, G., Towards a risk management process for autonomous underwater vehicles. Discussion paper, draft of 29 May Podder, K., Sibenac, M., Thomas, H., Kirkwood, W. and Bellingham, J., Reliability growth of Autonomous Underwater Vehicle Dorado. Proc. Oceans 2004, Kobe, Japan. MTS/IEEE. Pp

63 ANNEX A Adrian Jenkins requirements for the Pine Island Bay Glacier cruise of February 2007 Thanks again for taking the time to update me on the current situation with Autosub. It will be very disappointing not to go next year, but taking the long-term view, I would rather wait (knowing that NERC approval/funding is in place) for an opportunity to really push the science forward, than rush into a cruise now with the risk of the scientific gain being much less. Of course, we could run open water and sub-sea-ice missions on the outer Amundesn shelf, but if that is all we could do, it would not address the fundamental questions in the proposal. It is also impossible to guess where the sea ice will be, so really difficult to state the mission length and type for exploring parts of the continental shelf north of the ice fronts. For this reason I list only the Pine Island Glacier missions that would be required. They would be: 1) 60 km open water mission to depth of 200 m, close to the ice front. 2) 60 km open water mission to depth of 600 m, close to the ice front. 3) 60 km open water mission to depth of 1000 m, close to the ice front. 4) 3 x 60 km sub-ice-shelf missions to depth 600 m, in the outer half of the cavity. 5) 3 x 120 km sub-ice-shelf missions to depth of 1000 m, to the "minimum headroom" limit of the cavity (at least one of these is likely to be somewhat shorter than 120 km). If fast ice were in Pine Island Bay it might be necessary to add ~60 km to all of the mission lengths and they would then all be beneath (effectively) permanent ice. In this case I guess none of the missions would be achievable at present. An alternative could be Dotson Ice Shelf. A similar set of missions to 1)-5) beneath Dotson could address similar questions. If we could only run the open water missions, I do not think we would gain enough over the ship-based work to justify the effort of shipping Autosub. If we could run 3 x 60 km missions in the outer half of the cavity, two to depths of 600 m and one to a depth of 1000 m, I think we would come away with a very worthwhile dataset. Therefore the break-even point comes at 4). If the risk of 3 x 60 km missions were acceptable, I think it would be worthwhile going. I suspect that this will come out well beyond the acceptable risk. If it were only just beyond, and 2 x 60 km missions was deemed OK, I would have a serious think. If just one 60 km mission were acceptable I would prefer to wait until a later date. Having thought about this at length now, I have practical question about the implementation of the RMP. Supposing that your analysis of the probability of loss for 3 x 60 km open water missions and 1 x 60 km sub-ice-shelf mission happened to be less than Ed Hill's acceptable probability of loss, but that 3 open water plus 2 or more sub-ice-shelf missions put the probability of loss as too high. Presumably I would be allowed to go and undertake the four approved missions. Lets suppose they went ahead without problem. What now? Does the sub now sit on deck for the rest of the cruise? Or is the process interactive? Even without the new statistics (which would improve things) one more 60 km mission beneath the ice shelf would now have an acceptable risk. So would that be approved at this stage? And if the sub came back, would one further mission be approved as acceptable under the RMP? If the new statistics were included, even a 120 km mission might become acceptable? So it might be possible to achieve much more than was originally approved?? Similarly if Autosub had problems on its open water missions, they would presumably feed into the RMP at this stage and potentially exclude the sub-ice mission? These questions are hypothetical only. I have no doubt from our discussion yesterday that the current assessment will be that the only possible option is postponement. But I would be interested to know the answers. 63

64 ANNEX B PROBABILITY OF LOSS USING SIMPLE STATISTICS B.1 PER MISSION STATISTICS In Table 5 the per mission HIU fault rates ( ) are 0.2 and 0.31 respectively for the optimistic and pessimistic cases for loss under an ice shelf. Where relevant, the open water weighting of 0.1 and the sea ice weighting of 0.3 are applied to the probability of loss calculated using the under shelf ice HIU fault rates. Table B.1 Probability of loss for the proposed Jenkins scenarios using optimistic hazard rate (left) and pessimistic hazard rate (right). Simple Poisson statistics based on per mission hazard rate and weighting P(loss) = (1-exp(-.N))*W where is the hazard rate, N the number of missions in this category and W the weighting factor Optimistic open water weight 0.1 Pessimistic hazard rate 0.2 per mission sea ice weight 0.3 hazard rate 0.31 per mission ice shelf weight 1 Scenario 1 Scenario 1 Mission no weight P loss Scenario 1 weight P loss Mission no Overall Overall Scenario 2 Scenario 2 Mission no weight P loss Mission no weight P loss a-6a a-6a Overall Overall Scenario 3 Scenario 3 Mission no weight P loss Mission no weight P loss Overall Overall Scenario 4 Scenario 4 Mission no weight P loss Mission no weight P loss a-9a a-9a Overall Overall

65 B.2 DISTANCE-BASED STATISTICS Table B.2 Probability of loss for the proposed Jenkins scenarios using optimistic hazard rate (left) and pessimistic hazard rate (right). Simple Poisson statistics based on per km hazard rate and weighting P(loss) = (1-exp(-.d))*W where is the hazard rate, d the mission distance and W the weighting factor Optimistic open water weight 0.1 Pessimistic hazard rate per km sea ice weight 0.3 hazard rate per km ice shelf weight 1 Scenario 1 Scenario 1 Mission no distance (km) weight P loss Mission no distance (km) weight P loss Overall Overall Scenario 2 Scenario 2 Mission no distance (km) weight P loss Mission no distance (km) weight P loss a a a a a a Overall Overall Scenario 3 Scenario 3 Mission no distance (km) weight P loss Mission no distance (km) weight P loss Overall Overall Scenario 4 Scenario 4 Mission no distance (km) weight P loss Mission no distance (km) weight P loss a a a a a a a a a a a a Overall Overall

66 ANNEX C PROBABILITY OF LOSS USING THE KAPLAN MEIER METHOD Calculating the Kaplan Meier Statistics 1. Reorder the missions into distance order and against each note whether it failed (with HIU), or not (in which case, mark as censored). 2. At each failure, calculate the probability of survival for the subsequent interval as number of survivors at this distance (i.e. total number of entries below this failure, N) divided by N+1 (i.e. the number of survivors plus the failure that has just happened). 3. The probability of survival to any point is the cumulative probability of surviving each preceding distance interval, calculated as the product of preceding probabilities. Probability of loss is one minus the probability of survival. Table C.1 Probability survival with distance calculated using the optimistic (left) and pessimistic (right) fault assessments for the Autosub3 missions Kaplan Meier analysis Optimistic Pessimistic Cumulative Cumulative Mission Distance Fail or censor? P(survival) P(survival) Mission Distance Fail or censor? P(survival) P(survival) fail fail fail fail censor censor censor censor fail fail censor censor censor fail censor censor censor censor censor censor censor censor censor fail censor censor censor censor censor censor censor censor censor censor censor censor censor censor censor fail censor censor censor fail censor censor fail fail censor censor 66

67 Table C.2 Probability of loss for the proposed Jenkins scenarios using optimistic (left) and pessimistic (right) Kaplan Meier statistics. Kaplan Meier analysis Scenario 1 Scenario 1 Mission no distance (km) weight P loss Mission no distance (km) weight P loss Overall Overall Scenario 2 Scenario 2 Mission no distance (km) weight P loss Mission no distance (km) weight P loss a a a a a a Overall Overall Scenario 3 Scenario 3 Mission no distance (km) weight P loss Mission no distance (km) weight P loss Overall Overall Scenario 4 Scenario 4 Mission no distance (km) weight P loss Mission no distance (km) weight P loss a a a a a a a a a a a a Overall Overall

68 ANNEX D PROBABILITY OF LOSS USING THE WEIBULL METHOD Table D.1 Probability of loss for the proposed Jenkins scenarios using optimistic (left) and pessimistic (right) Weibull statistics. Weibull analysis P(loss) = 1 exp x where x is the mission distance Optimistic open water weight 0.1 Pessimistic sea ice weight ice shelf weight Scenario 1 Scenario 1 Mission no distance (km) weight P loss Mission no distance (km) weight P loss Overall Overall Scenario 2 Scenario 2 Mission no distance (km) weight P loss Mission no distance (km) weight P loss a a a a a a Overall Overall Scenario 3 Scenario 3 Mission no distance (km) weight P loss Mission no distance (km) weight P loss Overall Overall Scenario 4 Scenario 4 Mission no distance (km) weight P loss Mission no distance (km) weight P loss a a a a a a a a a a a a Overall Overall

69 ANNEX E PROBABILITY OF LOSS AFTER MITIGATION Table E.1 Probability of loss for the proposed Jenkins scenarios after mitigation measures described in Section 5 Weibull analysis P(loss) = where x is the mission distance D306 fault removed Six short missions added (no faults) Simpson distances repeated (no faults) Optimistic open water weight 0.1 Conditional y = 25 km monitored first sea ice weight ice shelf weight 1 Scenario 1 x Mission no distance (km) weight P loss P loss Overall Scenario 2 Mission no distance (km) weight P loss P loss a a a Overall Scenario 3 Mission no distance (km) weight P loss P loss Overall Scenario 4 Mission no distance (km) weight P loss P loss a a a a a a Overall

70 4) from Helen Beadman, NERC, to Adrian Jenkins, BAS, on 04/08/06, reporting the outcome of the risk assessment procedure: Dear Adrian To confirm our telephone discussion earlier today: Due to the large gap between the acceptable risk (as set by the Director NOC) and the current reliability of Autosub3 is has been decided that Autosub3 should not be deployed under the Pine Island Bay Glacier. In order to improve the reliability of Autosub3, so that it is ready for deployment in the event that you are able to secure ship time on a suitable ice-breaker, funds have been made available to run a trials plan (this financial year). I will keep you up to date on progress with the trials plan. In these exceptional circumstances NERC will allow a delay to the start date of your grant NE/C510608/1, and would need to review the situation in a year or so to see if a realistic opportunity had been identified for the deployment of Autosub3 beneath the PIG. I will investigate the necessary procedure for the delayed start date and will get back to you shortly with the details of how this will happen. Best wishes Helen Dr Helen Beadman Marine Planning Senior Science Programmes Officer Natural Environment Research Council Polaris House North Star Avenue Swindon SN2 1EU 70

71 5) Revised risk assessment prepared by Gwyn Griffiths, NOCS, 14/04/07, following Norway trial and two scientific cruises in summer 2007: National Oceanography Centre, Southampton Professor Gwyn Griffiths Underwater Systems Laboratory National Marine Facilities Division 14/04/07 David Lewis Head, NMFD University of Southampton, Waterfront Campus, European Way, Southampton SO14 3ZH, UK t +44 (0) f +44 (0) e gxg@noc.soton.ac.uk switchboard +44 (0) Dear David, Autosub3: A revised risk assessment for the proposed Jenkins cruise to Pine Island Bay following the March 2007 Norway trials. Based on the fault history provided to me by Steven McPhail, and the methodology set out in the document Fault assessment and estimated probability of loss under ice shelf for Autosub3 dated 31 July 2006, I have revised the risk of loss for the four scenarios provided by Dr Jenkins. I have also revised the owner s calculation (next page, example for scenario 4), taking into account the service days that the vehicle has already seen (29) and deducted these from its initial service life. The results are: 1. Min. set, no sea ice in front of glacier P(loss) = 9% Accept 10% 2. Min. set + 30km of sea ice in front of glacier P(loss) = 24% Accept 20% 3. Desired set with no sea ice in front of glacier P(loss) = 16% Accept 17% 4. Desired set + 30km of sea ice in front of glacier P(loss) = 30% Accept 23% This calculation is based on the motor fault on D306 being retired, as demonstrated by the Norway trials, but with the new high impact fault on those trials included (the rudder actuator). These results are based on a risk mitigation strategy that requires the vehicle to be operated for 25km in an area where recovery could be achieved prior to each and every mission. The results are also based on modelling the increased reliability of the vehicle after each successful mission as set out by Jenkins. The acceptable risk varies between scenarios as the vehicle is used with a different mix of risk and for different periods. Following the vehicle s performance during the Norway trials, the risk from Jenkins scenarios 1 and 3 are lower than the responsible owner s threshold; scenario 2 is sufficiently close that I suggest the owner should not make a decision solely based on this difference. Scenario 4 is the only one with significantly greater risk than that advised to the owner. I suggest that the transition from Minimum to Desired sets if sea ice is present is only undertaken after a revised risk assessment taking into account performance on the cruise and the details in the full engineering logs. Given that the risk weighting for sea ice is empirical and that the cruise is on the Nathaniel B. Palmer, a ship with a greater icebreaking capability than the James Clark Ross, my recommendation is that the Autosub3 should be provided to Dr Jenkins. However, I do advise that a small ROV (similar to a Seaeye Falcon) should be available on board to effect a recovery should the vehicle become stranded beneath sea ice, as this further mitigates the risk in scenarios 2 and 4. This risk assessment is generic, in that it is equally applicable to Pine Island Bay Glacier, Thwaites Glacier, Smith Glacier, or others with broadly similar ice shelf cavities. The PI should have the flexibility of choosing the final target glacier in light of local circumstances, e.g. ease of access, sea ice and wind conditions etc. Autosub3 will be used in open water between now and the likely date of this cruise. Statistics and engineering performance from those missions will be used to update the risk assessment. Yours sincerely Gwyn Griffiths 71

72 Worksheet to assist AUV responsible owner in determining acceptable risk User input in blue cells only. Calculated parameters are in buff cells. Owner Initial Inputs Initial AUV Capital Cost GBP Replacement cost charge 33.3 % of daily operations cost Required probability of 10 % reaching end of life Cost of Daily operations Direct AUV charge 4000 GBP per day Fraction science team 1000 GBP per day Fraction of ship 3700 GBP per day Total 8700 Replacement cost charge GBP/day Risk subsets % service life Relative risk Uniform Risk-based Open water Under sea ice Under shelf ice Š Š Calculated parameters Required service life 431 operational days Service life to date 29 To April 2007 [Trials, Burkill, Simpson, Trials] Service life remaining 402 Acceptable Hazard rate ( ) per day Campaign details Number of service days 20 days Risk subsets % campaign Relative risk Open water Under sea ice Under shelf ice 50 1 Š Š Accepable probability of 23 % loss for the campaign G Griffiths 14 April

73 6) from Helen Beadman, NERC, to Adrian Jenkins, BAS, on 02/07/07, reporting the outcome of the risk assessment procedure: Dear Adrian Following the Autosub trials, in Norway earlier this year, Gwyn Griffiths has undertaken a new assessment on the risk of loss of Autosub for the four scenarios which you provided. Following this assessment a discussion was held with Ed Hill, as responsible owner of Autosub, to determine if the acceptable level of risk is within that which would allow your scenarios to be undertaken. As a result of this we are pleased to offer you the use of Autosub for your mission sets 1-3, with a number of conditions. Mission set 4 would require a revised risk assessment taking into account performance on the cruise and the details in the full engineering logs. (Mission set 1: Min. set, no sea ice in front of glacier) (Mission set 2: Min set + 30km of sea ice in front of glacier) (Mission set 3: Desired set with no sea ice in front of glacier) (Mission set 4: Desired set + 30km sea ice in front of glacier). The conditions which are associated with this offer (to achieve the acceptable level of risk for missions 1-3) are: - the vehicle is operated for 25km in open water at the start of each and every mission - the vehicle is operated from the NB Palmer, with its better icebreaking capability - it is desirable that a small ROV (similar to a Seaeye Falcon) should be onboard to assist recovery should the vehicle become stranded beneath the sea ice (further mitigates the risk in mission sets 2 and 4). We are currently investigating the potential costs associated with taking a small ROV. I have attached the letter which Gwyn Griffith sent to David Lewis and explains in more detail the calculations that were undertaken and how the level of risk was determined. I realise that when your grant was first funded you had arranged time on the NB Palmer, but that the slot for this cruise has now passed. Please could you let me know what your plans are for re-arranging ship time, and when you might anticipate the Autosub cruise going ahead. I appreciate that this will be very conjectural at the moment. Please give me a call if you would like to discuss this further. Best wishes Helen Dr Helen Beadman Marine Planning Senior Science Programmes Officer Natural Environment Research Council Polaris House North Star Avenue Swindon SN2 1EU Tel: +44 (0) Fax: +44 (0) habe@nerc.ac.uk 73

74 Appendix 2. Mission Summaries Mission Summary Sheet M427 Campaign: Palmer09 Mission No.: M427 Operating Area: Pine Island Bay, Antarctica Vehicle Configuration and sensors used Sensors used Sub configuration RDI workhorse ADCP 150kHz downwards. RDI Workhorse ADCP 300kHz upwards Seabird SBE9+ CTD with dual TC Rear winglets set at 5.5 º pitched downwards. 9.9kg Positive buoyancy. 54 battery packs sensors, SBE43 oxygen and Wetlabs transmissometer. PHINS INS Kongsberg EM2000 Multibeam looking downwards Mission Objectives & Description Objectives: Mission for testing Autosub3 systems and sensors at up to 890 m depth. Mission Conditions Start of Mission End of Mission Time [logger] 17-Jan :16:06 17-Jan :03:23 Position [GPS] lat, long lat, long. Sea state Glassy smooth Wind speed 2.6m/s 157 degrees Battery Voltage 112V 102V Mission Statistics Mean motor power 394 [W] Mission duration seconds Mean water speed 1.22 [m/s] Distance travelled 37.5 [km] Maximum depth Mission Review Mission proceeded as planned and the sub was successfully recovered Mission completed successfully [yes/no]: Yes Sensor Data Quality CTD OK apart from intermittent T1 ADCP good EM2000 good when in range Faults arising during mission Motor appeared to stop intermittently fixed for M428 by increasing enable time from 15 to 30 seconds and making motor enable tick acknowledged and repeat. T1 on CTD appeared intermittent connectors cleaned and cable replaced for M428. Hotel Ground Fault reading high investigation showed that source was the CTD but no fault found. 74

75 Comments Mission Overview Picture Vehicle Setup Pictures 75

76 Mission Summary Sheet M428 Campaign: Palmer09 Mission No.: M428 Operating Area: Pine Island Bay, Antarctica Vehicle Configuration and sensors used Sensors used Sub configuration RDI workhorse ADCP 150kHz downwards. RDI Workhorse ADCP 300kHz upwards Seabird SBE9+ CTD with dual TC Rear winglets set at 5.5º pitched downwards. 9.9kg Positive buoyancy. 54 battery packs sensors, SBE43 oxygen and Wetlabs transmissometer. PHINS INS Kongsberg EM2000 Multibeam looking upwards. Mission Objectives & Description Objectives: Sub ice shelf Mission. Started with short (2 hr) test mission, followed by data retrieval and checking. Then send AUV under the ice shelf. 30 km at 200 m const altitude, then turn and 100 m up altitude. Mission Conditions Start of Mission End of Mission Time [logger] 19-Jan :23:52 20-Jan :12:12 Position [GPS] lat, long lat, long Sea state Wind speed Battery Voltage 104V 92V Mission Statistics Mean motor power 68.0 Mission duration seconds Mean water speed 1.30 Distance travelled Maximum depth 877.0m Mission Review Mission proceeded as planned and the sub was successfully recovered Mission completed successfully [yes/no]: Yes Sensor Data Quality CTD good but with offset on oxygen and noise on transmissometer. EM2000 good when in range. ADCP not good Faults arising during mission Poor ADCP Up profiling range. Comments 76

77 Mission Plan Mission Overview Picture 77

78 Vehicle Setup Pictures 78

79 Mission Summary Sheet M429 Campaign: Palmer09 Mission No.: M429 Operating Area: Pine Island Bay, Antarctica Vehicle Configuration and sensors used Sensors used Sub configuration RDI workhorse ADCP 150kHz downwards. RDI Workhorse ADCP 300kHz upwards Seabird SBE9+ CTD with dual TC Rear winglets set at 5.5º pitched downwards. 9.9kg Positive buoyancy. 54 battery packs sensors, SBE43 oxygen and Wetlabs transmissometer. PHINS INS Kongsberg EM2000 Multibeam Mission Objectives & Description Objectives: 2nd run under the PIG. 30 km in. 200 m alt, and 100 m off the ice coming out. Mission Conditions Start of Mission End of Mission Time [logger] 21-Jan :49:05 22-Jan :10:52 Position [GPS] lat, long lat, long. Sea state Wind speed Battery Voltage 96V 86V Mission Statistics Mean motor power 73W Mission duration seconds Mean water speed 1.27m/s Distance travelled 145.8km Maximum depth Mission Review Mission proceeded as planned and the sub was successfully recovered Mission completed successfully [yes/no]: Yes but only after 2 unsuccessful starts. Sensor Data Quality All good still an offset on oxygen sensor when using Palmer0901- SeabirdCCalsV2.con file. Faults arising during mission Bad crimps in fish cable meant that it was impossible to send an acoustic start command to the vehicle to send it on its way. Comments Successful mission completion. 79

80 Mission Plan Mission Overview Picture 80

81 Vehicle Setup Pictures 81

82 Mission Summary Sheet M430 Campaign: Palmer09 Mission No.: M430 Operating Area: Pine Island Bay, Antarctica Vehicle Configuration and sensors used Sensors used Sub configuration RDI workhorse ADCP 150kHz downwards. RDI Workhorse ADCP 300kHz upwards Seabird SBE9+ CTD with dual TC Rear winglets set at 5.5º pitched downwards. 13.9kg Positive buoyancy. 54 battery packs sensors, SBE43 oxygen and Wetlabs transmissometer. PHINS INS Kongsberg EM2000 Multibeam upwards Mission Objectives & Description Objectives: 3rd under ice shelf mission for Palmer km in at 200m altitude, returning at 100m from ice for multibeam. Mission Conditions Start of Mission End of Mission Time [logger] 23-Jan :22:41 23-Jan :22:41 Position [GPS] lat, long lat, long. Sea state Wind speed Battery Voltage Mission Statistics Mean motor power 44.0 Mission duration Mean water speed 1.50 Distance travelled Maximum depth Mission Review Mission proceeded as planned and the sub was successfully recovered Mission completed successfully [yes/no]: Yes Sensor Data Quality Good Faults arising during mission None Comments 82

83 Mission Plan Mission Overview Picture 83

84 Vehicle Setup Pictures 84

85 Mission Summary Sheet M431 Campaign: Palmer09 Mission No.: M431 Operating Area: Pine Island Bay, Antarctica Vehicle Configuration and sensors used Sensors used Sub configuration RDI workhorse ADCP 150kHz downwards. RDI Workhorse ADCP 300kHz upwards Seabird SBE9+ CTD with dual TC Rear winglets set at 5.5º pitched downwards. 13.9kg Positive buoyancy. 54 battery packs sensors, SBE43 oxygen and Wetlabs transmissometer. PHINS INS Kongsberg EM2000 Multibeam looking upwards Mission Objectives & Description Objectives: 4th under ice shelf mission for Palmer First 60 km Mission. Run 60km in at 100m altitude, turn when get to position or collision avoided and then run back at 120m up altitude Mission Conditions Start of Mission End of Mission Time [logger] 24-Jan :41:44 26-Jan :52:30 Position [GPS] lat, long lat, long. Sea state Wind speed Battery Voltage 95V 87V Mission Statistics Mean motor power 46.5m/s Mission duration seconds Mean water speed 1.34m/s Distance travelled 182.8km Maximum depth 978.4m Mission Review Mission completed successfully [yes/no]: No vehicle returned to safe waypoint with damage to nose section caused by collision with ice. Sensor Data Quality CTD OK up to collision but port plumbing was damaged causing no pumped flow through port sensors. ADCP OK? Little data from EM2000 due to range from ice over most of the mission. Faults arising during mission Vehicle systems were unable to cope with severe crevassing in ice however these conditions were beyond their design specification. Comments 85

86 Mission Plan Mission Overview Picture 86

87 Vehicle Setup Pictures 87

88 Damage to AUV front port side panel following collision with ice 88

89 Mission Summary Sheet M432 Campaign: Palmer09 Mission No.: M432 Operating Area: Pine Island Bay, Antarctica Vehicle Configuration and sensors used Sensors used Sub configuration RDI workhorse ADCP 150kHz downwards. RDI Workhorse ADCP 300kHz upwards Seabird SBE9+ CTD with dual TC Rear winglets set at 5.5º pitched downwards. 13.7kg Positive buoyancy. 54 battery packs sensors, SBE43 oxygen and Wetlabs transmissometer. PHINS INS Kongsberg EM2000 Multibeam looking down Mission Objectives & Description Objectives: PIG 5th Mission. 4 hour test mission, as check of systems following the collision Mission Conditions Start of Mission End of Mission Time [logger] 27-Jan :43:06 28-Jan :54:53 Position [GPS] lat, long lat, long. Sea state smooth Wind speed Battery Voltage Mission Statistics Mean motor power 40.8W Mission duration seconds Mean water speed 1.32 [m/s] Distance travelled 27.2 [km] Maximum depth 888.1m Mission Review Mission completed successfully [yes/no]: Mission ended after test mission segment in order to reconsider position of WP3 (the furthest into the ice). Sensor Data Quality All good. Faults arising during mission None Comments 89

90 Mission Overview Picture Vehicle Setup Pictures 90

91 Mission Summary Sheet M433 Campaign: Palmer09 Mission No.: M433 Operating Area: Pine Island Bay, Antarctica Vehicle Configuration and sensors used Sensors used Sub configuration RDI workhorse ADCP 150kHz downwards. RDI Workhorse ADCP 300kHz upwards Seabird SBE9+ CTD with dual TC Rear winglets set at 5.5º pitched downwards. 13.7kg Positive buoyancy. 54 battery packs sensors, SBE43 oxygen and Wetlabs transmissometer. PHINS INS Kongsberg EM2000 Multibeam looking downwards Mission Objectives & Description Objectives: Approximately 55km under ice shelf at 100 m altitude. EM2000 looking down. Turn on minimum headroom setting. Profile out the last 30 km. Mission Conditions Start of Mission End of Mission Time [logger] 28-Jan :27:46 29-Jan :08:16 Position [GPS] lat, long lat, long. Sea state 5 Wind speed 20m/s Battery Voltage 104V 91V Mission Statistics Mean motor power 46.2 Mission duration seconds Mean water speed 1.39 Distance travelled Maximum depth Mission Review Mission completed successfully [yes/no]: Vehicle turned around due to limited headroom as expected, but further limited headroom situations caused collision avoidance behaviour leading to a run back to the safe waypoint. Sensor Data Quality All good. Faults arising during mission Aft Argos antenna not working and forward antenna in bad condition so both were replaced. Comments 91

92 Mission Plan Mission Overview Picture 92

93 Vehicle Setup Pictures 93