28 July 2004 Representing Uncertainty on Charts: The “Hydrographic Crisis” “Beyond Safety of Navigation” Multibeam and Visualization Workshop

28 July 2004 The Gordon Reid The Canadian Coast Guard Ship (CCGS) Gordon Reid ran aground on an uncharted rock in Estavan Sound, off the coast of British Columbia on 28 September 2003 CHS Chart 3724: 1923 lead-line and sextant survey with 400 metres between sounding lines

28 July 2004 Research Undertaken Topic of the UNB Uncertainty Paper, “Bathymetric Uncertainty Representation on Nautical Charts” Research taken on by UNB & USM graduate students under the direction of Dr. Dave Wells

28 July 2004 The Uncertainty Situation Many users of nautical charts have no idea how uncertain is the information shown on the charts they are using This leads to inappropriate navigation decisions, groundings, and sometimes loss of life

28 July 2004 Uncertainty in Hazard Detection Bathymetric Uncertainties Positioning Uncertainties Incomplete coverage between survey lines (as was the fate of the Gordon Reid)

28 July 2004 Navigation Decisions Mariners have excellent positioning tools due to high accuracy GPS and ENCs This does not account for positioning uncertainty present at time of survey For example, over half of the inshore NOAA nautical charts were acquired by lead-line and sextant surveying prior to 1940

28 July Methods of Representing Uncertainty Two methods, a Source Diagram (SD), and a Reliability Diagram (RD) are graphical insets on a paper chart –Showing the geographical limits for each survey –A table describing the attributes of each survey area in the diagram Zone of Confidence (ZOC) methods used on Electronic Navigational Charts (ENC)

28 July 2004 Source Diagrams (SCD) Hydrographic organization Date of survey Scale of survey Direct line spacing information

28 July 2004 Model Source Diagram IHB M4

28 July 2004

NOS SD (A) 1990 to the present. Full bottom coverage. DGPS positioning used. (B1) 1990 to the present. Partial bottom coverage from single beam echo sounder. GPS or DGPS used. (B2) 1970 to Partial bottom coverage from single beam echo sounder. Primarily electronic radio-positioning. (B3) 1940 to Partial bottom coverage from single beam echo sounder. Visual positioning (B4) 1900 to Partial bottom coverage from lead line. visual positioning. (B5) Pre Partial bottom coverage from lead line. visual positioning. NOAA Approach

28 July 2004 CHS CHS

28 July 2004 RAN RAN

28 July 2004 Reliability Diagrams Give an assessment of accuracy as well as advising on preferred areas for navigation Examples of the attributes –Estimated accuracy of soundings –Distance between survey sounding lines –Classification of the survey (e.g. reconnaissance or incomplete; controlled; sounded by lead line; sounded by echosounder; shoals have been examined; has been sonar swept)

28 July 2004 Model Reliability Diagram IHB M4

28 July 2004 Zones of Confidence (ZOC) ZOC values assigned to areas on an ENC A1/A2: Full bottom ensonification with depths determined for all significant features B: Uncharted hazards may exist C: Uncharted depth anomalies are expected D: Large depth anomalies are expected

28 July 2004 ZOC Categories

28 July Active Directions Education for users Worldwide HO Survey Presentation on charts

28 July 2004 Worldwide HO Survey HOs from around the world have been contacted 5 questions were asked about their uncertainty policies

28 July 2004 Survey Questions Asked 5 questions about their uncertainty information –Both paper and electronic charts –Methods of communication to users –Satisfaction with current policy –Plans to change current policies, and barriers which prevent doing so Very diverse responses Australia, Canada, Denmark, Finland, Greece, Hong Kong, Iceland, Italy, Japan, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Saudi Arabia, South Africa, Sweden, Turkey, UK, US NGA, US NOAA

28 July : How is uncertainty represented on your paper charts? Some agencies do not use SDs or RDs at all (one agency removed SDs from their charts, since they were not kept updated) Some agencies use SDs only on large scale charts Most agencies are in the process of adding SDs to their charts. One agency claimed 100% of their charts had SDs

28 July Are ZOC values on your ENCs fully attributed? 40% have full ZOC attribution on all ENCs (over half of these use only ZOC values B and C, or in one case only B on all ENCs) 30% use only U attribution so far 30% are partway to full attribution Reasons for not having full ZOC attribution were lack of resources, lack of metadata upon which to base the ZOC, and the liability implied by assigning a ZOC

28 July How else do you communicate information on chart uncertainties to users? 60% use Notices to Mariners 25% used web pages, other nautical publications, and presentations to user groups. The Danish hydrographic office booklet “Behind the Nautical Chart” is free for downloading from its website. This booklet explains the uncertainty associated with hydrographic survey methods over the years (and is soon to be translated from Danish to English)

28 July Are you satisfied with your current policies and practices? 30% answered yes 60% answered with a qualified or unqualified no 10% dodged the question (e.g. “we will always try to improve”) The qualified no answers were based on a desire for better methods than SDs, RDs or ZOCs, and on liability issues associated with RDs and ZOCs

28 July Are you considering any changes to these policies and practices? 15% answered no 10% answered they would comply with any new international standards that might emerge 40% intend to work towards completion of SDs on all paper charts, and / or full ZOC attribution on all ENCs 35% seek improvements to their entire hydrographic data management strategy, uncertainty information

28 July 2004 Defining the end user Commercial Shipping Fishing and Natural resources Recreational users Military/Coast Guard

28 July 2004 Need based representation Establish product enhancements based on input from user groups. Common factor being improved situational awareness. What enhancements will most improve the safety and decision making of the Navigator?

28 July 2004 Why add the enhancement? How and under what circumstances will it be used? Voyage planning. Weather maneuvering. Shipboard medical emergency. Result: Time critical decision making requires clear depiction of reliability.

28 July 2004 What to add to the Chart? Source diagram – Voyage and approach planning.

28 July 2004 What to add to the Chart? Primary chart depiction - Best for time critical decision making. Rocks, soundings and depth contours printed in red (Gulf of Finland)

28 July 2004 Low-Density Data Sources Source: LT Shep Smith

28 July 2004 Uncertainty Modeling Traditionally, the measurement error of a given sounding is the value reported as the uncertainty of the depth. In other words: How good was that measurement? But, what mariners really want to know is: How well is the depth known at this location?

28 July 2004 When we shoal bias multibeam, we keep the least accurate measurements

28 July 2004 Fundamental Process Goal First: determine what is the true depth in the area of interest without any consideration of a final product, scale or ‘hydrographic’ concerns - i.e., how well do we know that depth? Then: make the appropriate products with due regard for the end-user requirements.

28 July 2004 Uncertainty Modeling Three basic methodologies: Forward error Backward error Interpolation

28 July 2004 Forward Error When applied to high-density MB bathymetry, each depth is assigned a predicted error based on: 1) the systems used to collect it 2) environmental conditions at the time of the survey

28 July 2004 Backward Error Uses the standard error of the measurements around the weighted mean. Limitation: difficult to distinguish between areas of steep slope, high seafloor irregularity, and high error.

28 July 2004 Uncertainty for Interpolated Areas Gaps between MB survey lines or SB soundings are a nagging concern. Typically, uncertainty interpolation: is related to the measurement uncertainty at the node where the measurement was made increases as a function of the distance to the nearest measurement is higher on a more irregular seafloor

Uncertainty Model of Clean Multibeam The uncertainty of the node is the greater of: – -the average uncertainty of the measurements –the 95% bound of the distribution of the measurements around the mean. Interpolated areas follow the sparse data rules High uncertainty is expected on steep slopes due to horizontal error.

28 July 2004 Other Types of Uncertainty Time-Dependent – dynamic seafloor areas may require a “changeability coefficient” to be assigned at every node. Superceding Data * – when superceding old data with new, some rules should be applied: –a model node with lower uncertainty supercedes one with greater –a newer node supercedes an older node –a shoaler node supercedes a deeper node * Primarily applies to navigation products

28 July 2004 Navigation Surface  Chart Product Generation Three steps are involved: Defocusing - apply the horizontal uncertainty of the model nodes to the model - at each node, adjacent nodes are adjusted in the shoal direction if they are deeper or fall into the horizontal error circle of the node. Generalization - for the intended product, use a “buffering” process Extract Cartographic Objects – e.g., contours, depth areas, and selected soundings

28 July 2004 UNB Recommendations Survey details including: - Date of survey - Survey and positioning technology used - Line spacing/amount of coverage Given by either exact survey details, or by a classification scheme to maximize clarity and simplicity

28 July 2004 UNB Recommendations Timeless, without the use of descriptive quality terms such as “modern standards” or “current technology” Be accompanied by a form of education for chart users Be part of a national standard, with the ultimate future goal of a global standard