Satellite Derived Bathymetry as Source Data for Navigational Charts

Introduction Review of Key findings from UKHO SDB Trial User group consultation Charting SDB Further work Satellite Derived Bathymetry (SDB) refers to depths processed from optical satellite imagery. The depth accuracy is where the technology has limitations and this needs to be understood UKHO have been assessing the accuracy of SDB and the most suitable methodology for representing SDB data on charts. How can this data be depicted on charts.

Satellite Derived Bathymetry? What Is Satellite Derived Bathymetry? Image © DigitalGlobe, 2013 …but dark patches may also be shallow Shallow=bright Deep=dark

Passive – affected by many uncontrollable factors Real-world sit Depth accuracy where technology has its limitations. uation is complex. More than one method of ‘unmixing’ the spectral signal from which depth can then be derived Passive – affected by many more uncontrollable environmental factors Passive – affected by many uncontrollable factors Spectral ‘unmixing’ is complex, more than one approach

Satellite Resolution Effect of resolution; improved resolution with advancing technology. Object detection. Good positional accuracy. Image © DigitalGlobe, 2013 Landsat program 30 metre resolution WorldView-2 Launch 2009 2 metre resolution

Nautical Charts: A Special Case The Cartographic consideration Use of SDB Nautical Charts: A Special Case Purpose of a chart is “to accurately represent nautical information necessary for safe navigation” ADMIRALTY Nautical Charts comply with International Convention for the Safety of Life at Sea Data assurance – an essential requirement Wide application of SDB (Environmental Management, Habitat Mapping & monitoring, Marine Engineering); why charts are different The Cartographic consideration

Antigua: A Case Study New Chart covering Southern Antigua, scale 1:25,000 New MBES harbour surveys already planned Existing surveys pre-1852! Antigua chart requirement already established & New Edition planned

Antigua SDB Trial Order 1a Survey carried out Sept-Oct 2013 WorldView-2 image acquired 27 Sept 2013 Tide gauge in place Horizontal GCPs Claims of accuracies of 10% of depth were not borne out by early experiences of SDB within the UKHO. UKHO decided in 2013 to conduct a controlled trial to fully understand the accuracy and reliability of SDB.

Antigua SDB Trial Several different SDB data sets were produced from the same image Small parts of MBES survey used to calibrate SDB data (ground truth) Majority of MBES data used as benchmark Results looked at in terms of: Overall accuracy (compared to MBES dataset) Ability to define critical soundings The imagery was processed using several different SDB processing methods, both in-house and through commercial companies, in order to compare them and understand the repeatability of this technology.   Results were looked at in terms of both overall accuracy (compared to MBES dataset) and SDB’s ability to define critical soundings. These factors were then used to assess a suitable CATZOC classification for SDB data.

Results Ground truth applied: Red = SDB datasets Blue = MBES Brief look at the results (representative of the whole area)

Results: Variance ±2m = SDB is deeper = SDB is shoaler ±3m +/- 3m quoted, but actually +/- 2.75m. Reliability didn’t necessarily relate directly to depth across the area Survey Standards vs. CATZOC Difference surfaces were created and coloured to show areas where the SDB data is within 2m and 3m of the MBES data. The difference surfaces were compared to the depth accuracy requirements for CATZOC C. ±3m

Results: Catzoc data passes CatZoc C = data fails No-ground truth applied: 78% data meets CatZoc C data passes CatZoc C data fails = more than 97% of the SDB data was within the CATZOC C depth accuracy requirements from the MBES data. Without ground truth data applied only 79% of the SDB data met CATZOC C requirements.   Though more than 97% of the “Commercial Company 1” data could be categorised as CATZOC C, the data that fails CATZOC C requirements is spread throughout the data set and would be difficult to delineate. accurate to approximately ±2-3m data, though much of the data was better than this. Ground truth applied: 97% data meets CatZoc C

Antigua SDB Trial SDB attributes: Good coverage Good horizontal accuracy Limited to ~15m depth Lesser depth accuracy than MBES, SBES and leadline, however 97.7% data within CATZOC C requirements Conclusion from Antigua SDB dataset Within its limitations it can provide useful information for the mariner

The Cartographic Question Important to get across uncertainty to the user Need some indication to the mariner as to how they should behave… SDB is not a self contained dataset, its merits may be relative to surrounding data. Eg. preferable for mariner to go through SDB rather than sparse leadline data…? Is existing symbology enough?

The Mariner’s perspective Usergroup Engagement The Mariner’s perspective

Establishing rules for representation on the chart Usergroup Engagement Establishing rules for representation on the chart Outline of options presented to mariner test group (not dwelling upon this!)

The Mariner’s perspective Usergroup Engagement The Mariner’s perspective Mariner has limited interest in how surveys are conducted, “that is UKHO’s job” Didn’t want more ‘clutter’ on the face of the chart No more symbology The general opinion was that they would like to know where SDB survey data is used on a chart, but there was a preference not to use any additional symbology. There was a preference to use the source diagram   no requirement to differentiate soundings derived from satellite bathymetry on the chart face itself.

Integrating with many other information sources Chart Compilation Integrating with many other information sources How SDB is considered as apart of chart compilation process

Chart Compilation New shoal depths found from SDB, not in original leadline survey Existing depths confirmed and repositioned for greater horizontal accuracy However… unconfirmed leadline shoal depths retained Example of how compiled chart benefited from SDB dataset

Accuracy of contouring improved Chart Compilation Accuracy of contouring improved Example of how compiled chart benefited from SDB dataset

BA2066 Southern Antigua Proudly announcing UKHO’s first published used of SDB on New Edition of BA2066. How SDB is communicated to end user

ENC Representation S-57 Standard for ENCs Tightly defined cartographic representation Existing S-57 attribution can already accommodate SDB data CATOC = C Data may be further qualified by object class M_QUAL sub-attribures as follows: POSACC = 10m SOUACC = 3m TECSOU = 11 “satellite imagery” Demonstration of how SDB already fits within the existing S-57 attribution schema

Conclusions & Implications for future charting Antigua chart not necessarily definitive way of representing SDB in future Limited control still required Currently SDB seen as a way of complementing & extending new MBES surveys… where conditions make this practical Building experience … growing confidence? Consider these as first baby steps…

WorldView-2 Satellite imagery credit: © DigitalGlobe, 2013 Thank you WorldView-2 Satellite imagery credit: © DigitalGlobe, 2013