EMODnet Thematic Lot n° 2
EMODnet Phase 2 – Annual (interim) Report Reporting Period: 16 October 2014 to 15 October 2015
Date: 18/11/2015
EMODnet Annual Report 2 – Lot 2 Geology
Table of Contents
List of abbreviations and acronyms ...... 3 1. Introduction ...... 3 2. Highlights in this reporting period ...... 6 3. Summary of the work done ...... 7 4. Challenges encountered during the reporting period ...... 8 5. Allocation of project resources ...... 10 6. Meetings held since last report ...... 14 7. Work package updates ...... 15 8. User Feedback ...... 52 9. Outreach and communication activities ...... 55 10. Updates on Progress Indicators ...... 62 Annex 1 ...... 66
2
EMODnet Annual Report 2 – Lot 2 Geology List of abbreviations and acronyms
EMSC – European-Mediterranean Seismological Centre EuroGeoSurveys – The Geological Surveys of Europe EUROSION – Project to provide the European Commission with recommendations for policy- making and information management practices to address coastal erosion in Europe, after thorough assessment of knowledge gained from past experiences and of the current status and trends of European coasts.
1. Introduction
This report presents the activities and progress of the EMODnet-Geology Project during the second year from October 2014 to October 2015. The report presents the highlights of the reporting period, a summary of the work done, the challenges encountered during the reporting period, the allocation of project resources, workpackage updates, user feedback, outreach and communication activities and updates on progress indicators. At this stage in the project, the EMODnet-Geology partners have identified suggestions for new information that could form part of the next phase of the EMODnet Programme, which are summarised in the workpackage updates in Section 7 of the report (see workpackages 3, 5 and 6) and in the User Feedback section (Section 8).
The EMODnet-Geology Project is one of seven that brings together information on the Geology, Chemistry, Biology, Physics, Bathymetry, Seabed Habitats, and Human Activities in the European marine environment. During the preparatory phase of EMODnet (2009-2012), 14 organisations from 14 countries demonstrated that geological information could be compiled and harmonised at 1:1,000, 000 scale to provide map information and supporting data for parts of the regional seas of Europe. In 2013, an expanded group of mainly geological survey organisations from 30 countries was successful in being awarded the contract to deliver similar information for the entire European seas (see Figure 1).
The current EMODnet-Geology Project started in October 2013 and will run for 3 years, ending in October 2016. The group consists of 36 partners who are able to provide geological information from all of the European seas shown in Figure 1 and, by including organisations from Iceland, Norway and Russia, to expand the information coverage into the North Atlantic Ocean and to the margins of the Arctic (Barents Sea and White Sea). The information that will be included in the project is principally that held by the project partners, although other organisations contribute to the geological mapping objectives in many of the participating countries. The geology data that were compiled in the preparatory phase and in the current project includes:
• Seabed substrate 3
EMODnet Annual Report 2 – Lot 2 Geology
• Sediment accumulation rate • Sea-floor (bedrock) lithology • Sea-floor (bedrock) stratigraphy • Coastal behaviour • Mineral occurrences (e.g. oil and gas, aggregates, metallic minerals) • Geological events and probabilities (e.g. earthquakes, submarine landslides, volcanic centres)
Figure 1. The European regional seas included in the EMODnet Programme.
The consortium includes the following organisations 1. Natural Environment Research Council – British Geological Survey (NERC-BGS United Kingdom); 2. Geological Survey of Finland (GTK); 3. Geological Survey of Sweden (SGU); 4. Geological Survey of Norway (NGU); 5. Geological Survey of Denmark and Greenland (GEUS); 6. Jardfeingi (Faroe Islands); 7. Iceland GeoSurvey (ISOR); 8. A.P Karpinsky Russian Geological Institute (VSEGEI); 9. Geological Survey of Estonia (GSE); 10. Latvijas Vidas Geologijas un Meteorologijas Centr (LEGMC; Latvia); 11. Lithuanian Geological Survey (LGT); 12. Polish Geological Institute (PGI); 13. Federal Institute for Geosciences and Natural Resources (BGR, Germany); 14. TNO – Geological Survey of the Netherlands; 15. Royal Belgian Institute of Natural Sciences (RBINS); 16. Bureau de Recherches Géologiques et Minieres (BRGM, France); 17. IFREMER (France); 18. Geological Survey of Ireland (GSI); 19. Geological Survey of Spain (IGME); 20. Instituto Português do Mar e da Atmosfera (IPMA, Portugal); 21. Istitituto Superiore per la Protezione e la Ricerca Ambientale. Servizio Geologico d'Italia (ISPRA); 22. Geological Survey of Slovenia (GeoZs); 23. Croatian Geological Survey (HGI); 24. Geological Survey of Montenegro (GEOZAVOD); 25. Geological Survey of Albania; 26. EKBAA- National Center for Sustainable Development (Greece); 27. Hellenic Center for Marine Research, Greece 4
EMODnet Annual Report 2 – Lot 2 Geology
(HCMR); 28. Institute of Oceanology – Bulgarian Academy of Science (IO-BAS); 29. National Research and Development Institute for Marine Geology and Geoecology (GeoEcoMar, Romania); 30. Geological Institute of Romania (GIR); 31. Prychornomorsrge (Ukraine); 32. Dokuz Eylul University (Turkey); 33. Geological Survey of Cyprus; 34. Continental Shelf Department of the Ministry of Transport and Infrastructure (Malta); 35 Centre for Environment, Fisheries and Aquaculture Science (CEFAS, United Kingdom); 36. University of Sussex (United Kingdom).
The partnership consists of the geological survey organisations of all of the maritime countries of the European Union. Twenty-four of the project partners are also members of the Geological Surveys of Europe (EuroGeoSurveys), which exists to promote the work of the geological surveys and therefore provides a long-term association under which the project partners collaborate.
As the principal holders of marine geological information, the partnership also ensures that data from all of the European regional seas (including the entire Baltic, Greater North Sea, Celtic Seas, Bay of Biscay and the Iberian Coast, Macaronesia and the Adriatic Sea) are provided to the project. The project will build on information primarily held by the project partners, but will also connect to other owners of information using Web Map Services (WMS) as for example by linking to the European-Mediterranean Seismological Centre (EMSC), which compiles information on earthquake activities. By doing so, the project will not recreate information that is held elsewhere. The development of a web delivery mechanism using open source standards also aims to ensure the long-term sustainability of the project by delivering the best available and up-to-date marine geological information held by the project partners.
A characteristic of the EMODnet-Geology Project is that the main focus is on harmonised interpreted map information rather than the underlying data that have been used to create the interpreted geological outputs. However, the web delivery system is designed to be able to access data catalogs of information held by each partner organisation and therefore has the potential to access more detailed information (maps and data) at national level. This aspect of the system is considered important to the long-term sustainability of the EMODnet-Geology outputs as it will provide access to the national sources of information where the organisations maintain their information resources.
5
EMODnet Annual Report 2 – Lot 2 Geology 2. Highlights in this reporting period
The second year of the project has been the main phase of compiling and delivering map information for each workpackage. Good progress has been made in all workpackages as described in more detail in Section 7. As more information becomes available there has also been an increase in awareness and use of the EMODnet-Geology outputs, although as reported in Section 10 we are unable to supply progress indicators related to the amount of geological information that has been accessed through the EMODnet-Geology portal. However there is strong evidence of that the EMODnet-Geology outputs are being identified as an important source of information for a wide range of government, industry and research users. There has also been a stronger focus on outreach and dissemination as the outputs are nearing completion.
The main highlights of the reporting period are:
• Assessment of all seabed substrate information available at 1:250,000 scale and the compilation of information at 1:1 million scale where the higher resolution information is not available; • Assessment of pre-Quaternary geology and the start of compilations of Quaternary information, and agreement on geomorphological information that should be included in the EMODnet-Geology outputs; • The development of guidelines for coastal behaviour information and the update of the EUROSION information for about 50% of the European coast; • New maps for geological events and probabilities including earthquakes, underwater volcanic centres, tectonics (active faults), tsunamis (origin and coastal areas affected); • Maps for 9 different mineral types including aggregates, hydrocarbons, gas hydrates, marine placers, phosphorite, evaporite, polymetallic sulphides, polymetalic nodules and cobalt-rich Fe-Mn crusts; • The relaunch of the project portal based on new software developments that continue to be based on the principles of free and open-source software. • Dissemination at national and international level and input to government planning processes; • The establishment of a number of national and regional networks based on the need for better co-ordination of information for example in Iceland (see Section 8), Italy and in the Adriatic and North Sea regions ((see Section 7; Workpackage 1 update); • Use of feedback from the users of the EMODnet-Geology outputs to provide recommendations for new or more detailed work in the next phase of EMODnet.
6
EMODnet Annual Report 2 – Lot 2 Geology 3. Summary of the work done
The second year of the EMODnet-Geology Project has been focused on the consolidation of the data products as outlined in the GANNT chart presented in the project workplan, which described this phase continuing from month 13 (November 2014) to month 24 (October 2015). Progress has been made in compiling information in all of the data product workpackages, namely workpackages 3 (seabed substrate), 4 (sea-floor geology), 5 (coastal behaviour), 6 (geological events and probabilities) and 7 (minerals). In particular, WPs 5 and 6 have now been brought to the same level as the others, as during the first year both were subject to delays in commencing the compilation of information.
As the information layers have developed, there has been a significant increase in outreach and dissemination activities to raise awareness of the EMODnet-Geology outputs and to engage potential users in gaining access to the information.
A number of sub-groups have been formed both nationally and internationally to help co- ordinate the delivery and harmonization of information from countries such as Italy and regions such as the Adriatic and North seas.
The third and final phase of the project (November 2015 to October 2016) will concentrate on the ‘convergence phase’ as outlined in the project workplan. This phase will focus on fine-tuning to ensure convergence with the other EMODnet portals, at the end of which the project team will deliver an analysis of user feedback, a draft Final Report 2 months before the end of phase 3 (August 2016) and a Final Report (October 2016).
7
EMODnet Annual Report 2 – Lot 2 Geology 4. Challenges encountered during the reporting period
Main challenge Measures taken
Delivery of the 1: 250,000 maps of the Work to compile 1:250,000 scale information seafloor of Europe. for seabed substrates demonstrated that only about 20% of the European seabed has information that allows it to be mapped at that scale. To ensure that as much information as possible could be presented on the EMODnet-Geology portal, the group has compiled information at the broader scale of 1:1 million (as used during the preparatory phase of EMODnet). A first version of this map was delivered to the Seabed Habitats Lot on June 30th 2015. Timeline to deliver broadscale 1:1 million The workpackage leader followed the same seabed substrate maps to complete map classification system as used for the coverage in areas where more detailed 1:250,000 maps (modified Folk system) to information was not available. speed the process of harmonisation, generalisation and compilation. Defining the features, both geological and In discussion about the type of seabed in geomorphological, that should be added to addition to its substrate (grain-size the EMODnet-Geology outputs. classification), the team agreed that a list of geologically important features/materials (e.g mobile sands, hard clays, tills) would provide useful additional information. This list of features can be defined in conjunction with the Working Group looking at Geomorphology and Landforms as many of the features will be common to both purposes. Coastline. An updated coastline was made available for all workpackages by the WP3 leader to have a common baseline for all information (with the exception of some coastal information – see Section 7). Representing geological events or features It was agreed that the entire feature should that have both an offshore and onshore be included in the EMODnet-Geology expression, such as volcanic centres. outputs. Similarities in the information being Contact between the EMODnet-Geology 8
EMODnet Annual Report 2 – Lot 2 Geology requested for mineral occurrences by the Workpackage 7 leader and the Co-ordinator EMODnet-Geology Project and where they of the Human Activities Lot has continued are being explored and exploited as throughout the year to avoid duplication of requested by the EMODnet Human Activities effort. Lot.
9
EMODnet Annual Report 2 – Lot 2 Geology 5. Allocation of project resources
At the end of the second year of the project, it is estimated that the main effort required to complete the data gathering workpackages is almost complete and that the thirs year will be for updates. There are some exceptions as explained in more detail in the following paragraphs. In workpackages such as Project Management (WP1), Dissemination (WP9) and Liaison with other EMODnet lots (WP10), the effort could be considered on a pro-rata basis for the overall project duration of 3 years, i.e 33% in the second year of the project and a total of 66 % of the effort to date. No time has been spent on Workpackage 11 (Project analysis and sustainability).
Workpackage 2 (Geological data specification and sourcing), which was initiated at the start of the project to source the information that would be delivered to the project has been more or less completed. It is unlikely that any major data sources will be identified in the final year that we are not already aware of. The allocation of the effort spent on compiling national datasets is therefore considered to be about 95% complete (90% at end of year1).
Each of the information workpackages (workpackages 3 to 7) have progressed at varying rates. Workpackage 3 (Seabed substrate) was the initial priority as the outputs required were essential to the progress of the Seabed Habitats Lot (see Section 7 of this report). The decision to include harmonised 1:1 million information in areas where 1:250,000 information was not available led to more effort in compiling information than had been anticipated during the second year. However the agreed classification scheme that was used for the 1:250,000 scale maps could also be used for the 1:1 million information, which helped to make rapid progress and to deliver a preliminary map for use by the Seabed Habitats Lot by June 2015. There will continue to be updates to both seabed substrate maps in the final year. The overall effort spent on Workpackage 3 is considered to be about 90% (50% at end of year 1).
Workpackage 4 (Sea-floor geology) has so far completed preliminary maps for both the pre- Quaternary and Quaternary geology information. As described in Section 7, the priority is the pre-Quaternary map, which is now well advanced and likely to have only minor changes in the final year. The work carried out by the ‘Geomorphology and Landform’ Working Group has established lists of features that it would be desirable to identify and this work will continue in the final year. However, the objective is to put a baseline in place for the Quaternary and geomorphology maps, both of which it is proposed would be the subject of a more detailed assessment of a future phase of EMODnet, therefore the project partners will deliver as much as possible withing the current timeframe without aiming for a complete information layer. The estimated percentage of effort spent on Workpackage 4 is about 80% (50% at end of year 1).
Workpackage 5 (Coastal behaviour) started later than the other workpackages due to the workpackage leader being unavailable to start the work until mid-2014. However, good progress has been made in delivering the Guidelines and information that updates the 10
EMODnet Annual Report 2 – Lot 2 Geology
EUROSION data. It is generally accepted by the project partners that compilation of coastal behaviour information is a complex subject and recommendations will be made about broadening the scope of this workpackage in the next phase of EMODnet. As it is estimated that about 50% of the EUROSION information has been updated during the EMODnet- Geology Project to date, and allowing for the absence of data in some areas, it is estimated that the effort spent on this workpackage is 60% (30% at end of year 1).
Workpackage 6 (Geological events and probabilities) is making use of existing information using Web Map Services where possible (e.g the European-Mediterranean Seismological Center as an up-to-date source of earthquake information). Section 7 of this report shows that good progress has been made in compiling information on submarine landslides and volcanoes, tectonics (active faults) and information about the origins and coastal impacts of tsunamis. The work to identfy morphological features in conjunction with the Workpackage 4 team will lead to new information being compiled in the final year. The estimated effort spend on this workpackage is 75% (40% at end of year 1).
Workpackage 7 (Minerals) has identified all of the types of information that the project partners should deliver from their national areas. Table 4 and the figures in the workpackage description in Section 7 shows good progress in all aspects of this WP. Compilation of all information layers are at an advanced stage, and are estimated to have taken about 90% of the effort required to complete the information (55% at the end of year 1).
Workpackage 8 (Web services and technology) establishes the mechanism for the delivery of the EMODnet-Geology web services. As this was necessary from an early stage in the project, much of the work has been completed. The recent release of a new version of the GeoNetwork code that is used to build the EMODnet-Geology portal will require an assesment of any new functionality and its application to the portal. The effort spent on this activity is estimated to be 90% of the total (65% at end of year 1).
In general, at the end of the second phase of the project (the ‘consolidation of the data products’ phase) the overall effort spent on the deliverables is mostly in line with the project workplan. The next phase, the ‘convergence phase’ from November 2015 to October 2016 will focus on fine-tuning the existing outputs and considering subjects that could be added or improved during the next phase of EMODnet. The delivery of information through the project portal, with the functionality requested during feedback from the project partners and external users, will also be a priority.
11
EMODnet Annual Report 2 – Lot 2 Geology
Partner WP1 WP2 WP3 WP4 WP5 WP6 WP7 WP8 WP9 WP10 WP11 1 19 0.15 3 0.6 0.45 0.76 0.76 25 5 10 0 2 0.4 0.15 10 0.6 0.45 0.76 0.76 0 0.8 0.66 0 3 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 4 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 5 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 6 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 7 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 8 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 9 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 10 0.4 0.15 0.7 10 0.45 0.76 0.76 0 0.8 0.66 0 11 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 12 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 13 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 14 0.4 0.15 0.7 0.6 10 0.76 0.76 0 0.8 0.66 0 15 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 16 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 17 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 18 0.4 0.15 0.7 0.6 0.45 0.76 10 0 0.8 0.66 0 19 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 20 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 21 0.4 0.15 0.7 0.6 0.45 10 0.76 0 0.8 0.66 0 22 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 23 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 24 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 25 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 26 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 27 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 28 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 29 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 30 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 31 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 32 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 33 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 34 0.4 0.15 0.7 0.6 0.45 0.76 0.76 0 0.8 0.66 0 35 0.4 0 5 0 0.45 0 0 0 0.8 0.66 0 36 0.4 0 0 0 5 0 0 0 0.8 0.66 0 Total 33 5 40 30 30 35 35 25 33 33 0
Table 1. Breakdown of percentage effort by partner and by workpackage (figures for second year of project only). See below for a list of partners and workpackage details.
12
EMODnet Annual Report 2 – Lot 2 Geology
Workpackages and lead organisation
WP1. Project Management. NERC-British Geological Survey (BGS); WP2. Geological data specification and sourcing. NERC-British Geological Survey (BGS); WP3. Sea-bed substrate. The Geological Survey of Finland (GTK); WP4. Sea-floor geology. Bundesanstalt fur Geowissenschaften und Rohstoffe – the Federal Institute for Geosciences and Natural Resources, Germany (BGR); WP5. Coastal behaviour. TNO-Geological Survey of the Netherlands; WP6. Geological events and probabilities. Institituto Superiore per la Protezione e la Ricerca Ambientale: ISPRA; WP7. Minerals. The Geological Survey of Ireland (GSI); WP8. Web services and technology. NERC-BGS and Bureau de Recherches Geologiques et Minieres, France (BRGM); WP9. Dissemination. NERC-British Geological Survey; WP10. EMODnet collaboration. NERC-British Geological Survey; WP11. Project analysis and sustainability. NERC-British Geological Survey;
Partner numbers
1. Natural Environment Research Council – British Geological Survey (NERC-BGS United Kingdom); 2. Geological Survey of Finland (GTK); 3. Geological Survey of Sweden (SGU); 4. Geological Survey of Norway (NGU); 5. Geological Survey of Denmark and Greenland (GEUS); 6. Jardfeingi (Faroe Islands); 7. Iceland GeoSurvey (ISOR); 8. A.P Karpinsky Russian Geological Institute (VSEGEI); 9. Geological Survey of Estonia (GSE); 10. Latvijas Vidas Geologijas un Meteorologijas Centr (LEGMC; Latvia); 11. Lithuanian Geological Survey (LGT); 12. Polish Geological Institute (PGI); 13. Federal Institute for Geosciences and Natural Resources (BGR, Germany); 14. TNO –Geological Survey of the Netherlands; 15. Royal Belgian Institute of Natural Sciences (RBINS); 16. Bureau de Recherches Géologiques et Minieres (BRGM, France); 17. IFREMER (France); 18. Geological Survey of Ireland (GSI); 19. Geological Survey of Spain (IGME); 20. Instituto Português do Mar e da Atmosfera (IPMA, Portugal); 21. Istitituto Superiore per la Protezione e la Ricerca Ambientale. Servizio Geologico d'Italia (ISPRA); 22. Geological Survey of Slovenia (GeoZs); 23. Croatian Geological Survey (HGI); 24. Geological Survey of Montenegro (GEOZAVOD); 25. Geological Survey of Albania; 26. EKBAA- National Center for Sustainable Development (Greece); 27. Hellenic Center for Marine Research, Greece (HCMR); 28. Institute of Oceanology – Bulgarian Academy of Science (IO-BAS); 29. National Research and Development Institute for Marine Geology and Geoecology (GeoEcoMar, Romania); 30. Geological Institute of Romania (GIR); 31. Prychornomorsrge (Ukraine); 32. Dokuz Eylul University (Turkey); 33. Geological Survey of Cyprus; 34. Continental Shelf Department of the Ministry of Transport and Infrastructure (Malta); 35 Centre for Environment, Fisheries and Aquaculture Science (CEFAS, United Kingdom); 36. University of Sussex (United Kingdom).
13
EMODnet Annual Report 2 – Lot 2 Geology 6. Meetings held since last report
Date Location Topic Short Description 18 January 2015 Paris, France. Working Group on Meeting to define Geomorphology and further steps towards Landforms. a common agreed list and definitions useable for both Workpackages 4 and 6. March 10-11 2015 Madrid, Spain. Full Project Meeting. Meeting to discuss all aspects and progress of the project. May 25-26 2015 Rome, Italy. Adriatic Sea Group Meeting to co- Meeting. ordinate specific geological information from the Adriatic region.
EMODnet-Geology Project Meeting, Madrid, Spain.
14
EMODnet Annual Report 2 – Lot 2 Geology 7. Work package updates
Workpackage 1. Project Management (led by the British Geological Survey)
Objectives: To manage the overall project, ensure delivery of the outputs and outcomes as agreed with the European Commission. To assess and evaluate the project and its results.
The EMODnet-Geology Project contract was signed on October 16 2013. Work on the project commenced immediately with partners starting the exercise of compiling inventories of available data for workpackages 3-7 (Workpackage 2). During the second year of the project, a full project meeting was held in Madrid, Spain on 10-11 March 2015 hosted by the the Instituto Geológico y Minero de España (IGME) at which all partners but one (Partner 32, Dokuz Eylul University, Turkey) were able to attend. The overall progress and plan for the following months were described by the Project Co-ordinator and the timescale for achieving the objectives according to the project workplan were presented along with financial information and other project management procedures. Each workpackage leader presented progress reports and summarised the actions required from each partner to meet the deadlines for delivery of the outputs. The meeting also included presentations by each partner summarising the status of the information that had been compiled at national level and submitted to the workpackage leaders.
Two sub-groups were formed during the year to address specific issues related to aspects of workpackages 4 (Seabed geology) and 6 (Geological events and probabilities) as described in the following progress reports for each workpackage.
The Project Co-ordinator has provided bi-monthly progress reports to the Secretariat and the European Commission according to schedule and has attended the EMODnet Steering Group meetings in Brussels (December 9/10, 2014) and Ispra, Italy (1 July 2015), as well as the INSPIRE meeting and open meeting with Joint Research Centre staff during the week in Ispra. The Project Technical Co-ordinator, Jonathan Lowndes (British Geological Survey) also attended the meetings. Towards the end of the reporting year, plans were advanced to participate in the EMODnet Jamboree held in Ostend, Belgium during the week commencing 19 October 2015.
The Project Co-ordinator attended the ‘Galway Statement Implementation – Atlantic Seabed Mapping’ meeting in Dublin, Ireland on December 2nd 2014 to present the EMODnet-Geology Project. Subsequently he was appointed as the Chair of the Working Group on Atlantic Seabed Mapping with European-US-Canadian participation. These activities contribute to the dissemination objectives and provide the opportunity to influence the development of a trans-Atlantic approach to seabed data collection and sharing, similar to the approach taken by the EMODnet-Geology partners.
There have been notable agreements or groups formed as a result of the EMODnet-Geology project, both at national and international level. For example, to advance co-operation in
15
EMODnet Annual Report 2 – Lot 2 Geology
Italy, a Memorandum of Understanding was signed between public institutions in order to deliver the most comprehensive national products to the project: the participants are the Istitituto Superiore per la Protezione e la Ricerca Ambientale, Servizio Geologico d'Italia (ISPRA ), the Italian National Institute of Oceanography and Experimental Geophysics (OGS), the CNR Institute of Marine Sciences (ISMAR), the Italian Ministry of Economic Development (MISE), the Italian National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA), and the Universities of RomaTRE and Palermo.
The project partner countries around the Adriatic Sea (Italy, Slovenia, Croatia, Montenegro, Albania and Greece) have created a working group on Adriatic Sea geology, which will work on all aspects of the EMODnet-Geology Project workpackages. The first meeting of the Adriatic Sea Group took place in Rome on May 25th and 26th 2015, and a second meeting will be held in Ostend, Belgium on the 19th October.
An informal regional network of North Sea partners from Denmark, Germany, the Netherlands and Belgium, has also grown out of EMODnet-Geology, and is gaining the interest of other disciplines. As reported in Section 10, EMODnet-Geology partners from Denmark, Belgium and the Netherlands have been invited to present Workpackage 3 developments at a workshop on habitats organised by the German Federal Agency of Nature Conservation at the end of November (Vlim, Germany).
Workpackage 2. Geological data specification and sourcing (led by the British Geological Survey)
Objectives: To prepare and provide access to the information required to deliver 1:250,000 maps of the seabed substrate (improving where possible the current resolution of the classes and data), the rate of accumulation and/or sedimentation on the sea-floor; geological boundaries and faults; geological events and event probabilities and minerals. For the coast, information will be provided on rate of sedimentation and erosion, and the coastal typology and/or behavioural description. The access to partners’ data will use the standards developed in the Geo-Seas Project. The data products (maps) will use the standards developed in the ur-EMODnet-Geology Project.
The sourcing of information, although mainly focussed in the early stages of the project, has continued throughout the second year as information has been identified and made available to the EMODnet information products. Progress is reported in the following sections describing workpackages 3-7.
Workpackage 3. Seabed substrate (led by the Geological Survey of Finland (GTK))
Objective: To compile and harmonise all available seabed substrate information at a scale of 1:250,000 to deliver the seabed substrate component of Section 2.4 of the tender documents and all available information on the rate of accumulation and sedimentation on the sea floor.
16
EMODnet Annual Report 2 – Lot 2 Geology
The aim of Workpackage 3 is to produce a fully populated GIS layer of harmonised seabed substrate information, including confidence information based on the density of the information used to create the geological interpretations (based on principles established during the preparatory phase of EMODnet from 2009-2012 - ur-EMODnet-Geology Project. In addition, a model for semi-automated classification of acoustic data for spatial modelling of substrate data has been provided by a case study led by the Centre for Environment, Fisheries and Aquaculture Science (Cefas, United Kingdom) (Annex 1).
The main objective of Workpackage 3 is to provide substrate information from the study area at a scale of 1:250,000. However, as only about 20% of the European sea area has been mapped at this scale (or more detailed), it was decided to produce a seabed substrate data layer at a broader scale i.e. 1:1 million to improve the coverage. The 1:1 million scale information is expected to have approximately 60% coverage. During the last year, Workpackage 3 has therefore focused on producing the 1:1 million layer of seabed substrate information, while continuing to update the 1:250,000 substrate information
Figure 2. Coverage of the seabed surface substrate data available for the EMODnet-Geology Project, with the scales of the available sea-bed substrate maps that have been used in the harmonised maps shown in Figures 4 and 5 (map prepared in June 2014).
17
EMODnet Annual Report 2 – Lot 2 Geology
as data become available, and to update the index of available mapping information. Workpackage 3 also started to collect information on the rate of accumulation and sedimentation as well as information on confidence of the substrate map. These datasets will be finalised during the 3rd year of the project.
Figure 3. Index map showing different classification schemes used at national level, which have been harmonised according to the agreed EMODnet schema.
The EMODnet-Geology partners have provided information on where and what kind of seabed surface substrate data/maps they have available for the project from their national waters including the EEZs (Figure 2). The index map was produced during the first year of the project and includes data from more than 50 organisations and more than 400 different map types. The individual patches describe areas that are congruent by scale and mapping technology. The GIS shapefile from which the map is derived contains metadata information that explains where the data is dreived.
Figure 3, which was produced early in the project (June 2014), shows the complexity of the classification schemes that were used by geological organisations in Europe and demonstrates how the harmonisation process that has been possible during EMODnet has significantly helped to present information that is comparable on a regional scale. There are more than 30 different classification systems used to compile national seabed substrate maps, which are summarised in Table 2.
18
EMODnet Annual Report 2 – Lot 2 Geology
Grain Size EMODn MNCR Udden - GTK& GEUS VSEGEI EGK Lithuania Ukraine Romania SHOM Larsonn Aloisi Augris Simplet Gautier Berné BRGM Hily Comments Wenthw Mm > Ø et Folk orth SGU Raukas eur (1990- (2011) (2009) (1986- (1976) 1981 (1979) 2013 1987) modif, Lesueur & Klingebi el (1986) modif >600 Bould Bould Boul Boul Bou Bould Bould Boulde Bould Caillou Caillo Boul Cobbl Cobbl Grav Galets ( Grave All defined boulder No information No information categories fall into this ers er (> der der lder ers (> ers > r (> ers > tis utis der e e el to >64) l( >2 ) group. Overlap with large (>256 256) (> (> (> 100) 1000 100) 10 (~peb (pebbl (>60 (>64) (>64) cob stones. Some national ) 256) 600) 200) ble, e), 0) ble, classifications include >200) Coquil > 2) smaller particles in -600 Larg Cobbl Cob category Boulders (GTK, e es les ble GEUS, VSEGEI, Ukraine)
Ston (100- (Shell) (>64 es 1000) , ) (>200) -256 Grave Cobbl Cob Gravel Gravel Almost all gravels belong here. Also stones, cobbles -200 l (2- e (64 ble Sma Sto (2- (2- and pebble sizes fit to -100 256) – 256) (64 - ll ne Cobbl Pebbl Cobble 200) 200) gravel grain sizes. Large 256) ston (20 stones overlap with es e (10- s (10 – boulder sizes. Some -64 Pebbl Peb es – (10 – 100) 100) Peb Pebbl Pebbl Gravel ( boulder sizes (GEUS, e (16 ble (60 200) 100) ble e (16- e (16- 4-64) VSEGEI, Ukraine) overlap – 64) (4- – (20- 64) 64) with gravel category. 64) 200) 64) -60 Gra -20 vel Gra Grav -16 Grave (2 – vel Cobbl el(2- Grav Grav -10 l (4 – 60) (2 – Grav Gran Gravel es (10 20) el (2- el (2- 16) 20) el (2 ule (1-10) – 100) 16) 16) -4 Coars Gra – 10) (1- Gravel e nule 10) (2-4) Sand (2 – (1 – 4) 4) -2 Sand Very Coa San Sand Sand Sand, Coars Coa Coar Coars Coars Coar Coarse Coars Quite allright. Some problems with fine sand (0.062 coar rse d (0.05 (0.062 (0.5-2) e sand rse se e e se sand e grain sizes as with MNCR 5-2) se san (0.0 – 2) 5-2) (0.5-2) san sand sand sand sand (0.5-2) sand Coarse sand grains. VSEGEI san d 6 – d (0.5- (0.5- (0.5- (0.5- (0.5- sand includes smaller d (1 (0.0 2) (0.5 2) 2) 2) 2) 2) particles. – 2) 6 – -2) -1 Medi Coa 2) Sand Coarse Sand -0.6 um rse Me (0.1- sand (0.1- Sand san diu 1) (0.5-1) 1) (0.25 d m – 1) (0.5 san – d 1.0) (0.2 -0.5 Me – Mediu Fine Fine Fine Med Medi Medi Med Mediu Fine diu 0.6) m sand sand sand san ium um um ium m sand sand ( m (0.25- (0.05- (0.2- d sand sand sand sand ( 0.25- 0.062 san 0.5) 0.5) 0.5) (0.2 ( ( ( ( 0.5) 5-0.5) d - 0.25 0.25- 0.25- 0.25 (0.2 0.5) -0.5) 0.5) 0.5) -0.5) 5 – 0.5) -0.25 Fine Fine Fine Fine Fine Fine Fine Fine Fine -0.2 Sand san san sand Very Ver sand sand sand sand sand (0.062 d d (0.1- fine y (0.2 (0.25 (0.25 (0.2 (0.25- 5 – (0.1 (0.0 0.25) sand fine 5- - - 5- 0.125) 0.25) 25 – 6 – (0.05- san 0.12 0.125 0.125 0.06 0.25 0.2) 0.2) d 5) ) ) 25) ) (0.0 -0.125 Very Silt Coarse Silt 4- Very Very Very Very fine (0.00 silt (0.003 0.2) fine fine fine fine san 5- (0.05- 9- sand sand sand sand d 0.1) 0.1) 0.062 (0.0 (0.06 (0.06 (0.0625- (0.0 5) 625- 25- 25- 0.125) 625 0.12 0.125 0.125 – 5) ) ) 0.12 5) -0.063 Mud Mud Mu Silt Silts, Silt Silt Silt Silt Silt Mud All mud, silt and clay grain sizes belong here. Part of -0.06 (< (< d (< Silt Silt (0.01- (<0.06 (0.0 (0.00 (0.00 (0.0 (0.002- (0.062 the fine sands (GTK, SGU, 0.062 0.062 0.06 (0.0 (0.0 0.1) 25) 02- 2- 2- 02- 0.0625) 5) GEUS = difference only -0.05 5) 5) 25) 02 – 02 – Silt Fine Clay Lutite 0.06 0.062 0.062 0.06 0.0025 mm, VSEGEI=difference -0.04 0.06 0.06 (0.00 silt (<0.05 (<0.05 Mud 25) 5) 5) 25) 0.0125mm) overlap here. ) ) 5 – (0.01- ) ) (<0.0 EGK and Ukraininan silt -0.02 4) 0.05) 0.05) classes overlap with sand and mud categories. -0.01 Clay (< Clay (< 0.01) 0.01) -0.005 -0.004 Clay Clay (< -0.002 Clay Clay (< Clay 0.0039 Clay Clay Clay Clay Clay (< (< 0.005 (<0.0 ) ( (<0.0 (<0.0 (<0. (<0.002) 0.00 0.00 ) 05) <0.0 02) 02) 002) 2) 2) 02) Baltic and North Sea Black Sea French classifications
Table 2. The grain-size limits used in different classification systems applied to national seabed substrate information.
Seabed substrate information at 1:250,000 scale
The 1:250,000 seabed substrate dataset produced during the first year of the project has been updated with new data from the project partners. The latest update of the 1:250,000 seabed substrate data shown in Figure 4 is from October 2015. Many partners only started to collect spatial data from existing archives and/or to digitize seabed substrate data as a result of the EMODnet-Geology project. In some cases the work has been time consuming therefore the data was not available for the first version of the seabed substrate dataset that was made available to the EMODnet Seabed Habitats Lot in June 2014. However this 19
EMODnet Annual Report 2 – Lot 2 Geology
has now been included in updates provided during the second year. The seabed substrate maps/data on a scale of 1:250 000 for the European Seas will be produced using hierarchies of 16, 7 and 5 classes. It is not possible to produce seabed substrate map based on 16 sediment class covering all areas shown on the index map (Fig 2) as original data do not include this detailed information. The final seabed substrate map will therefore be based on 5 sediment classes, which will allow a map to be produced for all areas shown in Figure 2. Some partners have active mapping programmes, and as they produce new information on seabed substrates these will be included in the WP3 output.
Figure 4. The EMODnet-Geology seabed substrate data at 1:250,000 scale based on 15 sediment classes plus areas of rocks and boulders at the seabed.
Seabed substrate information at 1:1 million scale
The 1:1 million scale seabed substrate data is an additional output of Workpackage 3, which was agreed at the EMODnet-Geology meeting in Malta (October 2014) so that the project partners could deliver broadscale seabed substrate dataset from the European seas that extends the coverage given by areas mapped at 1:250,000 scale. Compilation of the 1:1 million seabed substrate information has been the main focus of Workpackage 3 during the second year of the project. 20
EMODnet Annual Report 2 – Lot 2 Geology
The work to compile the 1:1 million information followed the Guidelines used in the previous phase of the project (for 1:250,000 information) and included several phases:
1. Harmonisation. The national seabed substrate data were translated into the EMODnet substrate classification system; 2. Generalisation. The maps were generalised at the target scale if not originally at that scale; 3. Compilation. The national seabed substrate maps were compiled into a European seabed substrate map/EMODnet-Geology substrate map; 4. Update of the map. The project partners delivered their harmonised (and generalised datasets) to the workpackage leader, GTK, who then combined the information into a single map (Figure 5). The resulting map was then sent to the EMODnet Seabed Habitat Lot on June 30th 2015. This version of the seabed substrate data will be updated during the final year of the EMODnet-Geology project as more data become available.
Figure 5. The EMODnet-Geology seabed surface substrate data at a scale of 1:1 million for the European Seas, based on 4 sediment classes plus areas of rocks and boulders at the seabed.
21
EMODnet Annual Report 2 – Lot 2 Geology
Figure 6. The Folk sediment triangle and the hierarchy of combined Folk classification developed for the EMODnet-Geology project.
The 1:1 million data uses the same modified Folk classification system as the 1:250,000 data (Figure 6). If the national datasets were not originally classified under the Folk system, they were reclassified on the basis of samples or/and expert-based prediction. Because of the challenging timeline and the diversity of the national datasets, the substrate reclassification scheme is simple and sometimes provides only a rough estimate for the substrate material from the uppermost 30 cm of the sediment column.
It was recommended that the most detailed possible Folk classification (16-7-5 Folk classes) should be used. Because there are fundamental differences in some national datasets (e.g. grain-size limits), it is not always possible to make one-to-one translation of the national substrate category into the Folk category. Thus the resulting class might be more of a "compromise" that includes the majority of the substrate variation in that class.
Confidence assessment 22
EMODnet Annual Report 2 – Lot 2 Geology
The British Geological Survey (BGS) leads the work to produce a confidence analysis of the seabed substrate data. BGS distributed the guidelines on Confidence Methodology to all partners in August 2015. The confidence analysis is based on the index map (Figure 2), which is used to assign a confidence score (0-4) using the “3 step confidence method” developed by the UK’s Joint Nature Conservation Council (JNCC) (Three-step confidence assessment framework for classified seabed maps, H. Elwood, JNCC). It is expected that the first version of the confidence dataset will be available by the end of November 2015. An example of the type of confidence information from the UK seafloor is shown in Figure 7.
Figure 7. An example of the confidence data from the UK marine area.
Sediment accumulation
The work to collect data on the rate of accumulation and sedimentation started during the second year of the EMODnet-Geology Project. The Workpackage 3 leader (GTK) presented the approach to collect the sedimentation data at the EMODnet-Geology Project meeting in October 2014, and the Guidelines on the data preparation were then distributed to the project partners in December 2014. Information was provided to the Workpackage 3 leader by April 2015 and was combined into the map shown in Figure 8 by October 2015. Primarily the data have been compiled as point-source information. Initially it was expected that a contoured map may be possible; however even in areas where the data points are relatively dense, such as in the Baltic Sea, it is now our view that a contoured map would be 23
EMODnet Annual Report 2 – Lot 2 Geology misleading both spatially and temporally as it would infer information in areas where there is no evidence to support accumulation rates. The user of such information would derive greater confidence from the values shown at individual points. . The information provided is mainly derived from published sources and during the final year of the project there will be further effort to include as much information as possible.
Figure 8. Information on sediment accumulation rates for recent sediments.
Geological features
In addition to providing the seabed substrate information classified according to the modified Folk system, the project team has discussed providing a list of geologically important surface features/materials that are not included in the Folk classification. This would include for example features such as mobile sands, hard clays and till areas. The Workpackage 3 team will produce a preliminary list of features in November 2015, and the project partners will contribute information to the geological features output during the final year of the project.
Coastline
The EMODnet Geology Project has used a coastline adopted by the European Environment Agency (EEA) on a scale of 1:100 000 (last upload 4th of July 2013), which is expected to be the standard coastline adopted by all of the EMODnet projects. The coastline has been available for partners on the WP leadfer’s ftp-site.
24
EMODnet Annual Report 2 – Lot 2 Geology
During the compilation it was obvious that the EEA coastline was not accurate enough in some regions, for example in Russia, the Canary Islands and Malta, therefore adjustments were made using national information. The usefeulness of the coastline has been discussed at project workshops and the majority of the partners have confirmed that the detail is adequate, although others, such as Italy have found that a lot of work is required to fit national data to the map currently used in the EMODnet-Geology Project. A definitive and detailed coastline for all EMODnet projects is therefore still seen as a future requirment.
Case Study on Quantitative spatial prediction of seabed sediment composition
Workpackage 3 includes a module that models the substrate and other terrain derivatives (e.g. morphology and seabed dynamics) from acoustic data layers to serve as case studies on deriving biologically relevant (and EUNIS compatible) datasets. This case study has been perfomed by the UK Centre for Environment Fisheries and Aquaculture Science (Cefas) and is presented in Annex 1.
Recommendations for future work on sediment dynamics and seabed sediment geochemistry
Sediment dynamics is an important driver of many physical, chemical and biological processes, and the distributing agent of chemical substances and biological organisms. Key to understanding the associated environmental effects is to understand sediment sources, transfer and storage. Pan-European mapping of large-scale sediment transport pathways and regional sediment fluxes will require the combination of geological, sediment and geomorphological data with hydrodynamics and hydrology. Data products from EMODnet- Geology, EMODnet-Bathymetry and EMODnet-Physics, produced in the current phase, are highly useful to map source and sink areas. Seabed-sediment patterns and gradients (Geology) reflect the major erosional and depositional sedimentary systems resulting from the current-derived (Physics) regional sediment transport pathways. Where quantitative grain-size data are available they can be combined with current data to provide estimates of seabed mobility. In areas with abundant sand, net sand transport can also be derived from the asymmetries of bedforms (Bathymetry). Finally, sediment fluxes and budgets are available from the literature. A pan-European source-to-sink approach is important to understand the sediment budgets in a supraregional framework, and thus to predict transnational sedimentation and erosion patterns affecting habitats and food chains.
The EMODnet-Geology partners have for several decades gathered a considerable amount of seafloor sediment chemistry data. For example in the UK more than 9000 seabed samples have been analysed for up top 38 metals and have been published on the BGS website (see http://www.bgs.ac.uk/discoverymetadata/13605645.html). Although the EMODnet- Chemistry Lot has been made aware of the data, there is a need to provide seafloor sediment geochemistry interpreted information to the EMODnet portal. The EMODnet- Geology consortium proposes that such data and interpretattions are included as a new 25
EMODnet Annual Report 2 – Lot 2 Geology
parameter in the next phase of EMODnet. This data has traditionally been produced by geologists and the EMODnet-Geology partners therefore have a good understanding of what is required to prepare this data for EMODnet. Such data is of high value in many marine activities and in marine spatial planning, which is important also when considering implementation of the maritime spatial planning directive (2014/89/EU).
Workpackage 4. Sea-floor geology (led by the Federal Institute for Geosciences and Natural Resources, Germany (BGR)).
Objective: To compile and harmonise all available seabed geology (outcrop and sub- Quaternary) information at a scale of 1:250,000 where possible.
Workpackage 4 continues to compile and harmonise pre-Quaternary and Quaternary geological information at 1:250,000 scale whereever possible. In the EMODnet-Geology project proposal, the group proposed to compile information on the Quaternary geology of the sea floor (sediments deposited during the last (approximately) 2 million years). These data were excluded from the ur-EMODnet-Geology project and not specified in the tender documents for the current EMODnet programme, however these are known to be of relevance to a wide range of users of the seabed environment. It was acknowledged that although the scope of the project would not allow full integration and harmonisation of the Quaternary geology of the European offshore area, the merger of national information would provide a basis for targeting future research (see Figure 11).
During the year, the Workpackage 4 leader produced a set of Guidelines describing the procedures for compiling information for both pre-Quaternary and Quaternary geology information. The leaders of Workpackage 4 and 6 (Geological Events and Probabilities; led by the Instituto Superiore per la Protezione e la Ricerca Ambientale, Italy: ISPRA ) also established a Working Group on ‘Geomorphology and Landforms’ during the EMODnet- Geology Project meeting in Malta in October 2014, to agree the definition of features that will be used to construct the Quaternary maps for the project. A meeting of the group was held in Paris, France on 18th February 2015.
Guidelines for collection and compilation of pre-Quaternary and Quaternary information
An inventory was produced of all available map datasets of both Quaternary and pre- Quaternary geology at each of the partner institutions, including the scales of the maps. To assist this process, a questionnaire was distributed to each organisation and the results were presented in preliminary maps.
A practical procedure (data model, geology classifications) was developed by the WP leader to demonstrate how to collect and deliver the Workpackage 4 geology data in a format conformable to INSPIRE.
26
EMODnet Annual Report 2 – Lot 2 Geology
These procedures were described in a Technical Guidance document “EMODnet–Geology - Workpackage 4: Seafloor Geology. Semantic Transformation and Vocabulary” . The document contains a description of the procedure to transform data according to OneGeology-Europe/INSPIRE standards and also includes the necessary vocabulary (plus definitions) to describe the data with the following properties:
• age (geochronology); • lithology; • structures/faults; • genesis (event environment, environmental process).
Pre-Quaternary geology
The pre-Quaternary geology data has been collected, integrated into the GIS and updated (Figure 9) . Due to the use of the common vocabulary presented in the Technical Guidance document, the collected datasets were largely semantically harmonised.
Figure 9. Status of sea-floor geology (pre-Quaternary lithology) available for Workpackage 4.
Quaternary geology and the Geomorphology and Landforms Working Group
27
EMODnet Annual Report 2 – Lot 2 Geology
The EMODnet-Geology Geomorphology and Landforms Working Group met in Paris on 18th February 2015 at the Maison de la Géologie (Figure 10). Since the group’s foundation at the EMODnet-Geology meeting in Malta in October 2014, almost 200 terms and definitions that are potentially important to describe submarine landforms had been collected and discussed by mail.
The workshop, which was attended by 16 EMODnet-Geology project members and the scientific Quaternary Map of Europe (IQUAME) advisor Phil Gibbard, Cambridge University, had the objective of defining further steps towards a common agreed list and definitions useable for both Workpackages 4 and 6. The discussions took into account the various national geologists’ view and the approaches of both workpackages, following which there was an agreed reduction of the number of terms to less than 100 definitions and a subdivision into the classes: • physiographic features; • biogenetic features; • landforms.
In addition to streamlining the EMODnet-Geology terms, synergies with the project to “Review of the Quaternary Geological Map of Europe” (IQUAME 2500) were discussed in Paris. Landforms/geomorphology are also subjects included in the IQUAME data descriptions.
Figure 10. Participants at the Geomorphology and Landforms Working Group meeting in Paris.
28
EMODnet Annual Report 2 – Lot 2 Geology
Figure 11. Sea-floor geology (Quaternary). Age.
Workpackage 5. Coastal behaviour (led by TNO – Geological Survey of the Netherlands).
Objective: To classify the coastal behaviour and typology of each country represented in the project partnership. A classification system for coastal types will be applied and areas of erosion and sedimentation in the coastal zone of each participating country will be shown on the resulting map layer.
At the start of the reporting year, the Project Co-ordinator and the Workpackage 5 leader met in Edinburgh (November 2015) to discuss the plan for the next 12 months. Subsequent discussions took place between the WP leader and other project members to agree on responsibilities. The project partner from the University of Sussex was allocated responsibility for working on coastal cliff vulnerability, sensitivity and an erosion index for the project. A matrix of features to be mapped was prepared and distributed to the project team.
Progress during the year has since focussed on the compilation of coastal migration information. The process involves using the EUROSION project data (http://www.eurosion.org/) as a baseline for pan-European coastal information and to update the data with information from the EMODnet-Geology partners. A recent version of the map of coastline migration is shown in Figure 12, although it should be noted that this does not include information that was received during October 2015. It is estimated that 29
EMODnet Annual Report 2 – Lot 2 Geology about 50% of the EUROSION information has been updated by the EMODnet-Geology partners (see Table 3 for details).
The Guidelines for compiling information for Workpackage 5 were presented by the WP leaders at the EMODnet-Geology Project Meeting in Madrid (March 2015) and distributed to each partner on 10th March.
The Guidelines aim at easy and uniform delivery of partner contributions to coastal behaviour, where ‘behaviour’ refers to the movement of the coastline in a landward (erosion) or seaward (accretion) direction. Two behaviour-related output parameters are defined: migration (subdivided into direction, rate and associated volume) and resilience. Migration is a prime indicator of behaviour, as it describes coastline changes caused by erosion and accretion. Resilience is the ability of a coastline to absorb and recover from erosion before a critical state is reached. Along with resistance, the ability to stop or resist change, it is a measure of vulnerability and provides a potential link to the risk faced by the coastal-zone population. The term resilience is most commonly used for coastlines formed by unconsolidated sediment. It is also applied to coastal cliffs and bluffs, though most of the literature that relates to, or includes, rock coasts refers to resistance, sensitivity and vulnerability rather than to resilience per se.
The key difference between rock coasts composed of consolidated sediment or lithified rock, and soft coasts, composed of unconsolidated sediment, is that there is no way to restore a rock coast once it has been eroded. The episodic nature of cliff and bluff recession is a key attribute when it comes to assessing vulnerability and risk. Resilience is particularly relevant in densely populated or economically valuable areas. Here, a critical state would result not just in significant coastline reorganization but also in major socio-economic impact.
In the Workpackage 5 Guidelines, the parameters used to describe coastal behaviour are defined and discussed. The structure of the attribute tables providing the metadata needed for harmonised data delivery to the EMODnet-Geology portal is also presented, along with the data-delivery methods necessary to allow seamless visualization of all coastline data supplied by the project partners, including easily accessible data available from non-partner institutes. Finally, a pilot study on coastal resilience was proposed, in which metadata and data on this parameter are requested on a voluntary basis. The GIS output will be published as OGC and INSPIRE compliant Web Map Services (WMS).
Progress
During the last year, and throughout the final year of the project, the Workpackage 5 leader has compiled, collated and presented each partner’s information on coastal behaviour using a standardized INSPIRE-compliant format. Migration direction, distilled for the most part from EUROSION (a former EU project on coastal erosion: www.eurosion.org), is the only mandatory parameter. Rate and associated volume of coastline migration, as well as
30
EMODnet Annual Report 2 – Lot 2 Geology
WP5 updated % updated Availability in Available in Remarks data EUROSION EMODnet dataset
Norway 100 No Yes
Lithuania 100 Yes Yes
Estonia 100 Partial Yes Only small area completed ; other areas have no information.
Poland 100 Yes Yes
Ireland 10 Yes No Pending update on EMODnet due to data formatting.
Belgium 100 Yes No Delivered as map.
Spain 100 Yes Yes
Ukraine 100 No Yes
Albania 100 No No Pending update due to error on dataset.
Sweden 5 Partial No Pending update on EMODnet due to data formatting.
Italy 100 Yes Yes
Russia 100 No Yes
Netherlands 100 Yes No Delivered as map.
31
EMODnet Annual Report 2 – Lot 2 Geology
Denmark 100 Yes No Pending update on EMODnet due to data formatting.
Portugal 10 Yes Yes Pending update on EMODnet due to data formatting.
Table 3. Updates to the EUROSION database according to information made available from countries who have delivered new information to the EMODnet-Geology Project. resilience, are solicited on a voluntary basis. Although harmonised and standardized layers for publication will be delivered using map web services, experience shows that the underlying data have been collected using different methods that generate output characterised by a multitude of specifications. The pluriformity of the data reflects a range of reference planes, coastline indicators, and time periods of monitoring.
Unlike the other workpackages, the provided and collated data concern only the EMODnet participant states’ coastlines and not their offshore territories. Various coastlines have been used as a frame of reference in different European projects. In this light, partners are encouraged to use the original coastline linked to the delivered data, and not to project these data onto the coastline presently used by other EMODnet-Geology workpackages. The resulting mismatch is not a problem at this stage. On the contrary, visualization of mismatches emphasizes that there is much work to be done in the next EMODnet phase. Visualizations will also show data and map coverages and gaps.
In order to identify and map areas of erosion, deposition and different degrees of vulnerability, each partner has been asked to provide all publicly available information on the following indicators:
• direction of coastline migration (mandatory) i.e. landward, seward or stable (imperceptible change, taken in practice as <0.5 m/yr mean change) • rate of coastline migration (voluntary) m/yr (mean landward or seaward change over period of monitoring; • volume of coastal erosion/accretion (voluntary) m3/m/yr (mean sediment gain or loss per metre of coastline over period of monitoring, as measured between a lower and an upper reference level); • coastal vulnerability (voluntary) (measure of risk of and resilience against erosion and flooding); many different indices are used in Europe, assessed on the basis of parameters such as wave height, coastal slope, river discharge, geomorphology, coastline behavior. There is no single standard.
32
EMODnet Annual Report 2 – Lot 2 Geology
The workpackage leader has also been working on a document on coastal vulnerability. It shows that there is no single way to portray coastal vulnerability for all of Europe in a meaningful way, especially for soft-sediment coasts. A case in point is a recent study outlining the different approaches for the US Pacific and Atlantic coasts, following years of study. Even where the same parameters can be used, the classes used in the vulnerability assessment have different boundaries. Otherwise, you would end up with a product that would work well for the Pacific but deem almost all of the Atlantic highly vulnerable, or work well for the Atlantic but deem almost all of the Pacific invulnerable. For management purposes, such uniform products are not helpful.
Recommendations for future work on coastal behaviour
The main issues in coastal behavior to be addressed in the next phase of EMODnet are (a) coastal vulnerability and (b) coastal sediment budgets. Compilation of European coastal- vulnerability maps will contribute to a pan-European tool for assessing the capacity to cope with the adverse effects of sea-level rise and coastal erosion. These maps visualize the socio-economic relevance of coastal-geological change as captured in the previous EMODnet phases. Sediment budgets, the second issue, quantify the exchange of sediment along and perpendicular to the coast. An inventory of existing sediment-budget studies is needed to establish a link between the different coastal behavior data products and those from other EMODnet topics such as seabed dynamics and bathymetry. It will help end users to explain and predict past, present and future coastal change and vulnerability. For data- availability reasons, it will not be possible to work at a common scale or with a common legend. Studies have shown, however, that such harmonization is not necessary and not even desirable from a pan-European perspective. Common discovery is more important. Aside from these new issues, existing data products on coastal typology and behavior at scales smaller than 1:250,000 should be added to the EMODnet-Geology inventory. They will add relevance in areas where 1:250,000 products will not be used in decision making because better information is available.
33
EMODnet Annual Report 2 – Lot 2 Geology
Figure 12. The EUROSION data of coastal migration updated with information from the EMODnet-Geology Project partners.
Workpackage 6. Geological events and probabilities. (led by the Istitituto Superiore per la Protezione e la Ricerca Ambientale. Servizio Geologico d'Italia (ISPRA)).
Objective: To identify and map the locations of all significant geological events and provide information on probabilities of occurrence where available. These will include sites of submarine landslides, earthquake epicentres and volcanic centres located in the offshore study areas as specified in Section 2.4 of the tender documents.
At the end of 2014 (December) the workpackage leader distributed the Guidelines for the compilation of all information on geological events and probabilities. The features to be 34
EMODnet Annual Report 2 – Lot 2 Geology
included in Workpackage 6 are submarine landslides (Figure 15), earthquakes (through the European-Mediterranean Seismological Centre – EMSC, Figure 13 and see below), volcanoes (Figure 16), tectonics (Figure 17) and tsunamis (Figure 18). Each feature is identified by characteristics that are detailed in the attribute tables of the GIS layers. At the joint Workpackage 4/6 Geomorphology and Landforms Working Group meeting held in Paris in February 2015 (see above), it was agreed that only active faults will be represented within WP6. In the final versions to be delivered at the end of the project, each map presented in Figures 15-18 will be accompanied by a confidence map delineating information gaps (similar to the index maps shown in Workpackage 3).
The first Workpackage 6 shapefiles were received by the WP leader at the end of April 2015, when the process of harmonization was then able to commence; additional terms have subsequently been proposed to describe landslides, which were approved after verifying their compliance with INSPIRE.
In summary, the method used is that each partner identifies the location and type of geological event in their national waters. The characteristics of the geological events include:
• Feature: name of the item; • Status: sets if the item is mandatory or not; • Format: identifies the format (numeric, text, etc.); • Definition: defines a data domain, a value range, a measure unit or a single code; • Description: provides tips and information to fill in the item; • Reference: references used for the definition; • Remarks: details, comments and any explanation.
Regarding landslides and volcanic centres, a shape file for each geometric feature (polygons, lines, points) are expected in order to represent any occurrence of geological events. The use of different geometric features is related to the peculiar characteristics of each occurrence as well as to the scale of representation. For example, at 1:250,000 scale, a landslide area larger than 60 hectares (about 600 km2) can be represented by a polygon; a landslide area smaller than 60 hectares can be represented by a point.
As agreed at the EMODnet-Geology meeting in Malta (October 2014) we will rely on the catalogue of the European-Mediterranean Seismological Centre (EMSC) to provide up-to- date earthquake information to the EMODnet-Geology Project . However not all events will be identified (e.g. submarine landslides) depending on the available information (see Figure 14). It is also important to distinguish areas of no occurrences from areas of no data. This information will be provided by means of confidence layers in which each partner will be asked to outline areas where there are no data in the final version of the maps. Since geological events are scattered, in areas where information is available, blank spaces will indicate no occurrences.
35
EMODnet Annual Report 2 – Lot 2 Geology
Figure 13. The European-Mediterranean Seismological Centre’s web portal containing information about earthquake activity in Europe.
Figure 14. Comparison of earthquake information received from EMODnet-Geology partners (left) with data held in the EMSC database (right).
36
EMODnet Annual Report 2 – Lot 2 Geology
Figure 15. Submarine landslides in the European seas.
Figure 16. Underwater volcanic centres.
37
EMODnet Annual Report 2 – Lot 2 Geology
Figure 17. Tectonic map showing active faults (defined as any fault that has documented activity during the last 2.58 million years or a ‘tectonic structure that is expected to move within a future time span of concern to society’).
Figure 18. Tsunamis. The information shows the source of tsunamis that originate beneath the sea and the coastal areas affected by the tsunamis.
38
EMODnet Annual Report 2 – Lot 2 Geology
Suggestions for future work in EMODnet
While collecting and harmonising data for Work Package 6 the project partners have found that a few features can be evidenced in different ways. For instance, it is sometimes easier to identify an area of continuous mass transport by drawing the boundaries of a canyon where such process occurs. However, the canyon itself is a geomorphological feature rather than an event. It would therefore be useful to represent separately processes (as events, e.g. occurrence of a landslide or a volcanic eruption), deposits (as seabed substrate, e.g. mixed sediments derived from a landslide or volcanic deposit derived by an eruption) and landforms (as geomorphology, e.g. mound constituted by the accumulation of material transported by a landslide or a seamount of volcanic origin).
Collecting information regarding tectonic elements was necessary during the compilation of Work Package 6 information since many events (earthquakes, landslides and volcanic eruptions) are caused by or connected to fault activity. Faults were also included in WP4 in the tender and have also been considered as potential additional information within WP3. However, in each case the tectonic elements to be included would be chosen according to different characteristics (age, activity, etc.). Furthermore, faults might have been active for very long time intervals spanning across deposits of different age or might have been re- activated in recent times, affecting Quaternary deposits and coastal areas, despite their original range of activity being much older.
In order to complement comprehensive information, it would be useful to consider tectonic elements as a separate layer, collecting all types of elements, which can then be selected by attributes according to each specific need. Features to be considered might include faults, folds, antiforms, synforms, morphological and volcanic lineaments under the definition "Structural geology". Knowledge provided by structural geology would be of value for planners and managers for the protection of coastal areas, considering the connection between tectonics and volcanic activity and earthquakes as well as tsunami hazards. Indeed, this layer would complement information concerning probabilities of events, which is otherwise difficult to assess. These aspects are particularly relevant in tectonically active areas characterized by volcanic activity and frequent earthquakes. Detailed data might not be available everywhere; however, information could be also derived from high-resolution morpho-bathymetries and seismic data.
Workpackage 7. Minerals (led by the Geological Survey of Ireland (GSI)).
Objective: To identify and map areas of minerals (including aggregates, oil and gas and metalliferous minerals) in each of the participating countries based on information available to the project partners, including publicly-available information (published scientific papers etc.) To arrange web-mapping services with holders of data on mineral resources.
Following the EMODnet-Geology meeting in Malta (October 2014), an updated Task Guide for Workpackage 7 was issued by the WP leader on the 3rd November 2014. The WP7 metadata schema is based on INSPIRE guidelines and terrestrial EU minerals projects, such 39
EMODnet Annual Report 2 – Lot 2 Geology as Promine (http://www.promine.com/; see Figure 19), and FP7 minerals programmes, to ensure that the offshore minerals data are comparable.
During the year, the workpackage leader continued to receive minerals information that were standardised, merged and modelled to produce a series of maps showing different mineral types and a compiled version of all mineral occurrences. Following the process of merging data received from project partners, and as a result of feedback up to March 2015, it was decided to reformat the deposit type schemas. A number of adjustments were made to the original Workpackage 7 data specifications including:
• Data schemes were altered, to reflect common methods of reporting metadata (see below and Table 5);
• Reference fields were added to all nine data schemes;
• Style files were created, to synchronise with INSPIRE guidelines and projects that detail onshore metal occurrences (e.g. the ProMine project and portal);
• WP7 Task Guide was reproduced to reflect all changes to the data schemes, additions of style files and added to the File Sharing System (FSS) via the EMODnet Central Portal;
• Empty shapefiles, with all attribute fields set up as directed in the Task Guide, were created and made available to all partners via the FSS for any further submission or data updates;
• An email update was sent to all partners on May 18th 2015 detailing all changes with: o an iteration of the Workpackage 7 Task Guide;
o link to FSS with login details so that partners could access empty shapefiles;
o notification that the merged data have been sent as WFS to BGS for publication on the EMODnet-Geology portal;
o a request that all partners QC their data on the portal and let the WP leader know if they have any problems or issues.
Data have now been submitted for all 9 mineral deposit types. Examples of mineral deposit layers within the WMS service are shown in Figures 20 to 24. The latest assessment of mineral occurrences by country are shown in Table 4.
40
EMODnet Annual Report 2 – Lot 2 Geology
Aggregates Hydrocarbons Gas Marine Phosphorite Evaporite Polymetallic Polymetallic Cobalt Rich Hydrates Placer Sulphides Nodules Fe-Mn Crust
Albania Denmark Portugal Latvia Portugal Spain Iceland Estonia Portugal
Estonia France Russia Poland Spain Portugal Latvia Spain
Finland Ireland Ukraine Spain Spain Russia
France Lithuania Spain
Germany Netherlands
Ireland Norway
Latvia Poland
Netherlan Romania ds
Norway Russia
Poland Spain
Portugal Sweden
Russia Ukraine
Sweden
Ukraine
Table 4. A summary of deposit types submitted by country (September 2015).
41
EMODnet Annual Report 2 – Lot 2 Geology
A complete up-to-date Workpackage 7 Marine Minerals dataset is presented as the WP 7 WMS, available via the EMODnet-Geology portal at http://www.emodnet- geology.eu/emodnet/srv/eng/search#fast=index&from=1&to=50&orgName=Geological%20 Survey%20of%20Ireland
In order to ensure the project would capture key metadata associated with each deposit type; unique data schemes have been devised, that detail key components of each deposit type as well as metadata relating to the data source and data contributing agencies. These data schema include features such as deposit type; units of measurement; status (e.g. licence, exploration, exploitation etc); operator; data provider etc. An example of the metadata schema for marine aggregates is shown in Table 5. Similar metadata tables are available for all 9 mineral types.
Field alias FIELDNAME Format Information
OBJECTID FID Number Feature ID. An internally generated identification number for each feature. Automatically generated within shape file
Shape SHAPE Geometry Polygon
Code CODE Text (2) Two letter county code, which corresponds to ISO3166-code. E.g FR, IE
Deposit Type DEPOSIT_TY Text (30) Aggregates – this exact wording must be entered in bold type.
Deposit Sub Type DEP_SUB_TY Text (100) e.g. sand, gravel
Units UNITS Text (50) Quantitaive measurement of deposit if available
Status STATUS Text (250) e.g. Licence, exploration,
42
EMODnet Annual Report 2 – Lot 2 Geology
exploitation etc.
Operator OPERATOR Text (250) Operating company
Exploitation Type EXPLOIT_TY Text (250) Mining method
Block Name BLOCK_NAME Text (250) Name of block, concession or deposit area
Block Number BLOCK_NO Text (250) Number of block, concession or deposit area
Data Provider DATA_PROVI Text (150) Name of organisation providing data
Data Provider DATA_CONT Text (150) The data providing Contact organisation contact details – email is required
Area extend (sq. km) AREA_KM2 Number (Double) Area of deposit in (11,4) kilometres squared
Depth to seabed (m) DEPTH_TO_S Text (50) Depth to deposit from sea surface
References REFERENCES Text (250) DOI &/or author, year and title. Or link to PDF etc if available.
Table 5. Metadata schema for marine aggregates. Similare metadata are available for all mineral types compiled in Workpackage 7.
Co-operation with EMODnet Human Activities Lot
Due to subject matter crossover regarding the mineral and hydrocarbon theme; the WP7 leader has maintained contact with the EMODnet Human Activities Lot in order to clarify data required by both projects and to inform our partners. So far we have assisted the Human Activities Lot in assessing their need for aggregate and hydrocarbon extraction data and hydrocarbon licencing information. We have contacted our partners in order to explain the EMODnet-Geology Project role in mapping the geology i.e. marine mineral deposits, and
43
EMODnet Annual Report 2 – Lot 2 Geology the role of the Human Activities Lot in collating data on activities relating to the exploration and extracting of economically valuable deposits. We have provided our partners with contact information for COGEA, the Humans Activities Lot Co-ordinator, and requested that they assist COGEA with their compilation where possible.
Figure 19. Minerals information from the European land area available through the ProMine Portal The EMODnet-Geology offshore minerals information has a similar metadata schema to allow comparison between land and marine minerals data.
44
EMODnet Annual Report 2 – Lot 2 Geology
Figure 20. The distribution of marine aggregates.
Figure 21. Hydrocarbons.
45
EMODnet Annual Report 2 – Lot 2 Geology
Figure 22. Gas hydrates.
Figure 23. Marine placer deposits.
46
EMODnet Annual Report 2 – Lot 2 Geology
Figure 24. Composite map of all occurrences of minerals compiled by WP7.
Workpackage 8. Web services and technology (led by the British Geological Survey).
Objective: To further develop the web portal providing access to the existing maps available in the OneGeology-Europe portal, as developed during the ur-EMODnet-Geology Project. The Geo-Seas data discovery and access system will also be integrated into the OneGeology- Europe portal to provide access to geological and geophysical data and metadata. The EMODnet-Geology portal will therefore provide access to data, metadata and data products that include both marine and terrestrial geological information for the whole of Europe.
As reported at the end of the first year of the project, it was decided that the EMODnet- Geology portal would be built using ‘GeoNetwork’. http://geonetwork-opensource.org/, an open-source catalog application to manage spatially referenced resources. The revised portal was re-launched on December 8th 2014 and partners were invited to provide feedback to the WP leader throughout the year to make improvements to the functionality. Recommendations for changes to standardise the appearance of the website with the other EMODnet web pages were also received from the EMODnet Project Secretariat. We are aware that the system does not yet provide the functionality that is required to deliver the EMODnet-Geology Project outputs, but a new version of the GeoNetwork software has just been released and will be assessed by the workpackage leader to ensure that we can respond to the feedback received to date. Figure 25 a-d provide some examples of map information that can be viewed on the EMODnet-Geology portal. The new software is more 47
EMODnet Annual Report 2 – Lot 2 Geology
intuitive and will provide a traditional map viewer aliong the lines of similar online map systems to EMODnet, making selection and display of the map layers easier for the user. Downloads will be feasible and a test on Workpackage 3 layers will take place in early 2016. Google Analytics are running on the current system, but because the system is based on remote access we are not able to create statistics at the moment; however the new system will allow the normal range of statistics to be derived during the final year.
The EMODnet-Geology portal
Users of the website can access online data using a Catalog to select the information of interest and then display on a map interface. As the map layers are updated these will be accessed from each of the WP leaders servers using Web Map Services (WMS). The portal allows the user to:
• search for WMS services of interest; • view information on the service; • view and interact with the service via a map interface; • obtain a link to the service to incorporate into a desktop GIS or other mapping application.
Links to the WMS, which can be used to incorporate into a desktop GIS or other mapping application, can be found on the Metadata information page and the Map page.
Frequently Asked Questions (FAQs), Tutorials, Reference Manuals and other information on the use of GeoNetwork Opensource can be found on the Documentation Center on The GeoNetwork Opensource Community website.
There are many different ways to search the catalog for maps and other geographic data. The most common way is via the Catalog or the search bar. The search bar allows you to enter a free text term and also offers a list of options depending on the text entered. Clicking on the search button (the icon of a magnifying glass) takes you to the Catalog page showing the results of your search. Clicking on the 'show advanced search' button opens additional search options. This allows for detailed searches based on:
• What? - searches are based on data content: o Keyword - select from a list of keywords. More than one keyword can be selected; o Organisation - select from a list of organisations who have data available. More than one keyword can be selected; o Kind - select the type of information to be searched for. More than one keyword can be selected;
• INSPIRE - searches are based on INSPIRE metadata within the catalog:
48
EMODnet Annual Report 2 – Lot 2 Geology
o INSPIRE annex - select a specific INSPIRE annex. The INSPIRE annexes for a metadata are based on the INSPIRE theme keywords assigned to it; o INSPIRE theme - select from a list of keywords from the INSPIRE themes thesaurus; o Service type - select from a list of INSPIRE service types.
• Where? - searches based on spatial extent: o Select an area of interest on the map by first selecting the extent tool.
• When? - searches are based on temporal extent: o Metadata change date - select a date or range where the metadata record has been updated; o Temporal extent - select the time period covered by the content of the dataset; o Date type - select from the creation, modification or publication date of the dataset.
Note: The term data refers to datasets, maps, tables, documents, etc, in fact anything that can be linked to the metadata record that describes it. The results of a search are dependent on the quality of the metadata associated with the resource.
The GeoNetwork search engine allows you to search on the following fields: • Title; • Abstract; • Keywords; • Location; • Free text (any metadata field).
Catalog
The simplest way to find data of interest is to go to the Catalog and use the filter option. The Catalog provides access to EMODnet-Geology Project metadata and consists of three main panels: • Filter panel; • Results panel; • Map preview.
From here is it possible to filter the metadata Catalog and view data layers on the map.
Filter panel
The filter panel provides a list of options to help find data of interest. The filter options are derived from the current metadata list displayed in the results panel. To filter simply click on the option required. 49
EMODnet Annual Report 2 – Lot 2 Geology
Results panel
The results panel displays a list of records based on the search and/or filter criteria. Clicking on the result title opens the full metadata record. Included are a number of options depending on the type of record: • View in Google Earth - downloads the metadata record as a kml. file; • Add to Map - displays the Web Map Service in the map window; • Image Link - displays any images associated with the record, usually the legend for the WMS service.
Map preview
The map preview displays a small overview map, which includes: • Boundary outlines of all records in the current results list; • Any WMS layers currently added to the map.
The icon in the top right of the preview map toggles between the full sized map.
Map
The map window displays any WMS services that have been added to the map. While it is possible to add WMS services from within the map it is easier to do so via the Catalog.
Metadata
The metadata screen displays a full metadata record for the selected result. The contents of this screen are dependent on the level of metadata associated with the selected result and can vary from record to record. Commonly included are:
• A description of the record; • Links to associated resources; • Information on the dataset; • Contact information.
50
EMODnet Annual Report 2 – Lot 2 Geology
a. EMODnet-Geology portal home page. b. Seabed substrate information for the northern European seas.
c. Seabed geology information. d. Minerals information (hydrocarbons). Figure 25. Screenshots of information layers taken from the EMODnet-Geology website.
Workpackage 9. Dissemination (led by the British Geological Survey).
Objective: To ensure widespread dissemination of the EMODnet-Geology products, through targeted contact with stakeholders and other users of marine geological information.
The dissemination activities of the EMODnet-Geology Project are ongoing throughout the period from 2013-2016. Activities during the reporting period are listed in Section 9 of this report.
51
EMODnet Annual Report 2 – Lot 2 Geology
Workpackage 10. Liaison with EMODnet lots (led by the British Geological Survey).
Objective: To ensure that the EMODnet-Geology project is fully aware and complementary to the objectives of other marine science initiatives within European waters. To prepare for better and linked marine data that will have an immediate impact on the planning of environmental policy and mitigation measures within the European Union and to facilitate impact assessments and scientific work.
The EMODnet-Geology Project Co-ordinator has attended all meetings of the EMODnet Steering Committee at which the co-ordinators discuss joint activities. The Workpackage 3 leader has delivered seabed substrate information according to achedule provide by the Seabed Habitats Lot and will deliver further updates in the final year of the project. As described in the Workpackage 7 progress report, close collaboration with the Human Activities Lot has been necessary to avoid duplication of effort when gathering information about minerals such as hydrocarbons.
The Project Co-ordinator and other members of the team have been active in establishing links with a number of complementary programmes as described in sections 7 (see Project Management report), 9 and 10.
Workpackage 11. Project analysis and sustainability (led by the British Geological Survey).
Objective: To analyse each phase of the project and to provide a report on the lessons learned.
An analysis and recommendations for sustainability will be provided at the end of the project. 8. User Feedback
Provide a complete record of feedback received from user (formal and informal) on your portal, your activities or those of other EMODnet projects/activities. Also provide any suggestions you have received for EMODnet case studies and/or future products/activities/events.
A wide range of feedback has been received both formally and informally during the last year. It has proven difficult to capture all of the feedback, which will be addressed in the final year of the project by identifying ways in which the views of the user community can be more easily recorded. Section 10 (indicator 7) demonstrates the varied use and impact that the EMODnet-Geology project and the EMODnet Programme in general has achieved, which culminated in the interest shown at the EMODnet Jamboree that took place just after the current reporting period.
52
EMODnet Annual Report 2 – Lot 2 Geology
The only formal feedback reported below is that from the North Sea Checkpoint. In general, the EMODnet-Geology partners are reporting strong interest in the project with various government, commercial and research organisations expressing their interest in the use of the EMODnet-Geology outputs. Based on this feedback, the project partners have identified new types of geological information that could be included in the next phase of EMODnet, or information that could be delivered in more detail. These generally fall under the headings of new information on sediment dynamics, geomorphology, structural geology, geochemistry and improved capture of a wider range of information in the coastal zone. Recommendations under each of these topics is included in Section 7 of the report under the progress reports for workpackages, 3,5 and 6.
In discussions with the European paleolandscapes research community (specifically the group responsible for preparing the European Marine Board’s Position Paper on ‘Land Beneath the Waves; submerged landscapes and sea level change – a joint geoscience humanities strategy for European Continental Shelf Prehistoric Research’ Position Paper 21 (see http://www.marineboard.eu/publications-full-list), it was recognised that there is a need for improved information as described below, which fits well with the development of the geomorphological information that has progressed under Workpackage 4.
Other specific examples of feedback come from industry groups, such as the European Marine Sand and Gravel Group (EMSAGG) who, following a presentation at their conference in Delft, the Netherlands in June 2015 by Maria Judge (Geological Survey of Ireland), expressed interest in accessing the minerals information held in the EMODnet-Geology system. Likewise universities (such as University College Dublin and University College Cork in Ireland) have asked for information that can inform their course work or to help in other European projects such as MINATURA 2020 (see http://minatura2020.eu/).
A link with EMODnet was part of a bid led by Heriot-Watt University (UK) named ATLAS. The ATLAS project has at the time of writing been invited to negotiate a contract with the EC and will include links to EMODnet through participation of the Project Co-ordinator and the EMODnet Secretariat. ATLAS will create a dynamic new partnership between multinational industries, SMEs, governments and academia to assess the Atlantic’s deep-sea ecosystems and Marine Genetic Resources and create the integrated and adaptive planning products needed for sustainable Blue Growth. ATLAS will gather and compile diverse information on sensitive Atlantic ecosystems (including vulnerable marine ecosystems (VMEs) and ecologically or biologically significant marine areas (EBSAs)) to produce a step-change in our understanding of their connectivity, functioning and responses to future changes in human use and ocean climate. This is possible because ATLAS takes innovative approaches to its work and interweaves its objectives by placing business, policy, and socioeconomic development at the forefront with science. ATLAS not only uses trans-Atlantic oceanographic arrays to understand and predict future change in living marine resources, but enhances their capacity with new sensors to make measurements directly relevant to ecosystem function.
53
EMODnet Annual Report 2 – Lot 2 Geology
Recommendation from the paleolandscape community for information in EMODnet- Geology
During the last one million years the land area of Europe was at times 40% larger than at present, and was usually 10-20% larger because of the global volumes of water locked up in ice-caps several kilometres thick on land and the resulted lowering of the global sea-level by up to 120-130 meters. The presently submerged, prehistoric terrestrial landscape of the European continental shelf is of great value because (i) it hosts evidence for the early migration and settlement of prehistoric humans, (ii) can provide significant knowledge on the past sea-level changes and (iii) bears geological deposits, e.g. marine aggregates and rear-earth deposits, of high economic interest.
The EMODNET-Geology Lot should seek to provide access to geological surveys with the aim to harmonize maps and provide reconstructions of the submerged landscapes of the European continental shelf at various time-frames (e.g. Last Glacial Maximum (LGM) and older low sea-level stages), with particular focus on:
1. shorelines and coastal environments and deposits (lagoons, dunes, estuaries etc., marine terraces, beachrocks); 2. valleys and riverbeds, terraces and associated deposits; 3. river-deltas and delta-clinoforms; 4. Submerged water points, e.g. Submarine Groundwater Discharges (=submerged springs), and freshwater lakes; 5. Thickness of Holocene deposits above LGM landscape; 6. Flora and fauna on the submerged landscapes.
North Sea Checkpoint Report
The North Sea Check Point project (Growth and Innovation in the Ocean Economy - Gaps and Priorities in sea basin and observation data MARE/2012/11: North Sea) has produced reports on findings to date on Data Adequacy for the ‘Wind farm challenge’ and the ‘Marine Protected Areas (МРА) challenge’ by identifying data needed to undertake the work, finding it, assessing it for its relevance, and making an initial assessment of its usefulness and gaining access to the data.
The Wind Farm Challenge report used seafloor geology as one of the environmental parameters to assess the suitability of an area of seabed for the construction and running of a wind farm and its likely energy, noting that such data would only be expected to provide an initial impression of a site’s suitability and that more accurate data would be gathered from potential sites following the initial site selection process. The report states that ‘the geology data available from the EMODnet OneGeology portal provided only a rough indication of the character of superficial sediments in the area (North Sea). Although a higher resolution of data would have been available from the British Geological Survey (BGS) for UK waters, this would have provided partial coverage of the study area. Focused geotechnical sampling would be
54
EMODnet Annual Report 2 – Lot 2 Geology expected to be undertaken in advance of any licence application being pursued. The data shows that the majority of the North Sea is characterized by fine grained sediments with banks of coarser gravel beds running out towards the Dogger Bank.’ The report also states that only a rough indication of superficial sediments is present in the area with no indication of sediment depth. Under the question of ‘Are the Commercial Terms Acceptable’ the report said that ‘commercial seabed geology products produced by the BGS for the North Sea………would be necessary and expected in the context of a real marine licence application.’ In the discussion and conclusions, the report noted that the EMODnet-Geology data was not provided in a format which allowed it to be integrated into GIS for analysis alongside other data, but was available as a kml file, which had to be viewed alongside the outputs of the Wind Farm challenge after the analysis of core data.
The North Sea Checkpoint Synopsis Report commented that it was not clear how the full dataset of sediment data could be access via the data holders (the British Geological Survey). The Checkpoint reports are available at: https://webgate.ec.europa.eu/maritimeforum/en/node/3723
Comment from EMODnet-Geology Project The scale of data compilation in the EMODnet-Geology Project is 1:250,000 wherever possible and 1:1 million in other areas (see WP3 report). The question raised about access to higher resolution information at organization level (in this case British Geological Survey data) is one that the partners are aware of and propose is addressed in a future phase of EMODnet by improved links to sources of information or the use of WMS. The closer linkages between information at national level and the harmonised information provided by EMODnet are key to delivering the best available geological information available through the network of national geological survey organisations and their collaborators. In the meantime, the project will investigate ways to help users of the EMODnet-Geology portal to navigate to information currently held only on individual organisations’ websites and the conditions that apply to access.
9. Outreach and communication activities
Date Media Title Short description and/or link to the activity October 2014 Science event Exploring our marine geological resources in EMODnet-Geology referenced in the fifth dimension: About 3D voxels, 4D poster and abstract by Van impact models and uncertainty. Lancker, V.; De Mol, L.; De Tré, G.; Maljers, D.; Stafleu, J.; Van den Eynde, D.; Van Heteren, S. (2014). In: Mees, J. et al. (Ed.) (2014). Book of abstracts – VLIZ Young Scientists’ Day. Brugge, Belgium, 7 March 2014. VLIZ Special Publication, 67: pp. 111. (poster) 55
EMODnet Annual Report 2 – Lot 2 Geology
October 2014 Study day Interactive management of marine resources EMODnet-Geology referenced in in the southern North Sea, a long-term presentation by Van Lancker, V.; perspective. Francken, F.; Terseleer, N.; Van den Eynde, D.; De Mol, L.; De Tré, G.; De Mol, R.; Missiaen, T.; Hademenos, V.; Maljers, D.; Stafleu, J.; Van Heteren, S. (2014). in: De Mol, L. et al. (Ed.) (2014). 'Which future for the sand extraction in the Belgian part of the North Sea?'. Study day, 20 October 2014, Belgium Pier - Blankenberge. pp. 89-94. December 2014 Conference Presentation including EMODnet-Geology in Workshop to develop the seabed session on ‘Progress on Large Scale Seabed mapping aspects of the Galway Mapping Data Integration & Exchange Statement on Atlantic Research Initiatives Overcoming data integration Co-operation. challenges, making data relevant, and accessible’ January 2015 Conference EMODnet-Geology was referenced at the European Innovation Partnership on Raw Materials Annual Conference and Horizon 2020 Brokerage Event on Raw Materials, Brussels, Belgium. January 2015 Conference Using sediment data to create series of Poster presentation at the (poster) surrogate-based habitat suitability maps for Quaternary Research Association the southernmost North Sea meeting in Edinburgh, UK. Vera Van Lancker and Sytze van Heteren. January 2015 Conference Using sediment data to create series of EMODnet-Geology referenced in (poster) surrogate-based habitat suitability maps for poster by Van Lancker, V. & van the southernmost North Sea. Heteren, S. at QRA 2015 Annual Discussion Meeting “The Quaternary Geology of the North Sea basin and adjacent areas”. Edinburgh (UK). February 2015 Meeting EMODnet-Geology referenced at the Atlantic Research Alliance stakeholders meeting in Brussels on 23-24 February. February 2015 Conference EMODnet-Geology referenced at the INDABA Mining Conference, Cape Town, South Africa. INFOMAR and Ireland’s contribution to Maria Judge, Geological Survey of February 2015 Meeting European collaborative marine geological Ireland (GSI) presented the mapping projects. EMODnet-Geology project at the
Irish Geological Research Meeting (IGRM), Cork, Ireland (Ireland’s annual geological research meeting). February 2015 Training event Geomorfologie van Kust en Zee. Zeebad voor EMODnet-Geology referenced in 56
EMODnet Annual Report 2 – Lot 2 Geology
Leerkrachten 2015. oral presentation by Van Lancker, V. in Gent (BE). Initiatief VLIZ, in samenwerking met de Vereniging Leraars Aardrijkskunde VLA - werkgroep Oost-Vlaanderen. (http://www.vliz.be/nl/Zeebad- voor-leerkrachten-2015) April 2015 Conference Seabed substrates and sedimentation rates of Kaskela, Kotilainen, Alanen, Poster the European Seas - EMODnet-Geology. Stevenson and EMODnet- Geology 2 Partners. Geophysical Research Abstracts, Vol. 17, EGU General Assembly 2015, Vienna, Austria. April 2015 Meeting EMODnet referenced at ‘The Atlantic – our Shared Resource: making the vision reality’ meeting held in Brussels on 16-17 April. April 2014 Meeting Presentation of EMODnet Project at the Marine Geology and Georesources Unit-IPMA, Portugal and distribution of EMODnet leaflet May 2015 Conference Assembling data on seabed substrates of the Kaskela, Kotilainen, Alanen, European Seas. Stevenson and EMODnet- Geology 2 Partners. GeoHab 2015, Salvador, Brazil. May 2015 Conference Habitat change assessment in dynamic EMODnet-Geology referenced in shallow water systems – presentation at GeoHab Lessons learned and way forward. Conference (Marine Geological and Biological Habitat Mapping). Salvador de Bahia (Brazil), 3- 8/5/2015. (oral) May 2015 Workshop EMODnet referenced at the European Ocean Observing System meeting organised by the European Marine Board and EuroGOOS in Brussels on 12-13 May 2015. May 2015 Conference EMODnet-Geology referenced at the Irish Association of Economic Geology Conference, 16-17th May, Limerick, Ireland. June 2015 Science event Geological resource management of the EMODnet-Geology referenced in future: Drilling down the possibilities. poster by Van Lancker, V., Van den Eynde, D., De Mol, L., De Tré, G., Van Britsom, D., De Mol, R., Missiaen, T., Hademenos, V., Maljers, D., Stafleu, J. and van Heteren, S. World Ocean Day. Brussels (BE). June 2015 Letter sent to Portuguese Institutions and colleagues 57
EMODnet Annual Report 2 – Lot 2 Geology
introducing the EMODnet Project and requesting (third-party) data for WP-7 Minerals. Also sent as an example, attached to the letter, figures of visualization form of WP7 (map and attribute table) as well as a table form for data. June 2015 Conference EMODnet –Geology 2 – Seabed substrate and Kaskela, Kotilainen, Alanen, sedimentation rates of the European Seas. Stevenson and EMODnet- Geology 2 Partners. Baltic Sea Science Congress 2015, Riga, Latvia. June 2015 ICES Meeting TILES project on Transnational and Integrated EMODnet-Geology referenced in Long-term Marine Exploitation Strategies) presentation by Van Lancker, V. and the estimation of far field effects of (2015). pp. 75-76. In: ICES (2015). marine aggregate extraction in an offshore Interim Report of the Working sandbank environment. Group on the Effects of Extraction of Marine Sediments on the Marine Ecosystem (WGEXT), 20– 23 April 2015. Ostend, Belgium. ICES CM 2015/SSGEPI:06, 102 pp. June 2015 Conference EMODnet-Geology and specifically WP7 Marine minerals (with a focus on aggregates) was presented by poster and 5 min presentation at the European Marine Sand and Gravel Group (EMSAGG) Conference in Delft, the Netherlands on 4th June 2015 June 2015 Conference Geological Knowledge Base, a digital platform EMODnet-Geology referenced in for sharing and quantifying resource presentation by Van Lancker, V, information. Francken, F., Terseleer, N. , Van den Eynde, D., De Mol, L., De Tré, G., De Mol, R., Missiaen, T., Hademenos, V., Bakker, M., Maljers, D., Stafleu, J. & van Heteren, S. pp. 1-2. 5th EMSAGG conference: Marine sand & gravel - Finding common ground. Delft, the Netherlands . June 2015 Conference EMODnet-Geology referenced at the 8th EuroGeo Conference, Barcelona, Spain. June 2015 Science event Geological resource management of the EMODnet-Geology referenced in future: Drilling down the possibilities. poster by Van Lancker, V., Van den Eynde, D., De Mol, L., De Tré, G., Van Britsom, D., De Mol, R., Missiaen, T., Hademenos, V., Maljers, D., Stafleu, J. and van Heteren, S. (2014). NATO Science Day. Oostende, Belgium. 58
EMODnet Annual Report 2 – Lot 2 Geology
June/July 2015 Meeting 3-4D Geological & Sediment Mapping. Data, EMODnet-Geology referenced in concepts and first results. presentation by Van Lancker, V, Francken, F., Terseleer, N. , Van den Eynde, D., Hademenos, V., Missiaen, T., De Mol, R. De Tré, G., van Heteren, S., Bakker, M., Maljers, D., Stafleu, J., Busschers, F. & De Mol, L. ANR SunRISE Network meeting. Barcelona (ESP). July 2015 Colloquium EMODnet 2 - Geologie des Meeresbodens Presentation at the BGR internal europaweit: BGR-Beteiligung und erste colloquium, Hannover, Germany. Ergebnisse Asch, K. and Müller, A.M July 2015 Meeting European Marine Observation and Data Presentation at International Network Phase 2. Workpackage 4 - Sea-floor Geoscience Information geology: Compilation and harmonization. consortium meeting in Hannover, Germany. Asch, K., Gdaniec, P. and Müller, A.M. August 2015 EMODnet-Geology kokoaa tiedon Euroopan Kotilainen, A.T., Kaskela, A.M., merialueiden geologiasta (in Finnish). Alanen, U., Stevenson, A., 2015. Summary in English. Geologi, 3, 80-87. Article in Geologi magazine. http://www.geologinenseura.fi/g eologi-lehti/3- 2015/Geologi_3_2015_04emodn et.pdf September Meeting Activities of the ECODAM group (ECOsystem EMODnet-Geology referenced in 2015 Data Analysis and Modelling). poster by Baetens, K., Baeye, M., Barbut, L., Desmit, X., Dulière, V., Fettweis, M., Francken, F., Kint, L., Lacroix, G., Legrand, S., Luyten, P., Nechad, B., Ozer, J., Ponsar, S., Pringle, N., Ruddick, K., Terseleer, N., Van den Eynde, D., Van der Zande, D., Vanhellemont, Q. & Van Lancker, V. Science Day RBINS OD Nature. Temse, Belgium. September Workshop TILES Workshop 2: Discussie resultaten EMODnet-Geology referenced in 2015 grondstofmodellering met eindgebruikers. Van Lancker, V., Kint, L, Montereale-Gavazzi, G., Terseleer, Nathan, Van den Eynde, D., Missiaen, T., Hademenos, V., De Tré, G., De Mol, R., van Heteren, S., Cox, D., De Mol, L., Roche, M., Van Gaever, S., Van Quickelborne, E., De Nil, K., Matton, C., Stolk, A. & VandenEede, S. Belspo Brain-be project TILES (BR/121/A2/TILES). Gent, Belgium. September Meeting The Workpackage 5 leader 59
EMODnet Annual Report 2 – Lot 2 Geology
2015 attended the Darmstadt Copernicus Users Information Day on Sep 11 to see what data will become available for the monitoring of coastal change. September Seminar Geologia ei tunne rajoja - merigeologista EMODnet-Geology dissemination 2015 yhteistyötä jo yli 50 vuotta during the national seminar at Helsinki University (in Finnish); Aarno Kotilainen. September Geological International Geological Forum in 2015 Forum Odessa called "Problems and prospects of development of Geology: science and production". PrichornomorSRGE provided to the forum an informational report on their participation in the EMODnet- Geology Project, containing the project tasks and a link to the EMODnet Central Portal. September Press release GTK press release about 2015 releasing/publishing EMODnet- Geology National seabed substrate data (in Finnish), 28.9.2015. GTK toi tarkat merenpohjan karttatiedot kaikkien saataville Link to press release: http://www.gtk.fi/ajankohtaista/ media/uutisarkisto/index.html?y ear=2015&number=642&newsTy pe=PressReleases September Paper The EMODnet project: a European initiative Presentation by T. Medialdea, J. 2015 for the development of the Giménez-Moreno, L. Somoza, R. geological knowledge in the European marine León & F.J. González of the environment. In Spanish (abstract in English). Instituto Geológico y Minero de España at the Resúmenes sobre el VIII Simposio MIA15, Málaga. EMODnet was also referenced at the same meeting in papers on ‘Potential and difficulties in the susceptibility assessment of submarine landslides in the MIA from an event catalogue’ by R. León, T. Medialdea, L. Somoza, J. Giménez-Moreno & F.J. González and ‘First catalogue of submarine mineral deposits from Spain: EMODnet-Geology Project’ by F.J. González, T. Medialdea, G. Gómez-Ramos, L. Somoza, E. Marino & R. León October 2015 Paper Offshore faulting in the Aegean Sea; a Paper by Sakellariou D and Synthesis based on bathymetric and seismic Tsampouraki-Kraounaki K. profiling data. EMODnet bathymetry and 60
EMODnet Annual Report 2 – Lot 2 Geology
geology referenced in the paper from the Proceedings of the 14th International Conference, Thessaloniki, May 2015. Bulletin of the Geological Society of Greece. Volume XLVIII. October 2015 Conference EMODnet-Geology referenced in a talk by Alan Stevenson on ‘European Initiatives and Potential Links with ECORD’ at the 2nd Coastal and Marine Geology Symposium’ in Istanbul, Turkey. 15th October. October 2015 Book launch EMODnet-Geology referenced at the launch of Minerals Yearbook, at “Raw Materials Diplomacy Dialogue between the EU and the advanced mining countries” event. Brussels, Belgium. October 2015 Webzine article European Marine Observation and Data An article on EMODNET, Network (EMODnet): Making Fragmented EMODnet-Geology and the GSI Marine Data Relevant and Accessible. work as WP7 Marine Minerals leader has been generated and will be published on the webzine Earthzine: http://earthzine.org/ 2015 Book Van Heteren, S & Van Lancker, V (2015). EMODnet-Geology referenced in Chapter 8. Collaborative seabed-habitat the book chapter. mapping: uncertainty in sediment data as an obstacle in harmonization, pp. 154-176. In: Diviacco, P., Fox, P., Pshenichny, C. & Leadbetter, A. (eds.). Collaborative Knowledge in Scientific Research Networks. IGI Global. doi:10.4018/978-1-4666-6567-5 2015 Book Van Lancker, V., Francken, F., Kint, L., EMODnet-Geology referenced in Terseleer, N., Van den Eynde, D., De Mol, L., the book chapter. De Tré, G., De Mol, R., Missiaen, T., Chademenos, V., Bakker, M., Maljers, D., Stafleu, J., van Heteren, S. (subm.). Building a 4D Voxel-Based Decision Support System for a Sustainable Management of Marine Geological Resources. In: Diviacco, P., Leadbetter, A. & Glaves, H. (eds.). Oceanographic and Marine Cross-Domain Data Management for Sustainable Development. IGI Global. 2015 Lectures Raising awareness of EMODnet- Geology products to students during courses in the framework of the Master programme Oceans and Lakes (Free University Brussels, Ghent University and Antwerp University), as well as 61
EMODnet Annual Report 2 – Lot 2 Geology
Master in Geology, Ghent University.
EMODnet-Geology on digital media
Irish INFOMAR Programme news feed: http://www.infomar.ie/news/
INFOMAR facebook page: https://www.facebook.com/INFOMAR-595185173894900/
INFOMAR twitter: https://twitter.com/followtheboats/
INFOMAR Newsletter: http://www.infomar.ie/publications/Newsletters/Summer%202015%20Newsletter%20Issue %202.pdf
A link to the EMODnet Central portal has been placed on a number of partner’s websites. For example:
The participation of the Continental Shelf Department (Malta) in EMODnet has been published on the web-page of the Department (https://mti.gov.mt/en/Pages/Continental%20Shelf/Other-Continental-Shelf- Activities.aspx). The web-page includes a link with the EMODnet-Geology portal. The website of PrichornomorSRGE (Ukraine) includes a link to the EMODnet-Geology project and information has also been targeted at the Ukrainian Institute of Geological Sciences, Geophysical Institute, the Ukrainian State Geological Research Institute and the Department of Marine Geology and Sedimentary Ore Formation.
The EMODnet seabed substrate data from the Finnish waters was published in the GTK's data portal HAKKU. Link to portal: http://hakku.gtk.fi/fi/ (September 2015). Dissemination of EMODnet Project at IPMA’s webpage (http://www.ipma.pt/pt/investigacao/geologia/projetos.detail.html?f=/pt/investigacao/geo logia/emodnet.html). 10. Updates on Progress Indicators
Indicator 1 - Volume of data made available through the portal As the information being made available through the EMODnet-Geology Project portal is primarily map information, the volume of data made available through the portla is not the most applicable measure for the outputs. Examples of the current map layers that are available are shown in the figures thoughout the report.
62
EMODnet Annual Report 2 – Lot 2 Geology
Indicator 2 - - Organisations supplying each type of data based on (formal) sharing agreements and broken down into country and organisation type (e.g. government, industry, science). The information supplied to the EMODnet-Geology Project are almost entirely held by the geological surveys of Europe (government), although interpreted map products can include information derived from broader sources such as industry and the university sector. The index map for Workpackage 3 for example shows that mapping information has been provided by more than 50 organisations (including the project partners). Other examples are provided throughout the report, for example in the collaborations described in Italy (see Section 7), Malta and Iceland (see Indicator 7 below).
Indicator 3 - Organisations that have been approached to supply data with no result, including type of data sought and reason why it has not been supplied. Those organisations who have been approached have willingly provided information to the project; however the data required to compiled the EMODnet-Geology outputs are compiled at national level then harmonised according to the Guidelines provided by the workpackage leaders therefore the organisations who contribute at a national level are not specifically identified through the EMODnet process.
Indicator 4 - Volume of each type of data and of each data product downloaded from the portal The EMODnet-Geology outputs are not yet available to be downloaded from the project portal, although they can be downloaded as Google Earth (kml.) files or added to maps using Web Map Services. Functionality is being investigated in the new version of software that is used on the website to display the map outputs.
Indicator 5 - Organisations that have downloaded each data type
See Indicator 7 for examples of organisations who are involved in the use of the EMODnet- Geology outputs.
Indicator 6 - Using user statistics to determine the main pages utilised and to identify preferred user navigations routes
Information not available (see Indicator 4).
Indicator 7 - List of what the downloaded data has been used for (divided into categories e.g. Government planning, pollution assessment and (commercial) environmental assessment, etc.)
Government planning
63
EMODnet Annual Report 2 – Lot 2 Geology
Information on the structural geology of the Aegean Sea that has been compiled using EMODnet-Geology funding has been updated and presented as part of a PhD thesis (Konstantina Tsampouraki-Kraounaki) and in a publication (see Section 9: Outreach and communication activities). The maps will be submitted for use to the Committee for Seismotectonics of the Earthquake Planning and Protection Organization of Greece.
While working on an indicative model for the archaeological potential of the Dutch part of the North Sea for the Dutch Cultural Heritage Agency and the Ministry of Infrastructure and the Environment, using Quaternary geology data products, EMODnet was flagged as a way to extend such work in Europe as part of the next project phase.
The Continental Shelf Department (CSD) of Malta has been collaborating with the Malta Environment and Planning Authority (MEPA) on some aspects of the implementation of the Marine Strategy Framework Directive (MSFD). The Department has informed MEPA of its involvement in EMODnet-Geology with a view to exchange information and data once the project is completed.
The UK Centre for Environment Fisheries and Aquaculture Science (Cefas) Chief Scientist Prof. Stuart Rogers briefed the United Kingdom’s House of Lords European Union Committee on Regional Marine Cooperation on the importance of international transboundary initiatives like EMODnet for the regional management of the North Sea. The Final report recommended (No. 48) that the EC worked closely with EMODnet to raise awareness of its’ capabilities, and that consideration be given to increased funding for EMODnet for data quality assurance.
In the Baltic Sea, at the Gulf of Finland, the Regional Council of Kymenlaakso (in Finland) has made a regional plan for the trade- and sea area of the region. The Ministry of the Environment has notarized that regional plan for the trade- and sea area of Kymenlaakso 26th November 2014. EMODnet-Geology seabed substrate data is one of the datasets that geologists provided for planners and has been used in the regional plan. Finnish marine geologists (and biologists) interacted with planners in order to evaluate, what kind of geological and biological data is of practical relevance for spatial planning, what is available and what is missing. Cooperation is essential to make scientific data on sea areas more easily available and understandable to spatial planners and improve the interplay between science and planning officials. http://www.kymenlaakso.fi/maakuntakaava/maakuntakaavat/kauppa-ja-merialue
One of the roles of the Continental Shelf Department (Malta) is the issuing of permits for marine scientific research offshore Malta. Permit applicants are made aware of the involvement of the Department in EMODnet if this would be beneficial towards the planning and execution of the research.
64
EMODnet Annual Report 2 – Lot 2 Geology
Regular updates on the project are provided by the CSD (Malta) to the Ministry of Transport and Infrastructure. The Ministry is fully supportive of the involvement of the Department in EMODnet and was actively involved in the second project partners’ meeting that was held in Malta in September/October 2014.
An informal regional network of North Sea partners from Denmark, Germany, the Netherlands and Belgium, grown out of EMODnet-Geology, is getting the interest of other disciplines. EMODnet-Geology partners from Denmark, Belgium and the Netherlands were invited to present Workpackage 3 developments at a workshop on habitats, organized by the German Federal Agency of Nature Conservation at the end of November (Vlim, Germany).
Iceland joined the EMODnet-Geology project in its second phase in 2013. The mapping procedures established have in many cases been pioneering work because literally no geological maps have been published of the Icelandic shelf or economic zone, apart from a few small scale black and white maps and sketches. The EMODnet-Geology partner from Iceland has presented the EMODnet project to the Ministry and the Minister of Environment in Iceland. The project has also been mentioned and described in the annual reports of Iceland GeoSurvey. Contacts have also been established with the Iceland Marine Research Institute, which has released valuable data for the sea bed substrate map, e.g. grain-size measurements and descriptions of samples and multibeam data. Now they are eagerly waiting for the map for their own seabed habitat mapping. Contacts also have been established with the Icelandic Coastguard (Hydrographic Department) which has delivered bathymetric information (multibeam data) along with an old collection of seabed substrate information. In November the Quaternary map and the Seabed Substrate maps will be presented and in 2016 the whole product of the project will be presented for the Icelandic Natural History Society. These maps, data collections and data harmonization is of no doubt an important step forward in understanding the geology of the country and its surroundings and will be of use for both the scientific community, the authorities and the industry of Iceland.
Industry/commercial
Triggered in part by a new national law governing the public availability (and requirement of its use) of subsurface data collected by or for the governments, there has been an interest from industry to make such data available through the portal of the Geological Survey of the Netherlands. It has been an opportunity to highlight the added value of EMODnet, as many of the projects they are involved in are transnational.
Research
Workpackage 3 data are being used by scientists with different backgrounds (seal research, geochemistry). They are encouraged to refer to EMODnet in their acknowledgements. 65
EMODnet Annual Report 2 – Lot 2 Geology
The Continental Shelf Department (Malta) actively collaborates with the newly-set up Department of Geosciences at the University of Malta. The Department of Geosciences had contributed to various work pages of EMODnet raising the profile of this project among the academic and student communities.
Workpackage 3 data have been requested by naval research organisations.
Minatura 2020, the aim of which is to develop a concept and methodology for the definition and subsequent protection of mineral deposits of public importance in order to ensure their best use in the future (http://minatura2020.eu/) has contacted the WP7 leader to request information as has the Blue Mining Project, an international European consortium of 19 large industry and research organisations on various maritime fields of expertise, which aims to develop solutions that will bring sustainable deep sea mining a step closer
66
EMODnet Annual Report 2 – Lot 2 Geology Annex 1
C5818
EMODnet-Geology Phase II
Case study: Quantitative spatial prediction of seabed sediment composition
Markus Diesing 7th October 2015
67
Abstract
This case study explores emerging techniques for quantitative spatial prediction of seabed sediment composition (mud, sand and gravel fractions of surface sediments). Predictions are based on observations (measured content of sediment fractions from grab samples) and carried out for two cases: i) In the high-resolution case, bathymetry and backscatter intensity data (10 m resolution) collected with multibeam echosounders and secondary variables derived from these two datasets were employed. ii) A low-resolution case was based on the EMODnet-Bathymetry data layer (250 m) and derivatives thereof. Additional data layers included the distance to the nearest coastline and modelled wave and current statistics. These were utilized in both cases. All spatial predictions were carried out with the Random Forest machine learning algorithm.
The observation data set contained samples from various sources, most of which were collected decades ago and prior to the introduction of the Global Positioning System. These samples have low spatial accuracy and changes in sediment composition might have occurred during the time interval between the collection of the samples and the acoustic surveys, which were carried out between 2006 and 2010. Including these samples in the model training led to poor predictions. Models trained without these samples yielded significantly better results despite the fact that the sample data set was reduced to less than a quarter of its original size. An important message is that the quality of the samples (positioning accuracy in relation to resolution of predictor variables, time difference between acoustic and sampling survey) is of greater importance than the quantity of the samples.
The most important predictor variables were backscatter, backscatter-related derivatives and hydrodynamics in the high-resolution case and peak wave orbital velocity statistics and the distance to the coastline in the low-resolution case (for which no backscatter data were available). It is noteworthy that hydrodynamics were important variables in both the high and the low-resolution study despite the fact that these were only available at a very coarse resolution in the order of 10 km. It might be expected that further improvements in model performance could be achieved by utilizing higher-resolution hydrodynamic model data.
Due to the lack of independent test samples, it was not possible to rigorously assess the accuracy of the predictions. A visual assessment of the correspondence of the predictions with the local geomorphology indicated that the high-resolution model yielded the most plausible results, followed by the low-resolution model and the EMODnet-Geology substrate layer.
The results of this case study have a direct bearing on possible approaches for repeatable and accurate seabed substrate mapping in the future: High-performing models can be built for the spatial prediction of sediment composition, given that key predictor variables and response variables of sufficient quality are available. However, such an approach could not be EMODnet Annual Report 2 – Lot 2 Geology
implemented across all European sea basins in the near future, due to the limited coverage and availability of multibeam echosounder data and a lack of standardisation for the collection and processing of backscatter data. An interim approach might therefore make best use of the data that are available. Such an approach could build on the existing data infrastructure of the European Union, making use of products made available via EMODnet, the Copernicus marine environment monitoring service and Geo-Seas. Introduction
The aim of the EMODnet-Geology project is to compile fragmented marine data products and make them available through a single data portal. With respect to work package 3, the central aim is to compile and harmonize all available seabed substrate information at a scale of 1:250,000. Harmonization of data will be carried out on the principles established during the EMODnet-Geology Phase I project. It is without doubt that a harmonized and accessible substrate data layer covering all European sea basins is of great benefit for research, conservation, spatial planning and the blue economy. It should, however, be noted that the underlying data and derived products might have been collected and produced sometimes decades ago.
In the meantime, we have witnessed significant improvements in several areas: Since the introduction of the Global Positioning System (GPS), positioning accuracy has increased dramatically, especially when compared to older systems like the Decca Navigator System. Seabed mapping and surveying is nowadays frequently carried out with multibeam echosounders, which collect detailed data on bathymetry and acoustic reflectance (backscatter strength) at very high spatial resolution. Lastly, rapid progress has been made in acoustic data interpretation from expert interpretation ‘by eye’ to more repeatable and objective methods such as geostatistics (Lark et al. 2015), image analysis (Lucieer 2008) and machine learning (Stephens & Diesing 2014). It is therefore timely to investigate what could be achieved by applying state-of-the-art spatial prediction methods to highly-detailed multibeam echosounder data. This will be achieved in the form of a case study, for a site where datasets of sufficient quality and quantity are available to give meaningful results.
In this case study, we will not aim to compare different data interpretation methods, as such comparisons have already been carried out by others (Ierodiaconou et al. 2011; Che Hasan et al. 2012; Diesing et al. 2014; Stephens & Diesing 2014; Galparsoro et al. 2015; Calvert et al. 2015). Instead, we selected a versatile machine-learning method that has been proven to derive accurate predictions across a wide variety of science domains and apply it to different acoustic datasets, namely i) multibeam echosounder data processed to a spatial resolution of 10 m and ii) EMODnet-Bathymetry data compiled from various sources at a resolution of 250 m. The results of the predictions will be compared with the EMODnet-Geology substrate layer produced at a scale of 1:250,000. The reasoning behind these comparisons is that they
69
EMODnet Annual Report 2 – Lot 2 Geology
will enable us to better understand what could be achieved on a sea basin scale in the near future (e.g. EMODnet Phase III) as exemplified by the application to the already existing EMODnet-Bathymetry data layer and in the more distant future, should there exist sufficient spatial coverage of sea basins with multibeam echosounder data. A comparison of these two predictions against the status quo (the current EMODnet-Geology seabed substrate layer) will help us better understand to what extent such predictions improve the accurate depiction of seabed substrates.
We have chosen a quantitative spatial prediction method to derive estimates of sediment composition (the fraction of gravel, sand and mud of surface sediments in this case). Once we have estimated the sediment composition for every cell in the grid, suitable classification systems (e.g. Folk 1954; Long 2006) can be applied to convert quantitative information into sediment classes. This ‘predict first, classify later’ approach has a distinct advantage, in that it allows easy reclassification. Conversely, classified maps that have been derived in a ‘classify first, predict later’ approach (such as the EMODnet-Geology substrate layer) are not easily convertible from the Folk classification (as used in EMODnet-Geology) to the EUNIS scheme (as used in EMODnet-Seabed Habitats).
The specific aims of this case study are threefold:
• To explore the effect of sample data quality on prediction performance;
• To compare the EMODnet-Geology substrate layer (1:250,000) against:
– Predictions based on EMODnet-Bathymetry layer and seabed samples
– Predictions based on multibeam echosounder data and seabed samples;
• To demonstrate potential avenues for flexible, repeatable and accurate seabed substrate mapping in the future.
70
EMODnet Annual Report 2 – Lot 2 Geology
Study site
The study site is located in the North Sea, off the east coast of Scotland (Figure 2). It has a size of approximately 15,000 km2. Water depths range from approximately 0 to 220 m (Figure 2). According to the current version of the EMODnet-Geology substrate layer, there is a predominance of fairly coarse substrates ranging from sand to gravel. Muddy substrates occur in places, e.g. in deep channels in the north of the study site and close to the coast in the south-west. Rock and boulder substrates are spatially very limited.
71
EMODnet Annual Report 2 – Lot 2 Geology
Figure 1. Location of the study site in the North Sea (inset) and substrates according to EMODnet- Geology overlaid on a hillshade image depicting the local geomorphology of the seabed. Data and Methods
The mapping of seabed substrate types requires the analyst to infer the most likely substrate at unsampled locations based on a spread of samples. Typically, such observations cover only
72
EMODnet Annual Report 2 – Lot 2 Geology
a small fraction of the seabed and inferences can be achieved in different ways: Traditionally, a skilled analyst would analyze the samples in conjunction with relevant ancillary data such as bathymetry, backscatter intensity (if available), geomorphology, current speed and direction etc. An understanding of the geological history and the implications for sediment characteristics and distributions is a crucial part of this methodology. The analyst exercises ‘expert’ judgment by interpreting these data in the context of the geological and hydrodynamic environment. This is the approach on which most, if not all, underlying maps of the EMODnet- Geology substrate layer are based.
Geostatistical approaches exploit Tobler’s first law of geography, which states that "everything is related to everything else, but near things are more related than distant things” (Tobler 1970). Geostatistics is used for modelling spatial data by applying probabilistic methods to regionalized variables.
The approach taken here can be described as predictive modelling, which is a two-step approach: Initially, the relationship between a set of predictor variables and a response variable is modelled from observations (samples). The established model is then employed to predict the response variable at unsampled locations for which values of the predictor variables are known. Predictive modelling of complex datasets is arguably best achieved with machine learning, which can be defined as programming computers to optimize a performance criterion using example data or past experience (Alpaydin 2010). Here, we train a machine learning algorithm to make predictions at unsampled locations based on learned relationships between response and predictor variables established at sampled locations.
There is a large variety of machine learning algorithms available, and selecting the ‘best’ algorithm for the problem in hand is often not straight forward. However, Random Forest (Breiman 2001) has become one of the most widely used and successful statistical learning models for classification and regression, showing good performance in a large number of domains (Pal 2005; Prasad et al. 2006; Cutler et al. 2007; Chan & Paelinckx 2008; Chapman et al. 2010; Che Hasan et al. 2012, 2014; Huang et al. 2012, 2014; Oliveira et al. 2012; Lucieer et al. 2013; Stephens & Diesing 2014; Diesing et al. 2014).
Substrate observations
Substrate observations were collated from a number of sources (Table 1). The minimum requirement for inclusion in this study was that the percentages of mud (grain size d < 63 µm), sand (63 µm ≤ d < 2 mm) and gravel (d ≥ 2 mm) fractions were reported. Only those observations that intersected with all the predictor variables (see below) were used; this resulted in a total of 801 substrate observations (Figure 2). Note that no attempt was made to predict the distribution of rock and hard substrates. This is due to several reasons: First, previous knowledge indicated that such substrates rarely occur in the study site. Second, no
73
EMODnet Annual Report 2 – Lot 2 Geology
adequate observations could be found to predict the distribution of hard substrates. Third, such a spatial prediction would require a different approach than the one outlined here.
74
EMODnet Annual Report 2 – Lot 2 Geology
75
EMODnet Annual Report 2 – Lot 2 Geology
Figure 2: Overview of the study site showing the bathymetry and locations of seabed samples available for analysis. Refer to Table 1 for abbreviations of data sources.
Table 1: Summary of data sources for substrate observations.
Source No. of References samples
British 619 Extracted from Offshore GeoIndex Geological Survey (BGS) http://mapapps2.bgs.ac.uk/geoindex_offshore/home.html
Contains British Geological Survey materials ©NERC 2015
Joint Nature 148 Sotheran, I. & Crawford-Avis, O. 2013. Mapping habitats and Conservation biotopes to strengthen the information base of Marine Protected Committee Areas in Scottish waters. JNCC Report No. 503. (JNCC) http://jncc.defra.gov.uk/page-6764
Eggleton, J., Diesing, M. & Schinaia, S. 2013. Offshore seabed survey of Turbot Bank possible MPA. Cefas Report (C5817). dbSEABED 21 Seabed substrates data integrated with the dbSEABED system. Chris Jenkins, INSTAAR, University of Colorado
http://instaar.colorado.edu/~jenkinsc/dbseabed/
Federal 13 Valerius, J., Van Lancker V., Van Heteren S., Leth J.O., Zeiler M. Maritime and 2014. Trans-national database of North Sea sediment data. Data Hydrographic compilation by Federal Maritime and Hydrographic Agency Agency of (Germany); Royal Belgian Institute of Natural Sciences (Belgium); Germany TNO (Netherlands) and Geological Survey of Denmark and Greenland (Denmark). (BSH)
Pre-treatment of response data
The mud, sand and gravel percentages are compositional data, i.e. each fraction is part of a total and is constrained between 0 and 1 and the three fractions must sum to 1 (or 100%). Because of this, each component should not be considered in isolation from the others. Here we followed recommendations of Aitchison (1986) and transformed the data onto the additive log-ratio scale where they can be analysed as two continuous, unconstrained response variables which can assume any value. The outputs of analysis are permutation-invariant regarding the denominator (Weltje 2006); here we have chosen to use the gravel fraction.
= log = log( ) log ( ) 𝑚𝑚𝑚𝑚𝑚𝑚 𝑎𝑎𝑎𝑎𝑎𝑎𝑚𝑚 � � 𝑚𝑚𝑚𝑚𝑚𝑚 − 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔
76
EMODnet Annual Report 2 – Lot 2 Geology
= log = log( ) log ( ) 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑎𝑎𝑎𝑎𝑎𝑎𝑠𝑠 � � 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 − 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 The two additive log-ratios alrm and𝑔𝑔𝑔𝑔𝑔𝑔 𝑣𝑣alr𝑒𝑒𝑒𝑒s constitute two response variables and separate predictive models must be built for both individually.
Predictor variables
Predictor features were selected based on previous experience, their expected importance for explaining the distribution of shelf sediments and their availability. Initially, high-resolution multibeam echosounder data collected between 2006 and 2010 were obtained from the United Kingdom Hydrographic Office and data layers of bathymetry and backscatter were derived at a resolution of 10 m. Several derivative layers were obtained from these primary data layers. These included Lee filtered backscatter, Hue-Saturation-Intensity (Daily 1983) and band ratios based on the former data derived from backscatter and eastness, northness, slope, roughness, curvature (Wilson et al. 2007), vector ruggedness measures (Sappington et al. 2007) and bathymetric position indices (Lundblad et al. 2006) from the bathymetry data. Additionally, the Euclidean distance to the coastline, modelled depth-averaged mean current speed and statistics of modelled peak wave orbital velocity at the seabed (Holt & James 2001; Aldridge et al. 2015) were also employed. For more detailed information on predictor variables refer to the Appendix.
In the case of the lower resolution 250 m EMODnet-Bathymetry input data, the same derivatives were calculated from the bathymetry data; however no backscatter data were available and hence backscatter-related derivatives could not be derived. The same additional data layers as in the high-resolution case were also included.
Feature selection
Prior to fitting the model, a feature selection step was implemented to test the statistical significance of the predictor variables for the prediction of the two response variables. The Boruta algorithm (Kursa & Rudnicki 2010) is a feature selection wrapper (Guyon & Elisseeff 2003) based on the Random Forest model. The algorithm uses the feature importance score generated by Random Forest to test each of the predictor variables against the effect of random noise. Only features that have scores significantly higher than random are retained for use in the final models.
Model fitting
Random Forest is a machine learning approach that belongs to the family of decision tree learning. The goal of decision tree learning is to create a model that predicts the value of a
77
EMODnet Annual Report 2 – Lot 2 Geology
response variable based on several predictor variables. Decision trees are of two main types: Classification trees predict class membership while regression trees predict a quantity.
The Random Forest is an ensemble technique developed by Breiman (2001). The algorithm ‘grows’ a large number of decision trees, thereby portioning the data into increasingly homogeneous groups. It is called a random forest because two elements of randomness are introduced. Firstly, each tree is constructed from a bootstrapped sample of the training data. Secondly, only a random subset of the predictor variables is used at each split in the tree growing process. This has the effect of making each tree in the forest unique. The underlying principal of the technique is that although each tree in the forest may individually be a poor predictor and that any two trees could give very different answers, by aggregating the predictions over a large number of uncorrelated trees, prediction variance is reduced and accuracy improved (James et al. 2013: p. 316). Observations not included in each tree construction (the so-called ‘out-of-bag’ samples) are then used to create a form of cross- validated prediction error. Random Forest also provides a relative estimate of predictor variable importance. This is a measure of the average reduction in prediction error associated with each variable.
The Random Forest prediction algorithm was chosen as the modelling tool for the analysis because it has shown high predictive accuracy in a number of domains (see above), it is versatile and can be used without extensive parameter tuning, it can handle a large number of predictor variables and is insensitive to the inclusion of some noisy/irrelevant variables.
The randomForest package (Liaw & Wiener 2002) in R (R Development Core Team 2011) was used for the implementation of the model via the Marine Geospatial Ecology Tools (Roberts et al. 2010) in ESRI ArcGIS v10.1. Each forest consisted of ntree = 500 trees and the number of variables available for splitting at each tree node (mtry) was equal to the number of predictor variables divided by three (rounded down).
Model validation
The Random Forest implicitly carries out a form of cross-validation using the out-of-bag observations. This usually gives a reliable measure for real model performance assuming enough trees are grown (Liaw & Wiener 2002). The performance is assessed by calculating the mean of the squared prediction error:
1 = 𝑛𝑛 ( ) , 2 𝑀𝑀𝑀𝑀𝑀𝑀𝑦𝑦� � 𝑦𝑦𝑖𝑖 − 𝑦𝑦�𝑖𝑖 𝑛𝑛 𝑖𝑖=1 with n being the number of observations, yi the observed value and ŷi the predicted value. The ‘variance explained’ by the models is then calculated by taking the ratio of the MSEŷ to the 2 variance of the observed values (σy ):
78
EMODnet Annual Report 2 – Lot 2 Geology
= 1 𝑀𝑀𝑀𝑀𝑀𝑀𝑦𝑦� 2 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 − 𝑦𝑦 Model predictions 𝜎𝜎
The fitted models were employed to spatially predict the two response variables based on the available predictor variables. To derive predictions of sediment composition, i.e. the fractions of gravel, sand and mud, it is necessary to back-transform the additive log-ratios:
( ) = ( ) + ( ) + 1 𝑒𝑒𝑒𝑒𝑒𝑒 𝑎𝑎𝑎𝑎𝑎𝑎𝑚𝑚 𝑚𝑚𝑚𝑚𝑚𝑚 𝑒𝑒𝑒𝑒𝑒𝑒 𝑎𝑎𝑎𝑎𝑎𝑎𝑚𝑚 (𝑒𝑒𝑒𝑒𝑒𝑒 )𝑎𝑎𝑎𝑎𝑎𝑎𝑠𝑠 = ( ) + ( ) + 1 𝑒𝑒𝑒𝑒𝑒𝑒 𝑎𝑎𝑎𝑎𝑎𝑎𝑠𝑠 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑒𝑒𝑒𝑒𝑒𝑒= 1𝑎𝑎𝑎𝑎𝑎𝑎𝑚𝑚( 𝑒𝑒𝑒𝑒𝑒𝑒+ 𝑎𝑎𝑙𝑙𝑙𝑙𝑠𝑠 )
The back-transformation was carried𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 𝑔𝑔out with− 𝑚𝑚𝑚𝑚𝑚𝑚 Raster𝑠𝑠 𝑠𝑠Calculator𝑠𝑠𝑠𝑠 in ArcGIS. The resulting predictions of sediment composition were subsequently classified according to the classifications of Folk (1954) and Long (2006) using a purpose-built R script.
Post-processing
The resulting maps of seabed sediment types have a fairly noisy appearance and some form of generalisation is desirable. We achieved this by applying a majority filter on image objects, which were derived by object-based image analysis (OBIA). OBIA is widely used in terrestrial remote sensing applications (Blaschke 2010; Blaschke et al. 2014), but was also successfully applied for mapping benthic habitats (Lucieer 2008; Lucieer et al. 2013; Diesing et al. 2014). OBIA is a two-step approach consisting of segmentation and classification. The aim of the segmentation is to divide the image into meaningful objects of variable sizes, based on their spectral and spatial characteristics. The resulting objects can be characterized by various features such as layer values (mean, standard deviation, skewness etc.), geometry (extent, shape etc.), texture and many others.
Here, only a segmentation of the predictor layer that was found to be most important (the Lee- filtered backscatter intensity) was carried out, using the multi-resolution segmentation algorithm in eCognition v9.1.0 software. This algorithm is an optimisation procedure, which locally minimizes the average heterogeneity of image objects for a given resolution of image objects. Starting from an individual pixel, it consecutively merges pixels until a certain threshold, defined by the scale parameter is reached. The scale parameter is an abstract term that determines the maximum allowable heterogeneity for the resulting image objects. The choice of an appropriate scale parameter was aided by the Estimation of Scale Parameter (ESP) tool (Dragut et al. 2010).
79
EMODnet Annual Report 2 – Lot 2 Geology
Figure 3. Illustration showing the process of generalizing the sediment predictions: Left – Lee-filtered backscatter intensity with outlines of image objects in red; Centre – Noisy classification results; Right – Generalized final results.
The resulting object polygons were exported as a shapefile for further analysis in ArcGIS. For each polygon, the most frequently occurring sediment class was determined with the Zonal Statistics tool and assigned to that polygon.
Map comparisons
The Folk maps derived from the high and low-resolution data prior to map generalization and the current EMODnet-Geology map were compared with the Map Comparison Kit (Visser & de Nijs 2006). A pixel by pixel comparison of map pairs was carried out and the matching number of pixels per substrate class was calculated. Based on these values the overall percentage agreement of two maps was derived. Further on, the software calculates the kappa coefficient as a measure of similarity for every mapped class. Kappa is the product of ‘kappa location’ (Pontius 2000) and ‘kappa histo’ (Hagen 2002). These statistics are sensitive to respective differences in location and in the histogram shape of all the categories. Given a fixed value for kappa, kappa location will increase if kappa histo decreases, and vice versa. Furthermore, if categories in both maps lie at identical locations, we have kappa location = 1 and kappa = kappa histo. If the frequency of categories in both maps is equal, we have kappa histo = 1 and kappa = kappa location (Visser & de Nijs 2006).
A pre-requisite to carry out such an analysis is that all maps have the same projection, resolution and extent. A cell size of 60 m was chosen, as this corresponds to the map resolution of 1:250,000 of the EMODnet-Geology substrate layer. All maps were clipped to the same spatial extent and projected to UTM 30N.
80
EMODnet Annual Report 2 – Lot 2 Geology
Results
Multibeam echosounder data
Initial trials of model fitting carried out with the full set of predictor variables and all seabed samples resulted in relatively poorly preforming models with 51.3% of the variance explained for alrm and 53.3% for alrs. We hypothesized that this might be due to the known low spatial precision of BGS samples, which were collected in the years 1967 to 1981 prior to the introduction of GPS. Additionally, changes in sediment composition might have occurred between the collection of the samples and the acoustic surveys (2006 to 2010). Removing the BGS samples from the response dataset resulted in a greatly reduced number of samples (n = 182) and spatial coverage; however the model performance increased considerably to explained variances of 83.4% for alrm and 85.2% for alrs. It was therefore decided to carry out all further analysis with the BGS sample data excluded.
The feature selection analysis revealed that about half of the predictor variables were not statistically significant and these were consequently removed from the analysis. Model fitting based on the selected features also showed a slight increase in variance explained for both response variables to 84.2% for alrm and 86.8% for alrs.
The analysis of the importance of predictor variables shows similar patterns for both response variables. Lee-filtered backscatter, followed by unfiltered backscatter are, perhaps unsurprisingly, the most important variables for prediction. The marked difference in importance between filtered and unfiltered backscatter also shows that a certain amount of smoothing of noisy backscatter data might be beneficial for quantitative spatial predictions. Some of the backscatter-related derivatives and hydrodynamics were also fairly important, whilst bathymetry and bathymetry-related derivatives scored lowest.
81
EMODnet Annual Report 2 – Lot 2 Geology
Figure 4. Predictor variable importance for the two response variables alrm and alrs. Refer to the Appendix for an explanation of the abbreviations of the predictor variables.
Despite the significant size of the datasets, each predictor layer had a size of approximately 840 MB, spatial predictions were derived in about 2 hours for each response variable. The resulting maps of the gravel, sand and mud fractions are shown in Figure 5. Note that the results are reported as fractions scaled between 0 and 1. To derive percentages, the values need to be multiplied by 100. Figure 6 shows the resulting Folk seabed sediment map.
82
EMODnet Annual Report 2 – Lot 2 Geology
Figure 5. Predicted fractions of A) - gravel, B) - sand and C) - mud.
83
EMODnet Annual Report 2 – Lot 2 Geology
84
EMODnet Annual Report 2 – Lot 2 Geology
Figure 6. Predicted seabed sediments according to the classification of Folk (1954) based on high- resolution multibeam echosounder data.
A visual assessment of the plausibility of the predicted sediment types was also carried out. This was achieved by overlaying the predictions on the hillshade relief map derived from the bathymetry, which depicts the local geomorphology. Areas where subaqueous dunes (sand waves) occur typically show up as sand and slightly gravelly sand (Figure 7). Coarser sediment such as gravelly sand and sandy gravel is predicted to occur where positive glacial landforms such as moraines, drumlins and ice-streaming features predominate. Deeper channels in the north of the study site are predicted to have a surficial sediment cover of muddy sand (Figure 6).
Figure 7. Visual assessment of the quality of the prediction: Left – Sand coinciding with subaqueous dunes; Right – gravelly Sand, gravelly muddy Sand and sandy Gravel coinciding with glacial landforms. Colours as in Figure 6.
EMODnet-Bathymetry
Similar to the high resolution case, including the BGS samples led to the construction of poorer performing models for both response variables with variances explained on the order of 50%. Consequently, these were excluded from the response dataset. The feature selection process indicated that most predictor variables were deemed significant, but those that weren’t (northness, some vector ruggedness measures and one bathymetric position index) were removed from the analysis. The variance explained did not differ greatly from those values achieved in the high-resolution case with 83.9% of variance explained for alrm and 84.7% for alrs. Peak wave orbital velocity statistics and the Euclidean distance to the coastline were generally the most important predictor features for both response variables.
85
EMODnet Annual Report 2 – Lot 2 Geology
Spatial predictions were carried out and the resulting maps of the two response variables were back-transformed to gravel, sand and mud fractions. Based on these results, a Folk map was derived (Figure 8). A visual comparison with the Folk map derived in the high-resolution case study shows similarities in areas further away from the coast and especially in the north-east of the site, but also marked differences closer to the coastline. Small-scale morphological features such as subaqueous dunes cannot be resolved due to the significantly lower resolution (and potentially the lack of backscatter data).
86
EMODnet Annual Report 2 – Lot 2 Geology
87
EMODnet Annual Report 2 – Lot 2 Geology
Figure 8. Predicted seabed sediments according to the classification of Folk (1954) based on EMODnet- Bathymetry data. Map comparisons
The pixel-by-pixel comparison of map pairs revealed that the Folk maps based on multibeam echosounder and EMODnet-Bathymetry data were the most similar pair with an overall agreement of 54.2%. The highest per-class agreement between the two maps was observed for sand with a kappa of 0.41, followed by gravelly sand (0.36). This agreement can mainly be attributed to high agreement in the number of pixels predicted as those classes in both maps (kappa histo), while the agreement in location in the maps is moderate. Conversely, there is low or no agreement for slightly gravelly muddy sand and gravelly muddy sand.
Table 2: Per-class similarity for the map pair multibeam echosounder derived data – EMODnet-Bathymetry derived data. Refer to Figure 8 for explanation of substrate codes.
Substrate 1.3.1 1.3.2 2.1.1 2.1.2 3.1.1 3.2.1 4.3.1
Kappa 0.25 0.00 0.41 0.13 0.36 0.23 0.03
Kappa location 0.38 -0.01 0.43 0.15 0.41 0.52 0.13
Kappa histo 0.67 0.17 0.93 0.84 0.89 0.44 0.24
The EMODnet-Geology map agreement was 48.0% with the multibeam echosounder-derived map. The highest per-class agreement was found for sand (kappa = 0.33), gravelly sand (0.28) and muddy sand (0.28), mainly due to agreement in kappa histo. Low agreement was observed for slightly gravelly sand, sandy gravel and gravelly muddy sand. Several substrates (rock, mud, sandy mud, slightly gravelly sandy mud, gravel and gravelly mud) were only present in the EMODnet-Geology map, while slightly gravelly muddy sand was exclusive to the multibeam echosounder-derived map.
Table 3: Per-class similarity for the map pair multibeam echosounder derived data – EMODnet-Geology. Refer to Figure 8 for explanation of substrate codes.
Substrate 1.3.1 2.1.1 2.1.2 3.1.1 3.2.1 4.3.1
Kappa 0.28 0.33 0.09 0.28 0.04 0.00
Kappa location 0.40 0.43 0.16 0.29 0.04 -0.01
Kappa histo 0.70 0.76 0.60 0.96 0.98 0.07
88
EMODnet Annual Report 2 – Lot 2 Geology
The EMODnet-Bathymetry-derived map has an overall agreement of 40.4% with the EMODnet-Geology map. Per-class agreement is generally low, but highest for sand (kappa = 0.21), muddy sand (0.19) and gravelly sand (0.16). Low or no agreement was observed for slightly gravelly sand, sandy gravel and gravelly muddy sand. Several substrates (rock, mud, sandy mud, slightly gravelly sandy mud, gravel and gravelly mud) were only present in the EMODnet-Geology map, while slightly gravelly muddy sand was limited to the EMODnet- Bathymetry-derived map.
Table 4: Per-class similarity for the map pair EMODnet-Bathymetry derived data – EMODnet-Geology. Refer to Figure 8 for explanation of substrate codes.
Substrate 1.3.1 2.1.1 2.1.2 3.1.1 3.2.1 4.3.1
Kappa 0.19 0.21 0.04 0.16 -0.01 0.00
Kappa location 0.46 0.26 0.06 0.19 -0.02 0.00
Kappa histo 0.42 0.82 0.75 0.84 0.43 0.41
89
EMODnet Annual Report 2 – Lot 2 Geology
Discussion
The presented results indicate the limitations of legacy sample data for quantitative spatial predictions. In both cases, the exclusion of BGS samples led to significant improvements of model performance. These improvements were achieved despite the fact that the set of samples was reduced to less than a quarter of the original size and the spatial distribution was very uneven across the study site. It could be expected that requirements on spatial accuracy are dependent on the resolution of the predictor variables. There were however no discernable differences between the high and low-resolution models in terms of performance. This might indicate that significant temporal changes in sediment composition have occurred during the time interval between sample collection and acoustic surveys. Alternatively, predictor variables with a spatial resolution of 250 m might still require a more precise positioning accuracy than can be achieved with Decca navigation systems. In fact, positioning accuracy might be worse than 400 m for all times and seasons for parts of the study site (Last 1992: Fig. 6). Our own experience shows that the inclusion of BGS samples alongside other samples led to reasonable results when predicting sediment composition in a similar way based on 500 m resolution predictor variables.
All models showed a very satisfactory performance as measured by the variance explained. The visual assessment of the high-resolution predictions also indicated that the resulting map was plausible and did agree with the geomorphology of the site to a large degree. This was however only partly the case for the low-resolution predictions, especially within approximately 20 km from the coastline. Despite that, agreement between predicted sediment class and geomorphology was higher than in the case of the EMODnet-Geology layer (Figure 1).
To test the accuracy of the three maps, it would have been desirable to have independent samples, which weren’t employed in any derivation of substrate types. This could be achieved by splitting the existing set of seabed samples into subsets for model training and validation. Such an approach was however deemed unfeasible, as the total number of samples was fairly low, especially when compared to the number of predicted classes and the fact that some classes are hardly represented by samples. A quantitative assessment of map accuracy was therefore not possible. Because of that the assessment of map quality must solely depend on the visual assessment, which indicated that the high-resolution model yielded the most reliable results. It is probably save to say that both models produced more plausible results when compared with the current version of the EMODnet-Geology substrate layer (Figure 1), which frequently appears to be unrelated to the morphological makeup of the site.
The overall agreement between the individual maps was fairly low, ranging between 40.4% and 54.2%. The highest per-class agreement was typically found for sand, sandy gravel and muddy sand, and this was mainly due to high kappa histo. The fairly low agreement between maps might be expected, as the study area is very complex and the method of prediction
90
EMODnet Annual Report 2 – Lot 2 Geology
(expert interpretation versus Random Forest) as well as the response and predictor variables (availability of backscatter, spatial resolution) varied to a degree. Notable differences were still found in a study that employed different prediction methods, but used the same response and predictor data (Diesing et al. 2014). These authors reported map agreements ranging from 68% to 87%, although only five substrate classes were mapped. There is a growing body of evidence that indicates that several factors such as prediction approach, quality and quantity of the response data, type, resolution and availability of predictor variables along with the general understanding of the underlying processes will affect the final outcome. At the same time, map users require a ‘definitive’ answer. As there is no such thing, it would be desirable to increase prediction accuracy whilst at the same time effectively communicate uncertainty in the predictions. A possible way of dealing with this could be a multi-model ensemble, whereby the same data are predicted with a suite of classifiers. Disagreement in predictions might be used as an indicator of confidence (Diesing & Stephens 2015). However, such an approach cannot deal with inadequacies in sample quality and quantity, sampling design, as well as availability of suitable predictor variables.
Despite the limitations mentioned above, it can be stated that the presented approach has several advantages: The Random Forest algorithm provides a versatile technique for dealing with inference problems. It has the ability to handle a large number of predictor variables even when the number of observations is relatively small. It is a non-parametric technique, i.e. no assumptions regarding the shape of distributions of the response and predictor variables are required (Cutler et al. 2007). It can handle complex non-linear relationships between predictor and response variables.
Based on out-of-bag observations, a cross-validated error estimate was derived. This usually gives a reliable measure for real model performance assuming enough trees are grown (Liaw & Wiener 2002). Furthermore, the presented method could be considered as repeatable, i.e. using the same input data and model parameters results in the same predictions. The predictions are quantitative, i.e. sediment composition instead of sediment class was predicted here. Quantitative outputs have the significant advantage that they easily allow reclassification of the data. For example, predicted mud, sand and gravel fractions could be classified following Long (2006) or Folk (1954) with minor effort. Although Random Forest is not fully objective as a certain amount of human intervention is always necessary, this intervention is minimal, as Random Forest essentially requires the user to select only two parameters, namely the number of trees (ntree) to be grown and the number of predictors tested at each node (mtry). In practice, the resulting model is usually fairly insensitive to parameters, provided ntree is large enough to stabilise predictions.
Additionally, Random Forest yields information on the importance of predictor variables and therefore could be used as an exploratory tool as well. Partial dependency plots (not employed in this study) could also be derived for further data exploration. Such plots indicate the
91
EMODnet Annual Report 2 – Lot 2 Geology
response of the dependent variable across the range of individual predictor variables of interest, while averaging out the effects of all other predictor variables. Partial dependency plots are useful for understanding the nature of the relationship between predictor and response variable, the shape of the underlying function and for identifying thresholds in predictor variables, which may be particularly important.
The variable importance plots indicate that backscatter strength and hydrodynamics are critical variables for a successful spatial prediction of sediment composition. This might be unsurprising, as backscatter is generally assumed to be an important proxy for sediment type and hydrodynamics are an important agent for shaping the sediment patterns of continental shelf seas. Based on such prior knowledge, those predictor variables were indeed chosen for inclusion in the models, and it is reassuring to see that this selection was justified by the model results. It is also noteworthy that hydrodynamics were important variables in both the high and the low-resolution study despite the fact that statistics of current speed and wave orbital velocity were only available at a very coarse resolution on the order of 10 km. It could be expected that further improvements in model performance are achievable by including higher- resolution hydrodynamic model data.
Spatial predictions based on gridded data inadvertently lead to somewhat noisy results. This disadvantage can be minimized with the application of OBIA and zonal statistics in ArcGIS, should a certain level of generalization be required. The level of generalization can be tuned by the selection of the scale parameter during the segmentation process. It should however be noted that generalization might lead to the loss of infrequently occurring sediment classes.
The results above have a direct bearing on possible approaches for flexible, repeatable and accurate seabed substrate mapping in the future. It has been demonstrated that high- performing models can be built for the spatial prediction of sediment composition, given that key predictor variables (backscatter strength, hydrodynamics) and response variables of sufficient quality are available. Ideally, the amount of samples should be large enough to allow for a split into training and validation data so that an independent test of prediction performance and accuracy assessment can be carried out.
It is however unlikely that such an approach could be implemented across all European sea basins in the near future. This is mainly due to the fact that only a small fraction of the European seas has been mapped with multibeam echosounders. Additionally, no common standards for the collection of backscatter data exist, although recent efforts initiated by the GeoHab (Geological and Biological Habitat Mapping) community (Lurton & Lamarche 2015) might be considered as a step in the right direction. Furthermore, it has to be expected that a significant amount of existing seabed sample data suffer from the same shortcomings as discussed above. Deriving highly-resolved, accurate, validated and quantitative maps of seabed composition on a sea basin scale would therefore require concerted efforts in terms of
92
EMODnet Annual Report 2 – Lot 2 Geology
acoustic seabed mapping and sampling. Some countries, notably Ireland and Norway, have already embarked on national seabed mapping programs; the areas of interest are however defined by the respective EEZ boundaries rather than physiographic units.
An alternative or interim approach might therefore make best use of the data that are available, having the limitations (as demonstrated here) in mind. Such an approach could build on the existing data infrastructure of the European Union: Relevant datasets for lower- resolution, quantitative spatial predictions of sediment composition are already available. These include the EMODnet-Bathymetry data layer at a resolution of ⅛ arc minute (ca. 250 m), hydrodynamic model hindcasts of current speed at a resolution of 7 km and satellite- derived estimates of suspended particulate matter at a resolution of 1 km (both accessible on the Copernicus marine environment monitoring service) as predictor variables. Relevant data on seabed observations (sediment composition) are generally available via the Geo-Seas portal, but might also become available as part of EMODnet-Geology Phase III. As stated earlier, the level of positioning accuracy of seabed samples will determine to some extent the achievable resolution of the models and initial trials should be carried out to determine this. In order to deal with the expected lower accuracy of the predictions when compared against the high-resolution case, it might be desirable to include a visual assessment and manual editing step in the final workflow.
93
EMODnet Annual Report 2 – Lot 2 Geology
References
Aitchison, J. (1986). The Statistical Analysis of Compositional Data. Chapman and Hall, London.
Aldridge, J.N., Parker, E.R., Bricheno, L., Green, S.L. & van der Molen, J. (2015). Assessment of the physical disturbance of the northern European Continental shelf seabed by waves and currents. Continental Shelf Research.
Alpaydin, E. (2010). Introduction to Machine Learning, 2nd Edition. MIT Press, Cambridge, MA.
Blaschke, T. (2010). Object based image analysis for remote sensing. ISPRS Journal of Photogrammetry and Remote Sensing, 65, 2–16.
Blaschke, T., Hay, G.J., Kelly, M., Lang, S., Hofmann, P., Addink, E., Queiroz Feitosa, R., van der Meer, F., van der Werff, H., van Coillie, F. & Tiede, D. (2014). Geographic Object-Based Image Analysis - Towards a new paradigm. ISPRS Journal of Photogrammetry and Remote Sensing, 87, 180–191.
Breiman, L. (2001). Random Forests. Machine Learning, 45, 5–32.
Calvert, J., Strong, J.A., Service, M., McGonigle, C. & Quinn, R. (2015). An evaluation of supervised and unsupervised classification techniques for marine benthic habitat mapping using multibeam echosounder data. ICES Journal of Marine Science: Journal du Conseil , 72 , 1498–1513.
Chan, J.C.W. & Paelinckx, D. (2008). Evaluation of Random Forest and Adaboost tree-based ensemble classification and spectral band selection for ecotope mapping using airborne hyperspectral imagery. Remote Sensing of Environment, 112, 2999–3011.
Chapman, D.S., Bonn, A., Kunin, W.E. & Cornell, S.J. (2010). Random Forest characterization of upland vegetation and management burning from aerial imagery. Journal of Biogeography, 37, 37–46.
Che Hasan, R., Ierodiaconou, D., Laurenson, L. & Schimel, A. (2014). Integrating Multibeam Backscatter Angular Response, Mosaic and Bathymetry Data for Benthic Habitat Mapping. PLOS ONE, 9, e97339.
Che Hasan, R., Ierodiaconou, D. & Monk, J. (2012). Evaluation of Four Supervised Learning Methods for Benthic Habitat Mapping Using Backscatter from Multi-Beam Sonar. Remote Sensing, 4, 3427–3443.
Cutler, D., Edwards, T., Beards, K., Cutler, A., Hess, K., Gibson, J. & Lawler, J. (2007). Random Forests for classification in Ecology. Ecology, 88, 2783–2792.
Daily, M. (1983). Hue-Saturation-Intensity Split-Spectrum Processing of Seasat Radar Imagery. Photogrammetric Engineering and Remote Sensing, 49, 349–355.
94
EMODnet Annual Report 2 – Lot 2 Geology
Diesing, M., Green, S.L., Stephens, D., Lark, R.M., Stewart, H.A. & Dove, D. (2014). Mapping seabed sediments: Comparison of manual, geostatistical, object-based image analysis and machine learning approaches. Continental Shelf Research, 84, 107–119.
Diesing, M. & Stephens, D. (2015). A multi-model ensemble approach to seabed mapping. Journal of Sea Research, 100, 62–69.
Dragut, L., Tiede, D. & Levick, S.R. (2010). ESP: a tool to estimate scale parameter for multiresolution image segmentation of remotely sensed data. International Journal of Geographical Information Science, 24, 859–871.
Folk, R.L. (1954). The distinction between grain size and mineral composition in sedimentary-rock nomenclature. Journal of Geology, 62, 344–359.
Galparsoro, I., Agrafojo, X., Roche, M. & Degrendele, K. (2015). Comparison of supervised and unsupervised automatic classification methods for sediment types mapping using multibeam echosounder and grab sampling. Italian Journal of Geosciences , 134 , 41–49.
Guyon, I. & Elisseeff, A. (2003). An Introduction to Variable and Feature Selection. Journal of Machine Learning Research, 3, 1157–1182.
Hagen, A. (2002). Multi-method assessment of map similarity. 5th AGILE Conference on Geographic Information Science, pp. 1–8.
Holt, J.T. & James, I.D. (2001). An s coordinate density evolving model of the northwest European continental shelf: 1. Model description and density structure. Journal of Geophysical Research: Oceans, 106, 14015–14034.
Huang, Z., Nichol, S.L., Siwabessy, J.P.W., Daniell, J. & Brooke, B.P. (2012). Predictive modelling of seabed sediment parameters using multibeam acoustic data: a case study on the Carnarvon Shelf, Western Australia. International Journal of Geographical Information Science, 26, 283–307.
Huang, Z., Siwabessy, J., Nichol, S.L. & Brooke, B.P. (2014). Predictive mapping of seabed substrata using high-resolution multibeam sonar data: A case study from a shelf with complex geomorphology. Marine Geology, 357, 37–52.
Ierodiaconou, D., Monk, J., Rattray, A., Laurenson, L. & Versace, V.L. (2011). Comparison of automated classification techniques for predicting benthic biological communities using hydroacoustics and video observations. Continental Shelf Research, 31, S28–S38.
James, G., Witten, D., Hastie, T. & Tibshirani, R. (2013). An Introduction to Statistical Learning. Springer.
Kursa, M. & Rudnicki, W. (2010). Feature selection with the Boruta Package. Journal of Statistical Software, 36, 1–11.
Lark, R.M., Marchant, B.P., Dove, D., Green, S.L., Stewart, H. & Diesing, M. (2015). Combining observations with acoustic swath bathymetry and backscatter to map seabed sediment texture classes: The empirical best linear unbiased predictor. Sedimentary Geology, 328, 17–32.
95
EMODnet Annual Report 2 – Lot 2 Geology
Last, D. (1992). The Accuracy and Coverage of Loran-C and of the Decca Navigator System - and the Fallacy of Fixed Errors. The Journal of Navigation, 45, 36–51.
Liaw, A. & Wiener, M. (2002). Classification and regression by randomForest. R News, 2, 18–22.
Long, D. (2006). BGS detailed explanation of seabed sediment modified Folk classification.
Lucieer, V.L. (2008). Object-oriented classification of sidescan sonar data for mapping benthic marine habitats. International Journal of Remote Sensing, 29, 905–921.
Lucieer, V., Hill, N.A., Barrett, N.S. & Nichol, S. (2013). Do marine substrates ‘look’ and ‘sound’ the same? Supervised classification of multibeam acoustic data using autonomous underwater vehicle images. Estuarine, Coastal and Shelf Science, 117, 94–106.
Lundblad, E.R., Wright, D.J., Miller, J., Larkin, E.M., Rinehart, R., Naar, D.F., Donahue, B.T., Anderson, S.M. & Battista, T. (2006). A Benthic Terrain Classification Scheme for American Samoa. Marine Geodesy, 29, 89–111.
Lurton, X. & Lamarche, G. (2015). Backscatter measurements by seafloor mapping sonars. Guidelines and Recommendations.
Oliveira, S., Oehler, F., San-Miguel-Ayanz, J., Camia, A. & Pereira, J.M.C. (2012). Modeling spatial patterns of fire occurrence in Mediterranean Europe using Multiple Regression and Random Forest. Forest Ecology and Management, 275, 117–129.
Pal, M. (2005). Random forest classifier for remote sensing classification. International Journal of Remote Sensing, 26, 217–222.
Pontius, R.G. (2000). Quantification error versus location error in comparison of categorical maps. Photogrammetric Engineering and Remote Sensing, 66, 1011–1016.
Prasad, A.M., Iverson, L.R. & Liaw, A. (2006). Newer classification and regression tree techniques: Bagging and random forests for ecological prediction. Ecosystems, 9, 181–199.
R Development Core Team. (2011). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing.
Roberts, J.J., Best, B.D., Dunn, D.C., Treml, E.A. & Halpin, P.N. (2010). Marine Geospatial Ecology Tools: An integrated framework for ecological geoprocessing with ArcGIS, Python, R, MATLAB, and C++. Environmental Modelling & Software, 25, 1197–1207.
Sappington, J.M., Longshore, K.M. & Thompson, D.B. (2007). Quantifying Landscape Ruggedness for Animal Habitat Analysis: A Case Study Using Bighorn Sheep in the Mojave Desert. Journal of Wildlife Management, 71, 1419–1426.
Stephens, D. & Diesing, M. (2014). A Comparison of Supervised Classification Methods for the Prediction of Substrate Type Using Multibeam Acoustic and Legacy Grain-Size Data. PLOS ONE, 9, e93950.
96
EMODnet Annual Report 2 – Lot 2 Geology
Tobler, W. (1970). A computer movie simulating urban growth in the Detroit region. Economic Geography, 46, 234–240.
Visser, H. & de Nijs, T. (2006). The Map Comparison Kit. Environmental Modelling & Software, 21, 346–358.
Weltje, G. (2006). Ternary sandstone composition and provenance: an evaluation of the ‘Dickson model’. Compositional Data Analysis in the Geosciences (eds A. Buccianti, G. Mateu-Figueras & V. Pawlowsky-Glahn), pp. 79–99. The Geological Society, London.
Wilson, M.F.J., O’Connell, B., Brown, C., Guinan, J.C. & Grehan, A.J. (2007). Multiscale Terrain Analysis of Multibeam Bathymetry Data for Habitat Mapping on the Continental Slope. Marine Geodesy, 30, 3–35.
97
EMODnet Annual Report 2 – Lot 2 Geology
Appendix
The primary predictor variables comprised bathymetry (multibeam echosounder data at a resolution of 10 m and EMODnet-Bathymetry at a resolution of 250 m) and backscatter intensity (mutibeam echosounder data only).
Derivatives of bathymetry are listed in Table 5, those of backscatter in Table 6 and other predictors in Table 7.
Table 5: Predictor variables derived from bathymetry data (MBFP).
Variable Description Unit Name Reference
Slope The maximum slope gradient degree slope Wilson et al. (2007)
Roughness The difference between m rgh Wilson et al. (2007) minimum and maximum of cell and its 8 neighbours.
Aspect Direction of steepest slope, eastness Wilson et al. (2007) expressed as eastness (sine of aspect) and northness northness (cosine of aspect)
Curvature Rate of change of slope. curvPL Wilson et al. (2007) Profile curvature (curvPR) is measured parallel to maximum curvPR slope; plan curvature (curvPL) is measured perpendicular to curv slope.
Bathymetric Position Vertical position of cell relative m BPI3, BPI5, Lundblad et al. (2006) Index (BPI) to neighbourhood (identifies BPI10, BPI25, topographic peaks and BPI30, BPI40, troughs). Radii of 3, 5, 10, 25, BPI50, BPI60, 30, 40, 50, 60, 70, 80, 90 and BPI70, BPI80, 100 pixels were used. BPI90, BPI100
Vector ruggedness Based on a geomorphological VRM3, VRM5, Sappington et al. measure (VRM) method for measuring vector VRM7, VRM9, (2005) dispersion. VRM11
98
EMODnet Annual Report 2 – Lot 2 Geology
Table 6: Predictor variables derived from backscatter data (BSFP).
Variable Description Unit Name Reference
Lee-filtered Reduces the speckle noise by applying a dB BSFP_Lee5 backscatter spatial filter to each pixel in an image, which filters the data based on local statistics calculated within a square window. The value of the centre pixel is replaced by a value calculated using the neighbouring pixels. A neighbourhood size of 5 pixels was chosen here after trials.
HSI A synthetic colour image derived by applying HSI_R Daily (1983) high and low pass filters in order to separate high and low frequency information. High HSI_G and low frequency information is then mapped to hue (chromatic) and intensity HSI_B (achromatic) respectively with a fixed saturation value. These HSI values are then transformed to red, green, blue (RGB) colour space.
Band ratios Band ratios of the three HSI layers were Ratio_RG derived by simple division: Ratio_RB Ratio_RG = HSI_R/HSI_G Ratio_GB Ratio_RB = HSI_R/HSI_B
Ratio_GB = HSI_G/HSI_B
99
EMODnet Annual Report 2 – Lot 2 Geology
Table 7: Other predictor variables.
Variable Description Unit Name Reference
Euclidean The straight-line distance to the nearest m EucDist distance to coastline. coastline
Average current Depth mean tidal and wind-driven currents m/s Av_current Aldridge et al. speed calculated using the POLCOMS model (Holt (2015) & James, 2001) with a grid resolution of approximately 11 km.
Peak wave orbital The WAM spectral wave model was used to m/s PkOrbVelMx Aldridge et al. velocity at the provide significant wave height and zero- (2015) seabed crossing wave period with a grid resolution of PkOrbVelMn approximately 11 km over the period 2000 to 2008. These were used in conjunction with a PkOrbVelSD bathymetric model at a resolution of 200 m to derive statistics of wave orbital velocity at the seabed. These included the maximum (Mx), mean (Mn) and standard deviation (SD).
100