Towards precision fishery: standardized modelling of smart services in the fishery domain

Transcription

1 Towards precision fishery: standardized modelling of smart services in the fishery domain Ståle Walderhaug Department for Software Engineering, Safety and Security SINTEF Digital Tromsø Norway Per Gunnar Auran Department for Energy and Transport SINTEF Ocean AS Trondheim Norway Arne Jørgen Berre Department for Software and Service Innovation SINTEF Digital Oslo Norway 1

2 Our research goal Investigate in a large scale how smart data-driven services for fisheries can be identified and modelled using a model-driven approach with Archimate 3.0 big data reference model from BDVA. 2

3 The problem domain Operate in harsh environments Limited communication coverage and with rapid changes in weather, fish prices and regulations. Weather changes heavily affect route optimization, fuel consumption as well as the feasibility, efficiency and safety of fishery operations. Norway Sweden Finland Russia 4

4 Fishing is data intensive The fishing fleet is increasingly sophisticated with numerous sensors. Sales data are routinely accessed to obtain information AIS (Automatic Identification System) tracking portals are used The governing bodies (EC and national EU and EEC member states) requires fishermen and landing sites by law to report catch data back to them for monitoring purposes. 5

5 Fishery Datasets Oceanic Tuna and Small Pelagic Fisheries Dataset type / Variety Dataset Volume (GB) Velocity (GB/year) Start date Comment Market Data Catch reports, economic figures < 1GB <0, data sources, catch data >50k but important Vessel Data ESAS Eros KingsBay LiegFi Echebastar A1 pilot data Echebastar UE Fleet data , ,8 18,8 20,5 19,6 17,1 5,4 602, Recently installed by DataBio 12bit x 4variables x 1,07e+11/year (minimum) EO, CMEMS, Met Ocean physcis(current/temp) Wave data (open ocean) Biology Wind Coastal waves (Met) Oceanic Tuna EO + research data , ,8 115,5 165,7 37,1 121, , bit x 4variables x 1,11e+11/year (minimum) SINMOD Oceanographic modeling (4km) NOAA atmospheric + SINMOD input 385 N/A N/A Operationalized + fishmodel by Databio Only keept for model output generation Hydroacoustics SIMRAD EK80 series-echosounder (SIMRAD SX90 Sonar Simrad SN90 Sonar +echosounder 3,3 N/A , Per cruise N/A N/A Logging module development in DataBio For information only, main fish sensor) Mitigation- Research cruise, manual logging Total All fishery pilot data assets 5004,2 5815,4 6

6 Context: The DataBio project with pilot demonstrations 7

7 DataBio Fisheries Pilots Overview Apply Big Data Technology in two fisheries and three thematic groups: 6 pilots Eros AS Kings Bay AS T3.2 T.3.3 T 3.4 8

8 Methods and materials Use the BDVA reference model with Archimate modelling as a tool for bridging the communication gap actors involved Identify stakeholders, their operation goals and strategies for implementing them Identify state of the art data sources and software components 9

9 The results Archimate Motivation Strategy Business models Application Extended diagrams Pipelines Role Lifecycle 10

10 Motivation 11

11 Strategy 12

12 Business 13

13 BDVA distribution 14

14 Common application pipeline Data visualization C1 Pilot B2 Pilot Data management Data processing and analytics C2 Pilot Common datasets/ infrastructure 15

15 Modelling results summary 26 pilot systems Extended modelling to include agriculture and forestry 450 design diagrams 5700 modelling elements 16

16 DataBio Fishery services Small pelagic pilots common picture 17

17 Historic catch data: fishing grounds, landing site catch composition 18

18 Individual Haul Analytics 19

19 {High Machine learned catch prediction classification Deep Nevral Network (ESUSHI whitefish data) Latitude,Longitude,Salinity,Temperature,BottomTemperature,Mo onphase,meanwavedirectionswell,meanwavedirectionwind,me anwaveperiodswell,peakwavedirection,peakwavedirectionswel l,peakwavedirectionwind,peakwaveperiod,peakwaveperiodswe ll,peakwaveperiodswim,significantswellwaveheight,significantw aveheight,significantwaveperiod,significantwindwaveheight,sto kesdriftxvelocity,stokesdriftyvelocity,wavedirection,winddirecti on,windspeed,distancetoland,nitrogensurface,nitrogenbottom, phosphatesurface,phosphatebottom,silicatesurface,silicatebotto m,flagellatessurface,flagellatesbottom,diatomssurface,diatomsb ottom,oxygensurface,oxygenbottom,microplanctonsurface,micr oplanctonbottom,mesozoonplanctonsurface,mesozoonplancton Bottom,detriusSurface,detriusBottom,DEPTSurface,DEPTBottom, SISSurface,SISBottom,AvgDepth,AvgChlorophyll f(,t)= C Catch Probability Average Expectations Low Catch Probability 20

20 Findings from modelling Fishery Big Data Services Fishermen from different fisheries and geographical areas share the majority of goals and drivers The strategy for reaching these goals require many common activities and will utilize a common set of resources (e.g. processing and data sets) Reuse and standardization may be useful There is a large variation in quality and availability of relevant data sources for predictive analytics Data spaces should be in focus 21

21 Concluding remarks Archimate and domain experts Motivation and strategy models useful for communication and design Business models in some cases Archimate and BDVA Reference model useful component classifications and application design Pipeline and lifecycle modelling more structured Promosing results, but more evaluations are requiered to conclude 22

22 Thank you for your attention! 23