Data-driven policies, sea-rescue operations - PrédiSauvetage

Transcript

1. Data-driven policies: sea-rescue operations in France PrédiSauvetage project @ EIG
2. Who am I? I'm Antoine Augusti, a software engineer/data scientist (distributed systems engineer, research master's in machine learning). Work: started the data team @ Drivy (acquired by Getaround for $300M), sea rescue @ maritime affairs, software engineer @ Etalab. Areas of interest: public policy, regulation, transparency, tests, software architecture, open source, open data.
3. Quizz • Do you need a license to go out to sea in France? • How large and significant is the sea territory of France? • Who represents France at sea? • What's the emergency radio channel at sea? • How can you contact emergency services at sea? • How do you do a distress call? • Who's in charge of sea rescue in France? • Who's going to save you if you're in distress at sea?
7. International sea rescue convention: SOLAS SAR at sea was introduced by the International Convention for the Safety of Life at Sea (SOLAS), adopted in 1979 in Hamburg. No matter where an accident occurs, the rescue of persons in distress at sea will be coordinated by a SAR organisation and, when necessary, by cooperation between neighbouring SAR organisations. Before that, no system was in place to perform search and rescue operations at-sea. In some areas, there was a well-established organisation able to provide assistance promptly and efficiently, in others there was nothing at all. Now, the IMO (International Maritime organisation) coordinates SAR at sea efforts.
8. SOLAS in practice The IMO publishes a manual (IAMSAR) to describe how states should perform SAR missions: organisation, management, mission coordination, mobile facilities. States should declare to the IMO: • Who's the SAR organisation • Describe maritime rescue coordination centers (MRCC) and search and rescue regions (SRR) • Describe contact options • Say which kind of assistance they can provide (kinds of vessels, helicopters, people, special equipment)
14. MRCC in France: les CROSS In France, MRCC are called CROSS (Centres régionaux opérationnels de surveillance et de sauvetage). They are armed with French Navy personnel, supervised by officers from Maritime Affairs (Ministry of Ecology) under the authority of Préfet Maritime. The first CROSS was introduced in 1967. They've got a wide range of missions: • maritime assistance and rescue missions • pollution prevention • fishing regulation • relay nautical information • respond to piracy, terrorism and attacks
15. MRCC in France: les CROSS France has 5 MRCC in the metropolitan area, 4 overseas. It is armed 24/7 with 300 people. Rescue missions are always free. Assistance missions must be paid. They can be contacted through VHF radio, emergency calls, satellites, telex, distress fire signals. 2018 statistics: • coordinated 12,915 missions, 53% between June and September • saved 5,577 people and assisted 14,235 people, had 293 deaths • engaged 10,434 vessels and 1,463 helicopters or planes See main administrations involved at this link. Notice how important is the SNSM for rescue boats.
19. SNSM Non profit organisation created in 1967. • Missions: sea rescue, beach surveillance, first aid at public events • 8,000 volunteers (3,350 dedicated to sea rescue), 79 paid staff • 218 boat stations: 10 to 40 people / station • Largest ship owner of France (145 boats, 400 inshore boats / jetskis) • Budget of € 21 million, 25% funded by state and 75% by private donations • Contract with MRCCs: departure in 10mn daytime / 15mn at night. Go up to 20 miles
20. What did you do? Broad goal: use data analysis and machine learning to improve SAR operations. What do they really mean? • Minimise the number of assisted / saved / dead people? • Minimise or eliminate assistance operations? (fuel error, engine problem, electricity failure etc) • Assist MRCCs when choosing what they'll send at sea? • Point out lack of boats / helicopters for specific regions? • Gather data to improve prevention campaigns? • Draft out new policies based on previous operations? • Identify upcoming risks in maritime sea rescue operations?
21. Who's interested by these improvements • MRCCs want to improve how they coordinate operations, organise trainings, spot trends in kind of operations, weather, risk factors • Insurances want to understand how the sector is growing, provide appropriate products and get a benefit in the end • Local authorities want to know what happens in their zones, undertake preventative measures, adapt buoys and signaling, spot new trends • SNSM wants to know how their rescue boats are used, financial stability • Associations and federations wants to learn more about the practice of their sport, associated risks • Maritime affairs want the sea to be safe with an acceptable budget and appropriate policies • Emergency services want more money and manpower if they go often at sea
22. What we decided It would be great to do a bit of everything. The two of us are not going to solve this. We're only here for 10 months. People need documentation, easy access to data, and training to improve how they work with data. We're going to focus on making sure processes are explained, data is made available to everyone, and existing collaborations are improved.
23. We're going to publish all sea rescue operations as open data. Online. Available to everyone.
25. Work done • Read about sea rescue conventions and big picture • Visited multiple MRCCs. Talked with officers, préfectures, sport federations, SNSM • Product decisions and planning • Open data, documentation, interactive map, BI dashboards, analysis • Blog posts, reach out to medias • Training in person and online • Visited again MRCCs to train and adjust features
26. DEMO • Interactive map: https://carte.snosan.fr • Analysis: number of operations / shift, general BI
27. How is it used? • MRCCs use it for reporting, understand their zone, how they work, training • SNOSAN (sea-rescue observatory for leisure activities) perform analysis and write policy recommendations • Sea professionals and federations use these tools to debate policy changes • Maritime affairs improve data collection and prepare upcoming system changes • Journalists sometimes perform and publish in-depth analysis • Everyone has got access to tools, documentation and data
28. Prédisauvetage implied the idea of making predictions for sea rescue. Why we didn't go this route? • Users are not used to manipulating data concepts • Missing key data: all boats at sea, ports, high confidence in sea rescue data • No operational lever: can't easily move rescue-boats, no anticipation for helicopters, can't send volunteers at sea with no reasons • Manageable number of operations. Good enough understanding of needs by workers. • Not done in other sectors (more mature, higher number of ops) like firemen And what about predictions?
29. •