Modelling historical distribution of trawling effort British steam trawlers in the North Sea
Motivation Why study fleet dynamics? North Sea: dramatic changes fish abundance For understanding, not enough to focus only on fish biology Adults of many fish stocks: F >> M More understanding needed of fish main predator fishing fleets 2
Motivation Why study historical fleet dynamics? We have extensive historical data, under-utilised Historical commercial data more reliable than current quota restrictions, mis-reporting, discarding, compliance Strong political effect may now drive effort distribution; historically, more purely market-driven Tackle old questions using new methods Past is key to present 3
Statistical Charts Catch, catch rate, effort By year month rectangle species North Sea 1913-81 except war years Other regions post-war years until 1981 4
Case study North Sea steam trawlers pre- and post-war English steam trawlers Before WWII: large fleet low stock levels After WWII: reduced fleet high stock levels 5
Specific research question Why do fishers trawl where, when and how much? Can we model distribution of trawling effort? Effort (hr fished) by rectangle explained by: Fishers habit previous year s effort Distance to port transport, time costs Cpue fish price = Vpue value per unit effort Other variables climate, total stock levels, vessel tonnage 6
The data Landings Catch rate Effort 5 main species: cod, haddock, plaice, sole, whiting 7
The data a limitation only global effort data for the entire steam trawler fleet no individual trip data for vessels separately 8
Initial results graphic exploration Distribution effort previous year and current year Habit, distance to port, or catch rates? 9
Initial results graphic exploration Distance to port 100% 80% 60% 40% 20% 0% 1928 1938 1947 1957 (Ramsgate) (Lond. Bill.) Lowestoft (Great Yarmouth) Grimsby Hull (Bridlington) Scarborough (Whitby) Hartlepool (Sunderland) North Shields Leith Aberdeen Landings in main ports by steam trawlers By far most important: Hull & Grimsby Aberdeen 2 distance variables in model 10
Initial results graphic exploration Distance to port Distance to Hull/Grimsby Pr. effort H/Gr trawlers Pr. effort H/Gr/Ab trawlers Obs. effort trawlers Latitude Latitude Latitude Latitude 1950 Longitude Longitude Longitude Longitude Distance is important, but how can it best be modelled? Effort 11 Area Fleet ~ Distance 2
Initial results graphic exploration Catch rates (cpue) 5 principal species Cod Haddock Plaice Sole Whiting 1950 But price differences need to be accounted for 12
Initial results graphic exploration Value per unit effort (vpue) 5 principal species Cod Haddock Plaice Sole Whiting 1950 1950 Species-specific vpue s can be summed index of profitability of rectangles 13
Initial results graphic exploration Total cpue and total vpue Cpue 5 species Vpue 5 species 1950 1950: cpue and vpue ~equally distributed North Sea more high-price fish in S than N (plaice, sole) little contrast in data, difficult to assess effects on effort How are temporal trends? account for retail price index 14
Trends cpue and vpue (corrected for retail price index) Cpue 1928 Vpue 1928 Cpue 1938 Vpue 1938 15 Cpue 1947 Vpue 1947 Cpue 1957 Vpue 1957
Initial results modelling GLM effort r, y+1 = (vessels y+1 / vessels y ) (α effort r,y +* β vpue r,y +* γ distance r 2 ) + ε log-transformed ln effort r, y+1 = ln (vessels y+1 / vessels y ) + α ln effort r,y + β ln vpue r,y + γ ln distance r + ε Refining of model construction currently in progress 16
Initial results modelling Effort y+1 GLM Effect of Effort y on Effort y+1 very closely related may suggest conservative behaviour fishermen multicollinearity maybe model change in effort? Distance to port (if effort y excluded): generally (quite) strong negative effects Hull/Grimsby > Aberdeen (in accordance with fleet size) suggests fuel, time constraints are strong 17
Initial results modelling Effort y+1 Vpue Effects of Vpue (if previous year s effort excluded from model): 1928, 31 no effect 1937 positive 1947, 48 no effect 1950, 56 positive 1954 no effect 1955 (positive trend) 1977 no effect Note low contrast in vpue less likely to find effects Should species be accounted for? do fishers target species or total catch composition? Should more species be included? candidate species with high value hake, turbot, brill may significantly affect vpue distribution 18
Acknowledgements For information on statistical charts, see: www.cefas.co.uk/publications/techrep/tech128.pdf Feedback by many, including Laurie Kell Colin Bannister Trevor Hutton Beatriz Roel John Pinnegar Cefas library staff In commemoration of my Father and his tales about life at sea 19