Modelling fuel consumption of fishing vessels for predictive use. (10th June 2014)

• Record Type: Journal Article
• Authors: Davie, Sarah; Minto, Cóilín; Officer, Rick; Lordan, Colm; Jackson, Emmet
• Is Part Of: ICES journal of marine science. Volume 72:Number 2(2015)
• Page Start: 708
• Page End: 719
• Publication Date: 2014-06-10
• Subjects: fishing gear -- fishing vessels -- fuel consumption -- fuel cost predictions -- fuel price -- GAM -- modelling; Ocean -- Periodicals; Fisheries -- Periodicals; Fishes -- Periodicals; Marine biology -- Bibliography -- Periodicals; 551.4605
• DOI: 10.1093/icesjms/fsu084
• Languages: English
• ISSNs: 1054-3139

Abstract

Abstract: Fuel costs are an important element in models used to analyse and predict fisher behaviour for application within the wider mixed fisheries and ecosystem approaches to management. This investigation explored the predictive capability of linear and generalized additive models (GAMs) in providing daily fuel consumption estimates for fishing vessels given knowledge of their length, engine power, fleet segment (annual dominant gear type), and fuel prices. Models were fitted to half of the Irish fishing vessel economic data collected between 2003 and 2011. The predictive capabilities of the seven best models were validated against the remaining, previously un-modelled, data. The type of gear used by a fleet segment had an important influence on fuel consumption as did the price of fuel. The passive pot gear and Scottish seine gear segments indicated consistently lower consumptions, whereas dredge and pelagic gears showed consistently higher fuel consumptions. Furthermore, increasing fuel price negatively affected fuel consumption, especially for more powerful, larger vessels. Of the formulated models, the best fit to training data were a GAM with a gear main effect and two smooth functions; standardized vessel length and engine power interacting with fuel price. For prediction, overall, this model showed the closest predictions with the least bias, followed by three linear models. However, all seven models compared for predictive capability performed well for the most sampled segments (demersal and pelagic trawlers).