Provider: Ingenta Connect
Database: Ingenta Connect
Content: application/x-research-info-systems
TY - ABST
AU - Walsh, William A.
AU - Howell, Evan A.
AU - Bigelow, Keith A.
AU - McCracken, Marti L.
TI - Analyses of observed longline catches of blue marlin, Makaira nigricans, using generalized additive models with operational and environmental predictors
JO - Bulletin of Marine Science
PY - 2006-11-01T00:00:00///
VL - 79
IS - 3
SP - 607
EP - 622
N2 - Five generalized additive models (GAMs) were developed and evaluated to analyze blue marlin, *Makaira nigricans* Lacépède, 1802 catches reported by fishery observers in the hawaii-based longline fishery from March 1994 through February 2004. The coefficients from
three gAMs were applied to the corresponding predictor variables in logbooks from unobserved fishing trips to predict catches (n = 97,557 longline sets). Results demonstrated that application of an overparameterized GAM (7 predictors, 40 degrees of freedom per predictor) yielded an inaccurate
(26.6% greater than corrected logbook data) and imprecise (breadth of 95% prediction interval (PI): 105% of the point estimate) unobserved catch estimate. The same operational and environmental predictors allotted 66%–70% fewer degrees of freedom predicted unobserved catches accurately
(apparent error:−3.4%, 6.5%) and with reasonable precision (breadth of 95% prediction intervals: 21%, 24% of the estimated catch). Several extrinsic factors (e.g., hooks per float) not previously evaluated were significantly associated with blue marlin catches when used in another explanatory
GAM. Results are discussed relative to the need to utilize fishery observer data effectively with billfishes, which are generally taken as by-catch or incidental catch in longline fisheries, because observer data are likely to be the most accurate available information.
UR - http://www.ingentaconnect.com/content/umrsmas/bullmar/2006/00000079/00000003/art00014
ER -