Modelling intra-annual variation in abundance of squid Loligo forbesi in Scottish waters using generalised additive models

This paper reports a preliminary application of generalised additive models (GAMs) to describing intra-annual variation in squid abundance in Scottish waters. Models were constructed using data for the years 1983, 1988, 1989 and 1991. Cubic spline smoothers and Gaussian distributions were used to build up the models. Since reliability of local fishery abundance estimates is arguably proportional to fishing (sampling) effort, in addition to standard (unweighted) models, further models incorporating weighting by fishing effort were also fitted. The variables finally included in the GAMs were sea surface temperature (SST) and month. In the weighted models there is no temperature effect on abundance below 13 degreesC whereas the unweighted models show peaks of abundance in the temperature range 8-13 degreesC. Possible reasons for this discrepancy, and advantages and disadvantages of applying GAMs to data from this fishery are described. Further models are suggested including a two-stage procedure to deal the large number of zero observations in the abundance data. (C) 2001 Elsevier Science B.V. All rights reserved.