Comparing three indices of catch per unit effort using Bayesian geostatistics

Comparison between univariate and bivariate geostatistical models for estimating catch per unit of effort (cpue): A simulation study

“…The following bivariate spatial model, proposed by Pereira et al (2009) for the catch and effort joint modeling, was compared to a bivariate model without a spatial component.…”

“…(3) was the linear coregionalization model proposed by Gelfand et al (2004) and used by Pereira et al (2009). For this model one has a range (i.e., the distance beyond which there is practically no spatial correlation between data points) associated with each component of Y(·) if monotonic and isotropical correlation functions are used.…”

“…Provided catch and effort data are available for every sub-region within a fish species distribution, three cpue indices may be defined (cpue 1 , cpue 2 , cpue 3 ). Petrere et al (2010) and Pereira et al (2009) provide the following definitions: cpue 1 = C i /f i /n=(C/f ) which is the mean of ratios between catch and effort in different quadrats; cpue 2 = C i / f i =C/f which is the ratio between the total catch across all quadrats and the total effort across all quadrats; and cpue 3 = C i · f i / f 2 i = C · f /f 2 which is the ratio between the sum of the products of the catch by the effort in different quadrats, and the sum of the squares of the efforts.…”

“…the quantity C/f. However, if the data available are not simply represented as C/f but can be georeferenced as catch and effort pairs (C, f), an alternative would be to model the pairs of variates (C, f), as proposed by Pereira et al (2009). The current paper addresses this, namely: is it advantageous, when estimating an index of type cpue 1 , to model the pair (C, f) instead of the ratio C/f when imputing values for quadrats that were not observed?…”

“…Pereira et al (2009) used a bivariate model for (C, f), which was compared with another bivariate model that did not consider the spatial dependence between C and f. Geostatistical multivariate models in such studies have demonstrated a superior performance when compared to models that do not consider the spatial correlation; and it has also been shown that such multivariate geostatistical models are superior to those geostatistical models where the variates are mutually independent. Therefore the advantage of using a multivariate geostatistical model may be justified because such a model accounts for the cross covariance structure as well as the spatial dependence for each variable (C and f).…”