Using Delta‐Generalized Additive Models to Predict Spatial Distributions and Population Abundance of Juvenile Pink Shrimp in Tampa Bay, Florida

Peter J Rubec,Erin Leone,Luke Mceachron,Christi Santi,Richard Kiltie,Richard O Flamm

doi:10.1080/19425120.2015.1084408

Abstract

AbstractIn this article, we present an approach based on generalized additive models (GAMs) to predict species’ distributions and abundance in Florida estuaries with habitat suitability modeling. Environmental data gathered by fisheries‐independent monitoring in Tampa Bay from 1998 to 2008 were interpolated to create seasonal habitat maps for temperature, salinity, and dissolved oxygen and annual maps for depth and bottom type. We used delta‐GAM models assuming either zero‐adjusted gamma or beta‐inflated‐at‐zero distributions to predict catch per unit effort (CPUE) from five habitat variables plus gear type for each estuarine species by life stage and season. Bottom type and gear type were treated as categorical predictors with reference parameterization. Three spline‐fitting procedures (the penalized B‐spline, cubic smoothing spline, and restricted cubic spline) were applied to the continuous predictors. Two additive, linear submodels on the log and logistic scales were used to predict CPUEs >0 and CPUEs = 0, respectively, across environmental gradients. The best overall model among those estimated was identified based on the lowest Akaike information criterion. A stepwise routine was used to omit continuous predictors that had little predictive power. The model developed was then applied to interpolated habitat data to predict CPUEs across the estuary using GIS. The statistical models, coupled with the use of GIS, provide a method for predicting spatial distributions and population numbers of estuarine species’ life stages. An example is presented for juvenile pink shrimp Farfantepenaeus duorarum during the summer in Tampa Bay, Florida.Received February 10, 2015; accepted August 11, 2015

Highlights

We present an approach based on generalized additive models (GAMs) to predict species’ distributions and abundance in Florida estuaries with habitat suitability modeling
Our approach used habitat suitability models (HSMs) with georeferenced catch, effort, and environmental data derived from fisheries-independent monitoring (FIM) to predict the spatial distributions and abundance of estuarine species
With the delta-type modeling approach, the positive values are fitted by a generalized linear model (GLM) or a generalized additive model (GAM), and the probabilities of observing zero values are fitted by a GLM or a GAM for a binomial distribution (Stefánsson 1996; Ye et al 2001)

Summary

Introduction

We present an approach based on generalized additive models (GAMs) to predict species’ distributions and abundance in Florida estuaries with habitat suitability modeling. Our approach used habitat suitability models (HSMs) with georeferenced catch, effort, and environmental data derived from fisheries-independent monitoring (FIM) to predict the spatial distributions and abundance of estuarine species. To account for a high proportion of zero catches in fisheries data sets, scientists have developed models that split the data into two components: probability of zero occurrence and positive CPUE (Pennington 1983; Stefánsson 1996; Fletcher et al 2005). These have been variously labeled conditional, hurdle, two-step, and delta models (Maunder and Punt 2004). Combining two submodels complicates model interpretation, deltalognormal GLM models have been widely used to estimate bycatch and interannual indices of abundance (Maunder and Punt 2004)

Objectives

Methods

Results

Conclusion