Generalized additive models for categorical count data: An exploration of the decline of queen triggerfish Balistes vetula in the Bahamas and Turks and Caicos

Abstract

AbstractCitizen science is growing in importance for ecosystem management and long‐term monitoring. A large marine citizen‐science project operated by the Reef Environmental Education Foundation (2021, Reef environmental education foundation volunteer fish survey project database, World Wide Web electronic publication) collected logarithmic categorical data for species abundance across a number of otherwise understudied reefs in The Bahamas and Turks and Caicos during 1994–2020. We used several statistical models to estimate the presence and abundance of trends from these data. Variously specified abundance and presence‐absence models were fit to simulated count data, simulated categorized count data, and real‐world categorical data for Queen Triggerfish (Balistes vetula). These models produced simple patterns of presence and abundance from simulated data with minimal bias that were reasonable predictions based on cross‐validation. Based on model‐based estimates of presence and abundance, the Queen Triggerfish population decreased significantly in The Bahamas and Turks and Caicos during 1994–2020. This simple method for imputing abundance from size‐category counts at the level of individual diver observations, rather than aggregated across multiple observations, allows for higher resolution modeling of predictors of presence and abundance, with implications for other understudied reef‐dwelling species.