Evaluating long-term monitoring of temperate reef fishes: A simulation testing framework to compare methods

Abstract

A simulation testing framework was developed to evaluate the efficacy of detecting population trends of two sampling methods used to monitor inshore fish populations: angling and baited remote underwater stereo-video systems (stereo-BRUVs). The study is based on data collected as part of a long-term monitoring program in the Tsitsikamma National Park marine protected area, South Africa. As a test scenario, declining population trajectories of the most abundant species, Chrysoblephus laticeps, were simulated by introducing consecutive years of reduced recruitment over periods of 10 and 20 years applying an age-structured operating model. The operating model was designed to generate method-specific relative abundance indices and length–frequency data, using parameters derived from existing data collected in the long-term monitoring program. These were then fitted with an age-structured estimation model. Estimated spawner-biomass depletion was compared to the ‘true’ simulated population to quantify method-specific accuracy and bias using root-mean-squared error. Due to higher data variability and inherent size selectivity of angling, stereo-BRUVs provided more accurate spawner-biomass trends when describing a distinct population decline over 10 and 20 years. Additionally, spawner-biomass was found to be a more accurate population estimate than relative abundance indices due to the inclusion of population size structure information. The study demonstrates the potential of using simulation testing to evaluate sampling methods, given that the process generates the ‘true’ population with a known abundance and size structure.