A comparative study of optimization methods and conventional methods for sampling design in fishery-independent surveys

We have introduced and evaluated a procedure, the constrained spatial simulated annealing method, for developing an optimal sampling design for fishery-independent surveys. We used two criterion functions, minimization of the mean of the shortest distance (MMSD) and uniform distribution of point pairs for variogram estimation (WM), and three arrangements of the two criteria, all WM, all MMSD, and a combination of MMSD (2/3 of samples) and WM (1/3), to construct three optimized sampling designs (denoted as Designs I, II, and III, respectively). These three designs were compared in a simulation study with systematic sampling (Design IV) and stratified random sampling designs (Design V), commonly used in fishery-independent surveys. Three levels of sample size (small, medium, and large) were considered in the simulation study developed using a geostatistical approach. The results showed that for parameter estimation of the spatial covariance function, Design III was better than the other designs at relatively small sample size and Design II performed better than the other designs at relatively large sample size. For estimating fish stock abundance, the performance of the designs considered in this study can be ranked as follows: Design II > Design IV > Design III > Design V > Design I. It is clearly important to evaluate and improve sampling design based on historical survey data. Such a study allows us to identify an optimal sampling design to balance the quality of the data collected and the costs of the sampling programme, leading to the development and optimization of a sustainable and fishery-independent monitoring programme.