Biodiversity assessment in fisheries plays a crucial role in the conservation of marine ecosystem diversity. Accurate survey data serve as the foundation for ensuring the precision of quantitative analysis and the effectiveness of resource conservation measures. However, conducting fishery resource surveys in the ocean is costly and constrained by on-site conditions. Therefore, meticulous sampling design is a key factor in maximizing data quality and survey efficiency. While increasing the number of sampling stations can enhance result accuracy, it also escalates survey costs and may have adverse effects on the marine environment and ecosystems. Hence, sampling design needs to strike a balance between sample size and precision to meet management objectives and budget constraints. Optimizing fishery resource survey sampling stations design is widely recognized as an effective approach to enhance survey precision, and there have been numerous studies conducted on this topic both domestically and internationally. And recent studies have indicated that stratified sampling offers higher precision, making it a preferred choice for optimizing fishery resource survey site selection. However, most current research primarily focuses on optimizing stratification or sample allocation and pays less attention to methods that simultaneously optimize both stratification and sample allocation. This paper aims to provide valuable insights into the simultaneous optimization of stratification and sample allocation for biodiversity-focused fisheries surveys.
In this study, we take the Maoming as a case study and employs a method provided by the R package "SamplingStrata" to optimize the sampling design through stratified sampling under a multivariate scenario. This package, based on the use of a genetic algorithm, can determine the optimal stratification, sample size, and sample allocation that meet precision constraints in the presence of multiple stratification variables and multiple target variables. When using this package, it is essential to clearly define the stratification variables and target variables while predefining the precision requirements for the target variables. To maximize the efficiency of the samples, the selection of stratification variables should be based on their correlation with the target variables. Choosing stratification variables that are most correlated with the target variables can enhance the representativeness of the samples. Additionally, the precision requirements are expressed using the Coefficient of Variation (CV) for each target variable. The CV value reflects the magnitude of the sample estimate variance relative to its mean for each target variable.
In this paper, we use environmental data (water depth, dissolved oxygen, water temperature, pH, salinity) and water depth as stratification variables, and Shannon-Wiener diversity index (H′), Margalef richness index (D), Pielou evenness index (J′), and species number as target variables. Sequentially, the same maximum coefficient of variation (CV) is set for the target variables at 0.2, 0.15, 0.1, 0.05, 0.04, 0.03, 0.02, 0.01. Subsequently, samples were selected from the optimized stratification design, and the relative errors were calculated.
Results showed that biodiversity in the coastal waters of Maoming exhibits seasonal differences, with diversity indices being higher in autumn than in spring. Additionally, the number of sampling sites increases as the maximum coefficient of variation (CV) decreases. However, when the maximum coefficient of variation continues to decrease from 0.05, the number of sampling sites increases more significantly with each reduction of 0.01 CV, requiring one or more additional stations. At CV = 0.05, when stratified by environmental factors, spring requires 8 sampling points, while autumn and non-seasonal conditions require 7 and 11 sampling points, respectively. When stratified by water depth, spring requires 12 sampling points, while autumn and non-seasonal conditions require 14 and 21 sampling points, respectively. According to relative error analysis, when stratified by environmental factors and water depth, the mean relative errors for each index are 2.38, 2.02, 2.85, and 0.43, 3.14, 1.74, respectively, with stratification by water depth resulting in smaller mean relative errors for all indices except evenness.
In general, through stratified sampling and appropriate sample allocation, it is possible to reduce the number of sampling sites while maintaining data accuracy, thus minimizing the adverse impact on marine ecosystems. In the coastal waters of Maoming survey, setting a maximum coefficient of variation equal to 0.05 as the data precision requirement and using depth as the stratified variable prove to be suitable. The number of sampling sites for spring, autumn, and non-season-specific surveys is found to be 12, 14, and 21, respectively. This paper optimizes the fishery resource survey stations selection in the coastal waters of Maoming, offering an effective sampling stations optimization method for fishery resource surveys and providing guidance for future fishery resource survey stations design. |