Accurate assessment of marine biodiversity is fundamental to fisheries management and ecological conservation. Optimizing sampling design is crucial for improving the estimation precision of species richness as an essential biodiversity indicator. Based on seasonal bottom trawl survey data collected from August 2014 to May 2015 in the coastal waters of Guangdong, this study employed computer-simulated resampling techniques to compare the effects of simple random sampling (SRS) and stratified random sampling (StRS) on the estimation of fishery species richness under varying numbers of sampling sites and frequencies. The results showed that the species detection rate increased rapidly with the addition of sampling sites when their number was low, and the increasing rate gradually slowed as the number of sites increased. To achieve a 90% species detection rate, 44 sampling sites were required for year-round seasonal sampling; for an 80% detection rate, the minimum number of sites needed was 36, with the summer-autumn-winter or summer-autumn-spring seasonal combinations being the most efficient. Both the absolute values of relative estimation error (REE) and relative bias (RB) decreased as the number of sampling sites increased. StRS exhibited significantly lower absolute REE and RB values than SRS (P＜0.05). After excluding rare species, the number of sites required to achieve a 90% detection rate decreased to 20, and the absolute REE and RB values also declined, indicating that rare species increased sampling variability. The design effects (De) of 15 stratified sampling combinations were all below 1, demonstrating that StRS had higher sampling efficiency in Guangdong coastal fisheries surveys, particularly when the number of sampling sites was limited. This study provides a scientific basis for optimizing sampling designs in Guangdong coastal fisheries surveys, and suggests selecting site numbers and seasonal combinations based on monitoring precision requirements and cost constraints.