Abstract
We developed an efficient sampling design‐based approach using fishery‐independent surveys to estimate population abundance of pink shrimp Penaeus duorarum over time in Biscayne Bay, Florida. We initially implemented quarterly stratified random sampling (StRS) using nine habitat strata and determined that average pink shrimp density (numbers/m2) was highest in late fall and lowest in spring and late summer. Coefficient of variation of the quarterly surveys, expressed as percent standard error/mean density, ranged from 5.8% to 14.3%. We found StRS to be more efficient (i.e., with lower variance) than simple random sampling (SRS) in most seasons. Statistical analyses suggested that pink shrimp densities were dependent on the biophysical habitat variables of bottom substrate, depth, and salinity. We also noted ontogenetic shifts in these relationships that were particularly pronounced at the onset of sexual maturation. Poststratification analysis was used to further evaluate several alternative habitat‐based sampling schemes. Results showed that a five‐strata composite design that used all three habitat variables was similar in performance, but less complex, than the original nine‐strata design. In addition, the composite design outperformed both SRS and all other StRS designs indexed on single habitat variables. The new five‐strata composite design was implemented in late summer 1997 and achieved a significant reduction in coefficient of variation compared with the late summer 1996 survey. This new design did not perform as well as expected in late fall 1997, which we attribute to a mismatch between our seasonal sample allocation strategy and the timing of pink shrimp recruitment into Biscayne Bay in 1997. Finally, we show how statistical sampling designs that use stratifications based on relevant habitat covariates can yield high‐precision abundance estimates at low costs and provide a robust quantitative methodology for identifying habitat essential to fisheries production.