Many investigators sample a site or watershed only once and use that sample to characterize the location (e.g., indices of biotic integrity studies). Using real-world fish assemblages, we investigated the adequacy of using a single sample, site, or season to characterize the fish community at a site or watershed. We sampled twelve sites from the Little Choctawhatchee River watershed (416 km²) by electrofishing six times at each location over 18 months (n = 72). We estimated species richness at each site during each season and for the entire watershed using the Chao 2 estimator. We used these estimates to determine how many samples at the local level and how many sites at the watershed level were needed. Additionally, we determined the proportion of species found after 1–6 samples. Our results showed that to observe 80–100% of species, 5.6–47.4 samples, respectively, were needed at the local level over six seasons, 8.4–86.9 sites, respectively, were needed at the watershed level over a single season, and 9–70 samples, respectively, were needed for the entire watershed over six seasons. At the local level, we found a mean of 50–83% of the estimated species with 1–6 samples over six seasons. At the watershed level, we found a mean of 37% (SD = 3) with one sample and 76% (SD = 6) after six samples over one season. Our results indicate that in general a minimum of 5 samples is needed to detect at least 80% of the species present at the local (site) level, and a minimum of 8 sites is needed in watersheds of approximately 400–450 km².