Power and precision of invasive carp ichthyoplankton sampling along an invasion front

ABSTRACT Objective Detecting changes in the abundance of invasive fish species is valuable for assessing population trends and creating management plans. Although many programs focus on adult relative abundance to estimate population dynamics, assessments of early life stages can provide insights into reproductive success that are not detectable from adult surveys. However, early ichthyoplankton catches are often highly variable and guidelines are needed on detectable percentages of change and the sample sizes that are required for desired precision. We evaluated larval surveys of invasive carp (Silver Carp Hypophthalmichthys molitrix and Bighead Carp H. nobilis) across 7 years in the upper Mississippi River basin to assess the statistical precision and effectiveness of detecting fluctuations in the relative abundance of invasive carp larvae. Methods We sampled invasive carp in the upper Mississippi River and five tributaries from 2014 to 2019 and 2021. Sampling occurred biweekly from April through September at 30 sites spanning established, invasion, and presence fronts of the current carp distribution. Each site was sampled with three larval tows using a 0.5-m-diameter ichthyoplankton net. We evaluated the maximum annual larval densities (high-output spawning events) and mean densities (baseline spawning activity), calculating the coefficient of variation (CV) as our precision metric. We used generalized linear models to test how environmental factors affected sampling variability. We determined the relationship between the sample means and variability and then calculated the sample sizes that are needed to achieve CV = 0.30. Power analyses estimated the sampling effort that would be required to detect percentages of change in larval densities. Results We captured invasive carp larvae in 72 of 4,462 larval tows. The mean larval carp densities ranged from 1.4 to 2,869.3 carp/100 m3 (mean = 77.3, median = 7.5), with mean interreplicate variance of 60,461. Three tows per sample returned a CV of 1.46 (SE = 0.06), indicating low precision. Achieving CV = 0.30 required an average of 57 samples. Sampling CV was positively associated with river discharge and negatively associated with water temperature. Conclusions Sampling exhibited extreme variability with zero-inflated data, as most tows captured no larvae and 50% of the successful captures detected larvae in only one of three replicates. Environmental factors significantly influenced precision, with CV positively associated with discharge and negatively with temperature. Sampling location affected precision, with the lowest CV in established populations and the highest in invasion areas. Achieving acceptable precision (CV < 0.30) required approximately 60 tows per site. The values for maximum annual densities provided better statistical power than did those for the mean densities, reducing the required sample sizes by nearly 50% at some sites. However, both metrics showed low statistical power, with required sampling effort largely unattainable. We recommend that managers consider these statistical challenges when designing monitoring programs and supplement ichthyoplankton sampling with complementary methods like environmental DNA to improve detection efficiency, particularly in invasion-front areas.