Occurrence and Predicted Dispersal of Bighead Carp in the Mississippi River System: Development of a Heuristic Tool
Martin T. O’Connell, Ann U. O’Connell, and Valerie A. Barko
Abstract.—Bighead carp Hypophthalmichthys nobilis have become established within the Mississippi River system (MRS) and pose a serious threat to native fishes and aquatic ecosystems throughout North America. Determining their dispersal dynamics will be a key management tool for controlling their expansion. To better understand how bighead carp have spread through the MRS, we developed a simple diffusion model to be used as a heuristic tool to generate insights regarding dispersion patterns. First, we collected occurrence data from fish museums and government agencies spanning more than 30 years of sampling in the MRS and nearby rivers. These were then combined into a geographic information system database and used to create yearly occurrence maps for this species. We then developed a diffusion model for bighead carp using information on their movement and reproduction. The resulting model can be used to track the dispersal of hypothetical carp populations from different points of introduction within the MRS. With this model, we generated and compared four possible dispersal scenarios for bighead carp based on likely points of introduction. For each of these, we calculated Cohen’s kappa and sensitivity (measures of predictive success) to determine which dispersal scenarios were the most accurate in predicting bighead carp occurrence pattern after 30 years. We found significant agreement between the actual and predicted distributions of carp after 30 years of expansion for all four dispersal scenarios (Cohen’s kappa: range = 0.136–0.426, p < 0.05). The single introduction scenario (in the Arkansas River) had the lowest agreement with the occurrence data (Cohen’s kappa = 0.136, sensitivity = 32%) compared with the scenarios representing multiple points of carp introduction: introductions in the Arkansas and Missouri rivers (Cohen’s kappa = 0.370, sensitivity = 71%), introductions in the Arkansas and lower Ohio rivers (Cohen’s kappa = 0.391, sensitivity = 68%), and introductions in all three river locations (Cohen’s kappa = 0.426, sensitivity = 85%). The triple introduction scenario also had the highest sensitivity (sensitivity = 66%) when it was compared to the other three scenarios for accuracy on a year-to-year basis: Arkansas River only (sensitivity = 25%), Arkansas and Missouri rivers (sensitivity = 39%), and Arkansas and lower Ohio rivers (sensitivity = 46%). These results suggest expanding bighead carp populations in the MRS began from multiple origins rather than a single introduction. Other insights derived from the dispersal scenarios include evidence that bighead carp are possibly more widely dispersed than current occurrence data indicate and that the species is likely extending its range “under the radar” of standardized sampling. Finally, we used these dispersal scenarios to predict potential high carp density hot spots that could develop over the next 20 years in the MRS and should be targeted for control management.