Identifying information gaps in predicting winter foraging habitat for juvenile Gulf Sturgeon

The Gulf Sturgeon Acipenser oxyrinchus desotoi is an anadromous species that inhabits Gulf of Mexico coastal waters from Louisiana to Florida and is listed as threatened under the U.S. Endangered Species Act. Seasonal cues (e.g., freshwater discharge) determine the timing of spawning and migration and may influence the availability of critical habitat during winter months in six estuaries. Large information gaps, especially related to critical estuarine habitat for juveniles, hinder recovery efforts to protect these habitats and assess risks from emerging threats. Using Apalachicola Bay, Florida, as a model system, we developed and analyzed a preliminary Bayesian network model so that we could identify knowledge gaps (i.e., where expert knowledge was lacking) and data gaps (i.e., where data were unavailable) that limit the ability to assess the quantity of critical estuarine habitat for juvenile Gulf Sturgeon. The model hypothesized habitat availability per winter month in estuarine habitat under alternative scenarios of river discharge and length of the winter foraging season. A search for geospatial data sets revealed that the largest gap involved salinity, temperature, and oxygen (i.e., water condition) monitoring data, with data available only for Apalachicola Bay. For the Apalachicola Bay model, data gaps prevented the development of 53% of water condition geospatial data sets and a sensitivity analysis showed that water condition data most limited the ability to predict habitat availability. Expert knowledge was low, and conditional certainty scores showed that the relationships with the lowest certainty were abiotic suitability and habitat availability. Reducing information gaps could aid the development of a model that is appropriate for informing management. Future efforts could prioritize the expansion of water monitoring within critical habitat estuaries and predicting abiotic suitability and habitat availability. Bayesian network models can easily incorporate prior and new information for complex systems. Thus, our model could be updated as future research and monitoring efforts close these information gaps.