Integrated Bayesian models to estimate bycatch of sea turtles in the Gulf of Mexico and southeastern U.S. Atlantic coast shrimp otter trawl fishery

The Shrimp Trawl Observer Program, which is administered by NOAA Fisheries Southeast Fisheries Science Center (SEFSC) Galveston Laboratory, assigns observers to shrimp otter trawl vessels in both the Gulf of Mexico (GOM) and the waters off the U.S. east coast within the jurisdiction of the South Atlantic Fishery Management Council (SATL), and vessel participation has been mandatory since 2007. We applied integrated Bayesian models to the observer data to estimate sea turtle bycatch. We also estimated mortality, defined as the total number of sea turtles that were caught in shrimp trawls and died at the time of capture. The total bycatch mortality was estimated by multiplying the probability of mortality for turtles caught in shrimp trawl nets by the total bycatch estimated from a linear model of catch per unit effort (CPUE) per strata (area, season, depth zone, time period) multiplied by the total effort in each stratum. For rare species, the bycatch rate was estimated by modeling CPUE of all species together, and multiplying this by the species composition to get species-specific CPUE. Total bycatch mortality was estimated separately for the GOM and the SATL, and for standard shrimp otter trawl nets versus “try” nets, which are small nets fishers deploy in front of the primary nets to test catch rates. About 30% of seaturtles caught in standard nets were dead, while less than 1% of sea turtles caught in try nets were dead. Thus, although many Kemp’s ridley, loggerhead, and green sea turtles were caught in try nets in both regions, few of them were killed. For example, in the GOM in 2015, we estimated 95% credible intervals of 54-256 Kemp’s ridley, 173-495 loggerhead and 22-114 green sea turtles caught in try nets, but only 0-7 Kemp’s ridley, 0-17 loggerheads, and 0-3 green sea turtles were estimated to be killed. On the other hand, for standard nets in the GOM, we estimated 95% credible intervals of 63-369 Kemp’s ridley, 18-105 loggerhead and 75-226 green sea turtles captured, corresponding to mortality of 19-130 Kemp’s ridley, 5-36 loggerhead and 22-81 green sea turtles killed. In addition, we found 95% credible intervals of 24-99 turtles classified as unknown/other species killed in standard nets in the GOM. The unknown category includes sea turtles that were not identified by the observers, as well as leatherback and hawksbill sea turtles, which could not be modeled separately because there were only 3 leatherbacks and 1 hawksbill recorded. Total bycatch mortality of Kemp’s ridley and loggerhead sea turtles in standard nets decreased from 2007 to 2015 in the GOM, but green sea turtle bycatch stayed constant. In the SATL, the sample sizes were lower, fewer turtles were observed, and total effort was not as well estimated. Thus, total bycatch rates were estimated with wide credible intervals. There was no trend over time in the SATL from 2007 to 2016, and total mortality in standard nets in 2016 was on the order of 5-111 Kemp’s ridley turtles, 9-139 loggerhead turtles, 2-86 green sea turtles, and 13-168 of unknown/other species of turtles. These estimates of total bycatch mortality include only sea turtles that are dead at the time of capture; there may be additional mortality due to stress or injuries after live release.