Data up to 2016. Landings of vermillion snapper in the US southeastern Atlantic represent approximately 10% of total Vermillion snapper landings in FAO Area 31. The catch-age model included data from five fleets that caught vermilion snapper in southeastern U.S. waters: commercial handlines, commercial historic trawl, commercial other (pots, traps, diving, trawl, miscellaneous), recreational headboat, and general recreational. The model was fitted to data on annual landings (in numbers for the recreational fleets, in whole weight for commercial fleets); annual discards (in numbers, with a 0.41 release mortality rate applied for commercial lines and a rate of 0.38 applied for recreational discards); annual length compositions of landings and discards; annual age compositions of landings and the SERFS; three fishery dependent indices of abundance (commercial handline, headboat, MRFSS); and two fishery-independent indices of abundance (SERFS combined chevron trap and video gears, MARMAP Florida snapper trap). Several changes to the 2012 update assessment data sources were made for SEDAR55. Reproductive inputs to the assessment model (age at maturity, batch fecundity, spawning frequency) were updated based on new data (SEDAR55-WP07 2017). Age compositions from the SERFS trap data were extended back to 1990 whereas age compositions in the 2012 update began in 2002. The SEDAR55 assessment panel recommended using the FHWAR method rather than the SWAS method to develop historical recreational removals (SEDAR55-WP04 2017). The headboat and general recreational length compositions and the commercial length compositions (after 1992 when ages became available) were excluded by the SEDAR55 assessment panel due to conflicts with the available age data. Finally, the SERFS video data were included along with the chevron trap data for creating the fishery independent index. The video survey was initiated after the 2012 update assessment, and it was considered here as part of the TOR.

• Further investigate discrepancies between age composition data and indices of abundance. • Further develop methods to standardize and combine SERFS chevron trap and video gears for creating indices of abundance. • Evaluate sample size cutoffs and weighting procedures for age and length compositions. What should be the minimum standards, and how does this interplay with the number of age and length classes modeled in the assessment? • In stock assessment, various likelihood formulations have been used for fitting age and length composition data. The multinomial distribution and its robust versions have been the most widely applied. However, more recently the Dirichlet-multinomial and logistic-normal have attracted attention. A simulation study could shed light on the performance of these various likelihood formulations under sampling conditions realistic in the southeast U.S. • Vermilion snapper were modeled in this assessment as a unit stock off the southeastern U.S. For any stock, variation in exploitation and life-history characteristics might be expected at finer geographic scales. Modeling such sub-stock structure would require more data, such as information on the movements and migrations of adults and juveniles, as well as spatial patterns of larval dispersal and recruitment. Even when fine-scale spatial structure exists, incorporating it into a model may or may not lead to better assessment results (e.g., greater precision, less bias). Spatial structure in a vermilion snapper assessment model might range from the very broad (e.g., a single Atlantic stock) to the very narrow (e.g., a connected network of meta-populations living on individual reefs). What is the optimal level of spatial structure to model in an assessment of snappergrouper species such as vermilion snapper? Are there well defined zoogeographic breaks (e.g., Florida keys, Cape Hatteras) that should define stock structure? How much connectivity exists between the Gulf of Mexico and Atlantic stocks?