The South Atlantic Red Snapper Research Program

Congress indicated in the legislation that established the South Atlantic Red Snapper Research Program that age-2+ Atlantic red snapper “absolute abundance” needed to be estimated. However, “absolute abundance” simply means total abundance or population size in the study region. The word “absolute” distinguishes this concept from “relative abundance,” which only tracks population trends (increases or decreases) over time, without providing information on the actual population size.

There are many approaches in population ecology that can be used to estimate fish population size. A fishery stock assessment is a statistical model that utilizes landings data, fishery catch rate trends, life history data (e.g., growth, natural mortality, reproductive biology parameters, etc.), and fishery-independent relative abundance trends to track year classes or cohorts across time, estimate fishing mortality rates, and estimate spawning stock biomass. Notably, mark-recapture studies can also be utilized to estimate population size, including genetic mark-recapture like the CKMR approach in the SARSRP study.

Yet another approach to estimating population size is estimating the density (individuals per area) of fish in specific habitats and then scaling those density estimates to a population estimate based on estimates of the habitat distribution in the study region. This latter approach is the general approach of the Hierarchical Bayesian Integrated Modeling component of the SARSRP study on the Atlantic shelf from North Carolina to the Florida Keys, as well as the approach previously taken in the Gulf of Mexico Great Red Snapper Count study. It should be stressed that in neither study was there (or will there be) an attempt made to actually count each individual, or census, red snapper in the study regions.

There are two approaches being undertaken by the study team to estimate age-2+ red snapper population size (i.e., abundance or “absolute abundance”) in the US Atlantic Ocean. Both approaches will produce population size estimates independent of the SEDAR stock assessment for Atlantic red snapper. The word independent in this case means separate from the assessment, which will be done with approaches other than stock assessment modeling. However, to make the estimates meaningful for fishery management, they will eventually need to be integrated into the stock assessment model itself, which is the final objective of the overall study.

Mark-recapture is a population estimation technique in which an artificial mark is applied, or a natural mark is detected, on individuals within a population of interest. Then, the marked individuals are allowed to mix with unmarked individuals. In a subsequent sampling event, marked as well as unmarked fish are captured and the proportion of recaptured marks in this second sample of individuals is assessed. If 100 fish were originally captured and tagged, and then during the second sampling event 10 tagged fish were recaptured among 100 total fish (i.e., 90 did not have marks), then 0.1 is the proportion of tagged fish in the second sample.

Dividing the number of tagged fish originally released in the population by this recapture proportion (i.e., 100/0.1) produces an estimate of population size (N = 1,000). This is a simple example and omits various assumptions that need to be met (e.g., tagged fish are well-mixed in population, fish survive after being tagged, tags are retained, tags are recognized and reported when recaptured, etc.), but it illustrates the general concept of mark-recapture. The lower the proportion of tags recaptured in the second sample, the higher the estimated population size.

In the case of CKMR, the tag is a natural tag, namely DNA markers (microhaplotypes) within individual red snapper, and the statistical modeling approach involves the number of genetically related individuals observed relative to the number of individuals captured. This is accomplished by collecting tissue samples (fin clips) from many individuals (>10,000) in the Atlantic red snapper population, sequencing DNA for each individual, and assessing how similar each individual is to every other individual at the DNA sequence level. In simplest terms, the larger the number of related individuals observed for a sample containing a given number of individuals (say, 10,000), the smaller the population size would be estimated to be.

This generally is how CKMR analysis works. In the current study, the research team is spending considerable effort sampling adult red snapper fin clips throughout the study region from North Carolina through the Florida Keys over multiple years to ensure as robust a sample as possible to produce a reliable estimate of Atlantic red snapper population size.

Bayesian modeling is a type of statistical analysis in which information known ahead of time can be utilized to fit models to data. This information is called prior information, and it is included in the model by defining a distribution (aka, a prior) for each parameter that reflects our current knowledge.

If we have strong information on a given parameter before fitting a model to data, then the distribution for that parameter’s prior will often be narrow or informative. Likewise, parameters for which we have little information will be given loose or uninformative priors, which allows the model more solution space to estimate the posterior distribution for those parameters. In the model-fitting exercise, the goal is to fit a model that has the highest likelihood or best fit to the data, and in the process of fitting a Bayesian model, posterior distributions (i.e., posterior to or after model fitting) on all parameters are estimated. In this way, the posterior distributions, which are ultimately the quantities of interest, combine information from the priors and from the data being fitted by the model.

The integrated part of the modeling for the SARSRP study describes a modeling approach in which several different processes (e.g., habitat-specific fish density, detection probabilities, capture rates, etc.) are incorporated into a single model to produce an estimate of population size. Because the study team will include priors on certain parameters in some components of the model, it is therefore a Bayesian analysis.

There are vast areas of the US Atlantic shelf that have not been surveyed, thus the habitat in those areas of the shelf is unknown. In some cases, we have a general idea that hardbottom or reef habitat broadly exists in a given region, but those habitat types are often interspersed with sand habitat. Other areas are thought to be almost exclusively open sand habitat.  The term “unconsolidated” generally refers to this sand habitat, which gets that name because it is not consolidated into sedimentary rock or reef habitat.

The reason the study team needs to conduct remotely operated surveys in those unconsolidated areas is 1) to estimate what percentage of sand habitats actually contain some hardbottom habitat mixed in with the sand habitat, and 2) to estimate the occurrence and density of red snapper in open sand habitat. In the Gulf of Mexico Great Red Snapper Count study, a substantial percentage of adult red snapper in the western Gulf of Mexico was estimated to occur on unconsolidated habitats, while the fish were mostly associated with hardbottom habitat in the eastern Gulf of Mexico.

A key difference in the bottom type between those two regions is the predominance of muddy sediments in the western Gulf of Mexico, while quartz silica or carbonate sands predominate in the eastern Gulf of Mexico off Florida. While the Atlantic shelf in the SARSRP study region is more similar to the eastern Gulf of Mexico, it is still important to test the habitat association of red snapper in the Atlantic.

There are questions about the accuracy of discards estimates in the Atlantic recreational red snapper fishery, which are estimated via the MRIP program run by NOAA Fisheries. Tagging may be a way to estimate discards independently, with the two main approaches being conventional and genetic tagging. Conventional tagging involves attaching reward tags to fish and then have those reported by fishers who capture and discard the fish. Genetic tagging would involve taking fin clips from fish that we can identify by sequencing their DNA. The simulation study is designed to test the efficacy of either approach, as well as estimate the number of conventional or gene tags (fin clips) scientists would need to collect to estimate the magnitude of discards effectively.

Three dimensional (3D) telemetry is being conducted to estimate the effective sampling area of cameras attached to fish traps utilized by the Southeast Fishery Independent Survey (SEFIS) program. This involves deploying dozens of acoustic receivers on the seabed approximately 200 m apart from each other in an array that surrounds natural reef habitat. Acoustic tags that transmit a unique ping are then attached externally to red snapper, which are released back into the array. The spacing of acoustic receivers and pressure (depth) sensors in tags allows acoustically tagged red snapper fish to be tracked in three dimensions. Trap-cameras and remotely operated vehicles (ROVs) are deployed on reefs within each array and tagged fish location and movement are tracked with telemetry. Therefore, we can test whether fish present in the array are detected on trap-cameras or with the ROV’s camera. These data are then utilized to estimate the effective sampling area of each of these gears. Ultimately, this will allow counts of red snapper observed with cameras on natural reefs to be converted to red snapper density (i.e., fish per area), which can then be scaled up to the reef habitat in the region to estimate red snapper population size.