Fox et al.: Juvenile Agpenser oxyrinchus desotoi in the Apalachicola River in Florida 247 
array received routine maintenance and data were down- 
loaded from each receiver. Routine maintenance included 
replacing batteries and cleaning of any biofouling. Sup- 
plemental telemetry data were obtained by periodically 
sweeping the study area with a portable receiver and 
hydrophone (Vemco VR100, Innovasea Systems Inc.) 
to detect fish between receivers or outside of the array. 
Between December and May in 2016, our receiver array 
was restructured to include more receivers outside of the 
Brothers River, leaving a gate (i.e., receivers that covered 
the river channel) where the Brothers River flowed into 
the Apalachicola River; in the spring of 2017, another 
gate was added at the cut connecting the upper Brothers 
River to the main stem Apalachicola River. The purpose 
of these gates was to detect fish as they left or reentered 
the aggregation site. 
Data from the acoustic receiver array were carefully 
checked for any spurious detections (e.g., simultaneous 
detections of a single fish at disparate receiver stations), 
and such detections were removed. Raw detection data 
were converted into detection days (one detection per fish 
per receiver per day). If a fish was not detected on a study 
day but was known to be alive because of detections on 
subsequent days, that fish was assigned the location of 
its last known detection until it was detected elsewhere 
(this method was applied by Fox and Peterson, 2019, to 
telemetry data from tagged juvenile Atlantic sturgeon). 
We used these telemetry data to determine whether and 
how often tagged fish left the Brothers River during our 
mark-recapture sampling season or to determine if fish 
tagged elsewhere in the system entered the Brothers 
River—the lack of such movements into or out of the 
Brothers River would indicate closure of the aggregation 
site. Additionally, we used telemetry data to calculate 
overwinter survival, as described later in the “Survival 
analysis” section. 
Abundance estimation 
To estimate the annual abundance of age-1 juvenile 
Gulf sturgeon, we fit a Huggins closed capture model 
(Huggins, 1989, 1991) to mark-recapture data from the 
gill-net surveys. Models were fit to data by using the 
RMark package (Laake’”) in statistical software R (vers. 
3.5.1; R Core Team, 2018). Each week of the study was 
considered a sampling period in the capture history for 
each year, and in each sampling period, an individual 
was classified as either absent or present, regardless of 
the actual number of times it was captured that week. 
The model includes the assumption that the popula- 
tion was closed, meaning that there were no births or 
deaths and that no immigration or emigration occurred 
throughout each summer in the study period, and the 
assumption that tags were not lost or overlooked. We 
® Laake, J. 2013. RMark: an R interface for analysis of capture- 
recapture data with MARK. AFSC Process. Rep. 2013-01, 25 p. 
Alaska Fish. Sci. Cent., Natl. Mar. Fish. Serv., Seattle, WA. 
[Available from website.] 
primarily used telemetry data to look for violations to 
the assumption of closure, but we also used the program 
CloseTest (vers. 3; Stanley and Burnham, 1999) for an 
alternative way to investigate closure. 
After we compiled individual capture histories for each 
juvenile caught during the study, each fish was assigned 
to an age group (.e., age 1, age 2, or age 3+) on the basis 
of its length. A set of candidate models was produced to 
estimate juvenile cohort abundances in each of the 6 study 
years. The candidate set of models and their main settings 
were as follows: 
M,, in which capture probability was constant; 
M,, in which capture probability varied with time; 
M,, in which capture probability varied by fish age; 
M,,,.,, in which capture probability varied by an addi- 
tive effect of time and age; and 
e M,«,,in which capture probability varied by an inter- 
active effect of time and age. 
In all models, capture probability was set as equal to 
recapture probability because no evidence indicates that 
capture history had any influence on recapture proba- 
bility for sturgeon—this method has been used in other 
studies in which recruitment of sturgeon was estimated 
to age 1 (e.g., Bahr and Peterson, 2016). Akaike informa- 
tion criterion (Akaike, 1973), corrected for small sample 
size (AIC,) (Hurvich and Tsai, 1989), was then used to 
evaluate the relative likelihood of each model. For each 
year of the study, all models were averaged by using 
Akaike information criterion model weight to estimate 
abundance of age-1 Gulf sturgeon, to reduce bias in the 
event of several candidate models being plausible (i.e., 
the top model carried a weight <0.90) (Burnham and 
Anderson, 2002). 
Survival analysis 
We estimated overwinter survival by comparing the num- 
ber of acoustically tagged age-1 juveniles detected leav- 
ing the summer aggregation sites in the fall and winter 
with the number of fish from that cohort that returned 
the following spring. Detection of the fish that by then 
were age 2 was done through acoustic detection or phys- 
ical capture. Survival was calculated as the percentage 
of tagged age-1 fish that were confirmed to still be alive 
at age 2 (or an older age). Because this survival analysis 
does not include detection probabilities, it has a neces- 
sarily conservative approach, in that survival is likely 
underestimated because it is not possible to distinguish 
between fish that died and fish that remained alive but 
were not detected. 
Results 
Over 6 years of sampling, we set 1834 nets for a total of 
2205 net-hours of sampling effort within the Brothers 
River. Because anchored gill nets did not sample Gulf 
