hoffmannsarahlouise
commented
3 years ago Woohoo a signal, and one that makes sense to boot! Not so exciting for the turtles, but certainly a trend that seems to fit some of their hypotheses.
Which of course begs the question of whether the population is increasing, decreasing or stable.
NOAA https://www.fisheries.noaa.gov/species/leatherback-turtle estimates a 40% global decline in leatherbacks over “the past three generations”, with fishing interaction being the primary threat. Leatherbacks in particular are at risk of fishing interaction because their foraging grounds overlap significantly with longline/pelagic fisheries. It’s generally assumed that increase long-lining pressure, particularly in the north-western Atlantic gyre, has been a major contributing factor to population declines, especially on the population of turtles that nest in Florida. This paper http://seaturtle.org/library/ChevalierJ_1999_In9thextraordinarymeetingofthesocieta_pxx-xx.pdf (1999) is specific to French Guiana but also predicts the same population risk from fishing.
Another interesting wrench to throw in there is the feminization of the population (mostly reported in other species) due to increased nest temperatures (sea turtles have temperature dependent sex determination or, as we fondly refer to it, “hot chicks, cool dudes”). Some beaches in Australia reported 100% nest feminization in 2019+, but that might be another can of worms, particularly since it will be ~25y until those sea turtles come to nest. It’s pretty hard to get the permitting to sex juvenile leatherbacks because the only method (laparoscopy) is so invasive, and the turtles have to be reared in a lab for a few months before they can undergo the procedure. Leatherbacks don’t do well in labs because their pelagic drive is so great: we had to leash them to the center of the tank to prevent them from swimming repeatedly into the tank walls. Long story long, I’m not sure that anyone has a good handle on how the sex ratio has changed in leatherbacks specifically as nest temperatures increase, though it’s a major threat documented in other species.
I don't know what sort of population-dynamics / PVA type modeling the Marinelife folks have done.
To my knowledge, this is the first long term analysis of leatherback nesting data on the most densely nested beach in the western hemisphere. There was a bit of a kerfuffle/lawsuit/data embargo with these data that was only recently resolved as of Jan 2021, which is how the data come to us (well, you 😊). If you think there is an opportunity/bandwidth for population modelling, it would be novel to this nesting beach and could have really significant implications!
I’ll be out in the field next week, but do you want to shoot to meet sometime the week of the 24th to talk through some of the findings, ecology, and future directions?
Thanks and have a great weekend! Here’s a leashed juvenile leatherback, for your viewing pleasure: https://www.4ocean.com/blogs/blog/meet-our-partner-florida-atlantic-university-marine-research-lab)
ebuhle
Hold onto your :turtle: @hoffmannsarahlouise, we have our first real signal amidst the noise. Whether it's a positive signal from a conservation standpoint is another matter.
This one's dead simple. I categorized each female as a neophyte or remigrant in each year, then aggregated by year to produce a data set that's somewhat interesting in its own right:
I then fit these data with a Bayesian logistic regression aka binomial GLM. The base model is intercept-only, the first-order model adds a logit-linear time trend. These models, for a change, fit just fine; there is no indication that the counts are overdispersed.
This plot is pretty much the whole story:
The proportion of new breeders is declining steadily and almost linearly, with a couple of anomalies in 2011 and 2014. The dark inner shading represents posterior uncertainty in the fitted relationship, while the light outer shading is the posterior predictive distribution (PPD) -- again, the distribution of hypothetical replicate data generated from the model, after having been updated by the real data. The credible interval of the PPD is wider in some years than others because the total count varies; proportions estimated from smaller counts are more uncertain (likewise reflected in the confidence intervals around the observed proportions).
The trend model clearly outperforms the null model in terms of estimated out-of-sample predictive skill. The approximate leave-one-out information criterion (LOO) is a fully Bayesian analog of AIC; it estimates the same thing, but takes into account parameter uncertainty and requires fewer assumptions. This shows that the trend model (
glm_neo1) is far better than the null model, and the difference is greater than 2 SE (which AIC doesn't provide but LOO does).I haven't had as much time as I'd hoped to dig into sea turtle ecology, but this is consistent with what Gouvello et al. found with South African loggerheads. I find their explanation of that somewhat confusing; they do mention a possible transient response to release from bycatch mortality a la Crowder et al., which makes sense, but in general declining recruitment in a long-lived species is not a good thing. Which of course begs the question of whether the population is increasing, decreasing or stable. I haven't tried to tackle that question, but I'd be very interested in talking about it. I don't know what sort of population-dynamics / PVA type modeling the Marinelife folks have done, if any, but with such rich detection history and nest success data it would be low-hanging fruit, and presumably quite management-relevant.
We'll get another piece of the population-dynamics puzzle tomorrow when I look at nest success.