As of right now, this research note will have two figures. One will be a map of banding locations and the other will summarize the results from the model.
Figure 1: Distribution of eastern mallard banding
locations from 2022-2025. Hexagons include at least one banding site and
were used to reduce specificity regarding waterfowl hotspots in the
winter.
Figure 2: Violin plots and predictive plots depicting posterior distributions and predicted daily survival probability. Subplots (a) and (b) show posterior distributions of all banders and days after banding respectively on effects on daily survival. Negative and positive effects are colored based on a p-tail less than or equal to 20%. Subplots (c) and (d) show predicted daily survival for weight at banding and the five day rolling minimum temperature (degrees C), for the average state, average year, and the grand mean of the posteriors for day and bander.
For those of you that don’t work in the Bayesian world as much, I’m going to add a bit more about what we’re seeing in this graph. In subplot (a), I plotted all banders. This is to showcase how a majority of banders have little to no effect on the daily survival of the mallards. Yet, there are some specific banders who do show a negative effect.
In subplot (b), I graphed the effect size of days since banding. I categorized the effects as negative, positive (again, < 20% of posterior samples above or below zero), or little to no effect. Although I hypothesized that there would be a sort of increasing trend (lower daily survival at the beginning and then increasing), we don’t necessarily see that. Day 1 (first day after banding) is negative which makes sense - more stressed from the transmitter, preening more, etc. But, after day 1, the results show variation in effects.
In subplot (c), you can see the predictive daily survival probability depending on weight would be. Weight had 5.3% of the posterior samples below 0, indicating that it has a strong positive effect on survival. In other words, 94.7% of the time, this effect size would be positive. That being said, the heavier the bird, the more likely it is to survive after banding.
Lastly, in subplot (d), you can see a similar graph but with the temperature covariate instead. This effect was quite strong as 100% of the posterior samples were above 0 (this is not indicated in the graph - it’s from just the basic results). As the rolling average increases, the bird is more likely to survive. This credible interval is quite wide, and that is most likely due to a lower sample size of birds with rolling average around -20.