Data set used for the analysis of https://doi.org/10.1101/2023.05.03.539219. Open the RDS file with readRDS() in R-CRAN. See script and JAGS-model in manuscript appendix.

Distance data were prepared for analysis using a cutoff 200 m and a bin width of 20 m. This resulted in 10 distance intervals, where each observation was categorized into a specific distance interval (1-10). During the 48 years in which distances were available we included 3531 distances less than 200 m, corresponding to an area of approximately 55 km^2.

We developed a hierarchical state-space model in JAGS  following. The detection probability was computed in a multinomial model of cell probabilities, assuming a half-normal detection function. The distances for all lines and years were pooled to estimate the detection probability, and no covariates were added to the detection equation. The process model comprises two parts. The first part calculates the initial density in the first year to estimate the observation error. The second part models the population dynamics of subsequent years using a Gompertz model to estimate annual process variation. 

We used Markov Chain Monte Carlo (MCMC) methods to estimate the posterior distribution of parameters and adult density estimates. The model was run for 100 000 iterations with a burn-in of 25 000. The thinning was set to 2, and three chains were used to evaluate convergence. The package MCMCvis was used to evaluate the trace plots and posterior distribution of the model parameters. All parameters had an R-hat value less than 1.1. Parameter estimates are presented with a 95% credible interval (Cr. I), and the plot of the estimated time series of adult densities show the 80% highest posterior density. All analyses and graphics were performed in R  and the JagsUI package was used as an interface to JAGS from R .