Rieger, M. R.; Grueneberg, C.; Oberhaus, M.; Trautmann, S.; Parepa, M.; Anthes, N.

Population trends derived from systematic monitoring programmes are essential to identify species of conservation concern and to evaluate conservation measures. However, monitoring data pose several challenges for statistical analysis, including spatial bias due to an unbalanced sampling of landscapes or habitats, variation in observer expertise, frequent observer changes, and overdispersion or zero-inflation in the raw data. An additional challenge arises from so-called \'rolling\' survey designs, where each site is only visited once within each multi-year rotation cycle. We developed a GAMM-based workflow that addresses these challenges and exemplify its application with the highly structured data from the Ecological Area Sampling (EAS) in the German federal state North-Rhine Westphalia (NRW). First, we derive a routine that allows informed decisions about the most appropriate combination of distribution family (Poisson or negative binomial), model covariates (e.g., habitat characteristics), and zero inflation formulations to reflect species-specific data distributions. Second, we develop a correction factor that buffers population trend estimates for variation in observer expertise as reflected in variation in total bird abundance. Third, we integrate model weights that adjust for between-year variation in the representation of habitat or landscape types within the yearly subset of sampled sites. In a consistency check, we found good match between our GAMM-based EAS trends and TRIM-based trends from the standard German common Bird monitoring scheme. The study provides a template script for R statistical software so the workflow can be adapted to other monitoring programmes with comparable survey designs and data structures.