Model-based integration of citizen-science data from disparate sources increases the precision of bird population trends

Abstract

AimTimely and accurate information on population trends is a prerequisite for effective biodiversity conservation. Structured biodiversity monitoring programs have been shown to track population trends reliably, but require large financial and time investment. The data assembled in a large and growing number of online databases are less structured and suffer from bias, but the number of observations is much higher compared to structured monitoring programs. Model-based integration of data from these disparate sources could capitalize on their respective strengths.LocationGermany.MethodsAbundance data for 26 farmland bird species were gathered from the standardized Common Breeding Bird Survey (CBBS) and three online databases that varied with regard to their degree of survey standardization. Population trends were estimated with a benchmark model that included only CBBS data, and five Bayesian hierarchical models integrating all data sources in different combinations. Across models, we compared consistency and precision of the predicted population trends, and the accuracy of the models. Bird species body mass, prevalence in the dataset and abundance were tested as potential predictors of the explored quantities.ResultsConsistency in predicted annual abundance indices was generally high especially when comparing the benchmark models to the integrated models without unstructured data. The accuracy of the estimated population changes was higher in the hierarchical models compared to the benchmark model but this was not related to data-integration. Precision of the predicted population trends increased as more data sources were integrated.Main conclusionsModel-based integration of data from different sources can lead to improved precision of bird population trend estimates. This opens up new opportunities for conservation managers to identify declining populations earlier. Integrating data from online databases could substantially increase sample size and thus allowing to derive trends for currently not well-monitored species, especially at sub-national scales.

Large‐scale spatial variability in urban tolerance of birds

“…Still, the greater magnitude of interspecific over intraspecific variation supports the use of mean trait values in trait‐based ecology, at least for species–environment relationships. Ultimately, growing citizen science data will likely form a major component of future understanding of species–environment relationships for both basic (Soroye et al, 2018) and applied (Hertzog et al, 2021) research. Since species differences may drive evolutionary and ecological change within biological communities (Trevail et al, 2021), citizen science data offers an excellent opportunity to further our understanding of variability in species–environment relationships and how it impacts species' responses to anthropogenic pressures.…”

Outstanding challenges and future directions for biodiversity monitoring using citizen science data

“…However, simulation‐based studies have shown that poor estimates can be obtained when biases in presence‐only data are unknown and hence not accounted for (Ahmad Suhaimi et al, 2021; Simmonds et al, 2020). When estimating population trends, data integration of large CS list data and structured data can reduce bias (Boersch‐Supan & Robinson, 2021; Hertzog et al, 2021; Pagel et al, 2014).…”

Effects of input data sources on species distribution model predictions across species with different distributional ranges

“…Although both are valuable for addressing a wide range of socio‐ecological research questions and biological modelling and monitoring approaches (Chandler et al, 2017; Neate‐Clegg et al, 2020), doubts still exist about data quality from citizen science surveys (Aubry et al, 2017; Jiménez et al, 2019). The data combination from different sources also proved to be a useful tool in ecological modelling (Miller et al, 2019; Suhaimi et al, 2021), specifically to improve estimates of species distributions (Isaac et al, 2020; Robinson et al, 2020), abundance and population trends (Boersch‐Supan et al, 2019; Hertzog et al, 2021). However, there is still much to explore on the data quality (Johnston et al, 2019; Van Eupen et al, 2021) and usefulness of each input data source through the modelling process (Kobori et al, 2016; Kosmala et al, 2016), as well as the uncertainty hosted in each modelling step should also be controlled more exhaustively.…”