Effectiveness of hierarchical Bayesian models for citizen science data with missing values: A case study on the factors influencing beach litter in Shimane Prefecture, Japan

Citizen science can play an important role in addressing the issue of marine debris. However, citizen science data are often composed of inconsistent methods compared to data collected by experts. In this study, we applied beach cleanup data, collected in different survey years at different survey sites, to a hierarchical Bayesian model to elucidate the factors affecting the distribution of beach litter. The results showed the model accounting for differences between years had a smaller Watanabe-Akaike Information criterion than the model that did not account for it, indicating better accuracy of the model. The amount of beach litter was influenced by current velocity and bay openness, and these effects varied across years. The results indicate that citizen science data, which may contain missing values due to various constraints such as economic and human resources, can make an important contribution toward solving marine debris issues by flexible statistical analysis methods. •The model accounting for year differences indicated better accuracy.•The amount of beach litter was influenced by current velocity and bay openness.•The effects of current velocity and bay openness varied between years.•We developed a model that can utilize citizen science data with missing values.