Forecast the response of forest birds to climate change and forest management: does citizen science data provide accurate predictions?

Abstract

In the context of global changes and biodiversity mass extinction, species distribution models (SDMs) are of major importance for conservation and management. In particular, such models can be used for mapping spatial distribution of endangered species and forecasting their response to climate and land-use change. In Sweden, intensive forestry has caused a strong decline of forest biodiversity. Climate change is also expected to cause range contractions for northern-boreal species which are at the limit of their range boundaries. As SDMs require a large amount of data, ideally collected over large spatial and temporal scales, citizen science based on volunteer reporting of species can constitute a promising alternative. However, citizen science data (CSD) also have several drawbacks (e.g. presence only, uneven sampling effort). Furthermore, CSD still lack validation by comparison to systematically collected data. The aim of our study is to assess the reliability of CSD for forecasting species occurrence in response to various management and climate scenarios. We compare the predictions obtained by two different and independent citizen science datasets, opportunistic reports (OR) from Artportalen, and systematically collected data (SC) from Svensk Fågeltaxering. Both datasets cover the same large spatial area (whole of Sweden) and time period (2000-2013). As the latter have a well-defined sampling design and protocol and engage experienced observers, we used this dataset as reference to assess the accuracy of models based on OR. Absence data have been inferred from OR based on questionnaires sent to observers evaluating their skills and reporting habits. We then built species distribution models (logistic regression) according to climate and environmental predictors (allowing non-linear effects and interactions). We further used these models to forecast the responses of multiple species to various climate scenarios (based on IPCC projections) and forest management scenarios (simulations in Heureka Forestry Decision Support System). The assessment focused on eight forest bird species that encompass a diversity of functional traits and ecological preferences. This allowed us to test how the models performances vary between species. We also expected different species responses to climate and management scenario with potential trade-offs to conciliate the different species requirements. More generally, our study provide further evidences for the relevance of using CSD to forecast species distribution. In particular, our results highlight that OR and SC have different strengths and limitations. Opportunistic reports seems provide more accurate predictions for rare or cryptic species, whereas SC perform better in sparsely populated areas.