Predicting among species variation in amphibian responses to climate change from dynamic hierarchical occurrence models
Consolidated data sets where species occurrence or abundance is recorded using a spatially replicated design, especially over multiple years, are rare. The dearth of such data presents a significant challenge when the goal is to study the dynamics of species distributions, especially for taxa where large-scale systematic surveys are non-existent. A natural alternative is to aggregate data from many individual studies to make general inferences regarding range dynamics. Aggregating data from species occupancy studies (i.e., studies using the design proposed by MacKenzie et al., Ecology 2003) provide a natural data type for developing comprehensive databases that can be used to study species dynamics. Systematic multi-year occupancy studies have grown in frequency and allow for a degree of control in dealing with study-specific differences in rates of observation errors. We demonstrate how these data can be used to make inferences about species occurrence dynamics using two examples, both involving amphibian response to climate. Data are analyzed using simple two-state Markov models that incorporate site dynamics and are fit using Bayesian hierarchical models.
We first examine responses of wood frogs to droughts in ~700 wetlands from 14 study areas throughout the northeast United States. We show that the effect of drying on occurrence dynamics shows a strong spatial pattern that interacts with local wetland characteristics. These patterns play out in predictable ways, where drought has the largest effect on persistence of wood frogs in more southern sites and in shallower wetlands. The results highlight the importance of understanding the interaction between range position and physical environment when making predictions about climate change effects.
We extend the single-species analysis to then demonstrate how this approach can be used to examine community responses to climate change. As an example we use data from a large-scale amphibian occurrence database that includes observations for 86 species in > 6000 sites across 62 unique study areas from throughout the United States and Canada. We illustrate modeling approaches that can be used to examine the relationship between species traits and responses to a suite of climate variables. Combining detection non-detection data across studies is a potentially fruitful method for understanding range dynamics in species where broad-scale, systematic surveys are not available.