OOS 7-2 - popler: A new R package for synthesis of ecological time series from the LTER network

Tuesday, August 8, 2017: 8:20 AM
Portland Blrm 254, Oregon Convention Center
Tom E. X. Miller1, Aldo Compagnoni1, Andrew J. Bibian1, Brad Ochocki1 and Kai Zhu2, (1)BioSciences, Rice University, Houston, TX, (2)Department of Biology, University of Texas, Arlington, Arlington, TX

A central goal of ecology is to understand the drivers of fluctuations in population size. Long-term data from ecological time series provide the essential raw materials for achieving this aim. The Long-Term Ecological Research (LTER) network records high-quality ecological time series from a diversity of taxa and abiotic environments, yet these data area surprisingly under-used by ecologists for the study of population dynamics. LTER data are particularly valuable because they often include spatial replication of ecological time series (e.g., multiple plots or transects). Spatial replication facilitates the statistical inferences of 'process error' in demographic parameters and provides better resolution of signal vs. noise. To promote the use of LTER time series data, particularly for the study of population dynamics, we developed a database that contains all population time series from the LTER network and an R package ('popler') to browse, search, and retrieve data in an open-source computing environment. We developed hierarchical Bayesian population models to illustrate how popler can be used for comparative demographic analyses, leveraging the powerful spatial replication that is common to many LTER studies.


The popler database of ecological time series includes 26 LTER sites throughout North America and over 300 independent studies containing 8500 unique taxa, mostly plants and animals. The average time series length is 11 years and the longest time series include more than 50 years of observations. A majority of popler time series come from censuses of multi-species communities, which allows for the study of community-level fluctuations in abundance and the contributions of inter-specific interactions. We illustrate a complete workflow of data analysis, including data search, retrieval, and hierarchical Bayesian population modeling to quantify demographic parameters and their spatio-temporal variability. The popler package is freely accessible and the R code used for the analyses in this presentation is available from the authors.