What is PB-GJAM?

[website under construction]

Predicting Biodiversity with generalized joint attribute models (PBGJAM) is an interactive web-based tool that provides accurate and timely forecasts of species and community responses to climate change.



The Generalized Joint Attribute Model (GJAM)

GJAM is a generative model that fits individual species at the community scale, i.e., all species jointly. GJAM resolves the three serious challenges for forecasting species distributions using biodiversity data that are 1) multivariate, 2) multifarious (measured in different ways and on different scales), and 3) mostly zeros. Probabilistic inference in GJAM admits multivariate data, disparate scales and massive zero inflation.

Assimilation of biodiversity and habitat data

We have synthesized data from taxonomically diverse communities, including mammals, beetles, ticks, mosquitoes, birds, and vascular plants. For habitat prediction, we developed automated workflows using Google Earth Engine's python API to gather environmental data related to soils, climate, topography, land cover, and NASA remotely-sensed ecosystem attributes.

Predictions of species abundances & forecasts of climate vulnerability

By combining multiscale analysis of remotely sensed data with biodiversity networks, we deliver a new generation of biodiversity risk predictions. We model important community attributes including the distributions and abundances of each species, the aggregate number of species (species richness), and their association into communities, based on shared responses to the environment and to each other.

Explore

To visualize and download model inputs/outputs, including species abundances/distributions for insects, birds, mammals, and plants across the United States use our PBGJAM tool below.

Using your own data

For researchers interested in using their own data, see the following tutorials for help on downloading environmental data, fiting models using GJAM, performing variable selection, and interpreting model outputs. Examples will be posted on GitHub [link coming soon].

  1. Tutorial for downloading/installing/uploading data to GEE [link coming soon]
  2. Tutorial for downloading environmental and remote sensing data from GEE [link coming soon]
  3. Tutorial for merging files downloaded from GEE [link coming soon]
  4. Tutorial for inference and prediction using GJAM
  5. Tutorial for inference and prediction using GJAM for a time-series [link coming soon]
  6. Tutorial for inference and prediction using MASTIF to model mast seeding

Publications from our team related to PBGJAM

Clark, J.S., D. Nemergut, B. Seyednasrollah, P. Turner, and S. Zhang. 2017. Generalized joint attribute modeling for biodiversity analysis: Median-zero, multivariate, multifarious data. Ecological Monographs, 87, 34-56. Clark2017EcolMonogr appendixs1

About our research group

Our team includes ecologists, statisticians, remote sensors, and data scientists in the Nicholas School of the Environment at Duke University. The PBGJAM tool is supported by NSF and NASA's Advanced Information Systems Technology program, award number AIST-16-0052. More information about our funded project can be found here

Christoph Hellmayr,
postdoc
Brad Tomasek,
PhD candidate

To contact us

Please use the form below to contact us for help, to suggest an improvement, or to send us feedback. We will get back to you as soon as possible.