Transcript Cavner_BridgingSpeciesNicheModeling

Bridging Species Niche Modeling and
Multispecies Ecological Modeling and Analysis
Jeffery Cavner, J.H. Beach, Aimee Stewart, CJ Grady
[email protected], [email protected] ,[email protected], [email protected]
Biodiversity Institute University of Kansas
Species Diversity
LmRAD (Lifemapper Range and Diversity)




Biodiversity - describe, visualize and analyze different aspects of the
numbers and abundances of taxa in time and space.
Patterns of species richness - constituent species ranges sizes and
spatial locations of those ranges.
Patterns related to species associations, co-occurrence, and species
interactions requires testing against randomized distributions.
Species richness and species range can be summarized and linked
by one basic analytical tool, the presence/absence matrix (PAM).
Lifemapper as an overarching
architecture
• LmRAD is built on top of the existing Lifemapper architecture
• Lifemapper is an archival and species distribution modeling
platform consisting of a computational pipeline, specimen data
archive, predicted species distribution model archive
• Distribution models are built on-demand using openModeller.
• Inputs: climate scenario data and aggregated specimen occurrences
from GBIF and user provided occurrence points.
The Presence Absence Matrix (PAM)
Data Matrix
Grid
Most existing
indices of biodiversity
are simple
combinations of :
o
Vectors:
species richness
sizes of distributions
“dispersion fields”
“diversity fields”
o
Whitaker’s beta diversity
o
The dimensions of the PAM
Constraints
•
Construction of PAMs can be an extremely time consuming
data management task
•
Current methods for working with these matrices can be
computationally slow
Approach
• To overcome computational restraints we use a Python
implementation of the Web Processing Service standard
on a compute cluster, exposing spatial and statistical
algorithms.
• Allows a variety of species inputs
• Extendable clients including Quantum GIS (QGIS) and
VisTrails that share a common client library
Clients
Randomizing the PAM
•
•
•
To test the null hypothesis
By producing the same richness and range patterns while ignoring
realistic species combinations
Two Types of Randomization: Swap and Dye Dispersion
– Swap : keeps species richness and range size totals intact.
Additional Randomization methods
Dye Dispersion
–
–
–
–
Geometric constraints model
Assumes range continuity
Reassembles ranges
Keeps range size intact
QGIS is used as a WPS client
Using QGIS and WPS to construct a grid
The asynchronous nature of WPS combined with a
computational pipeline and compute cluster allow a user to
intersect hundreds of species layers at a time with the data grid
to populate the PAM.
Terrestrial Mammals
Proportional Species Richness
High
Yellow
Moderate Red
Low
Blue
Per-site Range Size
Statistical services provide diversity indices and plots
using WPS
By-species range-diversity plot
The plug-ins use a simple MVC pattern with QT threads for asynchronous
WPS requests and a client library for the communication layer
Conclusion
Jeffery Cavner, J.H. Beach, Aimee Stewart, CJ Grady
[email protected], [email protected], [email protected],
[email protected]
Biodiversity Institute University of Kansas