Lifemapper (Lm) is a high-throughput species distribution modeling system and set of multi-species analysis tools. Started in 1999, Lifemapper was created to compute, archive and web publish, species distribution models (“SDMs”, also known as species niche models and predicted habitat models) using all available online species occurrence data. Using the Lifemapper platform, known species localities georeferenced from museum specimens are combined with climate models to predict a species’ “niche” or potential habitat availability, under current day and future climate change scenarios. The LmSDM component computes these single-species predictions.
Lifemapper Range and Diversity, “LmRAD", extends Lifemapper from creating single-species predictions to large-scale multi-species, landscape-level analyses. The LmRAD component calculates multi-species biodiversity analyses by creating a binary presence-absence matrix (PAM) for a large number of species from SDM outputs or other species range maps, then computing a number of biodiversity measures from it.
The Lifemapper platform is a large but modular web services-based system of three core software components: (1) LmServer for data management and communications; (2) LmCompute for calculations; and (3) client applications including a Lifemapper plugin to the open-source geographic information system, QGIS, and a web site. The LmCompute and LmServer modules are both involved in modeling operations. LmCompute instances request jobs from LmServer, execute them, then post the model results back to LmServer where data are written to storage and metadata to the PostgreSQL database. Two applications on LmCompute underlie SDM calculations: openModeller and MaxEnt, while all LmRAD calculations are performed by Lifemapper code. From LmServer APIs and a Python client library, a user can obtain original and computed data using the Lm website or the QGIS workstation GIS environment.
The Lifemapper software can be logically thought of as having two primary resources: geospatial data and analysis tools.
The first, geospatial data, consists of a library of species and environmental data, including climate data (elevation, and observed climate, and predicted past, future climate), specimen occurrences (known locations where an organism has been collected), and computed maps of geographic locations with similar climate and elevation where the species could thrive. All of these data may be browsed, queried, and downloaded using the Lifemapper web services.
The second, analysis tools, are for analysis of both single-, and multi-species datasets. Single-species analyses are at the heart of the data library described above, and are encompassed in a module named Lifemapper Species Distribution Modeling, or LmSDM. Multi-species analyses are contained by the module named Lifemapper Range and Diversity, or LmRAD.
For information on data objects, terms, and workflow, see Using Lifemapper
The Lifemapper software can be installed on physical or virtual hardware that supports Rocks 6.2 cluster management toolkit (http://www.rocksclusters.org/) which is built on the CentOS 6.6 Linux operating system. A minimum of 16GB RAM is recommended, and storage space adequate for the computations you will execute. The default installation has low resolution (10 min) climate data, global extent, and a subset of species data from GBIF. It contains about 7 GB of input data and will create by default, about 50 GB of output data. Any compute nodes will need about XXGB of space for computations.
The most recent Lifemapper software is available as two Rocks rolls, one for the LmServer component, one with the LmCompute component. Each contains a README file with instructions. The third component, client applications, are intended to query a Lifemapper installation for data and analysis requests. The Python client library is this package. The Lifemapper plugin to the open source GIS package QGIS (www.qgis.org) can be downloaded from the QGIS plugin repository.
As part of a collaboration with San Diego Supercomputer Center through the Pacific Rim Analysis and Grid Management Assembly (PRAGMA), we recently ported Lifemapper to the Rocks Cluster Distribution to run on physical or virtual clusters. This change greatly stabilized and hardened our software, but our local resources struggle to handle much larger scale experiments. As part of an ongoing proof-of-concept experiment with Harvard University, we are running multiple SDMs on North American data for the kingdom Plantae. Our continuing work with SDSC has led us to the conclusion that deploying our software on a virtual cluster on SDSC’s Comet cluster would leverage both our existing Rocks-based software stack, and also the virtual cluster support unique to Comet.