/
2012.06.07 Range Maps
2012.06.07 Range Maps
12-06-07 Range Maps
Participants
John Donoghue, Edwin Skidmore, Nirav Merchant, Martha Narro, Jim Regetz, Mark Schildhauer
Agenda
Discussion topics document from John
- Current range modeling status
- Proposed range mapping deliverables
- Range modeling decision tree
- Script changes and need for retesting new scripts in the cluster
- Raster to vector conversion issues
- Scheduling a model run in the cluster
- Need to archive maxent results that are stored in temp path
- Scheduling subsequent analysis runs in the cluster
- Range mapping API
Notes
Mark and Nirav:
- NM: Parametric launcher (used to compute range models at TACC) is now in Python
- Previously was in shell.
- Ideal when have lots of independent tasks.
- MS: interested in connecting Kepler workflow system to iPlant big iron.
- Nirav discussed this with Matt Jones about 2 weeks ago.
- Wants to integrate the Kepler wfs into DE.
- Also talked to Regan about making iPlant data store visible in Kepler, which has hooks for it.
John:
1-3. Overview by John
4. Script changes and need for retesting new scripts in the cluster
- New models in the work flow, so those need to be added.
- Edwin will connect with Matt to see how to get the necessary R packages on TACC.
7. Need to archive maxent results that are stored in temp path
- Some of the files written in the tmp directories need to be captured.
- John will need to write a script to mine those tmp files to pull out the ones that are needed.
- Parameterize to rename based on each species.
- John will determine size of maxent output files.
5. Raster to vector conversion issues
- John will send Jim examples of raster files (er mapper format, binary so all 0 and 1’s) that do and don’t run when converting to vector.
- (goodle or grass)
6. Scheduling a model run in the cluster
- The run is a ways out, so can table this.
- Get the set up work done first.
8. Scheduling subsequent analysis runs in the cluster
- Once have range maps, will want to do some science on them – looking at how environmental variables impact them. In R.
- Dismo package is in R for modeling.
- Open modler (gbif uses it to provide range models). Is on source forge.
- The idea was to be able to plug different model algorithms into it.
- JR: Robert on raster packages. He’s good about making changes to fix bugs and optimizing.
9. API
- Out a ways. Discussion tabled for now.
Misc.
- Wiki page to document the project and instructions for how to execute the range model computations.
To Do List
- Jim will set up a wiki page on NCEAS Redmine under the BIEN project, then Martha will set up an iPlant wiki page to point to it.
- Jim will set up a git repository for the range map project (at NCEAS).
- John send Edwin a list of updates to R packages and preferred versions.
- John will write a script to mine the tmp files to extract files others in BIEN need. Includes parameterizing to rename based on each species.
- John will finalize the scripts for the next round of modeling.
- John will determine size of maxent output files and let Edwin know.
- John will send Jim examples of raster files that do and don’t run when converting to vector. (may not be a problem any more)
- Martha will set up a regular meeting every 2 weeks for Thursdays 1:15.
- Martha and John work out an organizational framework for the wiki.
- Martha – Question how should the license be structured?
- iPlant license on the R source code for map modeling. NCEAS license too?
- Then, decisions about where source code lives.