|
Jones, Zhang, Pereira, Higgins, Berkley
- Ricardo summary of work
- Finishing java bindings for OM, includes JNI code
- Looking at the ENM pipeline, thinks OM
- replaces rescale values actor (done internally now)
- 3 GARP actors replaced
- 4 step postprocessing optimization algorithm replaced too as one OM actor
- OM actors will take layers, a sampling grid, and occurrence data as inputs
- handles all clipping an re-projecting needed
- image outputs are currently GeoTIFF and bitmap
- current image converter needs to be able to create lower resolution images for display
- DIGIR query is currently a single call to DiGIR
- Need to do a DiGIR 'inventory' call to get list of species, then iterate across those species
- But note that EcoGrid does not provide the DiGIR inventory method
- Also note DiGIR results are not yet handled in Kepler
- Post processing steps are underspecified
- Action items
- finish up DiGIR integration (Rod)
- Need environmental layers in ecogrid with metadata (Deana? Matt will ask her)
- Ecogrid moves data layer files to nodes as needed by OM
- implement post processing, mostly GIS ops along with ROC threshold (Ricardo will further specify)
- SummaryStats -- pairwise intersections between Grid layers (current vs predicted)
- implement OM (Ricardo)
- GRASS routines need to be redone using JNI to better handle GRASS output (Jianting)
- verify all GIS operations are complete
- Clipping actor needs to gen a covex hull, check to be sure it does
- deal with data transfer issues for GRASS web service
- Distributed computation to apply more significant computational power across iterations (Dan)
- use map actor to refactor pipeline for simplification of iteration cycles
- 5 climate scenarios, 1000 species, 500 garp runs per species
- Make sure data is available on node when OM needs it for a given iteration
- Probably passing file-handles instead of data tokens (GDAL in OM depends on file handles)
|
