[kepler-dev] [Bug 1332] - GARP improvements and further implemenation

[bugzilla-daemon at ecoinformatics.org]

http://bugzilla.ecoinformatics.org/show_bug.cgi?id=1332





------- Additional Comments From berkley at nceas.ucsb.edu  2004-02-10 10:24 -------
This is a list of the actors that need to be implemented to complete the
preprocessing of the GARP pipeline.  These are actors that would be needed to
take a raw DiGIR query and get it into the form needed to submit to the current
GARP actors.

--DiGIRQuery: actor to query DiGIR for a set of species.  This should probably
be done via the ecogrid.

--EnvLayerMetadatExtractor: extracts the metadata from the env layers needed to
get user input.  the user's input includes: scaleValue, minSampleSizeThreshold,
SubSampleMethod (choose from medians, quartiles, infinite or manual),
numberOfReplicates (how many times to run the garp algorithm).

--RescaleValues: This actor rescales the presence data values from their current
form to a value between 1 and 254.  0 and 255 are reserved and not used.  this
could probably be implemented in java, though i'm not sure of the heterogeneity
of the input files....it might be better to use a gis system if there are a
variety of input formats.

--Clipping: For each species, this actor creates a minimum convex polygon of the
species plots (lats/longs).  It then buffers the polygon to n times the area of
the polygon.  the environmental layers are then clipped to the size of the new
polygon.  This should probably be implemented in a GIS system.

--Sampling: take in a list of species and occurance data.  randomly sample the
data into two sets: intrinsic data and extrinsic data.  the intrinsic data is
used in the actual garp pipeline, the extrinsic data is used to test the garp
output for ommission and commission error.

--Rescale (general binning utility): Take in a spatial file and rescale to a
specific scale.  this probably needs to be implemented in a gis...possibly could
write java to do it.

--SpeciesOccuranceCount: take in a list of species with occurance data that is
adjusted to the same scale.  output a vector of species and counts.

--SpeciesSelection: rescale a list of species coverages to a single scale. 
select the species that have an occurance count greater than a user defined
threshold.  this should output a list of rescaled species (with occurance data)
that have an occurance rate greater than the threshold.  This can probably be
implemented in a composite actor using existing (or soon to exist) actors.