[seek-dev] Re: workflow

Deana Pennington dpennington at LTERnet.edu
Mon Nov 22 14:39:04 PST 2004

-----Original Message-----
From: seek-dev-admin at ecoinformatics.org
[mailto:seek-dev-admin at ecoinformatics.org] On Behalf Of Matt Jones
Sent: Monday, November 22, 2004 2:44 PM
To: Deana Pennington
Cc: seek-dev
Subject: [seek-dev] Re: workflow

Thanks, Deana.  A few followups and a question:

Deana Pennington wrote:
> Comments on your revised workflow:
> 1) What you have to end up with is a layer stack of limited extent, 
> which is input for the GarpPresampleLayers actor, and a separate stack 
> of full extent, which goes to the GarpPrediction algorithm.  Right now, 
> it show one stack being generated using ConvexHull output and going to 
> both places.  There are 2 ways we could do this.  a)  put an interation 
> inside the IntegrateSpatialData actor, so it generates 2 stacks of 
> different extents (but then limits reuseability), or b) put in another 
> instance of the IntegrateSpatialData actor.  I think b is probably better.
So, where does the information come from for determining the limited 
extent?  Is it the same as the extent of the buffered convex hull?  Or 
something else?  Is the output of the current convex hull operation 
buffered?  How does the actor decide how big to make the buffer?

The limited extent is the same as the buffered convex hull, which is the
output from the convex hull operation.  I haven't actually looked at
Jianting's actor yet, but I think he put configuration parameters in that
allow the user to specify how big to make the buffer.  As I recall,
originally (when I thought this would be done as a raster operation in
GRASS) I suggested defaulting to some number based on desired cell
resolution, like 2x the cell width.  I think he came up with some other way
to do it that he liked better, and included both options.

> 2)  In the climate change scenario at the bottom, the IntegrateSpData2 
> actor must also have the hydro1k data as input.  So, put a relation in 
> after the Initial PrepofHydro1kData actor and use the output from that.
OK, I will do this.

3) I have a question about the convex hull operation.  In the process of 
incorporating Jianting's convex hull actor into the ENM workflow, I 
found that the actor requires three inputs: the filename containing the 
points, the filename for the output, and the number of points in the 
input file.  That third one is currently not hooked up in the workflow, 
which we need to fix.  My question: where's the right place to get the 
number of points?  To me, the possibilities are: 1) have the DarwinCore 
data source make it available on output, 2) open the file using a Kepler 
file reader, parse it and count the rows, then close it, 3) get the 
convex hull actor to count the rows itself.  I prefer (3) because that 
actor has to open the file anyway, but I'd like to hear other opinions. 

Well, so that's an interesting question.  The workflow, in its current
state, is designed to handle 1 species, and has no way of knowing how many
species were passed from the query.  E.g., if a list of 3 species with 100
points each is passed from DiGIR, there is currently no place in the
workflow to iterate through that list and compile the set of points for the
1st species, 2nd species, 3rd species.  It would seem like, to me, that once
the list is acquired, you would want to sort (if it doesn't come back that
way) and set up an array, get a count of species for the iteration, and at
the same time, you would get a count of points. Even if DiGIR passes the
info, you still have to have an actor to split the multiple point sets.


> Everything else looks good, and much more readable.
> Deana

Matt Jones                                     jones at nceas.ucsb.edu
http://www.nceas.ucsb.edu/    Fax: 425-920-2439    Ph: 907-789-0496
National Center for Ecological Analysis and Synthesis (NCEAS)
University of California Santa Barbara
Interested in ecological informatics? http://www.ecoinformatics.org
seek-dev mailing list
seek-dev at ecoinformatics.org

More information about the Seek-dev mailing list