[R-sig-Geo] R-sig-Geo Digest, Vol 80, Issue 4

Blair Christian blair.christian at gmail.com
Sun Apr 4 16:28:06 CEST 2010


Hello Wouter et al,

I have plenty to say about all aspects of this.  Currently at SAMSI we
are discussing similar topics in a subgroup of a spatial stat comp
working group...
http://sisla06.samsi.info/spatial/comp/doku.php

My feature list for spt is growing exponentially lately for spt
related things.  I welcome any and all interested in working on such a
library.  I spent a large amount of time reading up on different
date/time libraries, their differences, their inheritances, etc.  I'm
happy for you to bounce any ideas off me about what your constraints
are (eg data structure sizes, data structure processing times,
vizualization needs, etc)

I have been out of the R realm lately, working more with
python+gis+statistical modeling code, as well as with building a
geocoded data collection sensor (arduino + gps + sensor(s) of choice +
sd card).  I'm trying to gather some microclimate data and fit some
models (just a personal side project, but has some useful
applications).

If you have some very specific needs, I may already have the
alpha/beta classes/methods working.  Please contact me offline (or via
the listserve) with any specific needs/questions/comments.  If you can
write up some clear use cases/methods/classes you want, I could
probably put them together within a week or two...  Just let me know.

Best,
Blair



In response to:

Dear Wouter,

as to your question on related work, Blair Christian has started two
threads on this list:

https://stat.ethz.ch/pipermail/r-sig-geo/2009-September/006435.html
https://stat.ethz.ch/pipermail/r-sig-geo/2009-September/006491.html

and the resulting work is now at r-forge, called spt, see
http://r-forge.r-project.org/projects/spt

it is a mix of classes and methods for spatio-temporal data and
functions that are needed for his particular research domain. It has a
vignette.

For the two suggestions you make I cannot say much; if you have running
prototypes I'd be happy to look at them.

Simply adding an sp object to a time series (as attribute) [or a set of
time stamps to an sp object] means that you'll sacrifice the
multi-atribute space for multiple spatial entities [time entities], so
your objects become univariable. That might only work as a short-term
solution; next time you'll want to work with water chemistry data.

Best regards,
--
Edzer

Wouter Buytaert wrote:
>
> Hi all,
>
> this question is mainly aimed at the developers of sp, but any feedback
> is welcome.
>
> We are designing some classes for hydrological modelling in R (the
> r-hydro package on R-Forge) and facing the decision of how to design a
> class for georeferenced time series (e.g., a precipitation or river flow
> series, the first referring to a point and the second to an area).
>
> We're considering an xts time series with an sp object as attribute, but
> there may be other options (xts as data slot in an sp object for
> instance) that may be preferrable for reasons beyond my knowledge. Also,
> there may be similar efforts underway of which we have no knowledge.
>
> cheers
> wouter



More information about the R-sig-Geo mailing list