TODO before code freeze

From Intamap

Contents

1. Integration of copulas and projected sequential Gaussian processes.

Status: Copulas? PSGP - still some code problems with PSGP, but framework there. DC to follow up.

Components affected: R, (Java, WPS)

Responsibility: Hannes, Ben, Jon, Dan


2. Accepting SOS output

Status: Progress made, but parsing SOS output remains a problem - key issue is complexity of O&M response. Likely to accept a limited subset of SOS output. Hope to resolve issue of obtaining just one observation per location. Some information on what we accept at the moment is here: What_can_you_send_to_the_WPS

Components affected: WPS, Java?

Responsibility: Matt, Katherina, Dan, Edzer

3. Dealing with exceptions and errors

Status - we have an agreed set of responses to 'erroneous' requests (see http://wiki.intamap.org/index.php/Some_issues_to_discuss_before_the_meeting)

Components affected: WPS? (this is a big question - is this managed on the WPS, or in R - did we agree??)

Responsibility: depends on where it happens? Probably WPS in my view.

Jon: I have added some checking in R for most of these problems, R will now return with an error if the requests are bad, in terms of what we agreed on in Wageningen.

Matt: I have noticed the error messages, most useful. However, they are not propagated through to the WPS yet, I believe Jochen was working on that?

4. Passing parameters to interpolation methods

Status - proposal made, broadly accepted of passing name and values.

Components affected: WPS, Java, R

Responsibility: Matt, Jon, Katherina?

Matt This is complete now. An example set of parameters looks like this:


<int:MethodParameters xmlns:int="http://www.intamap.org" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
   <int:parameters>
      <int:Parameter>
         <int:name>NumberOfDesignPoints</int:name>
         <int:value>30</int:value>
      </int:Parameter>
      <int:Parameter>
         <int:name>CovarianceFunction</int:name>
         <int:value>Matern</int:value>
      </int:Parameter>
   </int:parameters>
</int:MethodParameters>

There is an ongoing issue about whether to wrap parameter values in speech marks (" ") but this should be shortly resolved.

5. Timing - methods needed for PSGP and copula (done in R I guess)


Status - framework in place - Olivier asks for important parameters which control the timing.

Components affected: WPS, Java and R (do we have a mechanism for requesting timing as in input to the service and a method for saying what the timings for the various methods are???)

Responsibility Olivier, Giannis, Ben, Hannes (Matt, Katherina and Jon?)

6. Automated choice of method (made in R I assume)


Status - unclear to me?

Components affected: R

Responsibility: Jon? all of us!

Jon: I think this is basically based on the timing function, when that has been finished, I can, together with Giannis and/or Olivier add to the interpolate function some tests that will choose the "correct" method.

7. Choice of projections and what to do with projections


Status - unclear to me?

Components affected: R? Depends on where this issue is detected?

Responsibility: Unclear?

Jon: Together with Matt I am trying to solve this by calling the function conformProjections in R. We can call this automatically, and it will do the following:

  • Only longlat given: throw an error
  • Only one of observations/predictionLocations has a projection: assume that also the other one has the same
  • One of observations/predictionLocations has a projection, the other is longlat: reproject
  • intCRS is given as optional argument: reproject (if possible) all spatial data to this projection before interpolating
  • targetCRS is given as optional argument: reproject the result. If observations and predictionLocations are longlat, reproject to targetCRS before interpolating

Status: R-function and possibility to call from WPS is there, some bug-fixes necessary.

Matt This is now working as far as I can tell. The errors are not propagated through the WPS yet so if it doesn't work the user won't know why - but this is another ongoing issue

8. Creating Web Coverage Service output from service


Status - unclear to me?

Components affected: WPS, Java, (R probably not?)

Responsibility: Katherina and Jan?