Crop Pollination - 10 vs. 1 for highest suitability

sjcsjc Member
edited July 2013 in Terrestrial Models
Hi there - I've completed a couple of successful runs using the Crop Pollination model, but I have a question about the required lulc.dbf. I noticed after running a model the cell in my lulc.dbf that held the floral resource suitability value for
"developed land" converted the 1 I assigned to a 10 (so, "developed land" should have been given a 1, indicating highest suitability, but instead it was assigned a 10). The model ran with the 10, and I'm wondering why I didn't get an error. Aren't the suitability values only allowed to range between 0 - 1?

I should add that when validating the model against field collected data I am getting R2 > 0.5 and p-values < 0.05 when using the 10 for the suitability of "developed land". When I change that suitability value of "developed land" to 1, I am getting R2 < 0.3 and p-values > 0.2. Quite the difference!
Post edited by sjc on

Comments

  • RichRich Administrator, NatCap Staff
    Hi sjc, the model isn't checking for valid inputs, and it probably should. That would explain why you didn't get an error.

    I think the high R and low p values are meaningless in the case when the suitability score is 10. That's an order of magnitude higher than a meaningful value, so I'm not certain that the R/p scores mean much.
  • sjcsjc Member
    I agree with you about the meaning of the results stemming from a suitability value of 10. Thank you for your response, Rich.
  • jdouglassjdouglass Administrator, NatCap Staff
    Hi sjc,

    Are you using the ArcGIS-based version of the model or the InVEST 3.0 version?

    While the ArcGIS-based model does not validate the inputs, the InVEST 3.0 version of the pollination model does do a lot of up-front error checking, including checking for floral resource values between 0 and 1.

    I would encourage you to try the InVEST 3.0 version of the pollination model if you are not already, as it is over 10,000 times faster than the ArcGIS version.
  • sjcsjc Member
    I am using the ArcGIS-based version. I'm doing so because eventually I'll be running a sensitivity analysis on it, and would like the convenience of setting that up in ModelBuilder. 10,000 times faster for 3.0? That's great!
  • jdouglassjdouglass Administrator, NatCap Staff
    Aha, yes that is a good reason to use it. Eventually, we'll have a scripting framework in place for running batches of any given model, but we're still in the early stages of testing it.

    Please do keep us posted as any other issues come up!
  • sjcsjc Member
    Hi again. I am having another issue. I decided to try the 3.0 version just to see the difference, but have gotten "Error 8: Not enough storage is available to process this command". I am running it on the same computer as the earlier runs, and should have ample space available. I've attached the log. Perhaps I'm overlooking something? Thanks!
  • jdouglassjdouglass Administrator, NatCap Staff
    edited August 2013
    Wow ... your landcover raster is pretty large. I could definitely see the model running out of addressable disk space on this.

    Any chance you're using 32-bit InVEST 2.5.4? If so (and you're on a 64-bit computer), you may want to check out the 64-bit version of InVEST (here's a link to 64-bit InVEST 2.5.5). The 64-bit version should allow you to use very large input rasters.

    Let me know how it goes!
    Post edited by jdouglass on
Sign In or Register to comment.