Water Yield comparison old version

Hi, 
Which improvement are in the version 3.2 with regard to the version 3.0 of Water Yield model.? I found large differences in the water yield results between these versions (more than 100%). The simulations show water yield results larger in all sub-basins with the version 3.2 than with version 3.0. Suggestions?
Does anyone else find this problem?
Thank you!
Natalia

Comments

  • DougDoug Administrator, NatCap Staff
    Hi Natalia,

    I believe the main changes to the model itself was handling LULC's such as open water, urban and wetland. In these instances, we use a different equation for the actual evapotranspiration (See Users Guide, Hydropower, eq 2). So in the Biophysical table, there is another column "LULC_veg" which dictates which equation is used. 

    We have also made many updates to our core gis algorithms, which I'd have to dig into.

    Do you think you could check your LULC codes and the LULC_veg column in the Biophysical table to make sure the data has been updated correctly?

    The User's Guide can be found here for further help on those fields:

  • NataliaNatalia Member
    edited December 2015
    Hi Doug, 
    Thank you for the response!
    The version 3.0.0 included this change in the LULC  and the column LULC_veg in the Biophysical table (in fact, the biophysical table of 3.0.0 versions is just like to the 3.2.0 version).
    Do you think that there is another update that change drastically the results between these model versions?
    Thank you!
    Natalia
    Post edited by Natalia on
  • DougDoug Administrator, NatCap Staff
    My apologies, I shouldn't have assumed that's when that change was made. I'll dig more into the differences and report back. Thanks for your patience.

    Doug
  • DougDoug Administrator, NatCap Staff
    Hi Natalia,

    One thing that did change was the pixel count for each subshed or zone seems to be slightly different. This will in turn effect the results. This is most likely due to the internal switch of our geoprocessing core algorithms. I'll converse with our lead on that and get back to you.

    Doug
  • Hi Doug, 
    I found an error in the input file, sorry!!
    The differences between both versions are now in the order of 5%. Only in two sub-basin I found differences around 15 % probably due to the change in the pixel count that you mentioned.
    Sorry again and thank you for your help!
    Regards,
    Natalia
  • DougDoug Administrator, NatCap Staff
    No worries! Thanks for bringing the question up, as it has sparked a discussion for how we can better track our changes.

    Cheers,

    Doug
  • RichRich Administrator, NatCap Staff
    Hi Natalia, and future thread readers, just an FYI, a 5-15% difference in output between the model versions seems extreme unless there's a bug that was patched.  Offhand, I suspect we patched a bug from 3.0 to 3.2 but didn't document it somehow.  Doug is traveling at the moment but he's going to follow up.  We've also made a formal issue for this in the repository that you can track if you like: 

  • Thank you Rich!! I will follow the updates of this issue.
  • DougDoug Administrator, NatCap Staff
    Hi @Natalia

    Just wanted to update this post with the latest on our investigation. It looks like between InVEST 3.0.0 and InVEST 3.1.0 there was a fix made in masking nodata values. The older version did not properly mask some nodata values, leaving behind pixels that should not have been there. In the newer versions that bug was fixed and the pixel count decreased, causing values to be slightly different.

    This should be the cause for the differences.

    I'm going to mark this as resolved, thanks!
This discussion has been closed.