GLOBIO unexpected output

Hi,

I’ve managed to run the GLOBIO model successfully and all the msa_x_ outputs look fine except for the msa_i_ file. Below is a screenshot of the msa_i_ output I generate.

The intermediate output for distance from infrastructure comes out blank too with a zero value. I’ve tried using both raster and vector layers in the input infrastructure directory folder but the outcome remains the same. I’ve attached the log report below. Would appreciate some help.

InVEST-GLOBIO-log-2020-05-04–14_35_10.txt (13.7 KB)

Hi @KarenL,

Thanks for posting about this issue. Would you be will to share your input data to help us debug this for you? Depending on the file sizes it could be uploaded here or shared via google drive, dropbox, or email. My email is ddenu@stanford.edu.

Best,

Doug

Hi @KarenL

When looking through some of your inputs I noticed that the pasture input raster was all 0 values with large no data holes. I also noticed that the potential vegetation raster was all 1 values except for a small nodata hole.

I’m trying to bring myself up to speed on the model, but wanted to make sure that those inputs were intended.

Doug

Hi @KarenL

There appears to be a bug in the GLOBIO model with how the infrastructure nodata values is handled. Could you try setting your infrastructure raster to have a nodata value of -1? I think that might work around the issue I’m seeing now, which could be causing the issues.

Thanks!

Doug

Hi Doug,

I’ve done that and the final MSA infra layer still remains blank although the intermediate output for the distance to infra comes up fine, which is definitely an improvement.

I’m actually using the GLOBIO in an urban setting so I’m not sure if that could potentially explain the odd output but I am going to try changing the values in the MSA parameter table for a local, urban species (if i can get that information) and see how that goes. Do you think it’d be worth a try?

Karen

Hi @KarenL,

I was able to patch the issue I mentioned above and continued the run using your data but also got more bad results. Upon further digging it looks like this set of data hits on a bug in a convolution function that we fixed a week ago, but that patch is living in our latest geoprocessing release which we are in the middle of getting into InVEST… All to say I’m not certain that changing the values in the MSA parameter table would help.

I’ve attached a file that shows the bug. The outline around that smoothed raster should not be there, and we definitely should not have negative values.

We’re actively working on getting this issue taken care of and I will respond back here when I have a new version of the model for you to try. Sorry about the inconvenience and thanks for helping us make InVEST better!

Following up on this topic as I never actually uploaded a file to show the bug!

The fix for this convolution issue is done and the uptake into InVEST is actively being developed.