Crop Production Tool - Yield values are very small

pamccord

I am working with the Crop Production tool and I'd like a little help understanding what seems to me to be a strange output.  

I have a study area that is about 700km by 400km. Within this study area there are multiple patches where soy is being grown for which I'm interested in obtaining the yield values. The first file that I've attached (entitled Soy1_Other0.JPG) is how I've set up the raster for the Crop Management Scenario Map. The pixels are about 10km by 10km in size and the pixels with a value of 1 represent the areas that I'm interested in where soy is being grown. Pixels with a value of 0 represent all other land cover types. However, when I run the tool with this as my Crop Management Scenario Map, I get extremely small yield values (ie, approximately 0.000000001 for most cells).

On the other hand, when I run the tool with the entire study area (700km by 400km) set to 1 (in other words, setting the whole area to 1 for my Crop Management Scenario Map) as I show in the second file that I've attached, I get more reasonable yield values. 

Is there any reason why I'm getting such small yield values when I limit the pixels that are producing soy within my Crop Management Scenario Map?

Thank you

Rich

Hi @pamccord, sorry for the delay, I've been swamped. First off, can you verify that you're running at least InVEST 3.3.3? That'll have the most recent version of the crop production model in it. Earlier versions did not work correctly.

If you are running 3.3.3 and you're still getting these issues, would you mind sharing your datastack with me via Dropbox to richsharp@stanford.edu? I'll take a look and see what I can do on my end.

pamccord

Thanks Rich,

I've sent you the data via Dropbox. I'm running version 3.3.3, and I've set up the parameters as follows:

Workspace: [user defined]

Results Suffix: [Empty]

Lookup Table: crop_lookup_table.csv

Crop Management Scenario Map: AOI_raster_all1.img OR SoyAreas_Morton6_presence2_reclass.img

Global Dataset Folder: [Using the global dataset sent when downloading InVEST]

Yield Function: observed

Percentile Column: yield_95th

All other fields are left empty.

When I run the model with AOI_raster_all1.img as my Crop Management Scenario Map everything seems to work fine. However when I use SoyAreas_Morton6_presence2_reclass.img, then I get very small values that seem to be erroneous. Please note that the SoyAreas_Morton6_presence2 dataset is slightly different from the one that I mentioned in my initial post because I was having trouble running the model with that initial dataset. The cell size for SoyAreas_Morton6 are smaller than those from AOI_raster_all1, but I would think the model would still run correctly (ie, without such small yield values).

Thank you!

Rich

Hi @pamccord, gosh I'm really sorry about this, we refactored and overhauled the crop production tool in a development version of InVEST, but haven't officially released it yet. It does a lot of things, including fixing all the bugs in the old version, drastically simplifying the user interface, making the data directory an order of magnitude smaller, and just in general running faster.

We're close to releasing InVEST 3.4.0 very soon and we have a release candidate here that mostly likely will work just like the final version: http://data.naturalcapitalproject.org/invest-releases/3.4.0rc2/InVEST_3.4.0rc2_x86_Setup.exe

Would you be willing to try it? Note there are two "crop production" tools now. One is "percentile" based that operates on the 110+ crops in our global database. There is a regression one too for predicting yields, but it works on a subset. 

Also, here's a much simplified users guide too: http://data.naturalcapitalproject.org/nightly-build/invest-users-guide/html/crop_production.html

Otherwise, I can verify what you're seeing is a bug in that model.

pamccord

Thank you @Rich! This addresses all of my concerns. I haven't tried the candidate version for 3.4 yet, but I will experiment with it soon and let you know if I have any questions at that time. Simply knowing that the previous results were due to a bug is very helpful.

Thanks again!