Hello. I am running HRA on InVEST 3.14.2 and am trying to include several threats that vary spatially in their intensity. However, it doesn’t seem that the variation in threat intensity is being interpreted correctly - or at all - in the calculation of Consequence or Risk.
My spatially explicit criteria layers will be a combination of both vector and raster. So far I have only tried vector files, where points in each file are assigned a ‘rating’ of 1-3 (in the attribute table) representing the intensity of the threat in that location. The ‘rating’ value field type is long integer.
When I run HRA with these spatially explicit criteria, I don’t see any evidence that the ‘rating’ is being incorporated into the resulting output files of Consequence (intermediate output) or Risk (final or intermediate outputs). HRA correctly “sees” those threat locations and buffers them according to my input file, but the intensity of the threat does not differ between the locations (points and their assigned buffer). Can you help me to understand what differences I should expect to see if my spatially explicit criteria files are being interpreted correctly?
Paths to the spatially explicit criteria files are placed in the “exposure_consequence_criteria.csv” file under the RATING column and are assigned to “C” as the Criteria Type.
I’ve attached a zipped file of my input files and a recent log file. Thanks for your help!
PSMV_HabitatRiskAssess_small.zip (3.5 MB)
Update: I also tried supplying the spatially explicit criteria vector files with tiny polygons instead of points. This did not resolve the issue. All the threat locations seem to be treated equally with no difference based on ‘rating’.
I suspect the rating info is either not supplied in a recognizable format (although I think I have followed all the user manual instructions) or that it is being lost in one of the intermediary steps.
Hi @magelcai and welcome to the forum! Thanks for your question - I have been looking at your data and will see if I can figure out your issue.
1 Like
Hi @magelcai,
I think we have figured out where the confusion is coming from! If I’m interpreting your question correctly, it seems like you’re expecting the different ratings for your different spatially-explicit point criteria to cover the entire area that is affected by the stressor (i.e., the ratings of points to affect the buffer space around the points). In reality, a spatially-explicit criteria vector only affects the underlying area, so one point will only define the rating for the one pixel. The model will just take your spatially explicit criterion (point vector) and rasterize it, so then after the E/C calculations, you’ll get a different score on that one pixel. Looking through the user guide, this is definitely not made clear enough so we will try to fix this!
As for your data, it looks like most of your points don’t directly overlap the habitat they are expected to impact, so it may be hard to see that the rating is actually having an effect. However if you zoom in on the one point in the ImperviousSurfaceQuartiles_PointSource shapefile that intersects the eelgrass polygons shapefile, you can see that the underlying pixel does have a different risk value.
If you want the different ratings in your consequence criteria to have a more spatially extensive impact and extend throughout the buffer area, you could use polygons instead of points.
Does this address your question? Am I interpreting your issue correctly?
1 Like
Oh, interesting. So the rating value from the feature is not factored into the E or C calculations in the buffer zone? That is good to know and definitely changes my approach to these input layers. Where I have a more intense threat I would expect the consequence to be higher throughout the buffered region and would like to represent that in InVEST. In my case, those datalayers represent watershed pour points with varying levels of imperviousness and industrial land use. We expect that more modified watershed likely deliver a great intensity of threats (nutrients, sediments, contaminants) throughout the zone of influence (the buffer).
So it sounds like I have two potential options for representing this spatial difference in rating throughout the buffered areas:
- Perform my own buffer of those points in GIS and create large polygons with the assigned rating value (rather than buffering the points in InVEST)
- Create a raster layer with the appropriate grid size that contains the rating value in the locations where I wish to represent those impacts
Can you confirm that both options would produce the desired model behavior? Can you think of any reason I should use one approach over the other?
Thanks, again, for your help!
Yes, the rating for a given vector feature will not affect the buffer around that feature, it will only impact the pixel(s) that spatially overlap the feature. I was also a bit confused by this, because while the shapefile referenced in the Habitat Stressor table and the spatially-explicit criteria in the E/C Criteria Scores table are the same file in your use case, they are actually separate components for the model, and the rating attribute is only read as part of the criteria score. (In the InVEST sample data for HRA, you will notice that the files referenced in the stressor table and criteria table are actually different, where the shapefile in the Criteria csv contains a ‘rating’ attribute and the shapefile referenced in the Stressor csv has no rating field). The stressor data is simply presence/absence.
So it sounds like I have two potential options for representing this spatial difference in rating throughout the buffered areas:
- Perform my own buffer of those points in GIS and create large polygons with the assigned rating value (rather than buffering the points in InVEST)
- Create a raster layer with the appropriate grid size that contains the rating value in the locations where I wish to represent those impacts
Can you confirm that both options would produce the desired model behavior? Can you think of any reason I should use one approach over the other?
Yes, if you want the area influenced by the rating to cover the current buffer area, you can use polygons or raster data. It would probably be simpler to use a vector approach and buffer the point shapefile you have currently (and ensure this new vector contains the rating field) and input this into your Criteria table, as you don’t need to worry about raster alignment and resolution. However, using a raster would give you more fine-grained control over where exactly different ratings are influencing risk and would allow you to explicitly encode spatial variation in intensity across the buffer zone.
For your convenience, here are the relevant user guide instructions on extracting rating scores from spatially explicit criteria layers:
“The Rating will be extracted from the spatial data as follows. If a raster file is used, its pixel values will be used as the Rating and therefore pixel values must be between 0 and the Maximum Criteria Score. If a vector file is used, the Rating value will be extracted from the attributes of the features. An attribute field “rating” must be present with values between 0 and the Maximum Criteria Score.”
Hope this helps!
1 Like