Crop Pollination: Model gives larger than expected edge effects when run at different extents

Hello,
we are currently running the InVEST crop pollination model at two different extents for some tests (150 x 150 km and 50 x 50 km). When we compare pixels at the edge of the small extent to the corresponding pixels at the large extent we get different values (sup_) up to 10 km from the edge (our max foraging distance is only 2.2 km). Does anyone know why the edge effects are so much larger than the max foraging distance? Based on the equation in the documentation it appears that it might be related to the Euclidean distance between cells (Dmx) that is considered ... do you know how this distance is determined? 
Many thanks!
Tatyana
Tagged:

Comments

  • jdouglassjdouglass Administrator, NatCap Staff
    Yes, it's definitely possible for pixels on the smaller extent to have different values than on the larger extent.  Neighborhood calculations (especially the foraging distance) are convolutions using an exponential decay kernel, so the raster with larger extents will have more nearby pixels (pixels that would be off the edge of the raster with the smaller extent) that will affect a target pixel.

    As mentioned in http://forums.naturalcapitalproject.org/index.php?p=/discussion/935/exponential-decay-function#latest, we use an exponential decay kernel with a radius of 5x the max foraging distance of the pollinator.  As we near the edge of this max foraging distance, the pixel values will generally be quite small (especially with large foraging radii), but will still influence the output for that pixel.
Sign In or Register to comment.