Ecosys_risk.tiff gives wrong impression when habitat polygons border / HRA model


When I was looking at the output maps from my HRA model I noticed that cells that include the border between two habitats show a higher risk than they should when looking at the Ecosys_risk.tiff. I assume this happens because the cumulative risk is calculated for each cell per habitat and the ecosys_risk map is created by summing these rasters across habitats. In cells that have more than one habitat ideally you would want a weighted average according to the % cover of habitat in that cell or in a simplified version just cumulative risk divided by number of habitats. But since it sums these values I think this output gives the wrong impression when you have small habitats with lots of borders.

Since there are the maps of cumulative risk per habitat this problem is not that important for me but I am wondering if this also affects the html  'ecosystem-risk-plot'? In that case I might be better off by running the HRA model with a very small cell size to calculate this plot.

Hope you are able to help or confirm,

Thanks, Tom


  • Hi Tom,

    Thanks for your comment.  I think I'm following what you're saying and have some ideas.  But it would help me to

    1) know the overarching question you are tackling with your analysis and
    2) see a visual of this issue of high risk along the borders of habitats/where they overlap.

    Can you write a couple of sentences about the question you're asking with the HRA model and attach a few screen shots: a) zoom in on maps of habitats to show example of overlap, b) ecosystem risk map and/or cumulative risk by habitats, c) ecosystem-risk plot and/or individual habitat risk plots.

  • Thanks Katie,

    1) The idea is to run the invest model for a marine area in Portugal, visualizing the risk posed to marine habitats (EUNIS classification) by human pressures. At this stage I was just running the model with some invented pressures.

    2) Regarding the ecosys_risk tiff, I added three images showing the problem. 

    -The habitat map.

    -The Ecosysrisk image shows the problem, the black lines are the borders between habitats where the model calculates a cumulative risk for both habitats and shows a summed value. Like this it seems as the highest risk exists at these places. 

    -The HRA_outputs_combined map: This map was created by drawing the cum_risk files for the different habitats. In this case the raster pixels that have two habitats in them will take the risk value of the habitat that is on top in the drawing order. (I didn't equalize the color ramps so it doesn't completely match the ecosys_risk image). 

    I hope this explains the problem.

    Thanks, Tom
    632 x 535 - 49K
    631 x 535 - 87K
    630 x 533 - 87K
  • Hi Tom!

    Quick question. How did you get the color ramps to show different shades. I usually just have high a low which only shows two colors. I would like it to represent the different risk values each cell has.

    Thanks, Valerie
  • Hi Tom,

    Thanks for sending your figures and the goal for your analysis.  I have a couple of comments and suggestions.

    1) Yes, you are interpreting the model correctly.  Ecosystem risk does sum risk for the different habitats.  I understand your thinking about the weighted average by % cover of habitats; however, in some cases in marine environments the habitats actually do overlap (e.g., canopy kelp and understory kelp).  While it is debatable, the idea behind ecosystem risk is that more habitats at risk may actually lead to a greater overall risk for the ecosystem.  Several approaches (e.g., Halpern et al. 2008 Science) also sum cumulative risk/impact by habitat to get at overall risk.

    2) However, I agree with you that this issue makes the ecosystem_risk output difficult to interpret. I much prefer exploring the risk by habitats separately.  The cumulative risk by habitat results are the ones we've found most useful in our applications to inform coastal and marine spatial planning (see Arkema et al. 2014 Environmental Research Letters, Arkema et al 2015 PNAS).

    3) Lastly, the best way to get around this issue of more than one habitat is to make sure you are running the model at a small enough resolution to differentiate between habitats where they in fact don't overlap.  While I usually run the model initially with a grid cell size of 1 km or 500 m to make sure it runs without errors, I ultimately use a resolution of 250 or 200 m.

    Hope this helps!
Sign In or Register to comment.