Public Lab Research note

Mapping Biodiversity Hotspots - Shannon Entropy Applied to Pixel Space (Case Study: Cape Cod, MA)

by podolsky | August 14, 2015 22:41 14 Aug 22:41 | #12144 | #12144

I developed Diversidad Software to quickly scan earth imagery (from planes, satellites, drones etc.) and find the areas of the image that have the highest spectral diversity. To find areas of maximum spectral heterogeneity, I use the same Shannon "diversity index" that ecologists use on species assemblages. But in my case I treat the pixels as species.

Here is an example on Cape Cod, MA: This is the raw SPOT Image for the Cape: Screen_Shot_2015-08-14_at_6.14.13_PM.jpg

Next, I filter the entire image and calculate the H' and H'max for every pixel in the image and convert that to a gray scale image where dark gray areas are high heterogeneity and light gray areas are low. Here is that image with the red polygons showing the areas of the image that achieved 60% or better of the maximum possible spectral diversity (by taking H'/H' Max • 100): Screen_Shot_2015-08-14_at_6.15.09_PM.jpg

Finally, to see where these areas of highest spectral diversity are found within the original image I merge the two results yielding this image final product:


Note: that most of the areas identified are in the coastal zone - and that is where the biological diversity is highest too. When one goes to these red areas and conducts a survey for most any taxa, the biodiversity within these polygons are significantly higher than it is outside of these polygons. The whole process can takes less than 1minute for several million acres (depending upon image resolution) and thereby provides a low cost, first-cut assessment of biodiversity hotspots.

Note2: the images shown here have been down sampled considerably from the original to allow easing posting here. Send me a message if you would like to examine the full res images.

Disclaimer: While areas of high biodiversity are a conservation priority there are many important species found in areas of high spectral homogeneity as well - just not as many of them.


Nice. In general, do built environments have higher or lower spectral diversity than natural environments?


Is this a question? Click here to post it to the Questions page.

Reply to this comment...

Great question Chris - indeed, built up areas are often spectrally very high diversity as one would expect given the myriad of surfaces we use. If I want to limit a Diversidad analysis to wild lands I mask out the urban or developed areas. But typically, I run everything as the results are always interesting to me. In some cases, biologically speaking, developed environments can have high diversity too though there are usually a lot of invasive species that occupy built environments.



Reply to this comment...

Thanks Richard,

Recently disturbed sites like a forest logged 50 years ago or a pasture abandoned 50 years ago will tend to have higher biological diversity than mature forest or shrubland (pioneer species are short lived and multiple early successional pathways tend to converge on fewer community types). Have you noticed that mature habitats have lower spectral diversity than disturbed habitats?


Is this a question? Click here to post it to the Questions page.

Reply to this comment...

Yes, I have noticed that mature habitats have lower spectral diversity than disturbed habitats.

Here is a recent Diversidad analysis of a 3D image from GoogleEarth of Yosemite Valley, CA. Note that there is no development at all in this image. The Diversidad hotspots are trained on "edges" or ecotones.


When I have conducted biological surveys on Diversidad hotspots they almost always support higher species diversity than non hotspot areas - especially for birds and mammals. Insect diversity does not correlate as well probably because they sample the environment in a much more "fine grained" way than do birds and mammals. In other words, for Diversidad to work on insect diversity we would need super high resolution that is sub meter in resolution.

If you want to send me an image, I would be happy to run a Diversidad Analysis for you. just email me a high res jpeg image.

Reply to this comment...

Login to comment.