by Joanna Hsu
November 7, 2012
Iryna Dronova, a graduating PhD student in Peng Gong’s group, dazzled us at Friday’s lab meeting about how remote sensing might be useful in ecology and restoration. A fly-by review of her presentation:
1. The problem: field work takes time, money, and often painful logistics. Furthermore, measurements in the field may not be accurate or scale up to relevant scales.
For example, I have spent a good deal of time squinting at plots to estimate percent of vegetation cover. It is one of the places in life I second guess myself and others the most. I have no doubt that you can identify grasses to the subspecies based on a fragment of last year’s flowering stalk, but can you also get the percent cover in this plot right? (And by “right”, I mean “the same answer as me”?)
2. Enter remote sensing and image classification, which can extract useful information from satellite images because different land and vegetation covers have different spectral signatures.
3. However, classifying blindly based on the spectral signatures of pixels alone often hides meaningful entities.
4. Adding scale-dependent, object-based image analysis helps considerably. This image segmentation allows us to more easily separate interesting entities from similar-colored background.
5. A cool example from Iryna’s research: wetlands research at Poyang Lake, China. Swaths of land dominated by different plant functional types – C3 grasses, C3 forbs, C4 tall grasses, C4 short grasses – can largely be separated based on their spectral signatures if object-based image analysis is applied at the right scale (figure below from Dronova et al. 2012).
A species’ “functional type” or “functional group” refers to its role and function in an ecosystem, as opposed to its taxonomy. Ecologists often group species based on their contributions and responses to their ecosystems – it may be more ecologically meaningful that a species is a perennial shrub than that it belongs to the Ericaceae family, subfamily Diapensiaceae.
6. As for the hard-to-classify land areas, a field visit can clarify the ambiguity. Sometimes these areas are the most biodiverse, composed of several plant functional types. Or, we could also improve our classification success rate by defining new optically distinguishable plant functional types, as these authors suggest. (Thought: then does the term “functional” become more about being functional for ecologists as opposed to functional for ecosystems?)
7. There are lots of ways questions in ecology and restoration could be addressed using remotely sensed data. Iryna’s list is below: