A Fundamental Challenge for Remote Sensing: What is a Forest?

The NASA Earth Observatory has a nice article, “Is That a Forest? That Depends on How You Define It”, accompanied by some nice images that illustrate the issue.

As discussed in a detailed letter in Nature Climate Change ([1], ask your librarian to help get access to the full article), one of the key concerns for climate science is the extent of forests on land. But, how shall we measure “a forest”?

Sexton and colleagues point out that there are more than 800 (!) definitions proposed for the concept of a “forest”. In the end, they boil down to a measure of the tree cover. The United Nations Framework Convention on Climate Change specifies measures that range from 10% to 30% tree cover, which is quite a range.

NASA is involved in this question because most of these definitions can be measured with Earth Observation satellites, giving us a global measure of the extent and health of forests over time. The Sexton group’s work is “the first global, Landsat-based map of tree cover (at 30-meter resolution)”. The post provides a nice comparative map, showing how much different the 10% versus 30% makes.

While there are some technical issues in measuring forests from space imagery, these uncertainties and errors are tiny compared to the differences from alternative definitions. Different estimates give as much as 25% difference in estimated forest cover, which can have extremely significant effects on policy.

In the end, the term “forest” may not be a useful concept for either scientific understanding or environmental policy. “But although ‘forest’ is an intuitive label for general discussion, the intuition transcends neither cultures nor ecosystems, and the term is neither directly relevant to carbon nor directly measured from satellite or in situ observations.” P. 4There are many measures of the characteristics of forests (e.g., canapy height, biomass, etc.) which can be unambiguously measured, and also apply to other biomes, including cities.

I would comment that this case resembles many cases in remote sensing from satellites and other devices–including UAVs.  Connecting the raw data to important (and actionable) concepts, such as Carbon budgets is complicated and requires semantic definitions that may be subjective or motivated by “non scientific” reasons, such as political or cultural definitions.

These difficulties are handled by careful and transparent discussions, and careful documentation of the semantics that are used in specific studies and conclusions. In other words, when done carefully, these semantic problems are handled in a rational way and in a way that lets others understand what was done (even if they want to do it a different way).

I make this point to caution amateurs who wish to look at images on the web and draw their own conclusions.  This may be how it works on TV, but it isn’t going to lead you to valid conclusions.

Similarly, the notion that slapping a web cam on a small drone is going to yield useful understanding of the environment is, well, unlikely in the extreme.


  1. Sexton, Joseph O., Praveen Noojipady, Xiao-Peng Song, Min Feng, Dan-Xia Song, Do-Hyung Kim, Anupam Anand, Chengquan Huang, Saurabh Channan, Stuart L. Pimm, and John R. Townshend, Conservation policy and the measurement of forests. Nature Clim. Change, advance online publication 10/05/online 2015. http://dx.doi.org/10.1038/nclimate2816


Space Saturday

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s