Malaysian forest clearing
We are currently working to refine a forest monitoring tool originally developed at the Center for Global Development. The tool, called FORMA, filters streaming satellite imagery toward a tractable estimate of forest clearing activity in the humid tropics at 1km spatial resolution. The intent of the system is to facilitate conservation payment schemes by releasing public and timely alerts on forest disturbance.
The FORMA alerts will be posted each month, and are meant to complement high resolution estimates of deforestation, which are updated less frequently. The New York Times, for example, recently featured the awesome work of researchers at SarVision, who have released detailed maps (approx. 50m resolution) of deforestation in the Malaysian state of Sarawak in two-year increments. SarVision researchers found that roughly 33 percent of coastal peatland forests in this area have been cut down since 2005 -- mostly to clear way for palm oil plantations. The SarVision maps are displayed below in a Google Earth gadget.
You can view a static image of the Sarawak data here, if you do not wish to download the Google Earth Plug-In.
FORMA is now operational for the full tropics at 1km resolution, with planned improvements to 250m resolution. We aimed the FORMA system at Sarawak in order to animate the monthly spread of clearing activity from December 2005 through December 2010. The video is embedded below; the slides are available for download here (pdf.zip, 6.8MB).
The brown pixels indicate areas that had been cleared between 2000 and 2005. This data set (Hansen, et al., 2008) serves as our training data -- the measure of clearing activity that we use to calibrate our detection algorithms. The yellow-to-red pixels that light up after Dec 2005 are FORMA estimates of clearing activity. Any pixel that lights up represents an area that had been subject to forest clearing activity between Dec 2005 and the specified month. Yellow pixels indicate areas with low signals of clearing activity -- but still sufficiently high to exceed our confidence threshold. Red pixels represent areas with strong signals.
We can now use these data to examine the short-term spatial dynamics of deforestation. For example, most of the clearing activity in this area took place on the periphery of previously cleared plots. We are currently working to better understand the economic drivers of deforestation through research on the formation of new clusters of deforestation. And we are relying on FORMA to flag the formation of clusters at high temporal resolution. More to come -- and soon!
REDD Metrics
info@reddmetrics.com
github.com/reddmetrics
