Remote Sensing

At the Northeastern Research Station FIA unit, one of our top priorities is the statistical analysis of forest inventory data.  In order to make the best possible estimates of the amount, quality, and health of our forests at county, state and regional scales, we statistically stratify the data.  We use remote sensing information to help us do this.  Below is an zoomed in classified satellite map of Connecticut, with some forest inventory plots on top.  The green areas are forested, and the red are nonforest.

A closer look at the data reveals how landscape complexity can affect the accuracy of our classification procedure and the characteristics of the classified map.  For example, the proximity of a plot to an edge between two classes or a road can affect our accuracy assessment procedure.  Other factors that might affect the classification accuracy of our technique include GPS and satellite georeferencing inaccuracy.  We have determined that the below schematic of our plot is a reasonable representation of zones of uncertainty surrounding the actual plot locations (gray and white squares = 30 m. Landsat pixels; dots = subplot centers; white circular area = actual plot area; dark, circular shaded area = potential GPs positional inaccuracy; lighter circular shaded area = potential satellite image positional inaccuracy).