Module 4 - Ground Truthing and Accuracy Assessment

In this week's lesson we learned about two major ground truthing methods used to assess land use/land cover classifications: in-situ and ex-situ measurements. For our lab assignment, we conducted an ex-situ ground truthing measurement to quantify the "overall accuracy" of the LULC classification codes we categorized in our last lab session (Module 3).

In our lab assignment, we were instructed to create 30 sample points within our study area. My method of selecting sampling locations began by first selecting a sampling pattern per the image depicted in our lab instructions (shown on the left). I decided to recreate a “stratified random” system by creating equally-sized 5x6 quadrants of our study area in Pascagoula, Mississippi using the Fishnet tool in ArcMap. Then, I decided on a method of choosing at least 1 sampling point in each quadrant while attempting to meet all of the following criteria: selecting sampling points for each of my land use/land cover classifications, selecting sampling points near a street (while still staying within the class boundary) so that it would be easier to verify it against Google Maps’ Street-Views, and by being consistent in how closely I zoomed into each quadrant so that my selections were consistent as well.

While conducting my accuracy assessment, I maintained notes in the attribute table of the Truthing.shp sample points. Out of my original LULC classes within my 30 sample points, class "121 - School" had the highest accuracy; and class "61 - Forested Wetland" had the lowest accuracy. The percent of sample points that were correct was: 50%. My final map product is at the top.