Module 2 - Land Use Land Cover Classification (LULC), Ground Truthing & Accuracy Assessment - Lab Assignment

 This module tied in the visual interpretation methods I learned in module 1 to a real-world application: Land Use Land Cover Classification (LULC) assessments. I was to apply visual interpretation and recognition elements to an aerial photo of Pascagoula, MS to identify and classify areas along the photo and construct an LULC map. Next, I was to develop a sampling plan for ground truthing and implement it to properly assess my LULC map for ground truthing.

To construct my LULC map of the aerial photo (provided as a .tif file), I first opened the photo in ArcGIS Pro. I looked at the photo both from large and small scales and developed a plan. I created a new feature class in my project's geodatabase with a polygon feature type and created fields for both the code I would classify each polygon, as well as a description of the land use or land cover. I began digitizing polygons, starting with larger scale features and working my way towards smaller ones. I did not get into extremely fine detail, as this process would take far too much time for the purposes of this class. What's more, I wanted to balance my minimum map unit (MMU) to the resolution of the photo. Once the entire area of the photo had been digitized and coded according to USGS standards, I began symbolizing the area for reader purposes. 

I accomplished ground truthing by using an overall accuracy method in which I digitized 30 points along the aerial photo. My method was a mix of random sampling and stratified sampling. It was stratified in the sense that I wanted to distribute the points over many of my LULC classifications and random in that I distributed points along each individual classification as random as possible. Once I had created all 30 points, I found the location of each point in Google Maps. I carefully looked at each point location in both Street view and Google Earth to determine whether the point fell into the correct classification. I found only four points out of the 30 digitized to be incorrect, resulting in an 87% overall accuracy assessment.

The image above shows the final map layout combining elements created from the LULC analysis and ground truthing assessment. The layout contains a fully digitized aerial photo of Pascagoula, MS with polygons symbolized by LULC code identification. The points represent ground truthing sample points and are symbolized by whether they were accurate (green) or not (red). The layout contains all standard map features.