Lab 2: Land Use Land Cover Classification & Ground Truthing

 This week in lab, I used the visual interpretation concepts we covered last week in lab and lecture to classify land areas in a aerial image of Pascagoula, Mississippi. I used a land classification scheme to code everything on the image to the second level - represented by a double digit code. Each classified polygon falls under a Level I classification - represented on the legend by a single digit number. The most difficult part for me was digitizing the wetlands and water features because there are so many smaller streams with complicated boundary lines. 

Then, I used ground truthing to verify my classifications. Since I didn't have access to this exact location to conduct a true ground truthing, I used Google Maps street view. To create a random set of 30 points for sites to verify, I created a new feature class, created a single polygon over the entire image, and used the create random points tool. I located each point on Google Maps and classified each point as 'accurate' if the exact point was the same classification of the classified polygon it was in or 'not accurate' if it was different. These points are shown as green and red points on the map. I calculated an accuracy assessment and found that 73% of my Land Use Land Cover classifications were accurate.