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

 

This week’s assignment focused on learning how to classify land use and land cover (LULC) from high-resolution satellite imagery and then evaluate how accurate those classifications were. We worked with an aerial image of Pascagoula, Mississippi, and created our own data by digitizing different land use types.

First, I created a new polygon feature class and digitized different land cover areas by identifying patterns of tone, texture, shape, and association. I assigned each polygon a Level II LULC code and a short description. My final categories included residential (11), commercial and services (12), deciduous forest (41), forested wetlands (61), non-forested wetlands (62), lakes (52), streams and canals (51), and bays and estuaries (54). Learning to recognize these features based on visual cues made me more aware of how land use reflects both natural environments and human activity.

Next, I conducted a ground truth accuracy assessment. I created 30 sample points around the map and then used Google Maps to check what was actually at each exact coordinate. For each point, I marked whether my original classification was correct (Y) or incorrect (N), and if incorrect, I updated the classification to what it should be. This step showed me how important precision is when mapping. Sometimes, the surrounding area matched my category, but the exact point did not, which still counted as incorrect.

My results showed that Commercial and Services areas were the most accurate in my original classification, likely because buildings and parking lots show up clearly. The category with the lowest accuracy was Streams and Canals, which can be hard to distinguish when the water is narrow or partially covered by vegetation. Overall, I was able to calculate my percentage of correct classifications and visually display true and false points on the map.