Pascagoula, MS Land Use/Land Cover Map

 

During this week's lab in Photo Interpretation and Remote Sensing, we focused on Florida Land Use and Land Cover. The main goal was to assign all of the areas shown in the provided aerial photo of Pascagoula, MS to the second level of the Florida Land Use/Land Cover codes. There are nine primary Level I Codes, specifically: 

1. Urban and Built Up Land
2. Agricultural Land
3. Rangeland
4. Forest Land
5. Water
6. Wetland
7. Barren Land
8. Tundra
9. Perennial Snow or Ice

Each of these class are then subdivided into Level II classes, which are then subdivided further into higher levels that become more specific at each level. For example, Urban and Built Up Land is further subdivided into the following classes:

11 Residential

12 Commercial and Services

13 Industrial

14 Transportation, Communications, and Utilities

15 Industrial and Commercial Complexes

16 Mixed Urban or Built-up Land

17 Other Urban or Built-up Land

In order to create the final map, a new feature class was created and named LULC. Then the aerial photo was analyzed and key features were used to divide the areas into classes by digitizing polygons around similar areas. I started with the water class as these areas seemed to jump out to me at first glance. The Lakes were traced first using the Edit tool, and then followed by the Stream and Canals, which is seen to the west of the photo in dark blue. Next the non-forested wetlands within the river were created by tracing the brown areas with similar texture, which I assumed were marsh grasses. This aerial photo was primarily residential, which I identified by small to medium buildings with driveway access to a single lane road. Also, these areas had only one or two cars parked nearby. Two classes that I had trouble determining were the differences between deciduous or mixed forested areas and forested wetlands, as forested wetlands contain deciduous or mixed tree species. I classified the forested wetlands based on their vicinity to water features their brown-ish color that contrasted with the green, upland forested areas. The industrial and commercial areas differed from residential areas because the buildings were larger, with larger parking areas or paved areas surrounding the buildings. Commercial areas seemed to be located close to the main road. Agricultural areas did not have paved areas, had little to no buildings present, and were also noticeable based on the parallel rows of crops that represented the agricultural ditches use for watering crops. One method I found was useful when drawing the polygons to exactly follow the adjacent polygon shape was the TRACE feature and making sure the lines "snapped" automatically to the line or vertex of the adjacent polygons. 

Once the classes were assigned, the next task was to "ground truth" or verify the areas. Thirty (30) sample points were created randomly using the CREATE RANDOM POINTS tool. We used Google Earth Street View to  virtually "visit" each sample point to determine if the original classification I assigned was accurate. The "true" points were symbolized as green and the "false" or incorrect locations were assigned as red. There were 4 "false points" and 26 "true points." The false sample points included small forested areas or transitional areas that included parcels that had been cleared of trees and left to regrow with no maintenance. Based on these sample points, the overall accuracy was calculated to be 83.33%.