Working with remotely sensed data to evaluate land cover classes in the Black Water National Wildlife Refuge

Problem:

The goal of this project is to use training polygons to perform a supervised classification of land cover classes in the Black Water National Wildlife Refuge.

Analysis Procedures:

ArcMap was used for this analysis. Tools the image classification toolbar, specifically the training sample manager, draw polygon and interactive supervised classification tools within the toolbar. Data used was an image file (.tif) of Black Water National Refuge.

The assignment consisted of two parts. The first step was to create training polygons that would be used in the supervised classification. This was done by opening the training sample manager and drawing polygons, at least two for each of the six land cover classes. Then an interactive supervised classification was performed using the training polygons created. The output file was then exported to a GRID in order to calculate areas for each of the classes. The output was also compared to the provided image so that areas that were not properly trained could be re-evaluated. This process was repeated to create a better output by adding additional training polygons for the areas that did not match up well to the original image.

Process Diagram showing the flow of the Analysis Procedure

Results:

• 

• 

Map on left shows the training polygons for each of the six land use types used for the supervised classification. Results of classification shown in map on right.

• 

• 

Map on left shows the training polygons for each of the six land use types used for the revised supervised classification. Results of improved classification shown in map on right. Notice the addition of the developed land use type.

Application and Reflection:

The assignment has taught me how to perform a supervised classification of land cover classes using remotely sensed data. This is a very important topic as remote sensing becomes more widely used. A situation that the skills from this assignment could be applied is through the classification of land cover that lies within flood plains. The data needed would be imagery (tif file) for an area close to the coast of North Carolina as well as a FEMA flood map layer. The creation of training polygons and subsequent interactive supervised analysis could be utilized to determine what type of land cover lies within these high risk areas. This can be especially important for land cover classes of developed areas as well as things like hog waste facilities which is an important topic in North Carolina specifically, This analysis would gain insight on the percentage of developed areas that lie within floodplains and the percentage that is a high risk water contamination possibility (i.e. the hog waste). This technique would be beneficial in this situation because the training could be performed on a small section and then be used to classify a larger area. This analysis would go one step further than our assignment and would include overlaying the floodplain layers with the results of the interactive supervised analysis.