Embedded Files
Identifying Land Coverage Classes for the Black Water National Wildlife Refuge
Problem and Objective
As a GIS Analyst for the Black Water National Wildlife Refuge, I have been tasked to develop a land cover dataset of the entire refuge aerial photography. I am to perform a supervised image classification using one aerial photograph of the refuge.
Analysis Procedures
To complete this task, I will use the ArcGIS Pro software. The only data I started this task with was an aerial photograph of the Black Water National Wildlife Refuge. Within the software, I used the Training Samples Manager and the Image Classification tools for the supervised classification. I also used the Build Raster Attribute table tool, the Add Field tool, and the Calculate Field tool to calculate the total area of each land cover class.
To complete this task I started by first importing the aerial photograph into the ArcGIS Pro project. I then began using the training samples manager. I created a new scheme and added the necessary land cover classes. From there I created polygons for each class that will act as training samples. Once I did this, I used the classification tool and the new training samples to classify the aerial photograph. With the resulting raster image, I used the Build Raster Attribute Table tool to create an attribute table. I then used the Add Field tool and Calculate Field tool to add and calculate a field for the area of coverage for each class. After reviewing these results, I determined I needed to edit the training samples and try to classify the image again. I went back to the training samples manager and added several more polygons for each land coverage class. I then repeated the steps to classify the image and calculate the area coverage. I then was able to compare the two classifications, both to each other and to the original photograph. And I noticed that the second classification was more accurate once there were more training samples used.
Results
The map above displays the first attempt at classifying the Black Water National Wildlife Refuge. The Developed, Herbaceous, Barren, and Cultivated classes to not classify well with these training samples. The Water, Wetland, and Forest classes did classify well.
The above map is the improved classification of the Black Water National Wildlife Refuge. After adding multiple samples for each class in the training samples set, all classes were classified well.
Application and Reflection
I learned the new skill of image classification from this task. This will be extremely valuable if I ever need to work with satellite data. Below I describe another situation where these skills would be needed.
Problem Description
I work as a GIS analyst for the city of Raleigh. They are interested in seeing how the land cover of the city has changed over a certain period of time. I am to create training samples for the city and then classify an old image of Raleigh and a more recent image of Raleigh. I then can see how the land use has change over the years.
Data Needed
I would need to download two satellite images of Raleigh: one from the beginning of the time period and the other being a more recent photo. I would use the USGS EarthExplorer to locate and download these images.
Analysis Procedure
Once I download the necessary satellite images and import them to ArcGIS Pro, I would create training samples before performing the supervised classification. I would utilize the NLCD2011 land cover classification legend to determine the necessary groups and their symbolgy. Once creating numerous samples for each group and saving the training samples, I would be able to perform the supervised classification for each image. From the classification, I would be able to calculate the area of each land cover class for both images and compare them. This would allow the city of Raleigh to know how the land use has changed over the years.
Page updated
Report abuse