Image Classification

Strategies: The strategies adopted for this analysis are explained as follows. I used ArcGIS Pro (Version 2.9) software. I used the following Modules/tools for the analysis: Add field tool, Export raster tool, and Classification tool (Image Classification). Data and data types – The data include a tiff file of aerial photography of the Black Water National Wildlife Refuge. Data sources – GIS 520, Fall Semester, NCSU, and Black Water National Wildlife Refuges. 

Methods: I added the tiff file to the map. I changed the band of the image from its true color to an infrared color by changing red to Band 4, green to Band 3, and blue to Band 2. I created training samples for each of the 7 land cover classes identified in the aerial photography by drawing at least two polygons per each land cover class. I merged all the polygons for each class to create a single multipart polygon per class. I saved the training samples as a shapefile. Then, I used the classify tool under the classification tool (image classification) to perform the image classification by selecting the aerial photograph from the content pane and using the saved training samples (shapefile). I compared the result to the orthophoto and identified the areas that were not correctly classified. On the second attempt, I created more training samples with at least 3 polygons per each land cover class. I also redefined the training samples by selecting specific areas that look different and are not well classified in each landcover class in addition to the previously selected areas. I merged each of the polygons for each class to create a single multipart polygon per class. I saved the training sample as a shapefile. I rerun the image classification tool (classify tool) by selecting the aerial photograph from the content pane and using the saved training sample shapefile as the training sample. I compared the classified image (the result) to the orthophoto. I exported each of the classified images as a GRID file using the export raster tool. I use the add field tool to add a new field to the attribute table of each of the classified images. I also used the calculate field tool to calculate the areas occupied by each land cover class in the study area based on the raster cell size. I changes the symbology as needed to produce layout maps for both the training samples and the classified images.

Problem Description: One of the primary stressors on water quality and quantity is the conversion of forest land to urban and agriculture. The purpose of this analysis is to compare the areas occupied by each landcover class in 2010 and 2020 within the McDowell Creek watershed using aerial photographs of the creek taken in both years. 

Data Needed: The data needed include aerial photographs of the McDowell Creek Watershed for 2010 and 2020 (in tiff format).

Analysis Procedures: I will add the tiff file to the map. I will change the band of the image from its true color to an infrared color by changing red to Band 4, Green to Band 3, and Blue to Band 2. I will create training samples for each of the land cover classes identified in the 2010 aerial photograph by drawing at least two polygons per each land cover class. I will merge the polygons for each class to create a single multipart polygon per class. I will save the training samples as a shapefile. Then, I will use the classify tool under the classification tool (image classification) to perform the image classification by selecting the aerial photograph from the content pane and using the training samples. I will compare the result to the orthophoto and identify the areas that are not correctly classified. I will redefine the training samples by selecting more polygons to correct areas not well classified and repeat the supervised image classification again. Then, I will create training samples with at least 3 polygons per each land cover class identified in the 2020 aerial photograph. I will repeat the process of merging, saving, classifying, reviewing, and reclassifying from the previous classification of the 2010 image classification. I will export each of the classified imagery from both years as a GRID field using the export raster tool. I will also use the add field tool to add a new field to the attribute table of each of the classified images. I will use the calculate field tool to calculate the area of each class in the classified imagery for both years. I will compare the result to identify what land cover has changed over the years