Create landcover classification of the Central Valley by:
Digitizing training polygons
Creating a signature file
Running a supervised classification
Our Data folder that we are going to work with.
Same Data folder as seen from ArcCatalog
Refer to the following website to get more info on the image data we from Landsat 7 that we are going to work with.
https://web.pdx.edu/~nauna/resources/10_BandCombinations.htm
Table 1 (from link above)
2011_imagery
2014_imagery
These images are mosaiced exports of multiple Landsat 7 scenes
Table 2
Band coloration
Looking at figure 1 and comapring it with values in Table 1 we know that BLUE is associated with BAND 1 (layer 1 in ArcGIS) and GREEN is associated with BAND 2 (layer 2) and RED is associated with BAND 3 (layer 3 in ArcGIS). but ArcGIS kind of assigned the wrong values to each BAND (layer) in our data so we need to correct that.
We will go to symbology and modify things accordingly and we will also set Background to transparent.
Results after Band Coloring correction for 2011_imagery
Results after Band Coloring correction for 2014_imagery
We will be using the Image Classification toolbar in ArcGIS
Using the imagery, we will create polygons that represent several land classes.
We will use these to tell ArcGIS which pixel values are water, active agriculture and fallowed land.
Open the Training Sample Manager on the Image Classification toolbar.
Use the Draw Polygons tool to draw a polygon in an area you know is active agriculture. This doesn't have to be an entire field, just a small, representative area.
Draw additional polygons that represent active agriculture. Pan around for other fields to ensure your polygons represent the all the agriculture in your imagery. Continue doing this until you feel that the active agriculture in the imagery is adequately represented in your class. What are we training ArcMap to do here?
Select all polygons you have created and use the Merge Training Samples tool to merge all your agriculture polygons into one class. Rename the class "ActiveAg".
Repeat this process for two additional land classes: water and fallowed land. Don't worry about urban areas or vegetation in the foothills, we will account for this later.
Create multiple polygons around agricultural areas
Group all polygons under one class called Agriculture
Repeat Same process for water bodies
We will save a signature file for later use.
Scatter plot indicates slight overlap in our classes but that's acceptable for now
We will tell ArcGIS to classify based on those created classes
And the results >>
These results aren't that good, it has detected class 1 (agriculture) and class 2 (Water) well but class 3 not so much, but it is a good first try.
we can modify the signature or create a new one to enhance our classification.
We will continue doing more analysis on this one soon.