Applying Supervised Classification Approaches to Landuse Cover

Machine learning is a powerful tool that can be leveraged to understand complex environmental patterns and make predictions. In this workflow, I use supervised classification (decision tree classifier) to identify landuse cover in Santa Barbara, CA through predictions. Land use cover is classified based off of a training data set containing small parcels of know land use cover type. I then use the spectral resolution data from Landsat 5 to identify spectral patterns from 6 bands associated with these parcels to create a predictive model. This is then applied to the entire region of Santa Barbara to predict whether a plot of land is: green vegetation, dry grass or soil, urban or water. This is workflow can fundamentally be applied to a variety of different environmental topics and is an invaluable skill that aid in addressign a variety of environmental issues.

Note: the data associated with this analysis is too large to include in the GitHub repo. Instead, download data from here. Unzip the folder and all the contents and store in your directory as follows.

Project Structure

houston_blackout
│   README.md
│   Rmd/html/Proj files    
│
└───data
    │   gis_osm_buildings_a_free_1.gpkg
    │   gis_osm_roads_free_1.gpkg
    │
    └───ACS_2019_5YR_TRACT_48_TEXAS.gdb
    |   │   census tract gdb files
    |
    └───VNP46A1
    |   │   VIIRS data files