Vegetation-health-report

Use this repo to generate vegetation health reports for the input AOI (Area of Interest) geojson or shapefile.
Input AOI --> generates NDVI --> performs kmeans clustersing --> generates PDF report.

Features

Uses cloud optimized GeoTIFF (COG) format to avoid downloading large files onto your local machine
Option to provide start date and end date
Automatically calculates ideal no. of clusters (kmeans) using elbow curve method
Generates an interactive folium map (html) with identified clusters
For each available date, generates a pdf report containing RGB, NDVI, NDVI classes, Kmeans clusters, and clusters superimposed on RGB
Script to download data can be used as an independent module with options to specify data source (sentinel, landsat, etc.), bands, cloud threshold, etc.
Applies a precautionary buffer of radius=0 at input AOI, to make sure polygons are valid

See

Please see Exploratory Data Analysis (EDA) readme for more details on data.

Sample results for a

Usage

Run the script main.py and pass the following arguments:

"--aoi": path to aoi vector file (geojson or shapefile)
"--clusters" [optional]: Number of clusters desired in output. Will auto compute ideal no. of clusters if left blank
"--start_date" [optioanl]: start date in YYYY-MM-DD format. Will take today's date if left blank
"--end_date" [optional]: end date in YYYY-MM-DD format. Will take today's date if left blank
"--out_dir" [optional]:path to directory where data will be generated
"--crs" [optional]: target CRS of the generated data. If skipped, CRS will match input aoi's CRS
"--cloud_threshold" [optional]: Cloud cover threshold in %. If skipped, the default value is set to 5
Example 1: python main.py --aoi resources/test_aoi_river.geojson
Example 2: python main.py --aoi resources/test_aoi_river.geojson --start_date 2021-08-17 --clusters 3 --cloud_threshold 15

Installation

These steps are for Ubuntu (might differ a bit for MacOS and Windows) Ensure that GDAL is installed in your environment. See this page for gdal's installation instructions.
Step 1: Clone the repo --> git clone https://github.com/seedlit/vegetation-health-report.git
Step 2: move into the repo --> cd vegetation-health-report
Step 3: Create a new virtual environment --> python3 -m venv venv
Step 4: Activate the virtual environment --> source venv/bin/activate
Step 5: Install the dependencies --> pip install -r requirements.txt
Step 6: Run tests --> python test.py

TODO:

remove dependency on gdal. Most of the gdal's functions that I am using can be completed by rastertio or some other library. This would help in creating a pypi package for this repo
create a package - conda or pypi
Incorporate the generated folium maps in generated pdf report (as screeshots). Existing solution use selenium to open the html in browser and take a screenshot. But ideally, I would like to avoid that.
Degub cloud cover checks issue: the cloud cover threshold doesen't seems to be working
Use multiprocessing to speed up
Make async. For example, no need to wait for data to get downloaded for all dates. Start wokring as soon as data is downloaded for a date. And remaining downloading can continue in background
Add area for each NDVI classes. For eg. 70% area belongs to moderate vegetation, etc.