Installation

Ensure that you have a C compiler (gcc), git, virtualenv and its required dependencies installed. The versions from the system reposity on Ubuntu 18.04 are sufficient. Then, this repository can be installed via:

git clone --recurse-submodules https://github.com/LyteFM/tree-cover-server.git
virtualenv --python=python3.7 venv
source venv/bin/activate
pip install -r requirements.txt

Compile the sqlite extension file into the tree cover directory. Instructions are given for Linux, for other OS check the extension-functions.c.

cd treecover
gcc -fPIC -lm -shared extension-functions.c -o libsqlitefunctions.so
cd ..

launch server

For local development in debug mode, run:

export FLASK_APP=server.py
export FLASK_DEBUG=1
flask run

For production use, you can use:

gunicorn wsgi:app

If you get an import error for flask_bootstrap, try deactivate and source venv/bin/activate again. Ensure that you're authenticated via earthengine authenticate. You can use the UI in two ways:

1. Upload a small CSV file with the columns longitude, latitude, Aridity_Zone and optionally a plot_id. An example is given in treecover/data/example_fetch_sentinel_input.csv.
2. Upload a CSV with already populated features by using the command line tool. Example: treecover/data/example_server_sentinel_input.csv or example_input_file_landsat.csv

The webapp without GEE retrieval can be viewed at: http://tree-cover.herokuapp.com/ Note that Heroku deployment only works with submodules when pushing from the command line.

command line tool for data retrieval

The command line tool located in treecover/sentinel.py can be used to fetch raw data for the locations given in a CSV file with columns longitude, latitude, Aridity_Zone and optionally a plot_id. If the compiled extension is not located in treecover/libsqlitefunctions.so, you need to pass its location as --libsqlite PATH/TO/EXTENSION for modes 1 and 3. You can use it in the following modes:

1. Only write the necessary features to a specified csv file
2. Only write the raw fetched data to a sqlite db
3. Write the raw fetched data to a sqlite db and export the features to a csv file

source venv/bin/activate
cd treecover
# Mode 1:
python3 sentinel.py data/example_fetch_sentinel_input.csv --output outfile_1.csv
# Mode 2:
python3 sentinel.py data/example_fetch_sentinel_input.csv --db data/my_small_test.db
# Mode 3:
python3 sentinel.py data/example_fetch_sentinel_input.csv --db data/my_small_test.db --output my_small_outfile.csv

In order to save some RAM, you can specify e.g. --chunk 10 in order to commit transactions/ append to the csv after 10 instead of the default of 100 processed rows. Then, a GCP f1-micro with enough disk space is sufficient.