Proof Of Concept, Generate Fire Boundaries using Google Earth Engine

About

This is a proof of concept to generate fire boundaries using Google Earth Engine.

The goal is to prove the utility of generating fire perimeter from satellite imagery, thus reducing the need for manual generation of perimeters using helicopters.

Issue: Current implementation dependant on cloud cover! Current algorithm doesn't work with lots of cloud and smoke. Smoke may cause holes in the perimiter. Consider using MODIS + VIIRS as source.

Issue: Need a government google account to productionize this.

Credit: The code that identifies fire is basead on code by https://github.com/ashlinrichardson available at: https://github.com/bcgov/wps-research/blob/master/py/gee/active_fire.js

Process / Components

flowchart LR

    subgraph Openshift
        job["Scheduled Job</br>Runs daily, generates fire perimeters for active fires > 90ha.</br>Uses publicically available data."]
        postgis[(PostGIS)]
        subgraph Servers
            pg_featurserv["pg_featurserv</br>A lightweight RESTful geospatial feature server for PostGIS."]
            rasterserv["Raster Server</br>Serve up GeoTIFF files."]
        end
    end
    WFS["https://openmaps.gov.bc.ca"]
    GEE["Google Earth Engine(GEE)"]
    s3[S3 Compliant, OCIO Object Storage Service]
    user("User with GIS software/Browser")

    job--"Query active fires > 90ha"-->WFS
    job--"Generate fire perimeter GeoTIFF"-->GEE
    job--"Store polygonized fire perimeter"-->postgis
    job--"Store fire GeoTIFF"-->s3
    pg_featurserv--"Serve features for GIS"-->postgis
    rasterserv--"Serve fire GeoTIFF"-->s3
    user--"OCG API"-->pg_featurserv
    user-->rasterserv

Demo

Proof of concept demo: https://wps-dev-featureserv.apps.silver.devops.gov.bc.ca/

pg-featureserv screenshot

GeoTIFF screenshot

General

• Create .env file (you can copy .env.example and set appropriate values)
• Assumes use of python poetry
• Google Earth Engine does not work on versions newer than 3.8.*

Ubuntu

pip install pygdal==3.0.4.10
poetry install
poetry run python -m pip install pygdal==3.0.4.10
make cli

Running

poetry run python3 -m fire_perimeter.cli

Using macports on m1

I had trouble using pyenv to install the version I need. So installing python with macports, and telling poetry to use the version I want.

sudo port selfupdate
sudo port upgrade outdated
sudo port install python38
sudo port install gdal
sudo port install proj9

Python 3.8 is installed to: /opt/local/Library/Frameworks/Python.framework/Versions/3.8

Find out where proj is installed, and set PROJ_DIR for your environment

port contents proj9

poetry env use /opt/local/Library/Frameworks/Python.framework/Versions/3.8/bin/python3
poetry run python -m pip install --upgrade pip
poetry install
poetry run python -m pip install gdal==$(gdal-config --version) --no-cache-dir

NOTE: --no-cache-dir is very important to make sure that gdal install doesn't skip numpy bindings. NOTE: order is very important here, you need to have installed numpy before gdal.

Openshift

Build image

Shortcut!

Taking some shortcuts! Skipping build in openshift, and pushing up from local

make build
docker tag wps-fire-perimeter:latest image-registry.apps.silver.devops.gov.bc.ca/e1e498-tools/wps-fire-perimeter:latest
docker login -u developer -p $(oc whoami -t) image-registry.apps.silver.devops.gov.bc.ca
docker push image-registry.apps.silver.devops.gov.bc.ca/e1e498-tools/wps-fire-perimeter:latest

Deployment

oc -n e1e498-dev process -f openshift/templates/perimeter_cronjob.yaml | oc -n e1e498-dev apply -f -