CNWI-GEE

Canadian National Wetland Inventory Google Earth Engie Random Forest Classifications. Provides high level access for doing standardized random forest wetland classifications.

Installation and Setup

This package is built to be used with the annaconda project. For best user experience use some recent version of conda. These walk though will be using a miniconda3

Dependencies

• Geopandas
• google earth engine

conda create -n cnwi-gee python=3.10 -c conda-forge earthengine-api geopandas pandas

# Step 2): activate new conda env and authenticate earth engine api
$ conda activate cnwi-gee
# authenticate earth engine api
(cnwi-gee) $ earthengine authenticate

Example Pipeline

from dataclasses import dataclass

import ee

from cnwi import inputs, rf, td, funcs, sfilters
from cnwi import prebuilt



def main():
    # load dataset
    dataset = ee.FeatureCollection("users/ryangilberthamilton/BC/williston/fpca/willistonA_no_floodplain")
    williston = prebuilt.WillistonA()
    
    # create a training object
    training = td.TrainingData(
        collection=dataset,
        label='cDesc'
    )
    
    # create s1 inputs
    s1s: List[ee.Image] = inputs.s1_inputs(williston.s1)
    
    # create s2 inputs
    s2s: List[ee.Image] = inputs.s2_inputs(williston.s2)
    
    # create elevation inputs
    elevation = inputs.nasa_dem()
    filter = sfilters.gaussian_filter(3)
    smoothed = filter(elevation)
    slope = ee.Terrain.slope(smoothed)
    
    # Create the inputs stack
    stack = ee.Image.cat(*s1s, *s2s, smoothed, slope)
    
    # sample the stack
    training.sample(
        stack=stack
    )
    
    # create the rf model
    model = rf.RandomForestModel()
    # train the model
    trained = model.train(
        training_data=training.samples,
        predictors=stack.bandNames(),
        classProperty=training.value
    )
    
    # classify the image
    classified_img = stack.classify(trained).uint8()
    
    # export image and samples to cloud
    return sys.exit(0)
    
    
if __name__ == '__main__':
    main()