Earth Engine Algorithms

This repository contains two main algorithms for the Earth Engine platform:

1. EKF: Extended Kalman Filter
2. BULCD: Bayesian Updating of Land Cover Classification

Installation and Prerequisites

1. Clone repository
2. Create new virtual environment: python -m venv {path/to/environment}
3. Activate environment: source {path/to/environemnt}/bin/activate
4. Install dependencies: pip install -r requirements.txt

Earth Engine Extended Kalman Filter

This algorithm implements an Extended Kalman Filter for the purpose of keeping continual track of expected values of satellite imagery.

Preparing New Test Run

1. Create a new study group

This can be achieved by:

1. Create a new folder in the lib > point_groups > groups > {new_group_name} directory

2. Add a json file in the new directory with the following structure:

   {
       "points": [
           [lon, lat],
           [lon, lat],
           ...
       ]
   }

   Where lon and lat are the coordinates of the point of interest.

3. Register the new group by adding the new path to the STUDY_POINT_GROUPS dictionary in lib/observations_points.py:

   NEW_GROUP_PATH = build_path("new_group_name")
   
   STUDY_POINT_GROUPS = {
       ...
       "new_group_tag": NEW_GROUP_PATH,
   }

4. Register the study area by adding the boundary coordinates to a new variable in lib/study_areas.py:

   NEW_STUDY_AREA_TAG = [
       (lon, lat),
       (lon, lat),
       (lon, lat),
       (lon, lat),
   ],

5. Register the image collection by adding the new collection to the COLLECTIONS dictionary in lib/image_collections.py:

   IMAGE_COLLECTION = gather_collections_and_reduce(
        {
            "L8dictionary": {
                "years_list": [2017, 2018],
                "first_doy": 150,
                "last_doy": 250,
                "cloud_cover_threshold": 20,
            },
            "L9dictionary": {
                "years_list": [2017, 2018],
                "first_doy": 150,
                "last_doy": 250,
                "cloud_cover_threshold": 20,
            },
            "default_study_area": (ee.Geometry.Polygon(NEW_STUDY_AREA_TAG)), # study area tag created above
            "band_name_reduction": "swir",
            "which_reduction": "SWIR",
            "day_step_size": 6,
            "verbose": False,
            "dataset_selection": {
                "L5": False,
                "L7": False,
                "L8": True,
                "L9": True,
                "MO": False,
                "S2": False,
                "S1": False,
                "DW": False,
            },
            "first_expectation_year": 2017,
            "verbose": False,
        }
    )
   
   COLLECTIONS = {
       ...
       "new_collection_tag": IMAGE_COLLECTION,
   }

Running the algorithm

Open the kalman_caller.py file and modify:

• STUDY_GROUP_TAG variable to the tag of the new study group created above.
• COLLECTION_TAG variable to the tag of the new image collection created above.

Then run the script.

Viewing Results

The results will be saved in the tests/kalman/{STUDY_GROUP_TAG}.../{datetime/ directory.