Load and deploy cloud function files from notebook

cloud_functions/sustainable_sourcing_layers/main.py

@@ -0,0 +1,42 @@
+import json
+import ee
+from flask import jsonify
+import functions_framework
+import logging
+import os
+import requests
+
+import google.auth
+import google.cloud.logging
+
+from suso_layers_2025a import get_suso_stats
+
+client = google.cloud.logging.Client()
+client.setup_logging()
+
+@functions_framework.http
+def main(request):
+  """Handle requests in a format (geojson) suitable for BigQuery."""
+  credentials, _ = google.auth.default(
+      scopes=['https://www.googleapis.com/auth/earthengine']
+  )
+  ee.Initialize(credentials, project=os.environ['GOOGLE_CLOUD_PROJECT'])
+  try:
+    replies = []
+    request_json = request.get_json(silent=True)
+    calls = request_json['calls']
+    for call in calls:
+      geo_json = json.loads(call[0])
+      try:
+        logging.info([geo_json])
+        response = get_suso_stats(geo_json)
+        logging.info(response)
+        replies.append(json.dumps(response))
+      except Exception as e:
+        logging.error(str(e))
+        replies.append(json.dumps( { "errorMessage": str(e) } ))
+    return jsonify(replies=replies, status=200, mimetype='application/json')
+  except Exception as e:
+    error_string = str(e)
+    logging.error(error_string)
+    return jsonify(error=error_string, status=400, mimetype='application/json')

cloud_functions/sustainable_sourcing_layers/requirements.txt

@@ -0,0 +1,6 @@
+earthengine-api
+flask
+functions-framework
+google-api-core
+google-cloud-logging
+requests

cloud_functions/sustainable_sourcing_layers/suso_layers_2025a.py

@@ -0,0 +1,112 @@
+import os
+
+import ee
+import google.auth
+from google.api_core import retry
+
+# First, initialize.
+credentials, _ = google.auth.default(
+    scopes=['https://www.googleapis.com/auth/earthengine']
+)
+ee.Initialize(
+  credentials,
+  project=os.environ['GOOGLE_CLOUD_PROJECT'],
+  opt_url='https://earthengine-highvolume.googleapis.com'
+)
+
+# See https://github.com/google/forest-data-partnership/tree/main/models.
+cocoa_2025a = ee.ImageCollection('projects/forestdatapartnership/assets/cocoa/model_2025a')
+coffee_2025a = ee.ImageCollection('projects/forestdatapartnership/assets/coffee/model_2025a')
+palm_2025a = ee.ImageCollection('projects/forestdatapartnership/assets/palm/model_2025a')
+rubber_2025a = ee.ImageCollection('projects/forestdatapartnership/assets/rubber/model_2025a')
+
+filter2020 = ee.Filter.calendarRange(2020, 2020, 'year')
+filter2023 = ee.Filter.calendarRange(2023, 2023, 'year')
+
+cocoa2020 = cocoa_2025a.filter(filter2020).mosaic().rename('cocoa_2020')
+cocoa2023 = cocoa_2025a.filter(filter2023).mosaic().rename('cocoa_2023')
+coffee2020 = coffee_2025a.filter(filter2020).mosaic().rename('coffee_2020')
+coffee2023 = coffee_2025a.filter(filter2023).mosaic().rename('coffee_2023')
+palm2020 = palm_2025a.filter(filter2020).mosaic().rename('palm_2020')
+palm2023 = palm_2025a.filter(filter2023).mosaic().rename('palm_2023')
+rubber2020 = rubber_2025a.filter(filter2020).mosaic().rename('rubber_2020')
+rubber2023 = rubber_2025a.filter(filter2023).mosaic().rename('rubber_2023')
+
+# See https://eartharxiv.org/repository/view/9085/.
+natural_forest2020 = ee.ImageCollection(
+    'projects/computing-engine-190414/assets/biosphere_models/public/forest_typology/natural_forest_2020_v1_0'
+  ).mosaic().divide(255).selfMask()
+
+# THRESHOLDS FOR DEMONSTRATION ONLY! Tune these to your needs.
+thresholds = {
+    'forest': 0.5,
+    'cocoa': 0.45,
+    'coffee': 0.96,
+    'palm': 0.89,
+    'rubber': 0.5
+}
+
+# A mini-ensemble of GDM and fodapa data products.
+ensemble = ee.Image.cat(
+  natural_forest2020.rename('forest'),
+  cocoa2020.rename('cocoa'),
+  coffee2020.rename('coffee'),
+  palm2020.rename('palm'),
+  rubber2020.rename('rubber')
+).unmask(0)
+
+# Threshold the probabilities.  THRESHOLDS FOR DEMONSTRATION ONLY!
+crop_names = list(thresholds.keys())
+threshold_values = list(thresholds.values())
+thresholded = ensemble.select(crop_names).gt(ee.Image(threshold_values))
+
+# Unclassified means no predicted presence at the specified thresholds.
+unclassified = thresholded.reduce('sum').eq(0).rename('unclassified')
+
+# Confusion means two or more classes predicted presence.
+confusion = thresholded.reduce('sum').gt(1).selfMask().rename('confusion')
+
+def get_suso_layers_2025a() -> ee.Image:
+    '''Returns the stack of probability images in separate bands.'''
+    return ee.Image.cat(
+        natural_forest2020.rename('natural_forest_2020'),
+        cocoa2020.rename('cocoa_probability_2020'),
+        cocoa2023.rename('cocoa_probability_2023'),
+        coffee2020.rename('coffee_probability_2020'),
+        coffee2023.rename('coffee_probability_2023'),
+        palm2020.rename('palm_probability_2020'),
+        palm2023.rename('palm_probability_2023'),
+        rubber2020.rename('rubber_probability_2020'),
+        rubber2023.rename('rubber_probability_2023'),
+    )
+
+def get_areas_image() -> ee.Image:
+    '''Returns data for area calculations in square meters.'''
+    return ee.Image.cat(
+      thresholded,
+      unclassified.rename('unclassified'),
+      confusion.rename('confusion')
+    ).multiply(ee.Image.pixelArea())
+
+@retry.Retry()
+def get_suso_stats(geojson):
+    """Get area stats for the provided geojson polygon."""
+    region = ee.Geometry(geojson)
+    feature_area = ee.Number(region.area(10))
+    suso_image = get_areas_image()
+    # Sum of pixel areas in square meters.
+    stats = suso_image.reduceRegion(
+        reducer=ee.Reducer.sum(),
+        geometry=region,
+        scale=10
+    )
+    # Gini index.
+    # See https://en.wikipedia.org/wiki/Decision_tree_learning#Gini_impurity.
+    crop_names = ['forest', 'cocoa', 'coffee', 'palm', 'rubber']
+    gini = ee.Number(1).subtract(ee.List(
+        [ee.Number(stats.get(c)).divide(feature_area) for c in crop_names]
+    ).reduce(ee.Reducer.sum()))
+    # Update the EE dictionary.
+    stats = stats.set('gini', gini).set('total_area', feature_area)
+    # Request the result to the client and return it.
+    return stats.getInfo()