Zarr Cloud-Native Workflow for Water Body Detection

Start Here

• New contributor or team member: go to Developer Onboarding
• Running workflows as a user: go to EOAP Users Onboarding
• Releasing/operating docs and workflows: go to Maintainers Onboarding
• Need STAC input contract details: go to Input STAC Requirements
• Need producer output semantics: go to Producer Outputs Contract
• Need hands-on commands for stac-zarr and TiTiler: go to How-to: stac-zarr Usage
• Need conventions and implementation details: go to STAC Zarr Best Practices
• Need current compliance status: go to GeoZarr Minispec Compliance
• Want an end-to-end runnable walkthrough: go to TL;DR

Overview

This module implements a cloud-native approach to detecting water bodies using Sentinel-2 data. The workflow uses the Common Workflow Language (CWL) for defining tasks and outputs data in the Zarr format, fully compatible with the STAC Datacube Extension.

By integrating Sentinel-2 data with Zarr and CWL, the workflow enables:

• Efficient parallel processing of geospatial data.
• Scalable cloud-native solutions for environmental monitoring.
• Interoperable metadata for easy integration with Earth Observation (EO) datalakes.

Workflow Steps

1. Crop Sentinel-2 Imagery

Crop Sentinel-2 imagery to the defined Area of Interest (AOI) using the input bounding box and Sentinel-2 bands.

Input:

• STAC item URL
• Bounding box (e.g., -121.399,39.834,-120.74,40.472)
• EPSG code (e.g., EPSG:4326)

Output:

• Cropped GeoTIFF images

• Compute NDWI

Apply the Normalized Difference Water Index (NDWI) formula to the cropped imagery:

Input:

• Cropped Sentinel-2 bands (green and nir)

Output:

• NDWI GeoTIFF

• Apply Otsu's Thresholding

Use Otsu's thresholding algorithm to convert the NDWI to a binary water mask.

Input:

• NDWI GeoTIFF

Output:

• Binary water mask GeoTIFF

• Convert to Zarr Datacube

Aggregate binary water masks over time into a Zarr datacube. The resulting dataset includes temporal and spatial dimensions.

Input:

• Binary water masks

Output:

• Zarr dataset

• Generate STAC Metadata

Produce a STAC catalog/collection for NDWI and water-bodies outputs, then convert to Zarr STAC metadata.

Input:

• Process outputs

Output:

• Zarr encoded result
• STAC catalog and collection metadata