2026-09-03 –, Conference Management Room6
Watershed delineation is essential for flood hazard mapping to minimize disaster impacts. Traditional approaches require time-consuming DEM preparation and preprocessing. This presentation introduces a serverless function that extracts watersheds from nationwide flow direction data hosted on cloud storage, comparing performance between COG and Zarr formats.
Background
Watershed delineation is essential for flood hazard mapping and disaster risk assessment. Traditional workflows require downloading large DEM datasets, performing preprocessing, and running computationally intensive algorithms—a process taking hours or days. This creates barriers for rapid assessment and limits accessibility for practitioners without specialized GIS infrastructure.
Approach
This presentation introduces a serverless watershed extraction system using AWS Lambda. Users click a point on a web map, triggering upstream cell tracing on cloud-hosted J-FlwDir (Japan's nationwide flow direction dataset). The system generates GeoTIFF and transparent PNG outputs, returning presigned S3 URLs within seconds.
Performance Comparison
This presentation compares COG and Zarr performance using real-world test cases on AWS Lambda. Initial benchmarks on large basins (millions of cells) demonstrate both formats process watersheds interactively, with measurements of total time and phase breakdown (extraction, GeoTIFF generation, PNG rendering, S3 uploads). Analysis identifies bottlenecks and discusses optimization opportunities such as chunk size configuration.
Zarr, GDAL/rasterio, J-FlwDir
I make my conference contribution available under the CC BY 4.0 license. The conference contribution comprises the abstract, the text contribution for the conference proceedings, the presentation materials as well as the video recording and live transmission of the presentation:
I am a geospatial engineer at MIERUNE Inc., working on QGIS plugin development and serverless WebGIS solutions. Previously, I contributed to iRIC Software, an open-source river analysis platform. I am interested in exploring how serverless architectures can be applied to practical hydrological and river engineering workflows.