This paper presents the design, implementation, and scalability evaluation of a serverless pipeline that automates processing of camera imagery for streamflow monitoring. Deployment of camera systems at scale requires timely image processing, consistent data organization, reliable cloud uploads, and minimizing human intervention. To address these challenges, an automated, scalable, and serverless pipeline was developed, requiring no manual triggering or server maintenance. The pipeline responds to data events instantly, executes image processing tasks independently, and scales dynamically. The pipeline was deployed at two field locations where it processed hourly image uploads over a multi-month period, demonstrating reliability, automation, and uninterrupted operation. In scalability testing, the pipeline maintained average execution time of 21. 8 seconds and stable per-invocation cost of ∼0. 00205. No performance degradation or failures were observed as the number of sites increased from a single site to 300 sites, demonstrating horizontal scalability to meet the demands of large monitoring networks. • A serverless pipeline was developed for camera-based streamflow monitoring. • Cloud functions are triggered on image upload for segmentation and stage estimation. • Hydrologic outputs and metadata are published to a centralized, queryable database. • Deployment was demonstrated using AWS Lambda, S3, and Aurora Serverless. • The pipeline processed data from 300 simulated sites with consistent performance.
Neupane et al. (Sun,) studied this question.