PROBES snapshot: River gauge reading extraction using object detection and generative AI

This one-page brief, developed by the Digital Transformation Accelerator (DTA) team at the Alliance of Bioversity International and CIAT, is based on a probe led by Kayathri Vigneswaran and Mariangel Garcia Andarcia, DTA AoW3 co-lead at IWMI. The probe explores whether computer vision and generative AI can be used to automatically extract water-level readings from photographs of river gauges. By combining object detection with multimodal AI models, the approach aims to provide a low-cost alternative to manual monitoring and expensive sensor-based systems. Implemented as an early-stage, safe-to-fail experiment, the probe focuses on learning how AI-powered image analysis performs under real-world field conditions and whether it can support more accessible, scalable, and community-driven hydrological monitoring. The brief synthesizes early insights to inform future development and potential integration into climate and water management decision-making systems.