Comment on "When are AI models ready for deployment? reassessing Google's global AI flood forecasting system through the lens of responsible modelling"

Le résumé fourni par la source

This preprint is a formal technical comment on the recent critique by Li et al. (2026) published in Journal of Hydrology X, which evaluated the global AI flood forecasting system introduced in Nearing et al. (2024, Nature). Recently, Li et al. (2026) evaluated our global AI flood forecasting system; however, their critique fundamentally misunderstands the system's intended purpose and the operational realities of early warning systems. This evaluation highlights a distinction in modern hydrology between physical process modeling and models that support humanitarian action. By examining the framing of operational readiness, thresholding, extreme event definitions, event windows, and benchmarking through a humanitarian lens, we demonstrate that the evaluation metrics chosen in Nearing et al. (2024) align with established disaster response protocols and end-user needs, whereas those suggested by Li et al. largely do not. Answering Li et al.’s call for rigor and transparency, the first principle of model evaluation is to assess a model for its intended purpose.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.