The paradox of AI advances and declining in situ networks in hydrometeorology

Artificial intelligence (AI) is transforming hydrometeorological forecasting, but are we overlooking the foundations on which these advances depend?

As AI offers unprecedented opportunities to improve the prediction, monitoring, and understanding of floods, droughts, and other hydrometeorological hazards, a fundamental challenge is emerging: the increasing capabilities of AI are not necessarily matched by sustained investment in the observational networks that underpin them.

Our paper highlights this critical paradox. While AI is advancing rapidly, the availability and accessibility of reliable in situ observations remain insufficient and, in some regions, are declining. These measurements are essential not only for training and calibrating AI models, but also for validating their predictions and ensuring that technological advances translate into trustworthy and operationally useful information.

We argue that AI innovation and observational infrastructure must advance together, rather than be treated as competing priorities. Strengthening monitoring networks, improving data accessibility, and sustaining long-term observations are fundamental investments in the future of hydrometeorological science.

Ultimately, this work calls for a more balanced approach to innovation: the future of AI-driven hydrometeorology depends not only on smarter algorithms, but also on the quality, continuity, and accessibility of the observations that make them possible.

More information about the publication:
Olsson, J., Du, Y., Pechlivanidis, I. et al. Building on quicksand: the paradox of AI advances and declining in situ networks in hydrometeorology. npj Nat. Hazards 3, 52 (2026). https://doi.org/10.1038/s44304-026-00237-0