@article{ZENG2026, 
author = {Qin ZENG},
title = {From Forecast Skill to Warning Effectiveness: MAZU and the Frontiers of End-to-End Early-Warning Science},
year = {2026},
journal = {Journal of Meteorological Research},
volume = {40},
number = {5},
pages = {1367-1372},
keywords = {MAZU (Multi-hazard, Alert, Zero-gap, and Universal), multi-hazard early warning, artificial intelligence (AI), extreme weather, warning effectiveness},
url = {https://www.sciopen.com/article/10.1007/s13351-026-6325-9},
doi = {10.1007/s13351-026-6325-9},
abstract = {Numerical weather prediction and artificial intelligence (AI) have steadily improved forecast skill, yet the toll of extreme weather remains high. Where forecasts are already skilful, early warning is limited less by accuracy than by whether warnings are acted on. MAZU (Multi-hazard, Alert, Zero-gap, and Universal) is an intelligent meteorological early-warning solution developed by the China Meteorological Administration in response to the United Nations Early Warnings for All (EW4All) initiative. Built on the cloud from components, it takes a multi-hazard approach and assembles observation, forecasting, risk assessment, warning, and response into configurable services. By mid-2026, customized versions had been deployed in 7 countries and cloud services trialled in more than 40 countries; and deployment in 30 countries is planned over the next five years. MAZU’s contribution to warning outcomes still awaits independent evaluation. Three questions remain open: how AI can forecast extremes reliably; how warning systems can be transferred and verified across observing networks and hazards; and how warning effectiveness can be evaluated and attributed along the chain from observation to action.}
}