Liberty Mutual Reinsurance (LM Re) and Paris-based climate intelligence startup Celest.Science have announced a collaboration to develop a fallback methodology for event characterization when key observational data is unavailable or delayed, addressing one of the more stubborn operational vulnerabilities in parametric risk transfer.
The speed and objectivity that make parametric products attractive depend entirely on trusted and timely data. A tropical cyclone’s track, intensity, and footprint must be characterized clearly enough to determine whether a pre-agreed trigger has been met. When that data is disrupted, whether by the storm itself, infrastructure failure, or funding cuts to government monitoring agencies, settlement can stall.
Mark Rüegg, CEO of Swiss parametric specialist CelsiusPro, has noted more broadly that while all parametric contracts typically include a fallback methodology, the market needs to review whether those fallbacks depend on the same single data source as the primary trigger. It is precisely that concern that LM Re and Celest.Science are now directly targeting. That is precisely the gap LM Re and Celest.Science are attempting to close for tropical cyclone exposures.
The two parties will explore how Celest.Science’s physics-based and AI-driven climate modelling can be combined with multiple independent data streams to reconstruct a credible picture of a cyclone event when one or more traditional sources are offline. The output, if validated, would function as a structured alternative for characterizing events when primary data is absent, rather than a replacement for it.
Celest.Science was co-founded in 2023 by Léo Lemordant and Pierre Gentine, director of the NSF’s Learning the Earth with Artificial Intelligence and Physics (LEAP) Center at Columbia University. The company’s approach integrates deep learning, generative models and physics-informed AI to analyse interactions across atmospheric, oceanic and land systems. That methodology is designed to perform under conditions where observational records are thin or absent.
“As the risk landscape evolves and the availability of critical observational data becomes increasingly uncertain, LM Re continues to explore innovative approaches intended to enhance the robustness of its parametric products,” said Jean-Christophe Garaix, head of parametrics and agriculture at LM Re.
Lemordant echoed the underlying challenge. “Extreme weather often unfolds where observations are fragmented or incomplete,” he said. “By combining climate science, physics-based modelling and AI, we aim to explore how we can reconstruct a robust picture of what happened.”
The collaboration remains in its exploratory phase. Any methodology developed would require technical validation and applicable governance approvals before being embedded in an LM Re product, and the terms of individual contracts would govern its use.
The question of what happens when parametric data sources fail has sharpened in recent years. When the National Oceanic and Atmospheric Administration (NOAA) ceased updates to its billion-dollar weather disaster database in 2025, AM Best warned that parametric catastrophe bond triggers, which often depend on NOAA data, may need to be redesigned if further data sources were to disappear. The National Hurricane Center’s tropical cyclone reports remain intact, but the episode reinforced how much of the parametric market’s operational reliability rests on a small number of public data sources.
LM Re has been expanding its parametric capabilities across perils and geographies. Earlier this year, it made two payouts under its ShakeNet Parametric earthquake product. One to a client in Peru 21 days after a magnitude 5.8 event in May 2026, confirmed as the world’s first earthquake parametric payment triggered by Safehub’s sensor-based Shake Network. A second to the Mexican federal government following a magnitude 7.3 earthquake in July 2026, the largest ShakeNet Parametric deployment to date. The Celest.Science collaboration applies that same focus on data resilience to the tropical cyclone market, where exposure volumes and contract complexity run considerably higher.
