Long-term changes in rainfall conditions triggering landslides in Calabria (Southern Italy) over 1921–2020 are analysed through a dynamic approach in rainfall threshold definition. Shifts in seasonality and a decreasing trend in triggering rainfall are observed over the years, suggesting an increased propensity of the territory to generate landslides, even with less severe triggering rainfall in recent decades. The work highlights the essential role of landslide data collection and archiving.

Reservoir operators need reliable forecasts during typhoons to reduce flood risks and secure water supplies. In this study, we developed a method that combines multiple weather forecasts with computer models trained on past rainfall and reservoir inflow records. The approach improved inflow forecasts up to three days ahead, helping reservoir managers make earlier decisions, improve flood control, and better prepare for future extreme weather.

Geohazards occur more often and at larger scale due to climate change. In this study, we used for the first time automated drone docks to monitor geohazards at 3 different sites in Norway and Switzerland. We present a monitoring workflow for automated Uncrewed Aerial Vehicle operations which can produce frequent, detailed maps showing surface change and displacement. Our results show how this approach can improve future hazard monitoring and early warning and help to protect communities at risk.

Floods cause major social and economic losses, but estimating risk is difficult because extreme events are rare. We used artificial intelligence to generate over a thousand realistic winter weather seasons and river flows for the Elbe basin. The approach reproduced observed patterns and produced a wider range of extreme storms, showing that artificial intelligence can expand plausible flood scenarios for improved risk assessment.

Landslides along highways pose safety risks and disrupt travel. Many warning systems use rainfall data, but this alone may miss key factors. Using satellite-derived soil moisture, precipitation records, and landslide observations from Alabama, we found that wetter-than-average soil conditions increase landslide risk. We developed a framework that combines rainfall and soil moisture to improve prediction and reduce false alarms.

Meteotsunamis are rare but dangerous anomalous waves triggered by atmospheric disturbances, they are not currently forecast in Northwest Europe. We analysed the strongest recorded event on June 18, 2022, which reached 1 m amplitude. We showed high-resolution, high-frequency coupled models can represent such events up to three days ahead and help better understand their atmospheric triggers. These models, together with improved observations, can enhance early warnings and coastal safety.