RHITA is both a Python algorithm and a public web tool that automatically detects and tracks four major weather hazards across Europe: heatwaves, cold spells, heavy precipitation and strong winds. It runs in near real time using open forecasts from the European Centre for Medium-Range Weather Forecasts, while simultaneously maintaining a consistent historical archive built from the ERA5 reanalysis, a gridded reconstruction of past weather stretching back to 1950. The result is a single framework that can tell a policymaker on a Tuesday morning how unusual an unfolding storm is, and can also tell a researcher how hazard statistics across the whole continent have shifted over seventy-five years.

The technical core of the system is elegantly simple. RHITA first scans three-dimensional gridded data, in longitude, latitude and time, and flags every grid cell that exceeds a local quantile threshold. For extreme heat, heavy rain and strong winds, that threshold is the 99th percentile of the local climatology; for cold spells it is the 1st percentile. Because the thresholds are quantile-based and computed at each gridpoint over the 1950 to 2023 reference period, a temperature that counts as extreme in Lisbon is not the same one that counts as extreme in Helsinki, yet both are judged against their own local norms.

Next, neighboring flagged cells are grouped into spatial clusters at each time step, with clusters smaller than a minimum area discarded to filter out noise. Finally, the centroids of surviving clusters are linked across consecutive days whenever they lie within a critical distance of each other, allowing the algorithm to follow an event as it drifts, and even to handle cases where one event splits in two or two events merge into one. Each reconstructed event is then described by a suite of metrics: mean and maximum intensity above the threshold, total impacted area, and duration. A return period, expressing how rare the event is compared with the full historical record, is estimated for each metric.

Crucially, the free parameters of the algorithm were not chosen arbitrarily. The team optimized them against EM-DAT, the international disasters database, using records from 2000 to 2023, a period chosen because earlier decades suffer from serious reporting gaps. A grid search over physically plausible parameter ranges sought the configurations that best detected events matching documented disasters, requiring both spatial overlap with an affected country and temporal overlap within a ten-day window. The optimization achieved detection sensitivities above 0.75 for all hazard types except cold spells, which the authors attribute partly to the fact that EM-DAT records cold spells two to three times more often than heatwaves, including many moderate events that a symmetric, physically based threshold is not designed to capture.

Applied to the full ERA5 archive, RHITA detected 760 heatwaves, 681 cold spells, 805 heavy precipitation events and 686 strong wind events over Europe between 1950 and 2024. The resulting climatology paints a stark picture of a continent where heat is changing fastest. Heatwave frequency rose significantly in all four European subregions, with the strongest trend in Southern Europe at nearly one additional event per year per decade, followed by Western, Northern and Central-Eastern Europe. Cold spells moved in the opposite direction, declining most steeply in Northern Europe by roughly 0.87 events per decade.

The heat signal extends beyond frequency. Maximum excess temperature above the local threshold climbed by 0.07 degrees Celsius per decade, and the average area impacted by each heatwave grew by about 80,000 square kilometers per decade, roughly twice the surface area of Switzerland. By contrast, the continental-scale picture for heavy precipitation and strong winds is far more heterogeneous. A significant increase in heavy precipitation events emerged in Central and Eastern Europe, and wind events increased in Western Europe, but no statistically significant trends appeared in precipitation intensity, duration or impacted area when averaged across the whole continent. The authors note that positive slopes for precipitation intensity came close to the significance threshold, hinting at a tendency toward more intense downpours that the IPCC also assesses as likely, but they caution that aggregating across seasons and a spatially diverse continent can mask distinct regional and seasonal signals.

The web interface, developed by the Institut Pierre-Simon Laplace, is designed for a broad audience rather than specialists alone. A real-time dashboard displays events from the past fifteen days as bubbles on an interactive map, sized by impacted area and colored by hazard type, alongside a table of event attributes and a bar chart of event counts. Clicking an event opens a page showing its spatio-temporal evolution through a Web Map Service, its full set of metrics, its start and end dates, and its estimated rarity relative to the ERA5 climatology. A historical archive section offers the complete 1950 to 2024 catalog with summary statistics, scatter plots of duration over time, and breakdowns by country and hazard type. The interface also supports exploration of compound events, such as concurrent heatwaves in different regions or successive storms, which are particularly consequential when multi-hazard interactions amplify impacts.

The team is candid about limitations. The 0.25-degree resolution of the underlying data limits the detection of small-scale convective phenomena such as localized thunderstorms, and the combination of quantile thresholds with a minimum three-day duration means some impactful but short-lived or moderately intense events may slip through. RHITA also deliberately focuses on the hazard component of disasters, leaving exposure and vulnerability to other analytical frameworks, and a systematic sensitivity analysis of alternative parameter choices is planned for future work. Yet the framework’s flexibility is arguably its greatest strength: because it operates on any gridded dataset, it can be applied directly to climate model simulations, opening the door to projections of how event frequency, intensity, duration and extent may evolve under future warming. With its code, data and dashboards all openly available, RHITA offers something the disaster-response community has long needed: a transparent, reproducible yardstick for measuring just how extraordinary the weather has become, delivered while the emergency is still unfolding.

Subject of Research: Real-time detection and tracking of extreme weather events across Europe

Article Title: RHITA: a web tool for real-time detection of extreme weather events

Article References: Cazzaniga, G., Akakpo-Numado, A., Brockmann, P., Burq, A., Vrac, M., & Faranda, D. (2026). RHITA: a web tool for real-time detection of extreme weather events. Natural Hazards and Earth System Sciences, 26(10), 4741-4752. https://doi.org/10.5194/nhess-26-4741-2026

Image Credits: AI Generated

DOI: 10.5194/nhess-26-4741-2026

Keywords: extreme weather, heatwaves, cold spells, heavy precipitation, strong winds, RHITA, ERA5 reanalysis, ECMWF forecasts, EM-DAT, climate trends, return periods, open-source web tool