Conference Agenda

Issuing warnings, be it hazard-based or impact-oriented, requires a data processing pipeline to generate a reliable warning product which can be distributed to the end-user. Each step along the chain, including both manual and automatic ones, offers its own challenges to refine the data.

Traditionally, hazard-based warnings are derived from initially complex, gridded weather forecast data and have to be simplified for easy understanding by the public. Yet, hazard warnings do not provide specific information regarding their consequences as, for example, physical damage to infrastructure, disruption of societal activities, or economic losses. Under the umbrella initiative Early Warnings for All (EW4A) the World Meteorological Organization advocates for the advancement of early warning systems, increasingly tailoring them to the needs of specific users, with a focus on impacts, informing actions to mitigate damage. Developing accurate and useful impact-based forecasts is challenged by limited data and information, lack of standardized technical protocols, issues sharing impact data and little knowledge on the needs of various user groups.

To ensure the quality of any warning system, verification is crucial. A dense network of measurements. Yet even if this is given, as simplifications are made to issue a pleasing product to the user, verifying warnings poses a number of challenges. Impact warnings are even more challenging to verify and guidelines are needed to do so.

This session aims to unite scientists, natural catastrophe modelers, weather forecasters, tool developers, stakeholders, and policy professionals, and discuss advancements and challenges related to the entire warning chain. We welcome inputs on the identification of extreme weather and impacts, the generation of hazard or impact warnings and forecasts, their verification, visualization, uncertainty, and user needs. The session features expert presentations and a panel discussion to allow the community to collaborate on developing storylines, marking a significant step forward in weather and impact modeling.

We present the development and application of the GVZ (cantonal building insurance Zurich) windstorm risk model to assess building damages in the Canton of Zurich, Switzerland. This model offers two key applications that support both rapid damage estimation immediately after a windstorm event and probabilistic risk modelling.

The first application uses high-resolution meteorological data from MeteoSwiss's ICON model to provide impact forecasting and post-event analysis. The model estimates the number of buildings affected and the potential damages at the municipal level, visualized through an interactive dashboard. This tool supports resource allocation and informed decision-making during and immediately after windstorm events.

The second application involves a probabilistic risk assessment based on a winterstorm hazard event set generated using the CLIMADA platform (ETH Zurich, Aznar-Siguan and Bresch, 2019) and the method described by Schwierz et al. (2010) and Welker et al. (2021). We estimate return periods for extreme events like Winterstorm Lothar. Our findings indicate that a storm of Lothar's magnitude today would result in approximately CHF 80 million in damages, corresponding to a 130-year return period.

Hail is a significant contributor to weather-related damages to buildings, cars, and agriculture in Switzerland, demanding actionable information on hail risks and forecasts across sectors. The research project scClim (https://scclim.ethz.ch/) addresses this demand by establishing a seamless model chain from observing and modeling hail events to the quantification of hail impacts, including simulations to compare hail occurrence in current and future climate. Within the project, we have developed a hail impact forecast prototype that has been co-designed with relevant stakeholders from public and private institutions and is running in a pre-operational fashion during the 4-year research project. Combining ensemble weather forecasts with exposure and vulnerability information, we use the open-source risk assessment platform CLIMADA to provide impact-based forecasts for buildings and different crop types in Switzerland. Furthermore, the platform provides post-event assessments of hail impacts based on operational radar data and/or crowdsourced hail reports. Finally, by means of high-resolution convection-resolving climate simulations using a pseudo global warming approach for a 3 K global warming level that have been conducted within the project, we discuss hail impact distributions in current climate and hail impact projections to future climate.

Increasingly intense rainfall events are causing serious damages to infrastructures and endangering human lives. To better protect them, early warning systems can be set up to evacuate people, move and protect cars or protect infrastructure by installing flood barriers. However, warnings must be issued with sufficient lead times (tens of minutes) before the flood event occurs in order to be useful.

Deterministic forecasts based on the advection of precipitation radar measurements can anticipate flash flood precipitation. However, these systems are subject to considerable uncertainties, especially for extreme convective events that tend to cause flash floods. These uncertainties include the growth and decay of the storm cell, as well as the estimation of the cell's displacement. The use of these deterministic forecasts leads to low detection probabilities for localized intense precipitation events. The generation of ensemble precipitation forecasts improves on deterministic forecasts by proposing several precipitation scenarios, some of which may lead to higher discharge forecasts.

As part of the Radar4Infra project, a flash flood forecasting and alert system based on NowPrecip1.0 radar forecast (Sideris et al., 2020) is being developed for several small catchments. This system builds on recent improvements of the weather radar network and data processing (Germann et al., 2022), sophisticated nowcasting algorithms (Sideris et al., 2020) and a state-of-the-art rainfall-funoff model adapted for Alpine catchments (Jordan, 2007). More precisely, it consists of a radar precipitation forecast, with a spatial resolution of 1 km², a temporal resolution of 10 min, a forecast horizon of 6 hours and a forecast update rate of 10 min. This forecast is then introduced into the Routing System rainfall-runoff simulation model (Schäfli et al., 2005; Jordan, 2007), also with a 10 min temporal resolution. The rainfall-runoff simulation model is calibrated on flow measurements, with input data from rain gauges or precipitation fields (CPCH from MeteoSwiss, Sideris et al., 2014).

The methodology followed consists of evaluating the quality of deterministic flow forecasts using the "probability of detection" and "false alarm ratio" indicators, calculated at several flow thresholds (Cosson, 2023). These benchmark forecasts are then compared with ensemble forecasts. The latter are derived from preliminary tests (NowPrecip2.0) carried out by MeteoSwiss for a few selected flood events.

Two examples of hindcasts showed promising results : the Anniviers event (catchment area 88 km² in the Swiss Alps), and the Cressier event (catchment area 4.5 km² in the Jura region). In the Anniviers example, the system predicted a flood peak almost two hours ahead of time, while the watershed response time was only one hour in this situation. In the Cressier example, the "NowPrecip2.0 RZC-based" ensemble forecast was not able to predict the peak discharge, but the system predicted a smaller flood one hour ahead, although the response time of the catchment area is only 20min. Moreover, in both cases, the deterministic NowPrecip1.0 RZC-based forecast was not able to predict any flood discharge at all.

Ensemble radar-based hydro-meteorological cascade allowed to predict flash floods better than deterministic radar advectionpredictions.

Accurate and precise weather forecasting is crucial for issuing timely weather hazard warnings. However, current numerical weather prediction (NWP) models often struggle to accurately represent extreme weather events due to limitations in spatial and temporal resolutions. Additionally, with only a few model runs typically initialized per day, the effectiveness of the data assimilation process in capturing rapidly changing environments and improving initial conditions is limited. These constraints prevent NWP models from capturing small-scale weather features, such as severe convective thunderstorms. Furthermore, standard fixed thresholds used in issuing weather warnings may not adequately account for the varying levels of risk associated with different locations and use cases. This uniform approach can lead to either underestimation or overestimation of the actual risk.

To address these challenges, Meteomatics has developed the operational high-resolution NWP model EURO1k. Characterized by a 1 km horizontal grid spacing, a 72-hour forecast horizon, and an hourly refresh rate across the pan-European domain, the EURO1k model strongly enhances forecast accuracy. This high resolution allows EURO1k to accurately represent small-scale weather patterns, resulting in precise forecasts of extreme weather events. Additionally, thanks to its hourly refresh rate and data assimilation capabilities, the EURO1k model can be utilized for nowcasting. In addition to assimilating standard data sources like radar, satellite data, weather stations, and radiosondes, the EURO1k model also integrates data from a network of Meteodrones—small unmanned aircraft systems (UAS) developed by Meteomatics that collect vertical atmospheric profiles up to 6000m in altitude.

Moreover, Meteomatics has developed a highly customizable weather warning system, where multiple weather variables can be combined, and specific thresholds selected to enable targeted warnings for specific locations. The integration of the high-resolution EURO1k model with this customizable alert system allows for more accurate and use-case-specific warnings, optimally addressing relevant local risks. By employing this advanced approach, Meteomatics substantially enhances the reliability and precision of weather hazard warnings, ultimately improving preparedness and response measures.