A study led by CIMA Research Foundation analyzes the Italian warning system by relating the return period of events (associated with their severity) to the issued warning level. The assessment confirms the difficulty of predicting intense localized precipitation but highlights the validity of a warning system that has recently celebrated its twentieth anniversary
The 20th anniversary of the Italian Warning Directive, which has defined actions, tools, and procedures to address hydrogeological and hydraulic risks on Italian territory, has been celebrated with an event organized by the Civil Protection Department
On our website, we have retraced the history and scientific developments that have characterized our warning system over time. Twenty years after its introduction, can we evaluate it? In other words, is it possible to determine if the issued alerts, produced by a combination of objective assessments (models) and subjective analyses (expert analysis), correspond to the event’s severity?
Addressing this question is a study led by researchers at CIMA Research Foundation, in collaboration with the Italian Civil Protection Department and the University of Florence, recently published in the Journal of Flood Risk Management.
The assessment
“The Italian warning system, defined in the 2004 Directive, plays a crucial role in predicting and responding to hydrogeological and hydraulic events, such as floods and landslides. One of its features is its hybrid nature, in the sense that strict modeling is combined with technicians’ experience,” explains Francesco Silvestro, researcher at CIMA Research Foundation and the study’s lead author. As we have reported, indeed, the issuance and management of warnings are not based solely on the model’s outcome but require an interpretation based on the operators’ skills and experience. “There are thus two elements, the objective one provided by the model and the subjective one by the technical analysis, which together contribute to establishing the alert level to be issued. Understanding how and how well this process is functioning then becomes crucial, both from a scientific point of view and for the public, to ensure that the warnings are as reliable as possible.”
Evaluating a hybrid system is not simple because it requires considering both the intrinsic uncertainties of the forecasting models and the variability linked to the human assessment component (conducted by technicians). Various studies, conducted in different countries, have presented evaluations of warning systems: some, for example, are based on cost-benefit assessments, others on post-processing techniques to estimate the forecasts’ uncertainty.
In their work, the researchers at CIMA Research Foundation employed a new approach, which relates the return period of river flows, estimated through hydrological modeling, to the issued warning level. “Practically speaking, we can say that the return period, which indicates the statistical frequency with which a flood occurs, provides a measure of the event’s severity (and rarity). Thus, comparing this data with the issued warning level (for floods, yellow, orange, and red) it is possible to assess whether the alert effectively reflected the risk – and, consequently, whether the civil protection measures were proportionate,” explains Dr. Silvestro. “To estimate this relationship, we based ourselves on a hydrological model, implemented at the national level, which provides a description of the ‘real world’; to also take into account hydrological uncertainty and calibration errors of the model, we applied a statistical analysis to associate river flows with their return period.”
Validity of the warning system
In general, the comparison between the issued warning levels and the return period of flood events shows a good correlation, indicating that the alert system is effective in most cases. This result highlights the importance, for an effective decision-making process in the issuance of warnings, of integrating human evaluations into the models.
“Very intense, localized, and fast precipitations remain difficult to predict. This is a historic challenge in the field of weather forecasting – and consequently also in geohydrological predictions. The small spatial scale at which they develop, the specific atmospheric conditions, and the sudden nature of these phenomena, in fact, make it difficult to represent them on forecasting models,” says Dr. Silvestro. “Work in this field is indeed very active, both in terms of scientific research to make predictions increasingly detailed and accurate, and in terms of civil protection, with the development of systems that can timely inform the population of a danger. Especially, since uncertainty related to these extreme events remains, it is essential that knowledge of the necessary good practices to protect oneself is widespread among all citizens.”
“The importance of this work, however, is not only in the conclusions of the evaluation but also in having employed a method of investigation that can be extended to other warning systems, in addition to the Italian one,” concludes the researcher. “The only necessary requirements are that the warning be issued based on predefined territorial structures (such as municipalities or provinces) and having a hydrological model sufficiently detailed and reliable, with a good historical reanalysis of events.”