The role of predisposing factors in determining the rainfall intensity necessary to cause flash floods in Portugal

Flash floods are often responsible for deaths and damage to infrastructure. The general objective of this work is to create data-based models to understand how the predisposing factors influence the triggering factor (precipitation) in the case of flash floods in the continental area of Portugal. Flash floods occurrences were extracted from the DISASTER database, which contains the location and date of historical flood events in the study region. Historical daily rainfall data was collected automatically from the Copernicus database. We extracted the accumulated precipitation for 3 days preceding each event and calculated the rainfall intensity. The predisposing factors were extracted considering the whole basin that corresponds to each flood event. The  analyzed predisposing factors were: accumulated flow, average slope, average elevation, predominant slope aspect, predominant lithology and soil properties (percentage of clay, coarse sand and coarse elements and field capacity). Elevation can often define different climatic and vegetation zones, while slope influences both the concentration and the infiltration capacity. The slope aspect can influence the amount and intensity of rainfall that affects the hillslope, as well as the amount and intensity of solar radiation. Lithology represents the properties of bedrock and the soils properties influence water infiltration and percolation. The Random Forest algorithm and the Leave-One-Out cross-validation technique were used to evaluate the model's performance and create a final model that identifies the relationship between the predisposing factors and the different rainfall intensities related to each flash flood occurrence. The final model obtained a root mean square error (RMSE) value of 3, an acceptable value for the objectives of the work. The percentage of coarse elements in the soil, average slope and field capacity were defined as the most important factors in the model for defining the amount of rainfall needed for flash floods to occur in mainland Portugal. The model developed can help to predict flash flood occurrence and future work involves combining the susceptibility model with the model created in this project to create a warning system that can be updated in real time, taking into account rainfall forecasts.

Acknowledgements: This work was financed by national funds through the FCT – Fundação Portuguesa para a Ciência e Tecnologia, I.P., under the grant to support the completion of the doctoral dissertation with the reference 2022.14473.BD.