A new Pluvial Flood Index (PFI) considering meteorological, hydrological and hydrodynamic processes for real-time flash flood forecasting

Abstract

Pluvial (flash) floods regularly cause significant damage in both rural and urban catchments. Such pluvial floods are usually caused by short-term local precipitation events of extreme intensity, resulting in infiltration excess and overland flow. In contrast to fluvial floods, the hazard of pluvial floods is mainly due to overland flow and flow in small ditches and creeks. Therefore, pluvial floods cannot be evaluated with common extreme value statistics, which are based on fluvial discharge records at river gages. On the other hand, pluvial floods are not only influenced by precipitation alone, but also by hydrological and hydrodynamic processes. Thus, precipitation statistics are not sufficient to evaluate and predict pluvial floods, either. Therefore, we suggest a new pluvial flood index (PFI), which evaluates the danger resulting from overland flow and surface flooding during pluvial floods and takes into account precipitation along with hydrological and hydrodynamic processes. The new PFI is based on the proportion of pluvial flood hazard areas (PFHA). We define PFHA as areas where pedestrians or vehicles are at risk because water depth, flow velocity or the combination of both (flow rate) exceed defined thresholds. Based on historical events and design events (combining different probabilities of precipitation and initial soil moisture), we defined thresholds of PFHA to generate four classes of PFI ranging from no flood danger to very large flood danger. Hence, PFI is a simple, dimensionless measure, which can convey valuable information about the occurrence and severity of a pluvial flood to the general public and authorities. PFHA and PFI for different events are determined from precipitation input, dynamic simulation of infiltration and saturation excess and hydrodynamic simulation of surface runoff concentration. Thus, simulation and forecasting PFI does not only require quantitative precipitation input and appropriate hydrodynamic overland flow models, but also adequate distributed, process-based hydrological models that consider infiltration excess and saturation runoff resulting from different initial soil moisture and land surface conditions. We demonstrate the application and usefulness of the new PFI in case studies for historical events and for a large-scale test area and show the potential using ML methods to allow real-time forecasting. We will also demonstrate that this information is much more valuable than rainfall warning alone. Moreover, the PFI can be linked to detailed local data to improve decision making of local municipalities. Therefore, the PFI is a valuable core piece for operational, real-time pluvial flood forecasting and early warning systems. The proposed system provides information on whether pluvial flooding will occur in a certain area on the scale of several hectares to square kilometers and how extreme this flood will be.