Rainfall thresholds derivation for warning pluvial flooding risk in urbanised areas

Aim of this work is the development of an operational tool for pluvial flooding warning in an urban area based on off-line rainfall thresholds derived by coupling a rainfall runoff modelling and a hydraulic routing. The critical conditions considered for issue flood warnings were not only based on the water stage, but also on the extension of the flooded area. Further, a risk assessment framework for quantifying the reliability of the rainfall thresholds has been included; rainfall thresholds used in pluvial flooding warning should be influenced by the uncertainties in the rainfall characteristics (i.e. rainfall duration, depth and storm pattern). This risk assessment framework incorporates the correlated multivariate Monte Carlo simulation method, an hydraulic model for the simulation of rainfall excess propagation over surface urban drainage structures, i.e. streets and pathways. Thresholds rainfall are defined using a number of inundation criteria, to analyze the change in the rainfall threshold due to various definitions of inundation. Starting from estimated water stages and flooded area from inundation simulation rainfall thresholds can be obtained according a specific inundation criterion, including, together, a critical water depth and a critical flooding area. Finally, the second phase concerns the imminence of a possible hydrological risk by comparing the time when cumulative rainfall and rainfall thresholds meet to each other. The developed procedure has been applied to the real case study of Mondello catchment in Palermo (Italy).


Introduction
Floods are natural phenomena which cannot be prevented; many studies show that the severity and frequency of floods have increased in recent years and underline the difficulty to separate the effects of natural climatic changes and human influences as land management practices, urbanisation, etc. [1][2][3][4][5][6].As consequence, interest towards careful methodologies for the assessment of flood-prone areas has significantly increased.
Floods, generally, are divided into two main categories: general floods, which are caused by precipitation, occurring during a long period over a given river basin, and flash floods, that usually affect basins less than 1000 km2, with response times of a few hours or less.Moreover, Mediterranean ephemeral streams have specific features compared to other river systems.These basins are small and highly torrential and may generate flash-floods [7].Runoff generation in semiarid zones is the result of many spatial and temporal complex processes related to hillslope and catchment scale.The complexity of the processes involved derives from great heterogeneity of rainfall inputs, surface and subsurface characteristics, and strong nonlinear dependency on antecedent wetness, which controls the infiltration capacity of the soil surface and the connectivity of surface and subsurface runoff pathways [8].
Moreover, the rapid transformation processes of urban areas induced, consequently, the increase of catchment imperviousness and the derived increase of surface runoff generated during rainfall events.The natural drainage network is, often, insufficient to convey such discharges and it is gradually substituted by artificial systems having the function to convey the runoff coming from urban areas towards the closest receiving water body.However, flooding events in urban areas occur quite frequently because of rain events of lower intensity than the design one, even in case of correct network dimensioning.
Large research efforts have been made on the flashflood analysis and modelling despite the lack of accurate measured rainfall and discharge data [5,9,10].Usually, flash floods take place in steep mountain streams or in urban areas [11,12].The lack of data on flash floods and in particular the lack of accurate discharge estimates, can often provide an obstacle to improvements in flood forecasting, warning, planning and emergency management.
Usually, flood warning systems are based on on-line hydrological and/or hydraulic models in order to provide forecasts of water stages or discharges at critical river sections [13][14][15].This procedure is inappropriate for flash flood warning in urban areas or in catchments with a small area.Therefore, it would be helpful for the flash flood alter to be in accordance with the observed or forecasted rainfall if exceeding a critical value, namely, the rainfall threshold.Generally, rainfall thresholds identify precipitation critical values in the context of landslides and debris flow hazard forecasting [16,17] and in the flood forecasting.A number of approaches have been proposed for the determination of the rainfall threshold for flood forecasting or warning [18-20, 13, 21, 22, 14, 23, 24, 15].In Europe, the Integrated Project FLOODSite [25], among others aims at assessing the advantage for using the rainfall threshold approach as an alternative to the traditional ones in the case of flash floods.
In these cases, the change of rainfall in time may be more important than the total rainfall accumulation for flood forecasting.According to the approach proposed by [19,15], in this study the rainfall threshold has been estimated in an urban area by coupling results of hydrodynamic model in terms of water stage and flooding area.Particularly, dependency of the antecedent soil moisture conditions has been neglected because urban areas are characterized by imperious surfaces.
Because incomplete processes understanding and uncertainties associated with the current prediction, flood prediction methods and analysis are, still, uncertain.Many studies have shown that sources of error propagate through the model and affect its output [26][27][28][29][30][31][32]