Summary In the context of flood management, this paper investigates the relationship between spatial rainfall and runoff production, based on 15 years of radar data, 16 raingauges and 12 flow stations from the 1400 km 2 Lee catchment, UK. Event-based, semi-distributed rainfall–runoff modelling is undertaken. Alternative rainfall estimators (radar data and raingauge networks of various density) are considered and their effects on simulated runoff evaluated as a function of rainfall type, catchment type and catchment scale. An index of spatial variability is defined, based on the difference between the reference rainfall (defined by the full raingauge network) and alternative rainfall estimators. A modified Nash–Sutcliffe efficiency criterion measures the performance of the simulated runoff with respect to reference simulated runoff. Results show a complex picture. The dominant effect is the spatial variability of the rainfall. No clear pattern emerges as a function of catchment scale, or response time, except that the impact of spatial variability is damped at the whole catchment scale. The sensitivity to spatial rainfall is enhanced on urbanised catchments.

The significance of spatial rainfall representation for flood runoff estimation: a numerical evaluation based on the Lee catchment, UK

Summary Flash floods cause some of the most severe natural disasters in Europe but Mediterranean areas are especially vulnerable. They can cause devastating damage to property, infrastructures and loss of human life. The complexity of flash flood generation processes and their dependency on different factors related to watershed properties and rainfall characteristics make flash flood prediction a difficult task. In this study, as part of the EU-FLASH project, we used an uncalibrated hydrological model to simulate flow events in a 27 km2 Mediterranean watershed in Israel to analyze and better understand the various factors influencing flows. The model is based on the well-known SCS curve number method for rainfall–runoff calculations and on the kinematic wave method for flow routing. Existing data available from maps, GIS and field studies were used to define model parameters, and no further calibration was conducted to obtain a better fit between computed and observed flow data. The model rainfall input was obtained from the high temporal and spatial resolution radar data adjusted to rain gauges. Twenty flow events that occurred within the study area over a 15 year period were analyzed. The model shows a generally good capability in predicting flash flood peak discharge in terms of their general level, classified as low, medium or high (all high level events were correctly predicted). It was found that the model mainly well predicts flash floods generated by intense, short-lived convective storm events while model performances for low and moderate flows generated by more widespread winter storms were quite poor. The degree of urban development was found to have a large impact on runoff amount and peak discharge, with higher sensitivity of moderate and low flow events relative to high flows. Flash flood generation was also found to be very sensitive to the temporal distribution of rain intensity within a specific storm event.

Flash flood prediction using an un-calibrated hydrological model and radar rainfall data in a Mediterranean watershed under changing hydrological conditions

DEM resolution effects on machine learning performance for flood probability mapping

U-FLOOD – Topographic deep learning for predicting urban pluvial flood water depth

Using machine learning models, remote sensing, and GIS to investigate the effects of changing climates and land uses on flood probability

Preparation of a flood probability map serves as the first step in a flood management program. This research develops a probability flood map for floods resulting from climate change in the future. Two models of Flexible Discrimination Analysis (FDA) and Artificial Neural Network (ANN) were used. Two optimistic (RCP2.6) and pessimistic (RCP8.5) climate change scenarios were considered for mapping future rainfall. Moreover, to produce probability flood occurrence maps, 263 locations of past flood events were used as dependent variables. The number of 13 factors conditioning floods was taken as independent variables in modeling. Of the total 263 flood locations, 80% (210 locations) and 20% (53 locations) were considered model training and validation. The Receiver Operating Characteristic (ROC) curve and other statistical criteria were used to validate the models. Based on assessments of the validated models, FDA, with a ROC-AUC = 0.918, standard error (SE = 0.038), and an accuracy of 0.86% compared to the ANN model with a ROC-AUC = 0.897, has the highest accuracy in preparing the flood probability map in the study area. The modeling results also showed that the factors of distance from the River, altitude, slope, and rainfall have the greatest impact on floods in the study area. Both models’ future flood susceptibility maps showed that the highest area is related to the very low class. The lowest area is related to the high class.

Mehdi Ramazanzadeh lasboyee

Papers

Using machine learning models, remote sensing, and GIS to investigate the effects of changing climates and land uses on flood probability

Spatial Prediction of Future Flood Risk: An Approach to the Effects of Climate Change

Spatial modeling of flood probability using geo-environmental variables and machine learning models, case study: Tajan watershed, Iran

Predicting temporal and spatial variability in flood vulnerability and risk of rural communities at the watershed scale.