Towards flash-flood prediction in the dry Dead Sea region utilizing radar rainfall information

Flash-flood warning models can save lives and protect various kinds of infrastructure. In dry climate regions, rainfall is highly variable and can be of high-intensity. Since rain gauge networks in such areas are sparse, rainfall information derived from weather radar systems can provide useful input for flash-flood models. This paper presents a flash-flood warning model which utilizes radar rainfall data and applies it to two catchments that drain into the dry Dead Sea region. Radar-based quantitative precipitation estimates (QPEs) were derived using a rain gauge adjustment approach, either on a daily basis (allowing the adjustment factor to change over time, assuming available real-time gauge data) or using a constant factor value (derived from rain gauge data) over the entire period of the analysis. The QPEs served as input for a continuous hydrological model that represents the main hydrological processes in the region, namely infiltration, flow routing and transmission losses. The infiltration function is applied in a distributed mode while the routing and transmission loss functions are applied in a lumped mode. Model parameters were found by calibration based on the 5 years of data for one of the catchments. Validation was performed for a subsequent 5-year period for the same catchment and then for an entire 10-year record for the second catchment. The probability of detection and false alarm rates for the validation cases were reasonable. Probabilistic flash-flood prediction is presented applying Monte Carlo simulations with an uncertainty range for the QPEs and model parameters. With low probability thresholds, one can maintain more than 70% detection with no more than 30% false alarms. The study demonstrates that a flash-flood warning model is feasible for catchments in the area studied.     

Hydrological impact and potential flooding of convective rain cells in a semi-arid environment

Abstract

t Spatio-temporal storm properties have a large impact on catchment hydrological response. The sensitivity of simulated flash floods to convective rain-cell characteristics is examined for an extreme storm event over a 94 km² semi-arid catchment in southern Israel. High space–time resolution weather radar data were used to derive and model convective rain cells that then served as input into a hydrological model. Based on alterations of location, direction and speed of a major rain cell, identified as the flooding cell for this case, the impacts on catchment rainfall and generated flood were examined. Global sensitivity analysis was applied to identify the most important factors affecting the flash flood peak discharge at the catchment outlet. We found that the flood peak discharge could be increased three-fold by relatively small changes in rain-cell characteristics. We assessed that the maximum flash flood magnitude that this single rain cell can produce is 175 m³/s, and, taking into account the rest of the rain cells, the flash flood peak discharge can reach 260 m³/s.

Editor Z.W. Kundzewicz; Guest editor R.J. Moore

Citation Morin, E. and Yakir, H., 2013. Hydrological impact and potential flooding of convective rain cells in a semi-arid environment. Hydrological Sciences Journal, 59 (7), 1275–1284. http://dx.doi.org/10.1080/02626667.2013.841315     

Peak runoff contributing area as hydrological signature of the probability distribution of floods

For the analysis of hydrological extremes and particularly in flood prediction, deeper investigation is needed on the relative effects of different hydrological processes acting at the basin scale in different hydroclimatic areas of the world. In this framework, the theoretical derivation of flood distribution shows a great potential for development and knowledge advancement. In addition, another promising path of investigation is represented by the use of distributed hydrological models via simulation modelling (including Monte Carlo, discrete event and continuous simulation). In this paper results of a theoretically derived flood frequency distribution are analyzed and compared with the results of a simulation scheme that uses a distributed hydrological model (DREAM) in cascade with a rainfall generator (IRP). The numerical simulation allows the reproduction of a large number of extreme events and provides insight into the main control for flood generation mechanisms with particular emphasis to the peak runoff contributing areas, highlighting the relevance of soil texture and morphology in different climatic environments. The proposed methodology is applied here to the Agri and the Bradano basin, in Southern Italy.     

Assessment on the function of reservoirs for flood control during typhoon seasons based on a distributed hydrological model

A grid‐based distributed hydrological model, PDTank model, is used to simulate hydrological processes in the upper Tone River catchment. The Tone River catchment often suffers from heavy rainfall events during the typhoon seasons. The reservoirs located in the catchment play an important role in flood regulation. Through the coupling of the PDTank model and a reservoir module that combines the storage function and operation function, the PDTank model is used for flood forecasting in this study. By comparing the hydrographs simulated using gauging and radar rainfall data, it is found that the spatial variability of rainfall is an important factor for flood simulation and the accuracy of the hydrographs simulated using radar rainfall data is slightly improved. The simulation of the typhoon flood event numbered No. 9 shows that the reservoirs in the catchment attenuate the peak flood discharge by 423·3 m³/s and validates the potential applicability of the distributed hydrological model on the assessment of function of reservoirs for flood control during typhoon seasons. Copyright © 2011 John Wiley & Sons, Ltd.     

Influence of rainfall spatial variability on flood prediction

This paper deals with the sensitivity of distributed hydrological models to different patterns that account for the spatial distribution of rainfall: spatially averaged rainfall or rainfall field. The rainfall data come from a dense network of recording rain gauges that cover approximately 2000 km² around Mexico City. The reference rain sample accounts for the 50 most significant events, whose mean duration is about 10 h and maximal point depth 170 mm. Three models were tested using different runoff production models: storm-runoff coefficient, complete or partial interception. These models were then applied to four fictitious homogeneous basins, whose sizes range from 20 to 1500 km². For each test, the sensitivity of the model is expressed as the relative differences between the empirical distribution of the peak flows (and runoff volumes), calculated according to the two patterns of rainfall input: uniform or non-uniform. Differences in flows range from 10 to 80%, depending on the type of runoff production model used, the size of the basin and the return period of the event. The differences are generally moderate for extreme events. In the local context, this means that uniform design rainfall combining point rainfall distribution and the probabilistic concept of the areal reduction factor could be sufficient to estimate major flood probability. Differences are more significant for more frequent events. This can generate problems in calibrating the hydrological model when spatial rainfall localization is not taken into account: a bias in the estimation of parameters makes their physical interpretation difficult and leads to overestimation of extreme flows.     

The use of agent based modelling techniques in hydrology: determining the spatial and temporal origin of channel flow in semi‐arid catchments

Information on the spatial and temporal origin of runoff entering the channel during a storm event would be valuable in understanding the physical dynamics of catchment hydrology; this knowledge could be used to help design flood defences and diffuse pollution mitigation strategies. The majority of distributed hydrological models give information on the amount of flow leaving a catchment and the pattern of fluxes within the catchment. However, these models do not give any precise information on the origin of runoff within the catchment. The spatial and temporal distribution of runoff sources is particularly intricate in semi‐arid catchments, where there are complex interactions between runoff generation, transmission and re‐infiltration over short temporal scales. Agents are software components that are capable of moving through and responding to their local environment. In this application, the agents trace the path taken by water through the catchment. They have information on their local environment and on the basis of this information make decisions on where to move. Within a given model iteration, the agents are able to stay in the current cell, infiltrate into the soil or flow into a neighbouring cell. The information on the current state of the hydrological environment is provided by the environment generator. In this application, the Connectivity of Runoff Model (CRUM) has been used to generate the environment. CRUM is a physically based, distributed, dynamic hydrology model, which considers the hydrological processes relevant for a semi‐arid environment at the temporal scale of a single storm event. During the storm event, agents are introduced into the model across the catchment; they trace the flows of water and store information on the flow pathways. Therefore, this modelling approach is capable of giving a novel picture of the temporal and spatial dynamics of flow generation and transmission during a storm event. This is possible by extracting the pathways taken by the agents at different time slices during the storm. The agent based modelling approach has been applied to two small catchments in South East Spain. The modelling approach showed that the two catchments responded differently to the same rainfall event due to the differences in the runoff generation and overland flow connectivity between the two catchments. The model also showed that the time of travel to the nearest flow concentration is extremely important for determining the connectivity of a point in the landscape with the catchment outflow. Copyright © 2007 John Wiley & Sons, Ltd.     

Data assimilation for precipitation nowcasting using Bayesian inference

This work introduces a new variational Bayes data assimilation method for the stochastic estimation of precipitation dynamics using radar observations for short term probabilistic forecasting (nowcasting). A previously developed spatial rainfall model based on the decomposition of the observed precipitation field using a basis function expansion captures the precipitation intensity from radar images as a set of ‘rain cells’. The prior distributions for the basis function parameters are carefully chosen to have a conjugate structure for the precipitation field model to allow a novel variational Bayes method to be applied to estimate the posterior distributions in closed form, based on solving an optimisation problem, in a spirit similar to 3D VAR analysis, but seeking approximations to the posterior distribution rather than simply the most probable state. A hierarchical Kalman filter is used to estimate the advection field based on the assimilated precipitation fields at two times. The model is applied to tracking precipitation dynamics in a realistic setting, using UK Met Office radar data from both a summer convective event and a winter frontal event. The performance of the model is assessed both traditionally and using probabilistic measures of fit based on ROC curves. The model is shown to provide very good assimilation characteristics, and promising forecast skill. Improvements to the forecasting scheme are discussed.     

Hydrological data assimilation with the Ensemble Square-Root-Filter: Use of streamflow observations to update model states for real-time flash flood forecasting

The objective of the study is to evaluate the potential of a data assimilation system for real-time flash flood forecasting over small watersheds by updating model states. To this end, the Ensemble Square-Root-Filter (EnSRF) based on the Ensemble Kalman Filter (EnKF) technique was coupled to a widely used conceptual rainfall-runoff model called HyMOD. Two small watersheds susceptible to flash flooding from America and China were selected in this study. The modeling and observational errors were considered in the framework of data assimilation, followed by an ensemble size sensitivity experiment. Once the appropriate model error and ensemble size was determined, a simulation study focused on the performance of a data assimilation system, based on the correlation between streamflow observation and model states, was conducted. The EnSRF method was implemented within HyMOD and results for flash flood forecasting were analyzed, where the calibrated streamflow simulation without state updating was treated as the benchmark or nature run. Results for twenty-four flash-flood events in total from the two watersheds indicated that the data assimilation approach effectively improved the predictions of peak flows and the hydrographs in general. This study demonstrated the benefit and efficiency of implementing data assimilation into a hydrological model to improve flash flood forecasting over small, instrumented basins with potential application to real-time alert systems.     

Evaluating flash-flood warnings at ungauged locations using post-event surveys: a case study with the AIGA warning system

Abstract

This article presents a comparison between real-time discharges calculated by a flash-flood warning system and post-event flood peak estimates. The studied event occurred on 15 and 16 June 2010 at the Argens catchment located in the south of France. Real-time flood warnings were provided by the AIGA (Adaptation d’Information Géographique pour l’Alerte en Crue) warning system, which is based on a simple distributed hydrological model run at a 1-km² resolution using radar rainfall information. The timing of the warnings (updated every 15 min) was compared to the observed flood impacts. Furthermore, “consolidated” flood peaks estimated by an intensive post-event survey were used to evaluate the AIGA-estimated peak discharges. The results indicated that the AIGA warnings clearly identified the most affected areas. However, the effective lead-time of the event detection was short, especially for fast-response catchments, because the current method does not take into account any rainfall forecast. The flood peak analysis showed a relatively good correspondence between AIGA- and field-estimated peak values, although some differences were due to the rainfall underestimation by the radar and rainfall–runoff model limitations.

Editor Z.W. Kundzewicz; Guest editor R.J. Moore

Citation Javelle, P., Demargne, J., Defrance, D., Pansu, J. and Arnaud, P., 2014. Evaluating flash-flood warnings at ungauged locations using post-event surveys: a case study with the AIGA warning system. Hydrological Sciences Journal, 59 (7), 1390–1402. http://dx.doi.org/10.1080/02626667.2014.923970     

Evaluation of remote‐sensing‐based rainfall products through predictive capability in hydrological runoff modelling

The emergence of regional and global satellite‐based rainfall products with high spatial and temporal resolution has opened up new large‐scale hydrological applications in data‐sparse or ungauged catchments. Particularly, distributed hydrological models can benefit from the good spatial coverage and distributed nature of satellite‐based rainfall estimates (SRFE). In this study, five SRFEs with temporal resolution of 24 h and spatial resolution between 8 and 27 km have been evaluated through their predictive capability in a distributed hydrological model of the Senegal River basin in West Africa. The main advantage of this evaluation methodology is the integration of the rainfall model input in time and space when evaluated at the sub‐catchment scale. An initial data analysis revealed significant biases in the SRFE products and large variations in rainfall amounts between SRFEs, although the spatial patterns were similar. The results showed that the Climate Prediction Center/Famine Early Warning System (CPC‐FEWS) and cold cloud duration (CCD) products, which are partly based on rain gauge data and produced specifically for the African continent, performed better in the modelling context than the global SRFEs, Climate Prediction Center MORPHing technique (CMORPH), Tropical Rainfall Measuring Mission (TRMM) and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN). The best performing SRFE, CPC‐FEWS, produced good results with values of R²_NS between 0·84 and 0·87 after bias correction and model recalibration. This was comparable to model simulations based on traditional rain gauge data. The study highlights the need for input specific calibration of hydrological models, since major differences were observed in model performances even when all SRFEs were scaled to the same mean rainfall amounts. This is mainly attributed to differences in temporal dynamics between products. Copyright © 2009 John Wiley & Sons, Ltd.     

Spatial patterns in thunderstorm rainfall events and their coupling with watershed hydrological response

Weather radar systems provide detailed information on spatial rainfall patterns known to play a significant role in runoff generation processes. In the current study, we present an innovative approach to exploit spatial rainfall information of air mass thunderstorms and link it with a watershed hydrological model. Observed radar data are decomposed into sets of rain cells conceptualized as circular Gaussian elements and the associated rain cell parameters, namely, location, maximal intensity and decay factor, are input into a hydrological model. Rain cells were retrieved from radar data for several thunderstorms over southern Arizona. Spatial characteristics of the resulting rain fields were evaluated using data from a dense rain gauge network. For an extreme case study in a semi-arid watershed, rain cells were derived and fed as input into a hydrological model to compute runoff response. A major factor in this event was found to be a single intense rain cell (out of the five cells decomposed from the storm). The path of this cell near watershed tributaries and toward the outlet enhanced generation of high flow. Furthermore, sensitivity analysis to cell characteristics indicated that peak discharge could be a factor of two higher if the cell was initiated just a few kilometers aside.     

雷达测雨误差及其对淮河流域径流模拟的影响

将雷达测雨数据与分布式水文模型相耦合进行径流过程模拟,分析雷达测雨误差及其径流过程模拟效果,研究雷达测雨误差对径流过程模拟的影响效应.在对淮河流域气象中心业务化的5种淮河流域雷达测雨数据进行误差分析的基础上,采用雷达测雨数据驱动HEC-HMS水文模型,模拟分析淮河息县水文站以上流域2007年7月1-10日强降雨集中期的径流过程.结果表明:利用雷达测雨数据的径流模拟结果与实测资料的模拟结果基本吻合,各种雷达测雨数据误差经过HEC-HMS水文模型传递后,误差明显减小.联合校准法对应的模拟效果最好,过程流量相对误差NBs'和洪峰流量相对误差Z'分别为-20.2%和-13.3%.     

A numerical study of the rainstorm characteristics of the June 2005 flash flood with WRF/GSI data assimilation system over south‐east China

The evolution and structure of rainstorms associated with a flash‐flood event are simulated by the Advanced Weather Research and Forecasting (WRF‐ARW) model of the National Center for Atmospheric Research and the Gridpoint Statistical Interpolation (GSI) data assimilation (DA) system of the National Oceanic and Atmospheric Administration (NOAA) of the United States. The event is based on a flash flood that occurred in the central Guangdong Province of south‐east China during 20–21 June 2005. Compared to an hourly mixed rain‐gauge and satellite‐retrieved precipitation data, the model shows the capability to reproduce the intensity and location of rainfall; however, the simulation depends on three conditions to a large extent: model resolution, physical processes schemes and initial condition. In this case, the Eta Ferrier microphysics scheme and the initialization with satellite radiance DA with a fine 4‐km grid spacing nested grid and coarse 12‐km grid spacing outer grid are the best options. The model‐predicted rain rates, however, are slightly overestimated, and the activities of the storms do not precisely correspond with those observed, although peak values are obtained. Abundant moisture brought by the south‐westerly winds with a mesoscale low‐level jet from the South China Sea or Bay of Bengal and trapped within the XingfengJiang region encompassed by northern Jiulian, southern Lianhua and eastern small mountains are apparently the primary elements responsible for the flood event. All simulated rainstorms were initiated over the southern slopes of the Jiulian Mountain and moved south or north‐eastward within the Xingfengjiang region. Meanwhile, the Skew‐T/Log‐P diagrams show that there is a fairly high convective available potential energy (CAPE) over the active areas of the rainstorms. The higher CAPE provides a beneficial thermodynamic condition for the development of rainstorms, but the higher convective inhibition near the northern, eastern and southern mountains prohibits the storms from moving out of the region and causes heavy rainfall that is trapped within the area. Copyright © 2010 John Wiley & Sons, Ltd.     

Assessing flood risk in Baiyangdian Lake area in a changing climate using an integrated hydrological-hydrodynamic modelling

ABSTRACT

Flood-risk is affected by both climatic and anthropogenic factors. In this study, we assess changes in flood risk induced by a combination of climate change and flood prevention sets in the Baiyangdian (BYD) Lake area of China. Extreme storm events are analysed by the bias-corrected climate data from global climate models. A hydrological model is implemented and integrated with a hydrodynamic model to assess flood risk under three scenarios. The streamflow into the BYD was validated against historical flash-flood events. The results indicate that the changing climate increased extreme precipitation, upstream total inflow and the flood risk at the core region of Xiong’an New Area (XNA), the newly announced special economic zone in the BYD area. However, flood prevention measures can effectively mitigate the climatic effect. The research highlights the severe flash-flood risk at BYD and demonstrates the urgent need for a climate-resilient plan for XNA.     

半干旱地区洪水模拟效果差异性分析及影响因子响应评估

中国北方半干旱地区的降水与下垫面条件具有明显的时空异质性,如何完整准确地描述该类区域的水文过程是当代水文学研究的难点之一.选择半干旱地区水文实验区域——绥德流域和曹坪流域,通过构建不同时空规律的降水场,并结合3种不同产流机制的水文模型,进行大型数值模拟实验,去探究时间、空间、产流机制等因素对半干旱地区洪水模拟的影响,为该类地区水文模型的研制工作提供借鉴.结果 表明:1)半干旱地区中小流域的产流对降雨强度较为敏感,因此降水输入的时间步长对洪水模拟效果的影响程度较大;相比之下,流域雨量站数量的增减,仅体现在降雨分布场的暴雨中心缺失以及面平均降雨量的微小差别,对洪水模拟效果的影响程度较小.2)水文模型能否准确描述主导水文过程是半干旱地区洪水模拟效果优良的关键,流域的尺度效应及其下垫面条件的空间异质性是半干旱地区不同水文模型研制和调整应当优先考虑的问题,无论时间步长、雨量站数量怎么组合,产流结构适宜的模型其模拟效果总是趋于较好的结果.     

Delineation of flood generating processes and their hydrological response

Knowledge about flood generating processes can be beneficial for numerous applications. Especially in the context of climate change impact assessment, daily patterns of meteorological and catchment state conditions leading to flood events (i.e., storylines) may be of value. Here, we propose an approach to identify storylines of flood generation using daily weather and snow cover observations. The approach is tested for and applied to a typical pre‐Alpine catchment in the period between 1961 and 2014. Five precipitation parameters were determined that describe temporal and spatial characteristics of the flood associated precipitation events. The catchment's snow coverage was derived using statistical relationships between a satellite‐derived snow cover climatology and station snow measurements. Moreover, (pre‐) event snow melt sums were estimated using a temperature‐index model. Based on the precipitation and catchment state parameters, 5 storylines were identified with a cluster analysis: These are (a) long duration, low intensity precipitation events with high precipitation depths, (b) long duration precipitation events with high precipitation depths and episodes of high intensities, (c) shorter duration events with high or (d) low precipitation intensity, respectively, and (e) rain‐on‐snow events. The event groups have distinct hydrological characteristics that can largely be explained by the storylines' respective properties. The long duration, high intensity storyline leads to the most adverse hydrological response, namely, a combination of high peak magnitudes, high volumes, and long durations of threshold exceedance. The results show that flood generating processes in mesoscale catchments can be distinguished on the basis of daily meteorological and catchment state parameters and that these process types can explain the hydrological flood properties in a qualitative way. Hydrological simulations of daily resolution can thus be analysed with respect to flood generating processes.     

Effect of rainfall uncertainty on the performance of physically based rainfall–runoff models

This paper analyses the effect of rain data uncertainty on the performance of two hydrological models with different spatial structures: a semidistributed and a fully distributed model. The study is performed on a small catchment of 19.6 km² located in the north‐west of Spain, where the arrival of low pressure fronts from the Atlantic Ocean causes highly variable rainfall events. The rainfall fields in this catchment during a series of storm events are estimated using rainfall point measurements. The uncertainty of the estimated fields is quantified using a conditional simulation technique. Discharge and rain data, including the uncertainty of the estimated rainfall fields, are then used to calibrate and validate both hydrological models following the generalized likelihood uncertainty estimation (GLUE) methodology. In the storm events analysed, the two models show similar performance. In all cases, results show that the calibrated distribution of the input parameters narrows when the rain uncertainty is included in the analysis. Otherwise, when rain uncertainty is not considered, the calibration of the input parameters must account for all uncertainty in the rainfall–runoff transformation process. Also, in both models, the uncertainty of the predicted discharges increase in similar magnitude when the uncertainty of rainfall input increase.     

Hydrometeorological controls and erosive response of an extreme alpine debris flow

On 29 August, 2003, an intense convective storm system affected the Fella River basin, in the eastern Italian Alps, producing rainfall peaks of approximately 390 mm in 12 h. The storm triggered an unusually large debris flow in the ungauged Rio Cucco basin (0·65 km²), with a volume of approximately 78 000 m³. The analysis of the time evolution of the rainstorm over the basin has been based on rainfall estimates from radar observations and data recorded by a raingauge network. Detailed geomorphological field surveys, carried out both before and after the flood of August 2003, and the application of a distributed hydrological model have enabled assessment of flood response, estimation of erosion volumes and sediment supply to the channel network. The accounts of two eyewitnesses have provided useful elements for reconstructing the time evolution and the flow processes involved in the event. Liquid peak discharge estimates cluster around 20 m³ s^?1 km^?2, placing this event on the flood envelope curve for the eastern Italian Alps. The hydrological analysis has shown that the major controls of the flood response were the exceptional cumulated rainfall amount, required to exceed the large initial losses, and the large rainfall intensities at hourly temporal scales, required to generate high flood response at the considered basin scale. Observations on the deposits accumulated on the alluvial fan indicate that, although the dominant flow process was a debris flow, sheetflood also contributed to fan aggradation and fluvial reworking had an important role in winnowing debris‐flow lobes and redistributing sediment on the fan surface. This points out to the large discharge values during the recession phase of the flood, implying an important role for subsurface flow on runoff generation of this extreme flash flood event. Copyright © 2009 John Wiley & Sons, Ltd.     

Integration of gauge and radar rainfall to enable best simulation of hydrological parameters

This study is about use of spatially distributed rain in physically based hydrological models. In recent years, spatially distributed radar rainfall data have become available. The distributed radar rain is used to precisely model hydrologic processes and it is more realistic than the past practice of distribution methods like Thiessen polygons. Radar provides a highly accurate spatial distribution of rainfall and greatly improves the basin average rainfall estimates. However, quantification of the exact amount of rainfall from radar observation is relatively difficult. Thus, the fundamental idea of this study is to apply hourly gauge and radar rainfall data in a distributed hydrological model to simulate hydrological parameters. Hence the comparison is made between the outcomes of the WetSpa model from radar rainfall distribution and gauge rainfall distributed by the Thiessen polygon technique. The comparative plots of the hydrograph and the results of hydrological components such as evapotranspiration, surface runoff, soil moisture, recharge and interflow, reflect the spatially distributed radar input performing well for model outflow simulation.

EDITOR D. Koutsoyiannis

ASSOCIATE EDITOR F. Pappenberger     

Identification of the controlling factors for hydrological responses by artificial neural networks

Identifying the controlling factors for hydrological responses is of great importance for artificial neural network-based flood forecasting models, which are often hindered by the lack of physical mechanisms. To explore the first-order controlling factors of hydrograph patterns, a hybrid neural network was designed to analyse the impacts of potential driving variables with different temporal and spatial resolutions on hydrograph patterns. The Jinhua River Basin in Southeast China was used as an example in this study. Flood events with different hydrograph patterns and six external factors denoting potential controlling factors were individually classified into specific clusters using self-organizing maps (SOMs). Based on the back-propagation neural network (BPNN) and leave-one-out cross-validation methods, the controlling factors of different flood patterns were identified by comparing the performances of flood simulation models trained with datasets before and after the potential controlling factor classification. The results showed that (i) the classification of controlling factors indicating various runoff regimes significantly improved the performance of data-driven models in flood simulation in terms of correlation coefficient, Nash-Sutcliffe coefficient, and normalized root mean square error; (ii) the spatial distribution of antecedent soil moisture and vegetation conditions as well as the temporal distribution of rainfall dominated different hydrograph patterns; and (iii) the transition of dominant rainfall-runoff processes could be identified in an individual flood event using the hybrid SOM–BPNN model, indicating the varying influence of potential controlling factors on streamflow. Overall, the hybrid neural network models trained with datasets classified by controlling factors provide a general analytical framework to identify the governing dynamics for different flood patterns and improve the accuracy of flood simulations. Additionally, more attention should be devoted to improving the time to peak error of hydrological models, which cannot be settled by data-driven models trained with different data-splitting strategies.