Event-based Rainfall-runoff Model Research Articles

Changes in flood-associated rainfall losses under climate change

Climate change is expected to impact the severity and frequency of floods, which has implications for flood risk management. Design floods and derived flood frequency curves obtained using event-based rainfall-runoff models are widely used in industry to assess flood risks for planning and design purposes. For these approaches it is necessary to have (a) rainfall inputs, and (b) rainfall losses specified, the latter representing the amount of rainfall that is either intercepted, stored on the surface, or infiltrated into the soil and does not contribute to the flood hydrograph. There is extensive research on changes in flooding under climate change that focus on projections of rainfall. However, there is little research into projections of rainfall losses under climate change, despite the knowledge that their changes will modulate the flood response. Here, we present one of the first studies seeking to quantify how rainfall losses, as represented by estimates of initial and continuing losses used in event-based models, are projected to change under climate change.We identify dependencies between rainfall losses and antecedent soil moisture in around half (over 200) of the largely unregulated catchments (i.e. watersheds) in Australia analysed in this study and use these relationships to project rainfall losses under climate change. Near universal increases in both the mean and variance of both initial losses and continuing losses are projected in these catchments, suggesting that increased rainfall losses could offset the impact of increased rainfalls for frequently occurring floods and result in an increased variance in flood responses.

Development of a New Event-Based Rainfall-Runoff Equation Based on Average Rainfall Intensity During an Event

Event-based rainfall-runoff models are practical tools commonly used to predict catchments’ response to a rainfall event. However, one of the main concerns is that the characteristics of rain events are neglected in the model development. This paper develops a novel event-based rainfall-runoff equation to incorporate rainfall characteristics into account. The performance of the new equation is evaluated based on the root mean square error, Nash–Sutcliffe efficiency coefficient, and per cent bias for 13,339 rainfall-runoff events between 2005 and 2020 over 23 catchments across New Zealand and Australia with oceanic, mediterranean, tropical, subtropical, and semiarid climates. Compared to the previous event-based models, the new equation shows an improvement in runoff estimation in almost all case studies. Furthermore, considering the new equation is simple, efficient, and takes the rain event duration into account, the new equation has the potential to become a robust alternative method to the conventional curve number method in hydrological engineering projects.

Modeling the Rainfall–Runoff Relationship with TOPMODEL in the Wadi El Kebir Watershed

The rainfall–runoff relationship was studied in the Wadi El Kebir watershed, located in Northeastern Algeria, using TOPMODEL (topography based hydrological model). This is a geomorphological and semi-distributed model which is used to predict the hydrological behaviour of watersheds and to calculate the water storage deficit of an aquifer in any location. It uses topographic information of the watershed to predict the extent of contributing areas in the production of runoff. TOPMODEL was applied with event-based rainfall–runoff modeling where 13 hourly rainfall series were used to predict the discharge at the basin outlet. A digital elevation model (DEM) was also used to define the contours of the basin and to map out the drainage directions and the topographic index. TOPMODEL was calibrated and validated using the measured discharges and various objective functions such as Nash (Nash-Sutcliffe) and coefficient of determination (R2). The TOPMODEL results showed a high-performance level. Indeed, after the calibration and validation procedure, the performance of the model oscillates between satisfactory and very good. For the calibration, Nash varied between 0.59 and 0.91, and R2 between 0.66 and 0.91. However, the values of these criteria coefficients were slightly reduced during the validation phase, Nash (0.53 to 0.84) and R2 (0.59 to 0.87). Also, the proposed model shows the weak contribution of groundwater flows in the hydrological response of the study area.

Performance Evaluation of Three Event-Based Rainfall-Runoff Models for a Small Tropical Watershed

Performance Evaluation of Three Event-Based Rainfall-Runoff Models for a Small Tropical Watershed

Effects of Microtopography on Runoff Generation in Plain Farmland: New Insights from an Event-Based Rainfall–Runoff Model

Plain farmland areas without significant topographic slope exhibit microtopographic features of different scales. Quantitative assessment of the effects of microtopography at different scales on runoff generation in typical farmland areas is of great significance for regional water resources management and flood disaster forecasting. The main objective of the study was to develop an event-based rainfall–runoff model based on the layered Green–Ampt model (LGAM) with the consideration of plot-scale microtopographic features in plain farmland areas. Our experimental field, located in Taihu Lake Basin, was classified into three types of topographic subunits (i.e., main field, rill, and ditch) according to the average elevation. To simplify the concentration process for three topographic subunits, the average concentration time method was employed. Here, various experimental scenarios were simulated, including two classical unsteady rainfall events in homogeneous soil, one ponding infiltration experiment, and two typical rainfall–runoff events in the experimental field. We found that the multilayered setting showed higher accuracy than the homogeneous setting for simulating infiltration in the ponding infiltration experiment in the field. The RMSE of simulated ponding water depth reduced from 0.166 cm to 0.035 cm and NSE rose from 0.988 to 0.999. The simulated hydrograph considering microtopography effects proved higher accuracy than that under unified topography assumption. After classifying topography, the RMSE and NSE of simulated hydrographs decreased and increased, respectively. The lower the topographic subunit, the earlier the outflow occurred. At the early stage, the runoff mostly originated from the relatively low topographic subunits. Infiltration-excess regime under saturated condition may initially dominate in the low-lying ditch under intense rainfall, with extremely high runoff coefficient. Concentration process in the plain farmland area was affected by both rainfall intensity and microtopography. The greater the rainfall intensity, the shorter the average concentration time. The concentration velocity under heavy rainfall was four times faster than that under light rainfall. The lower topographic subunit was characterized by shorter concentration pathway and average concentration time. Ditches reduced the peak flow and advanced the time to peak. This quantitative study provides new insights into effects of microtopography on runoff generation in plain farmland area as well as an effective alternative for plot-scale rainfall–runoff modeling.

Effects of Probability-Distributed Losses on Flood Estimates Using Event-Based Rainfall-Runoff Models

Probability distributions of initial losses are investigated using a large dataset of catchments throughout Australia. The variability in design flood estimates caused by probability-distributed initial losses and associated uncertainties are investigated. Based on historic data sets in Australia, the Gamma and Beta distributions are found to be suitable for describing initial loss data. It has also been found that the central tendency of probability-distributed initial loss is more important in design flood estimation than the form of the probability density function. Findings from this study have notable implications on the regionalization of initial loss data, which is required for the application of Monte Carlo methods for design flood estimation in ungauged catchments.

Comparison of seven simple loss models for runoff prediction at the plot, hillslope and catchment scale in the semiarid southwestern U.S.

Infiltration excess overland flow is the dominant mechanism for runoff generation in many dryland watersheds. Event-based rainfall-runoff models therefore partition precipitation into two components: loss and excess precipitation. The latter is then transformed into a runoff hydrograph. Numerous loss models have been developed over the past century ranging from simple empirical to sophisticated physically based methods. Complex models can lead to equifinality and associated uncertainty at larger spatial scales with varying soil and cover conditions. Simple models are therefore widely used in hydrologic practice. In the absence of measured data in many arid and semiarid regions, model parameters are often estimated based on laboratory or field infiltrometer tests. Given the documented importance of spatial scale on the runoff response in dryland catchments, it is not clear how models parameterized at the point or soil column scale will perform at the hillslope or catchment scale under real-world conditions. In this study, we compared the performance of seven simple loss models with three or less parameters: the Philip, Smith-Parlange, Horton, Kostiakov, curve number (CN), initial and constant (IC) and the linear and constant (LC) models. The latter is a modification of the IC model introduced in this study. We estimated parameters at the plot scale (2.8 m2) using rainfall simulation and then tested model performance at the hillslope (1.5–3.7 ha) and catchment scale (2.4–2.8 km2) based on measured rainfall-runoff data at two sites in New Mexico and Arizona, U.S. Results show that rainfall simulation can be used successfully to parameterize loss models at the hillslope scale. At the catchment scale, most models showed positive bias, suggesting that other losses (such as channel or transmission losses) play an important role in determining the catchment runoff response. Rainfall intensity and temporal distribution were found to be crucial for accurate runoff prediction. Models that are sensitive to rainfall intensity during the entire simulation (Philip, Smith-Parlange, Horton, Kostiakov, LC) therefore performed better than those with an initial abstraction term (CN, IC). During intermittent rain, the best results were achieved by methods expressing infiltration capacity as a function of cumulative infiltration (LC, Smith-Parlange).

Study on the applicability of the SCS-CN-based models to simulate floods in the semi-arid watersheds of northern Algeria

Algeria has experienced catastrophic floods over the second half of the twentieth century, causing many deaths and extensive material damage. This study was conducted to find a suitable event-based rainfall-runoff (RR) model for semi-arid conditions, where continuous data are not available in all regional basins. The study compared, based on data availability, the SCS-CN model based on the antecedent moisture conditions (AMC) and four modified SCS-CN models incorporating antecedent moisture amounts (AMA) in order to find the best model to reproduce the flood hydrographs in two catchments. The modified models were predominant over the SCS-CN method. Nonetheless, the Singh et al. (Water Resour Manag 29:4111–4127, 2015. https://doi.org/10.1007/s11269-015-1048-1 ) model (M4) and the Verma et al. (Environ Earth Sci 76:736, 2017a. https://doi.org/10.1007/s12665-017-7062-2 ) model (M5) were superior and demonstrated more stable structures. Coupled with the Hayami routing model, the models showed promising results and were able to reproduce the observed hydrographs’ shape. However, it was impossible to choose the preferred model since they each excelled as to a criterion. Therefore, the corresponding outputs were combined using the simple average (SA) method and the weighted average (WA) method. We found that the WA method showed better results in the two catchments and allowed a more accurate prediction according to the performance criteria.

Implications of event-based loss model structure on simulating large floods

Event-based rainfall-runoff models used for design flood estimation and operational flood forecasting commonly represent losses (the amount of rainfall which does not appear as runoff) using lumped, parsimonious models. In gauged catchments, these loss values are typically estimated by calibrating models to observed flood events. Studies which compare loss models tend to focus on which model best reproduces observed events, however, flood modelling often requires estimating events larger than those used for calibration. It is thus of interest to examine how loss models perform over a range of flood magnitudes, particularly larger floods. Here, we evaluate the efficacy of different loss models by exploring the dependence between event magnitude and model performance. We assess four loss models commonly used in flood modelling: Initial Loss Continuing Loss (ILCL), Initial Loss Proportional Loss (ILPL), Soil Conservation Service Curve Number (SCSCN) and Probability Distributed Model (PDM).The four loss models were incorporated into a lumped rainfall runoff model that was calibrated to observed storm events for 35 small, rural Australian catchments using a total of 876 unique storm events. In calibration, ILPL was found to best reproduce observed hydrographs, however the difference in performance between the four different models in terms of NSE was small. Dependence between event magnitude and model performance was explored by re-simulating observed events with the loss model parameters fixed to represent typical catchment conditions. These typical parameters were estimated from the median of the observed parameter ranges in each catchment with the largest five events excluded so that the impact of extrapolating model parameters to events larger than which they were calibrated could be explored. When these typical parameters were applied to large flood events the ILCL model displayed minimal dependence between model performance and event magnitude. In contrast, ILPL, SCSCN and PDM all clearly displayed an underprediction bias which increased with increasing event size. We explain this behaviour through the loss model structure of ILCL where ongoing loss (the loss which occurs after runoff begins) is constant and independent of rainfall fvmagnitude. In contrast, the three remaining models represent ongoing loss as a proportion of rainfall. As the difference in magnitude between small and large floods increases with climate change, extrapolation of flood models is likely to become a more pertinent issue in the future. We conclude that loss models which are independent of event magnitude are better suited to flood modelling applications where extrapolation is required.

Comparison of different methodologies for rainfall–runoff modeling: machine learning vs conceptual approach

Accurate short-term rainfall–runoff prediction is essential for flood mitigation and safety of hydraulic structures and infrastructures. This study investigates the capability of four machine learning methods (MLM), optimal pruning extreme learning machine (OPELM), multivariate adaptive regression spline (MARS), M5 model tree (M5Tree, and hybridized MARS and Kmeans algorithm (MARS-Kmeans), in hourly rainfall–runoff modeling (considering 1-, 6- and 12-h horizons). Their results are compared with a conceptual method, Event-Based Approach for Small and Ungauged Basins (EBA4SUB) and multi-linear regression (MLR). Hourly rainfall and runoff data gathered from Ilme River watershed, Germany, were divided into two equal parts, and MLM were validated considering each part by swapping training and testing datasets. MLM were compared with EBA4SUB using four events and with respect to three statistics, root-mean-square errors (RMSE), mean absolute error (MAE) and Nash–Sutcliffe efficiency (NSE). Comparison results revealed that the newly developed hybridized MARS-Kmeans method performed superior to the OPELM, MARS, M5Tree and MLR methods in prediction of 1-, 6- and 12-h ahead runoff. Comparison with conceptual method showed that all the machine learning models outperformed the EBA4SUB and OPELM provided slightly better performance than the other three alternatives in event-based rainfall–runoff modeling.

Towards an ensemble-based short-term flood forecasting using an event-based flood model- incorporating catchment-average estimates of soil moisture

Extreme floods pose a threat to life and property and accurate and reliable flood forecasts are required to mitigate the consequences of these events. In this paper, we propose a novel Monte-Carlo based framework that utilizes an ensemble-based short-term flood forecasting model. An event-based rainfall-runoff model is selected due to its simplicity and wide use by industry practitioners. However, a challenge with event models is that they are unable to account for the initial condition of the catchment wetness at the commencement of a particular rainfall event. To address this issue, we used independent catchment-average estimates of soil moisture to estimate catchment losses at the commencement and over the duration of the event. The adopted framework enables uncertainty in catchment losses and rainfall depth and patterns to be quantified and incorporated in forecasts. The accuracy and reliability of stochastic components of this framework are evaluated using ensemble verification measures. Later, these measures are used to evaluate four scenarios based on different combinations of deterministic and stochastic conditions. In the stochastic scenarios, the uncertainty of the flood forecasts is reliably quantified (88% reliability on average). The results show that operationally available estimates of catchment wetness can be used to account for uncertainty from event-based loss parameters, and that the uncertainty from average forecast rainfall depth is considerably more important than the uncertainty from forecast losses and the spatio-temporal patterns of the rainfall. This study has shown that event-based models have the potential to be applied within a probabilistic framework to operationally generate probabilistic flood forecasts.

Monitoring soil moisture at the catchment scale – A novel approach combining antecedent precipitation index and radar-derived rainfall data

Knowledge about soil moisture is important for event-based rainfall-runoff models but monitoring conditions at the catchment scale is not a trivial task. Soil moisture is highly variable in space and time, particularly in dry climates with seasonal and spatially heterogeneous rainfall. Point measurements are difficult to upscale, and remotely sensed (RS) data often lack in spatial or temporal resolution for local or regional studies. Longer latency periods – the time required before data becomes available – of some RS data make them less applicable to time-sensitive analyses such as flash flood forecasting. This study evaluated a novel approach for estimating catchment-scale volumetric soil moisture using an antecedent precipitation index (API) -based model. The model was trained and tested using in-situ soil moisture measurements collected during a 3-month field sampling campaign in a 142 km2 study area in central New Mexico. The calibrated model was applied at the catchment scale to produce soil moisture grids from radar-derived rainfall estimates. Model performance, resolution and latency were compared to satellite-based soil moisture estimates. Benefits of the proposed new method include high spatial resolution (1 × 1 km or less depending on the precipitation data source) and high prediction accuracy (root mean square errors 0.014–0.018 m3/m3). Given the short latency period for radar-derived rainfall data, the method has potential for use in operational flood risk assessment and forecasting.

Probabilistic Event Based Rainfall-Runoff Modeling Using Copula Functions

Multivariate probability analysis of hydrological elements using copula functions can significantly improve the modeling of complex phenomena by considering several dependent variables simultaneously. The main objectives of this study were to: (i) develop a stand-alone and event-based rainfall-runoff (RR) model using the common bivariate copula functions (i.e. the BCRR model); (ii) improve the structure of the developed copula-based RR model by using a trivariate version of fully-nested Archimedean copulas (i.e. the FCRR model); and (iii) compare the performance of the developed copula-based RR models in an Iranian watershed. Results showed that both of the developed models had acceptable performance. However, the FCRR model outperformed the BCRR model and provided more reliable estimations, especially for lower joint probabilities. For example, when joint probabilities were increased from 0.5 to 0.8 for the peak discharge (qp) variable, the reliability criteria value increased from 0.0039 to 0.8000 in the FCRR model, but only from 0.0010 to 0.6400 in the BCRR model. This is likely because the FCRR considers more than one rainfall predictor, while the BCRR considers only one.

Flood frequency analysis by an event-based rainfall-runoff model in selected catchments of southern Poland

The study evaluated the applicability of Event-Based Approach for Small and Ungauged Basins (EBA4SUB) for calculating annual peak flows with a specific return period (Q<sub>T</sub>) in southern Poland. Data used in the calculations in a form of observation series of annual peak flows were derived from the Institute of Meteorology and Water Management in Warsaw and covered a multi-year period 1971–2015. The data were statistically verified for their homogeneity, significance of monotonic trends, outliers and equality of variances. Peak flows with a given return period were estimated by a statistical method of Pearson Type III distribution, and by EBA4SUB model. The analysis showed that Q<sub>T</sub> for the investigated catchments was the most accurately matching the values derived from the statistical method when EBA4SUB model was employed. This was evidenced by the values of average relative errors that reached 34% for EBA4SUB model (with beta hyetograph). The results of the study demonstrated usefulness of EBA4SUB model for the estimation of Q<sub>T</sub> quantiles in catchments of the upper Vistula water region.

Comparison of SCS and Green-Ampt Distributed Models for Flood Modelling in a Small Cultivated Catchment in Senegal

The vulnerability to floods in Africa has increased over the last decades, together with a modification of land cover as urbanized areas are increasing, agricultural practices are changing, and deforestation is increasing. Rainfall-runoff models that properly represent land use change and hydrologic response should be useful for the development of water management and mitigation plans. Although some studies have applied rainfall-runoff models in West Africa for flood modelling, there is still a need to develop such models, while many data are available and have not still been used for modelling improvement. The Ndiba catchment (16.2 km2), which is located in an agricultural area in south Senegal, is such catchment, where a lot of hydro-climatic data has been collected between 1983 and 1992. Twenty-eight flood events have been extracted and modelled by two event-based rainfall-runoff models that are based on the Soil Conservation Service (SCS) or the Green-Ampt (GA) models for runoff, both coupled with the distributed Lag and Route (LR) for routing. Both models were able to reproduce the flood events after calibration, but they had to account for that the infiltration processes are highly dependent on the tillage of the soils and the growing of the crops during the rainy season, which made the initialization of the event-based models difficult. The most influent parameters for both models (the maximal water storage capacity for SCS, the hydraulic conductivity at saturation for Green-Ampt) were mostly related to the development stage of the vegetation, described by a Normalized Difference Vegetation Index (NDVI) anomaly. The SCS model performed finally better than the Green-Ampt model, because Green-Ampt was very sensitive to the variability of the hydraulic conductivity at saturation. The variability of the parameters of the models highlights the complexity of this kind of cultivated catchment, with highly non stationary conditions. The models could be improved by a better knowledge of the tillage practices, and a better integration of these practices in the parameters predictors.

Single- and two- step optimization of infiltration parameters and Manning’s roughness coefficients for a watershed using a multi-objective genetic algorithm

In the present study, two optimization methods, i.e. single-step optimization (SS) and two-step optimization with varying parameter range (TSVR) have been developed for optimizing four infiltration and two roughness coefficients of an event-based distributed rainfall run-off model using a multi-objective genetic algorithm (MGA). For this, an event-based rainfall run-off model is integrated with MGA to evaluate two optimization methods. Correlation coefficient (r) and Nash–Sutcliffe efficiency (NSE) are used as objective functions of MGA. Further, the MGA integrated run-off model has been tested on the Harsul watershed located in Maharashtra, India and the simulated results are compared with the observed data. From the simulation results of SS optimization method, it is observed that the average percentage error of the volume of run-off, peak run-off and time to peak are 40.31%, 23.44% and 22.74%, respectively. In case of TSVR optimization method, the average percentage error of the volume of run-off, peak run-off and time to peak are 38.91%, 14.34% and 29.23%, respectively. Results of this study indicate that the TSVR optimization method reduced the average percentage error in volume of run-off and peak run-off and increased the average percentage error of time to peak when compared to the SS optimization method.

Robust Initial Wetness Condition Framework of an Event-Based Rainfall–Runoff Model Using Remotely Sensed Soil Moisture

Runoff prediction in limited-data areas is vital for hydrological applications, such as the design of infrastructure and flood defenses, runoff forecasting, and water management. Rainfall–runoff models may be useful for simulation of runoff generation, particularly event-based models, which offer a practical modeling scheme because of their simplicity. However, there is a need to reduce the uncertainties related to the estimation of the initial wetness condition (IWC) prior to a rainfall event. Soil moisture is one of the most important variables in rainfall–runoff modeling, and remotely sensed soil moisture is recognized as an effective way to improve the accuracy of runoff prediction. In this study, the IWC was evaluated based on remotely sensed soil moisture by using the Soil Conservation Service-Curve Number (SCS-CN) method, which is one of the representative event-based models used for reducing the uncertainty of runoff prediction. Four proxy variables for the IWC were determined from the measurements of total rainfall depth (API5), ground-based soil moisture (SSMinsitu), remotely sensed surface soil moisture (SSM), and soil water index (SWI) provided by the advanced scatterometer (ASCAT). To obtain a robust IWC framework, this study consists of two main parts: the validation of remotely sensed soil moisture, and the evaluation of runoff prediction using four proxy variables with a set of rainfall–runoff events in the East Asian monsoon region. The results showed an acceptable agreement between remotely sensed soil moisture (SSM and SWI) and ground based soil moisture data (SSMinsitu). In the proxy variable analysis, the SWI indicated the optimal value among the proposed proxy variables. In the runoff prediction analysis considering various infiltration conditions, the SSM and SWI proxy variables significantly reduced the runoff prediction error as compared with API5 by 60% and 66%, respectively. Moreover, the proposed IWC framework with remotely sensed soil moisture indicates an improved Nash–Sutcliffe efficiency from 0.48 to 0.74 for the four catchments in the Korean Peninsula. It can be concluded that the SCS-CN method extended with remotely sensed soil moisture for reducing uncertainty in the runoff prediction and the proxy variables obtained from the soil moisture data provided by the ASCAT can be useful in enhancing the accuracy of runoff prediction over a range of spatial scales.

Choice of rainfall inputs for event-based rainfall-runoff modeling in a catchment with multiple rainfall stations using data-driven techniques

Input selection for data-driven rainfall-runoff models is an important task as these models find the relationship between rainfall and runoff by direct mapping of inputs to output. In this study, two different input selection methods were used: cross-correlation analysis (CCA), and a combination of mutual information and cross-correlation analyses (MICCA). Selected inputs were used to develop adaptive network-based fuzzy inference system (ANFIS) in Sungai Kayu Ara basin, Selangor, Malaysia. The study catchment has 10 rainfall stations and one discharge station located at the outlet of the catchment. A total of 24 rainfall-runoff events (10-min interval) from 1996 to 2004 were selected from which 18 events were used for training and the remaining 6 were reserved for validating (testing) the models. The results of ANFIS models then were compared against the ones obtained by conceptual model HEC-HMS. The CCA and MICCA methods selected the rainfall inputs only from 2 (stations 1 and 5) and 3 (stations 1, 3, and 5) rainfall stations, respectively. ANFIS model developed based on MICCA inputs (ANFIS-MICCA) performed slightly better than the one developed based on CCA inputs (ANFIS-CCA). ANFIS-CCA and ANFIS-MICCA were able to perform comparably to HEC-HMS model where rainfall data of all 10 stations had been used; however, in peak estimation, ANFIS-MICCA was the best model. The sensitivity analysis on HEC-HMS was conducted by recalibrating the model by using the same selected rainfall stations for ANFIS. It was concluded that HEC-HMS model performance deteriorates if the number of rainfall stations reduces. In general, ANFIS was found to be a reliable alternative for HEC-HMS in cases whereby not all rainfall stations are functioning. This study showed that the selected stations have received the highest total rain and rainfall intensity (stations 3 and 5). Moreover, the contributing rainfall stations selected by CCA and MICCA were found to be located near the outlet of contributing sub-catchments. This provides valuable information towards identifying the more contributing sub-catchments in catchments such as Sungai Kayu Ara where no flow measurement is available for sub-catchments.

Regional Design Storm and Flood Modelling—Risk Implications in Ungauged Catchments

Most planned developments in a catchment for control of excess water using a culvert, bridge or dam spillway are located at a site in a stream where there are no discharge measurements. Even though, for gauged catchments a number of established flood frequency models and rainfall-runoff models do exist, for ungauged catchments mostly regional flood frequency and event-based rainfall-runoff models are used, which depend on regional parameters. In this paper, a regional approach for design floods is presented and risk implication for design of drainage structures assessed. A case study in light of the above has been considered at four ungauged sites in the Limpopo Drainage Basin in north-eastern Botswana.

EVENT-BASED RAINFALL-RUNOFF MODELING USING ADAPTIVE NETWORK-BASED FUZZY INFERENCE SYSTEM

Population growth and transformation of agricultural or forest landscapes to built-up areas are the common phenomenon in the fast developing countries. Such changes have significant impact on hydrologic processes in the catchment which in turn may end up with an increase in both magnitude and frequency of floods in urban areas. Therefore, reliable rainfall-runoff models that are able to estimate discharge of a catchment accurately are in need. To date, several physically-based models are developed to capture the rainfall-runoff process; however, they require significant number of parameters which could be difficult to be measured or estimated. Beside these models, the artificial intelligence techniques have shown their ability to identify a direct mapping between inputs and outputs with less number of physical parameters. Adaptive network-based fuzzy inference system (ANFIS) is one of the well-practiced techniques in hydrological time series modeling. The aim of this study was to check the capability of ANFIS in event-based rainfall runoff modeling for a tropical catchment. A total of 70 rainfall-runoff events were extracted from twelve years hourly rainfall and runoff data of Semenyih River catchment where 50 of them were chosen for training and the remaining 20 for testing. An input selection method based on correlation analysis and mutual information was developed to identify the proper input combinations of rainfall and discharge antecedents. The results obtained by ANFIS model were then compared with an autoregressive model with exogenous inputs (ARX) as a bench mark. Results showed that ANFIS outperforms ARX model and has capabilities to be used as a reliable rainfall-runoff modeling tool.