Rainfall Input Data Research Articles

Adjusting rainfall inputs to hydrological models in a data-scarce area of Southern Africa

ABSTRACT A recent study in headwater sub-basins of the Zambezi River concluded that a machine learning model (HydroForecast) outperformed a deterministic monthly time-step water balance model (Pitman) when observed streamflow data were used as part of the training data for the machine learning approach. This was partly attributed to non-stationary bias in the input rainfall data that has a greater impact on a model that has mass balance constraints. The current study investigated methods of adjusting the input rainfall data based on an analysis of the streamflow simulation errors and found that the results for the water balance model could be improved so that they compared more favourably with the machine learning model. The improvements were achieved without changing the long-term statistical characteristics of the original monthly rainfall time series. It is acknowledged that there is some evidence that differences in daily rainfall patterns might explain some of the simulation errors.

Application of the Single Exponential Smoothing Method For Flood Disaster Prediction

The country of Indonesia is seen as one that is particularly vulnerable to natural catastrophes as well as calamities brought on by human activity. A disaster that may be brought on by both natural and human sources is a flood. Disasters brought on by flooding are unpredictable occurrences that frequently cause losses in the form of property damage, the theft of assets, and lost productivity at work and in school. Through this prediction information system, the people can find out the level of risk of flooding through excessive rainfall. in order to better anticipate and prepare for all possibilities that occur before the flood, the method used is Single Exponential Smoothing. This method was chosen because of the simple way the system works to find predictive values ??through past data. With this system, researchers can input rainfall data taken from the Meteorology, Climatology and Geophysics Agency, then the data is processed through the system and if rainfall gets high results. The risk of flooding will also be very high and a warning will be given to the public so that better prepared for the risk of flooding. The results obtained from this study are the results of an analysis of the exponential method single to obtain accurate rainfall prediction information with data MAD, MSE and MAPE.

A Neural Network-Based Hydrological Model for Very High-Resolution Forecasting Using Weather Radar Data

Many hydro-meteorological disasters in small and steep watersheds develop quickly and significantly impact human lives and infrastructures. High-resolution rainfall data and machine learning methods have been used as modeling frameworks to predict those events, such as flash floods. However, a critical question remains: How long must the rainfall input data be for an empirical-based hydrological forecast? The present article employed an artificial neural network (ANN)hydrological model to address this issue to predict river levels and investigate its dependency on antecedent rainfall conditions. The tests were performed using observed water level data and high-resolution weather radar rainfall estimation over a small watershed in the mountainous region of Rio de Janeiro, Brazil. As a result, the forecast water level time series only archived a successful performance (i.e., Nash–Sutcliffe model efficiency coefficient (NSE) > 0.6) when data inputs considered at least 2 h of accumulated rainfall, suggesting a strong physical association to the watershed time of concentration. Under extended periods of accumulated rainfall (>12 h), the framework reached considerably higher performance levels (i.e., NSE > 0.85), which may be related to the ability of the ANN to capture the subsurface response as well as past soil moisture states in the watershed. Additionally, we investigated the model’s robustness, considering different seeds for random number generating, and spacial applicability, looking at maps of weights.

The eco-hydraulics base as flood mitigation to overcome erosion and sedimentation problems: A case study in the Lae Kombih River, Indonesia

Erosion and sedimentation have a very big influence on flooding. Floods are strongly influenced by land use and population activities that change the river’s physical condition, including erosion and sedimentation. The river upstream was very steep, and the downstream was narrowing and meandering with high rainfall recorded. This study analyses erosion, sedimentation, and its handling using the eco-hydraulic base. The method involves input rainfall data, river hydraulics, land use, watershed area, and land cover. The analysis of hydrology, hydraulics, land use, flood discharge, and eco-hydraulic, inundation height, vegetation diameter, velocity reduced, and riverbank width measured in five bridges cross-sections along the river. The eco-hydraulic compares the width of existing riverbanks and design, high inundation, and the vegetation diameter to minimise flood discharge. Erosion in the right cliff is 22.73% and the left cliff is 37.04%, land erosion was 225.83 Mg∙ha –1∙year –1. The river’s bottom is formed by rocks of 0.18–1.30 mm. The plantation land used around the Lae Kombih River grows mainly an oil palm with a diameter of 0.5–0.7 m. The riverbank design on 100 m for vegetation diameter of 0.1–1.0 m can retain flood discharge for five years return period up to 72.3%, resulting in discharge of 112.04209.43 m 3∙s –1. The largest erosion and sedimentation on the river border is Dusun Silak, so it is recommended to plant Vetiveria zizanioides, Ipomoea carnea and Bambusoideae. An inundation height of 0.9 m can be recommended to design an embankment to be used as flood mitigation.

Development and Application of a Real-Time Flood Forecasting System (RTFlood System) in a Tropical Urban Area: A Case Study of Ramkhamhaeng Polder, Bangkok, Thailand

In urban areas of Thailand, and especially in Bangkok, recent flash floods have caused severe damage and prompted a renewed focus to manage their impacts. The development of a real-time warning system could provide timely information to initiate flood management protocols, thereby reducing impacts. Therefore, we developed an innovative real-time flood forecasting system (RTFlood system) and applied it to the Ramkhamhaeng polder in Bangkok, which is particularly vulnerable to flash floods. The RTFlood system consists of three modules. The first module prepared rainfall input data for subsequent use by a hydraulic model. This module used radar rainfall data measured by the Bangkok Metropolitan Administration and developed forecasts using the TITAN (Thunderstorm Identification, Tracking, Analysis, and Nowcasting) rainfall model. The second module provided a real-time task management system that controlled all processes in the RTFlood system, i.e., input data preparation, hydraulic simulation timing, and post-processing of the output data for presentation. The third module provided a model simulation applying the input data from the first and second modules to simulate flash floods. It used a dynamic, conceptual model (PCSWMM, Personal Computer version of the Stormwater Management Model) to represent the drainage systems of the target urban area and predict the inundation areas. The RTFlood system was applied to the Ramkhamhaeng polder to evaluate the system’s accuracy for 116 recent flash floods. The result showed that 61.2% of the flash floods were successfully predicted with accuracy high enough for appropriate pre-warning. Moreover, it indicated that the RTFlood system alerted inundation potential 20 min earlier than separate flood modeling using radar and local rain stations individually. The earlier alert made it possible to decide on explicit flood controls, including pump and canal gate operations.

Propagation of CMORPH rainfall errors to REW streamflow simulation mismatch in the upper Zambezi Basin

Study regionThis research is carried out in the Kabompo Basin, a headwater catchment of the Zambezi Basin which has an area of 67,261 km2. Kabompo River originates in North-Western Province of Zambia between the Zambezi and Congo River Basins. Study focusThis study focuses on error propagation of the National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center-MORPHing (CMORPH) rainfall product on simulated streamflow. Assessments are based on automated multi-objective calibration of the Representative Elementary Watershed (REW) model (2006–2012). Parameters of the model were optimized using the ɛ-NSGAII algorithm. Assessments on error propagation targeted streamflow modelling for hydrograph shape and volume, specific hydrograph characteristics, and water balance composition. New hydrological insights for the regionBy use of multiple objective functions, this study shows that uncorrected CMORPH results in substantial augmentation of rainfall error to streamflow simulation mismatch whereas bias corrected estimates result in attenuation of error. Analysis for water balance composition has great potential to improve application of satellite precipitation products in water management and decision making in the Zambezi basin. This study advises optimization of model parameters for each respective rainfall input data source so to identify outcomes and effects of respective rainfall data sources on the simulated water balance composition and closure.

An Optimal Rain-Gauge Network Using a GIS-Based Approach with Spatial Interpolation Techniques for the Mekong River Basin

Huynh, V.M.; Hong, S.; Kwon, Y.; Tran, V.T.; Huynh, V.T.M., and Kwon, S., 2021. An optimal rain-gauge network using a GIS-based approach with spatial interpolation techniques for the Mekong River Basin. In: Lee, J.L.; Suh, K.-S.; Lee, B.; Shin, S., and Lee, J. (eds.), Crisis and Integrated Management for Coastal and Marine Safety. Journal of Coastal Research, Special Issue No. 114, pp. 429–433. Coconut Creek (Florida), ISSN 0749-0208. Rainfall is one of the crucial types of data for hydrometeorological studies. In the context of climate change, which has become increasingly complex, hydrometeorological forecasting data are urgently necessary for further climatic simulation modeling. However, it is always a big challenge to achieve accurate rainfall data for such a minimum observation cost. Hence, an optimization design for the rain-gauge network in southern Vietnam was developed in this study, with a focus on the quantity and spatial distribution of rain gauges. The well-known geostatistics method combined with a spatial interpolation algorithm was applied using input rainfall data over the past 40 years. The selection of the appropriate spatial interpolation method affected the accuracy of the optimization process. The results of the interpolation analysis indicated the spatial variability of the rainfall patterns throughout the study area according to the seasons, as well as the geographic distribution of the rainfall, especially in the coastal area. The primary criterion for reducing the estimated data' uncertainty is improving the network density by installing additional rain-gauge stations. In this research, after the optimal number of stations was determined, a geographic information system-based approach and a multilayer overlaying technique were used to obtain the optimal rain gauges spatial distribution map. For future research, the proposed method is capable to apply in other related works to obtain more accurate rainfall estimation data.

Hydrological Models and Artificial Neural Networks (ANNs) to Simulate Streamflow in a Tropical Catchment of Sri Lanka

Accurate streamflow estimations are essential for planning and decision-making of many development activities related to water resources. Hydrological modelling is a frequently adopted and a matured technique to simulate streamflow compared to the data driven models such as artificial neural networks (ANNs). In addition, usage of ANNs is minimum to simulate streamflow in the context of Sri Lanka. Therefore, this study presents an intercomparison between streamflow estimations from conventional hydrological modelling and ANN analysis for Seethawaka River Basin located in the upstream part of the Kelani River Basin, Sri Lanka. The hydrological model was developed using the Hydrologic Engineering Centre-Hydrologic Modelling System (HEC-HMS), while the data-driven ANN model was developed in MATLAB. The rainfall and streamflows’ data for 2003–2010 period have been used. The simulations by HEC-HMS were performed by four types of input rainfall data configurations, including observed rainfall data sets and three satellite-based precipitation products (SbPPs), namely, PERSIANN, PERSIANN-CCS, and PERSIANN-CDR. The ANN model was trained using three well-known training algorithms, namely, Levenberg–Marquadt (LM), Bayesian regularization (BR), and scaled conjugate gradient (SCG). Results revealed that the simulated hydrological model based on observed rainfall outperformed those of based on remotely sensed SbPPs. BR algorithm-based ANN algorithm was found to be superior among the data-driven models in the context of ANN model simulations. However, none of the above developed models were able to capture several peak discharges recorded in the Seethawaka River. The results of this study indicate that ANN models can be used to simulate streamflow to an acceptable level, despite presence of intensive spatial and temporal data sets, which are often required for hydrologic software. Hence, the results of the current study provide valuable feedback for water resources’ planners in the developing region which lack multiple data sets for hydrologic software.

Mapping method of rainfall-induced landslide hazards by infiltration and slope stability analysis

By utilizing the Green-Ampt infiltration equation and the infinite slope stability model, a method for analyzing shallow slope failures caused by rainfall is developed. With rainfall intensity, soil characteristics, and topography, the modified Green-Ampt infiltration equation is used to estimate the rainfall infiltration capacity and depth of infiltration in a given slope. Assigning the calculated depth of infiltration as the depth of slip surface, the factor of safety of the slope is obtained through the infinite slope stability model. A time-series visualization map of the space-time varying factor of safety is generated when the method is implemented with the aid of Geographic Information System (GIS) software. The model is applied and validated with the landslides that occurred during the October 2019 Typhoon Hagibis in Marumori, Japan. The model results show good agreement with the reported time and depths of failure, and the analysis of the spatial distribution of predicted failures yielded receiver operating characteristic-area under the curve (ROC-AUC) value of 0.90. The applicability of the model can be extended for post-analysis, real-time, or projected assessment of slope stability, depending on the nature of input rainfall data (e.g., historical, real-time, forecast, hypothetical).

An Ensemble Flow Forecast Method Based on Autoregressive Model and Hydrological Uncertainty Processer

In the process of hydrological forecasting, there are uncertainties in data input, model parameters, and model structure, which cause a deterministic forecasting to fail to provide useful risk information to decision-makers. Therefore, the study of ensemble forecasting and the analysis of hydrological uncertainty are of great significance to guide the actual operation of reservoirs in the flood season. This study proposed a Bayesian ensemble forecast method, comprising of a Gaussian mixture model (GMM), a hydrological uncertainty processer (HUP), and an Autoregressive (AR) model. First, the GMM is selected as the marginal distribution function to estimate the uncertainty of observed and modelled data. Next, the AR model is used to correct the forecast rainfall data. Then, a modified HUP is used to deal with the uncertainty of hydrological model structure and rainfall input data. In the end, the ensemble flow forecast results are composed of the expected values of the posterior distribution obtained by HUP under different rainfall conditions. Taking the Three Gorges Reservoir (TGR) as a case study, the ensemble flow prediction in the forecast period is calculated by using the above method. Results show that the method proposed in this paper can improve the accuracy of runoff forecasts and reduce the uncertainty of the hydrological forecast.

Advanced hydrological streamflow simulation in a watershed using adjusted radar-rainfall estimates as meteorological input data

Among the input data of the watershed model for simulating changes of flowrate in the watershed, weather input data, especially input data related to rainfall, are the most important. Therefore, it is important to ensure the accuracy of rainfall input data to increase the accuracy of the watershed model results. Securing rainfall measurements with finer spatial and temporal resolutions is important in predicting flowrate variations at a sub-catchment, especially as they relate to global and local climate changes in weather conditions such as rainfall depth, rainfall intensity, etc. In this study, adjusted radar-rainfall estimates were suggested as alternative input data for watershed modeling. Through a statistical analysis of the representativeness of a ground rainfall measurement (10 km × 10 km grid), the necessity of radar-rainfall estimates (2 km × 2 km grid) was identified. By applying calibration factors to initial radar-rainfall estimates and comparing adjusted radar-rainfall estimates with ground rainfall measurements, it was proven that adjusted radar-rainfall estimates could be used as input data for watershed simulations (NSE > 0.92; n = 12). Adjusted radar-rainfall estimates and ground rainfall measurements were used as input data of the Soil and Water Assessment Tool model to predict flowrate variations at the outlets of a tributary and the entire watershed. As a result, the accuracies of the simulation results were improved for the outlets of a tributary and the entire watershed (NSE: 0.33 to 0.48 and 0.19 to 0.55, respectively). To obtain more reliable rainfall data, radar images easily accessible to users were applied, and the accuracy of the data was increased by applying simple equations to numerical data extracted from radar image processing. Additionally, the applicability of the adjusted radar-rainfall estimates was demonstrated by comparing the modeling results using the suggested rainfall data and existing ground-based rainfall data. The suggested methodologies are expected to contribute to more accurately predict the possibility of flood disasters in other regions and countries lacking infrastructure related to rainfall measurements and to establish appropriate countermeasures.

Flood Hazard Rating Prediction for Urban Areas Using Random Forest and LSTM

A flood hazard rating prediction model was developed that is based on a long short-term memory (LSTM) neural network and random forest. The target area was Samseong District in Seoul, which has a history of severe flooding. The Storm Water Management Model was used to generate training data for the LSTM model to predict the total overflow as the rainfall input data. Two-dimensional numerical analysis was performed to calculate inundation and flow velocity maps for training the random forest, which was used to generate a map of the predicted flood hazard rating of grid units given the total accumulative overflow of the target area. To confirm the goodness of fit, the proposed model was used to predict a flood hazard rating map for a rainfall event observed on July 27, 2011. The prediction accuracy for the flood hazard rating of each grid was 99.86% when the debris factor was considered and 99.99% when the debris factor was not considered.

Analysis of the effectiveness of existing pond to accommodate storm water runoff in UTHM

Rapid development within UTHM campus has changed the existing earth surface from pervious to impervious layer thus increasing the surface runoff. This study describe the application of XP SWMM software in determining the effectiveness of existing pond to accommodate the increase of storm water runoff based on designated rainfall event. Through this approach, the past studies, related journals, and Storm-water Management Manual (SWMM) are used as a reference for this study. Land and hydrographic survey is carried out in order to determine pond profile. The selected study principle using XPSWMM software is based on Rational Method as the catchment area is less than 80 ha and Manning Equation. The existing pond is designed to receive storm water runoff from two different catchments which are area around University Health Centre (PKU) and area starting from Centre for Diploma Studies (PPD) until Information Technology Centre (PTM). The input rainfall data is for 10 years rainfall event (2006-2016). It was found that there are no overflow from the pond, thus the volume for the pond (12356.39m^3) is effective to accommodate increasing storm water runoff from both catchment area.

Probabilistic Design Storm Method for Improved Flood Estimation in Ungauged Catchments

AbstractThe design storm approach with event‐based rainfall‐runoff models is a standard method for design flood estimation in ungauged catchments. The approach is conceptually simple and computationally inexpensive, but the underlying assumptions can lead to flawed design flood estimations. In particular, the implied average recurrence interval (ARI) neutrality between rainfall and runoff neglects uncertainty in other important parameters, leading to an underestimation of design floods. The selection of a single representative critical rainfall duration in the analysis leads to an additional underestimation of design floods. One way to overcome these nonconservative approximations is the use of a continuous rainfall‐runoff model, which is associated with significant computational cost and requires rainfall input data that are often not readily available. As an alternative, we propose a novel Probabilistic Design Storm method that combines event‐based flood modeling with basic probabilistic models and concepts from reliability analysis, in particular the First‐Order Reliability Method (FORM). The proposed methodology overcomes the limitations of the standard design storm approach, while utilizing the same input information and models without excessive computational effort. Additionally, the Probabilistic Design Storm method allows deriving so‐called design charts, which summarize representative design storm events (combinations of rainfall intensity and other relevant parameters) for floods with different return periods. These can be used to study the relationship between rainfall and runoff return periods. We demonstrate, investigate, and validate the method by means of an example catchment located in the Bavarian Pre‐Alps, in combination with a simple hydrological model commonly used in practice.

A Comparative Study of Various Hybrid Wavelet Feedforward Neural Network Models for Runoff Forecasting

Considering network topologies and structures of the artificial neural network (ANN) used in the field of hydrology, one can categorize them into two different generic types: feedforward and feedback (recurrent) networks. Different types of feedforward and recurrent ANNs are available, but multilayer perceptron type of feedforward ANN is most commonly used in hydrology for the development of wavelet coupled neural network (WNN) models. This study is conducted to compare performance of the various wavelet based feedforward artificial neural network (ANN) models. The feedforward ANN types used in the study include the multilayer perceptron neural network (MLPNN), generalized feedforward neural network (GFFNN), radial basis function neural network (RBFNN), modular neural network (MNN) and neuro-fuzzy neural network (NFNN) models. The rainfall-runoff data of four catchments located in different hydro-climatic regions of the world is used in the study. The discrete wavelet transformation (DWT) is used in the present study to decompose input rainfall data using db8 wavelet function. A total of 220 models are developed in this study to evaluate the performance of various feedforward neural network models. Performance of the developed WNN models is compared with their counterpart simple models developed without applying wavelet transformation (WT). The results of the study are further compared with - multiple linear regression (MLR) model which suggest that the WNN models outperformed their counterpart simple models. The hybrid wavelet models developed using MLPNN, the GFFNN and the MNN models performed best among the six selected data driven models explored in the study. Moreover, performance of the three best models is found to be similar and thus the hybrid wavelet GFFNN and the MNN models can be considered as an alternative to the most commonly used hybrid WNN models developed using MLPNN. The study further reveals that the wavelet coupled models outperformed their counterpart simple models only with the parsimonious input vector.

Coupling Numerical Weather Prediction and Hydrological Modelling for Runoff Forecast in Southern Brazil

Hydrological modelling is largely applied to forecast the effects of extreme rainfall on runoff dynamics. Little research has been done on coupling numerical weather prediction (NWP) and hydrological modelling to forecast runoff in southern Brazil given the limited hydrometeorological and NWP data available for the region. In this study, we evaluate a coupling method of NWP and a semi-distributed hydrological model to forecast runoff for a humid subtropical catchment (286 km²) in southern Brazil, state of Santa Catarina, municipality of Vidal Ramos. The HEC-HMS hydrological model was manually calibrated and validated using data of 2013. Four post-processing techniques of rainfall estimates were applied on rainfall products of two NWP models (BRAMS and GFS), providing rainfall input data for forecasting runoff. Then, a scenario of runoff forecast of an event in June 2014 was analyzed. We found the tested coupling method was able to forecast runoff with a lead time of 24 hours in which the best post-processing technique of rainfall is the “poor man ensemble” between the 90th quantile sampling of BRAMS and the mean of the GFS product. Hence, the validation of NWP products must be done carefully to avoid the propagation of uncertainties through a flood forecasting system.

Effects of hysteresis, rainfall dynamics, and temporal resolution of rainfall input data in solute transport modelling in uncropped soil

ABSTRACTNumerical simulations were carried out using the HYDRUS 1D model for three different types of uncropped soils. Input data were observed of rainfall during a 15-year period at three sites in Sweden. Simulations were run with and without consideration of hysteresis. Three different rainfall input time steps were used: 0.5, 4, and 24 h. Solute transport was evaluated using the centre of mass (zCOM) of breakthrough curves (BTCs) and largest depth where a pre-defined limit concentration was exceeded (zLC). Results showed that hysteretic flow was slower as compared to the non-hysteretic case. Results also showed that zCOM generally had a positive correlation with the number of high-intensity rainfalls and that the correlations were higher for the non-hysteretic case. The results show that it is important to use a sufficiently short rainfall input time step and an accurate description of the soil hysteresis when assessing the effects of a future higher-intensive rainfall climate with longer dry periods.Editor D. Koutsoyiannis; Associate editor D. Yang

Modelling Hydrologic Processes in the Mekong River Basin Using a Distributed Model Driven by Satellite Precipitation and Rain Gauge Observations.

The Mekong River is the most important river in Southeast Asia. It has increasingly suffered from water-related problems due to economic development, population growth and climate change in the surrounding areas. In this study, we built a distributed Geomorphology-Based Hydrological Model (GBHM) of the Mekong River using remote sensing data and other publicly available data. Two numerical experiments were conducted using different rainfall data sets as model inputs. The data sets included rain gauge data from the Mekong River Commission (MRC) and remote sensing rainfall data from the Tropic Rainfall Measurement Mission (TRMM 3B42V7). Model calibration and validation were conducted for the two rainfall data sets. Compared to the observed discharge, both the gauge simulation and TRMM simulation performed well during the calibration period (1998–2001). However, the performance of the gauge simulation was worse than that of the TRMM simulation during the validation period (2002–2012). The TRMM simulation is more stable and reliable at different scales. Moreover, the calibration period was changed to 2, 4, and 8 years to test the impact of the calibration period length on the two simulations. The results suggest that longer calibration periods improved the GBHM performance during validation periods. In addition, the TRMM simulation is more stable and less sensitive to the calibration period length than is the gauge simulation. Further analysis reveals that the uneven distribution of rain gauges makes the input rainfall data less representative and more heterogeneous, worsening the simulation performance. Our results indicate that remotely sensed rainfall data may be more suitable for driving distributed hydrologic models, especially in basins with poor data quality or limited gauge availability.

Hydrological and hydraulic modelling applied to the mapping of flood-prone areas

Overbank flooding caused by historically high flows, such as that in the Rio Mundau watershed (lying between the states of Alagoas and Pernambuco) in 2010, has been the cause of widespread damage. The purpose of work described in this paper was to propose a mapping of areas liable to flooding in the township of Rio Largo (Alagoas) in the Rio Mundau watershed by means of an “off-line” coupling of the hydrological/hydraulic models (MGB-IPH/HEC-RAS), through consideration of extreme floods with different return periods for discharge. The hydrological model had a parameterization appropriate for extreme floods, using as input rainfall data with different return periods. Calibration and validation of the hydrological model were adequate in drainage areas larger than 1500 km2, but were less acceptable in headwater drainage areas where different geology and soil cover gave rise to surface runoff. The hydraulic model showed good agreement with point observations of flood levels in 2010 in both rural and urban areas along the water-course (R2 = 0.99; RMSE = 1.41 m and CV (RMSE) = 0.04). In urban areas distant from the river, however, flood levels were over-estimated, indicating a need to use more detailed Digital Elevation Models. Flood events with return period greater than 50 years have the potential to cause great damage (floods exceeding 0.46 km2 in the urban area). The study showed that the use of coupled models was a viable approach for mapping areas liable to flooding, when it is not possible to analyse local flow frequencies in support of a hydraulic model.

A comparison between wavelet based static and dynamic neural network approaches for runoff prediction

Summary In order to predict runoff accurately from a rainfall event, the multilayer perceptron type of neural network models are commonly used in hydrology. Furthermore, the wavelet coupled multilayer perceptron neural network (MLPNN) models has also been found superior relative to the simple neural network models which are not coupled with wavelet. However, the MLPNN models are considered as static and memory less networks and lack the ability to examine the temporal dimension of data. Recurrent neural network models, on the other hand, have the ability to learn from the preceding conditions of the system and hence considered as dynamic models. This study for the first time explores the potential of wavelet coupled time lagged recurrent neural network (TLRNN) models for runoff prediction using rainfall data. The Discrete Wavelet Transformation (DWT) is employed in this study to decompose the input rainfall data using six of the most commonly used wavelet functions. The performance of the simple and the wavelet coupled static MLPNN models is compared with their counterpart dynamic TLRNN models. The study found that the dynamic wavelet coupled TLRNN models can be considered as alternative to the static wavelet MLPNN models. The study also investigated the effect of memory depth on the performance of static and dynamic neural network models. The memory depth refers to how much past information (lagged data) is required as it is not known a priori. The db8 wavelet function is found to yield the best results with the static MLPNN models and with the TLRNN models having small memory depths. The performance of the wavelet coupled TLRNN models with large memory depths is found insensitive to the selection of the wavelet function as all wavelet functions have similar performance.