Page 1
Development of Rainfall-Runoff Model Using Soft Computing Techniques
SYNOPSIS OF THE PROPOSED RESEARCH PLAN
SUBMITTED TO
CHHATTISGARH SWAMI VIVEKANAND TECHNICAL UNIVERSITY, BHILAI (INDIA)
FOR THE REGISTRATION OF THE TOPIC
FOR
DOCTOR OF PHILOSOPHY IN
THE FACULTY OF
COMPUTER & INFORMATION TECHNOLOGY
By
Pradeep Kumar Mishra
Enrollment no: AB5024
Year 2013
1
BHILAI INSTITUTE OF TECHNOLOGY, DURG
Page 2
Title of Research proposal : Design and Development of Artificial Neural Network Models for Long Range Forecast of Rainfall Runoff.
Name of the Research Scholar : Pradeep Kumar Mishra
Enrollment No. :
Email ID of the Research Scholar : [email protected]
Contact Details of Research Scholar : 27/9 Nehru Nagar (West), Bhilai,Contact No. 09926170794.
Name and Designation of Supervisor -I : Dr. Sanjeev KarmakarAssociate ProfessorDepartment of Computer ApplicationsBhilai Institute of Technology, Durg, Chhattisgarh, INDIA.
Name and Designation of Supervisor -II Dr. Pulak GuhathakurtaScientist‘E’ & DirectorIndia Meteorological Department (IMD),Pune, Sivaji Nagar, Pune, Maharastra, INDIA.
Research Centre : Bhilai Institute of Technology, Durg, Chhattisgarh.
Signature(Dr. SanjeevKarmakar)
Supervisor-I
Signature(Dr. PulakGuhathakurta)
Supervisor-II
Signature(Pradeep Kumar Mishra)
Research Scholar
Forwarded by Chairman DRC
SignatureName
2
Page 3
1. Introduction:
Accurate forecasting of Rainfall-Runoff (R-R)over a river basin through modeling has been
challenging for scientists and engineers of hydrology since decades and centuries. To overcome this
challenge the mathematical modeling and computation may play the significant role. Different
techniques had been employed, with various improvements, to get accurate runoff estimates.
However, the R–R analysis is quite difficult due to presence of complex nonlinear relationships in the
transformation of rainfall into runoff. However runoff analyses are extremely significant for the
prediction of natural calamities like floods and droughts. It also plays a vital role in the design and
operation of various components of water resources projects like barrages, dams, water supply
schemes, etc. Runoff analyses are also needed in water resources planning, development and flood
mitigations. A flood in every year loss of several lives and damage of property and crops of millions
of dollars in only because there was no assessment of runoff forecasting. Various types of modeling
tools had been used to estimate runoff. These techniques consist of lumped conceptual models,
distributed physically based models, stochastic models and black box (time series) models.
Conceptual and physically based models although try to account for all the physical processes
involved in the R–R process, their successful use is limited mainly because of the need of catchment
specific parameters and simplifications involved in the governing equations . On the other hand the
use of time series stochastic models is complicated due to non-stationary behavior and non-linearity
in the data. These models often require experience and expertise of the modelers.
From 1986, Artificial Neural Networks (ANNs) has emerged as a powerful computing system
for highly complex and nonlinear systems. ANNs belongs to the black box time series models and
offers a relatively flexible and quick means of modeling. These models can treat the non-linearity of
system to some extent due to their parallel architecture. However, various architecture of ANNs is
used in non-linear system. It is found that the architecture of ANNs is depending on the problem
space. By the broad literature review, it is found that the back-propagation and fuzzy based neural
system may be highly significant for the simulation of R-R. However, the proper design of their
parameters is rarely visible in the literature. Therefore,purposes of the proposed research work are to:
1. Successful applications of ANN models in the simulation of future R-Rs with high degree of
accuracy.
2. The generalization of ANNsfor R-Rmodeling over the Mahanadi basin
3. And evaluation of ANNs over the existing statistical methods.
Consequently, as per the above purposes the following objectives are proposed in this study:
1. To study in detail of ANN technique and its application especially for identification of
internal dynamics of non-linear dynamic system.
3
Page 4
2. Development of a Back-Propagation Neural Network (BPN) model for R-R modeling.
3. Development of a Fuzzy Neural Network (FNN) model for R-R modeling.
4. Test the BPN and FNN model.
5. To evaluate both the models over existing models.
6. To identify strength and limitations of BPN and FNN model in R-R.
The Mahanadi river basin at the appropriate scale is generally the most logical geographical
unit of stream flow analysis and water resources management. In the present study, Mahanadi River
basin has been selected as study area. The Mahanadi basin encompassed within geographical co-
ordinates of 80030' to 86050' East longitudes and 19020' to 23035' North latitudes as shown in Fig. 1.
The total catchment area of the basin is 1,41,600 km2. The average elevation of the drainage basin is
426 m with a maximum of 877 m and a minimum of 193 m. The river Mahanadi is one of the major
inter-state east flowing rivers in peninsular India. It originates at an elevation of about 442 m. above
Mean Sea Level near Pharsiya village in Raipur district of Chattisgarh. During the course of its
traverse, it drains fairly large areas of Chhatisgarh and Orissa and comparatively small area in the
state of Jharkhand and Maharashtra. The total length of the river from its origin to confluence of the
Bay of Bengal is about 851 km., of which, 357 km. is in Chattisgarh and the balance 494 km. in
Orissa. During its traverse, a number of tributaries join the river on both the banks. There are 14
major tributaries of which 12 are joining upstream of Hirakud reservoir and 2 downstream of it.
Approximately 65% of the basin is upstream from the dam. The average annual discharge is 1,895
m3/s, with a maximum of 6,352 m3/s during the summer monsoon. Minimum discharge is 759 m3/s
and occurs during the months October through June. Mahanadi basin enjoys a tropical monsoon type
of climate like most other parts of the country. The maximum precipitation is usually observed in the
month of July, August and first half of September. Normal annual rainfall of the basin is 1360 mm
(16% CV) of which about 86% i.e. 1170 mm occurs during the monsoon season (15% CV) from June
to September (Rao, 1993). The river passes through tropical zone and is subjected to cyclonic storms
and seasonal rainfall. In the winter the mean daily minimum temperature varies from 4°C to 12°C.
The month of May is the hottest month, in which the mean daily maximum temperature varies from
42°C to 45.5°C.
2. A brief Review of the work already done in the field:
The broad review of literature is done. The very important contributions have been found.
Accurate predictions of floods had been challenging for scientists and engineers since centuries. As
Different techniques had been employed, with various improvements, to get accurate flood estimates
(Bahremand, &Smedt, 2010; Bahremand&Smedt, 2010; Bekele& Knapp,2010; Bhadra, et al., 2010).
4
Page 5
Fig. 1. Location of Study Area (Mahanadi river basin)
The R–R analysis is quite difficult due to presence of complex nonlinear relationships in the
transformation of rainfall into runoff. However runoff analyses are very important for the prediction
of natural calamities like floods and droughts. It also plays a vital role in the design and operation of
various components of water resources projects like barrages, dams, water supply schemes, etc.
Runoff analyses are also needed in water resources planning, development and flood mitigations.
Various types of modeling tools had been used to estimate runoff. According to
Tingsanchali&Gautam, 2000; Bahremand&Smedt, 2010, these techniques consist of:
5
Page 6
1. Conceptual models.
2. Distributed physically based models.
3. Stochastic models and.
4. Black box (time series) models.
Nash & Sutcliffe, 1970, applied conceptual model technique. He suggested the necessity for a
systematic approach to the development and testing of the model is explained and some preliminary
ideas. Vandewiele, & Yu, 1992, studied monthly water balance models in Belgium. China, et
al.,1993, found Model for snowmelt runoff with remote sensing inputs is particularly useful in
Himalayan basins, ground surveys and lack climatologically and hydrological data networks. In this
model variables like temperature, precipitation and snow covered area are considered along with
some externally derived parameters like temperature lapse rate, degree-day factor .
. Franchinia, et al., 1996, developed a model based on R-R modeling for estimation of
exceedance probabilities of design floods and shown it is extended to the estimation of exceedance
probabilities of extreme design floods. Shamseldin, 1997, designed neural network technique to R-R
modelling. it was used different types of input information, namely, rainfall, historical seasonal and
nearest neighbour information. Using the data of six catchments, the technique is applied for four
different input scenarios in each of which some or all of these input types are used. The performance
of the technique is compared with those of models that utilize similar input information, namely, the
simple linear model (SLM), the seasonally based linear perturbation model (LPM) and the nearest
neighbour linear perturbation model (NNLPM). The results suggest that the neural network shows
considerable promise in the context of R-R modelling but, like all such models, has variable results.
JXia, et al. ,1997, focused on A non-linear perturbation model for river flow forecasting is
developed, based on consideration of catchment wetness using an antecedent precipitation index
(API). Catchment seasonality, of the form accounted for in the LPM, and non-linear behaviour both in
the runoff generation mechanism and in the flow routing processes are represented by a constrained
non-linear model, the NLPM-API. A total of ten catchments, across a range of climatic conditions and
catchment area magnitudes. It was found that the NLPM-API model was significantly more efficient
than the original the LPM. However, restriction of explicit non-linearity to the runoff generation
process.
Franchini,et al., 1997, Compared several genetic algorithm schemes for the calibration of
conceptual R-R. The analysis was conducted using an 11-parameter CRRM, called A Distributed
Model (ADM), applied to both a theoretical case without model and data errors and two cases of the
real world in which there are both model and data errors. Finally, assuming the same role as the GA
for the "Pattern Search" (PS) method in a two-step optimization technique (Hendrickson, et al., 1988),
the results of the two algorithms are compared, showing that, in the calibration of the ADM, the PS
may give a slightly superior performance .
6
Page 7
Dibike&Solomatine, 1999, studied River flow forecasting is required to provide basic
information on a wide range of problems related to the design and operation of river systems. The
availability of extended record of rainfall and other climatic data, which could be used to obtain
stream flow data, initiated the practice of R-Rmodeling. While conceptual or physically based models
are of importance in the understanding of hydrologic processes, there are many practical situations
where the main concern is with making accurate predictions at specific locations. In such situation it
is preferred to implement a simple “black box” model to identify a direct mapping between the inputs
and outputs without detailed consideration of the internal structure of the physical process. ANN is
one such technique with flexible mathematical structure which is capable of identifying complex non-
linear relationship between input and output data without attempting to reach understanding in to the
nature of the phenomena. In this study the applicability of ANNs for downstream flow forecasting of
Apure river basin (Venezuela) was investigated. Two types of ANN architectures, namely multi-layer
perceptron network (MLP) and radial basis function network (RBN) were implemented.The
performances of these networks were compared with a conceptual R-R model and they were found to
be slightly better for a particular river flow-forecasting problem.
Sajikumar&Thandavewara, 1999,proposedA non-linear R-R model using an artificial neural
network.in this model had been described. In this model ,A R-R model that can be successfully
calibrated (i.e., yielding sufficiently accurate results) using relatively short lengths of data, is desirable
for any basin in general, and the basins of developing countries like India, in particular, for which
scarcity of data is a major problem. An artificial neural network paradigm, known as the temporal
back propagation neural network (TBP-NN), is successfully demonstrated as a monthly R-R model.
The performance of this model in a “scarce data” scenario (i.e., the effects of using reduced
calibration periods on the performance) is compared with Volterra-type Functional Series Models
(FSM), utilising the data of the River Lee (in the UK) and the Thuthapuzha River (in Kerala, India).
The results confirm the TBP-NN model as being the most efficient of the black-box models tested for
calibration periods as short as six years.
Toth, et al., 2000, have compared of short-term rainfall prediction models for real-time flood
forecasting.This study compares the accuracy of the short-term rainfall forecasts obtained with time-
series analysis techniques, using past rainfall depths as the only input information. The techniques
proposed here are linear stochastic auto-regressive moving average (ARMA) models, artificial neural
networks (ANN) and the non-parametric nearest-neighbors method. The rainfall forecasts obtained
using the considered methods are then routed through a lumped, conceptual, R-R model, thus
implementing a coupled R-R forecasting procedure for a case study on the Apennines mountains,
Italy. The study analyses and compares the relative advantages and limitations of each time-series
analysis technique, used for issuing rainfall forecasts for lead-times varying from 1 to 6 h. The results
also indicate how the considered time-series analysis techniques, andespecially those based on the use
of ANN, provide a significant improvement in the flood forecasting accuracy in comparison to the use
7
Page 8
of simple rainfall prediction approaches of heuristic type, which are often applied in hydrological
practice .
Tingsanchali&Gautam, 2000, applied Two lumped conceptual hydrological models, namely
tank and NAM and a neural network model are applied to flood forecasting in two river basins. The
tank and NAM models were calibrated and verified and found to give similar results. The results
were found to improve significantly by coupling stochastic and deterministic models (tank and
NAM) for updating forecast output. The ANN model was compared with the tank and NAM models.
The NN model does not require knowledge of catchment characteristics and internal hydrological
processes. The training process or calibration is relatively simple and less time consuming compared
with the extensive calibration effort required by the tank and NAM models. The NN model gives
good forecasts based on available rainfall, evaporation and runoff data. The black-box nature of the
NN model and the need for selecting parameters based on trial and error or rule-of-thumb, however,
characterizes its inherent weakness. The performance of the three models was evaluated statistically .
Imrie, et al., 2000, developed a method for improved generalization during training by
adding a guidance system to the cascade-correlation learning architecture. Two case studies from
catchments in the UK are prepared so that the validation data contains values that are greater or less
than any included in the calibration data. The ability of the developed algorithm to generalize on new
data is compared with that of the standard error back propagation algorithm. The ability of ANNs
trained with different output activation functions to extrapolate beyond the calibration data is
assessed.
Abulohom& Shah, 2001, used Runoff modelling by using water balance equations.In this
modelling Statistical results showed that the model preformed well. Thecorrelation co-efficient
between the simulated and measureddata was on the range of 77% to 93%. Sivapragasam and Liong ,
2001, proposed rainfall and runoff forecasting with SSA–SVM approach . In this study, a simple
and efficient prediction technique based on Singular Spectrum Analysis (SSA) coupled with Support
Vector Machine (SVM) is proposed. While SSA decomposes original time series into a set of high
and low frequency components, SVM helps in efficiently dealing with the computational and
generalization performance in a high-dimensional input space. The proposed technique is applied to
predict the Tryggevælde catchment runoff data (Denmark) and the Singapore rainfall data as case
studies. The results are compared with that of the non-linear prediction (NLP) method. The
comparisons show that the proposed technique yields a significantly higher accuracy in the
prediction than that of NLP.
Hundecha, et al., 2001, designed a fuzzy logic-based rainfall-runoff model.R-R models are
used to describe the hydrological behaviour of a river catchment. Many different models exist to
simulate the physical processes of the relationship between precipitation and runoff. Some of them are
based on simple and easy-to-handle concepts, others on highly sophisticated physical and
mathematical approaches that require extreme effort in data input and handling. Recently,
8
Page 9
mathematical methods using linguistic variables, rather than conventional numerical variables applied
extensively in other disciplines, are encroaching in hydrological studies. Among these is the
application of a fuzzy rule-based modelling. In this model an attempt was made to develop fuzzy rule-
based routines to simulate the different processes involved in the generation of runoff from
precipitation. These routines were implemented within a conceptual, modular, and semi-distributed
model—the HBV model. The investigation involved determining which modules of this model could
be replaced by the new approach and the necessary input data were identified. A fuzzy rule-based
routine was then developed for each of the modules selected, and application and validation of the
model was done on aR-R analysis of the River.
There after G´omez-Landesa&Rango, 2002, developed Operational snowmelt runoff
forecasting in the Spanish Pyrenees using the snowmelt runoff model. The snowmelt runoff model
(SRM) was used to simulate and forecast the daily discharge of several basins of the Spanish
Pyrenees. They described a method for snow mapping using NOAA–AVHRR data and a procedure
to estimate retrospectively the accumulated snow water equivalent volume with the SRM. Real-time
snowmelt forecasts were generated with the SRM using area snow cover as an input variable. Even in
basins with a total absence of historical discharge and meteorological data, the SRM provides an
estimation of the daily snowmelt discharge. By integrating the forecasted streamflow over the
recession streamflow, snowmeltvolume is obtained as a function of time. This function converges
asymptotically to the net stored volume of water equivalent of the snowpack. Plotting this integral as
a function of time, it is possible to estimate for each basin both the melted snow water equivalent
(SWE) and the SWE remaining in storage at any point in the snowmelt season Spanish hydropower
companies are using results from the SRM to improve water resource management [18].
Mahabir& Hicks, 2003, proposed study of fuzzy logic to forecast seasonal runoff . In this
study, the applicability of fuzzy logic modelling techniques for forecasting water supply was
investigated. Fuzzy logic has been applied successfully in several fields where the relationship
between cause and effect (variable and results) are vague. Fuzzy variables were used to organize
knowledge that is expressed ‘linguistically’ into a formal analysis. For example, ‘high snowpack’,
‘average snowpack’ and ‘low snowpack’ became variables. By applying fuzzy logic, a water supply
forecast was created that classified potential runoff into three forecast zones: ‘low’, ‘average’ and
‘high’. Spring runoff forecasts from the fuzzy expert systems were found to be considerably more
reliable than the regression models in forecasting the appropriate runoff zone, especially in terms of
identifying low or average runoff years. Based on the modelling results in these two basins, it is
concluded that fuzzy logic has a promising potential for providing reliable water supply forecasts.
Gaume&Gosset, 2003, designed Feed-Forward Artificial Neural Networks (FNN) have been
gaining popularity for stream flow forecasting. However, despite the promising results presented in
recent papers, their use is questionable. In theory, their “universal approximator‿ property guarantees
that, if a sufficient number of neurons is selected, good performance of the models for interpolation
9
Page 10
purposes can be achieved. But the choice of a more complex model does not ensure a better
prediction. Models with many parameters have a high capacity to fit the noise and the particularities
of the calibration dataset, at the cost of diminishing their generalisation capacity. In support of the
principle of model parsimony, a model selection method based on the validation performance of the
models, "traditionally" used in the context of conceptual R-Rmodeling, was adapted to the choice of a
FFN structure. This method was applied to two different case studies: river flow prediction based on
knowledge of upstream flows, and R-Rmodeling. The predictive powers of the neural networks
selected are compared to the results obtained with a linear model and a conceptual model (GR4j). In
both case studies, the method leads to the selection of neural network structures with a limited number
of neurons in the hidden layer (two or three). Moreover, the validation results of the selected FNN and
of the linear model are very close. The conceptual model, specifically dedicated to R-Rmodeling,
appears to outperform the other two approaches. These conclusions, drawn on specific case studies
using a particular evaluation method, add to the debate on the usefulness of Artificial Neural
Networks in hydrology.
Agrawal&Singh, 2004, proposed Runoff modelling by using ANN.in this model Multi layer
back propagation artificial neural network (BPANN) models have been developed to simulate R-R
process for two sub-basins of Narmada river (India) viz. Banjar up to Hridaynagar and Narmada up to
Manot considering three time scales viz. weekly, ten-daily and monthly with variable and uncertain
data sets. The BPANN runoff models were developed using gradient descent optimization technique
and were generalized through cross-validation. In almost all cases, the BPANN developed with the
data having relatively high variability and uncertainty learned in less number of iterations, with high
generalization. Performance of BPANN models is compared with the developed linear transfer
function (LTF) model and was found superior.
Maria et al. 2004, compared ANN and Box & Jenkins techniques and concluded that ANN is
an improvement on Box & Jenkins model. Sohailet al., compared ANN with MARMA (multivariate
auto regressive moving average models in a small watershed of Tono area in Japan during wet and
dry seasons. They concluded that ANN models have shown better results during wet seasons when
the nonlinearity of R–R process is high.
Nayak&Sudheer, et al., 2005, analyzed the skills of fuzzy computing based R-R model in real
time flood forecasting. The potential of fuzzy computing has been demonstrated by developing a
model for forecasting the river flow of Narmada basin in India. This work has demonstrated that
fuzzy models can take advantage of their capability to simulate the unknown relationships between a
set of relevant hydrological data such as rainfall and river flow. Many combinations of input variables
were presented to the model with varying structures as a sensitivity study to verify the conclusions
about the coherence between precipitation, upstream runoff and total watershed runoff. The most
appropriate set of input variables was determined, and the study suggests that the river flow of
Narmada behaves more like an autoregressive process. As the precipitation is weighted only a little by
10
Page 11
the model, the last time-steps of measured runoff are dominating the forecast. Thus a forecast based
on expected rainfall becomes very inaccurate. Although good results for one-step-ahead forecasts are
received, the accuracy deteriorates as the lead time increases. Using the one-step-ahead forecast
model recursively to predict flows at higher lead time, however, produces better results as opposed to
different independent fuzzy models to forecast flows at various lead times.
Corani& Guariso,2005, proposed frameworkfirst classifies the basin saturation state
providing a set offuzzy memberships, and then issues the forecast exploiting a setof neural predictors,
each specialized on certain basin saturationcondition by means of a weighted least-square training
algorithm.The outputs of the specialized neural predictors are linearlyweighted, according to the basin
state at forecast time: The morethe training conditions of a predictor matches the current
basinsaturation state, the higher its weight on the final forecast. Theframework has been tested on an
Italian catchment and mayoverperform classical neural networks approaches.
Valença, et al., 2005, used a Constructive Neural Networks model (NSRBN) were used to
forecast daily river flows for the Boa Esperança Hydroelectric power plant, part of the Chesf
(CompanhiaHidrelétrica do São Francisco) system. This dam is located at Parnaíba River, in
theborderline between Maranhão and Piauí, two Brazilian States. Several studies have been dedicated
to the prediction of river flows with no exogenous inputs that are with the only use of past flow
observations. In the present work, Constructive Neural Networks are first used withoutexogenous
input that is without the use of rainfall observations. Only the last measured discharges areprovided as
input to the networks, analyzing the performance of the forecasts provided for the validation sets over
the varying lead-times. In the second type of application, the same optimal number of past discharges
is given as input to the ANN, along with exogenous inputs,that is past rainfall values, thus testing a
rainfall-runoff modeling approach. The NSRBN model approach is shown to provide better
representation of the daily average water inflow forecasting, than the models based on Box-Jenkins
method, currently in use on the Brazilian Electrical Sector.
Knebl,et al.,2005, proposed a model which was consists of aR-R model (HEC-HMS) that
converts precipitation excess to overland flow and channel runoff, as well as a hydraulic model
(HEC-RAS) that models unsteady state flow through the river channel network based on the HEC-
HMS-derived hydrographs. HEC-HMS is run on a 4!4 km grid in the domain, a resolution consistent
with the resolution of NEXRAD rainfall taken from the local river authority. Watershed parameters
are calibrated manually to produce a good simulation of discharge at 12 subbasins. With the
calibrated discharge, HEC-RAS is capable of producing floodplain polygons that are comparable to
the satellite imagery. The modeling framework presented in this study incorporates a portion of the
recently developed GIStool named Map to Map that has been created on a local scale and extends it to
a regional scale. The results of this research will benefit future modeling efforts by providing a tool
for hydrological forecasts of flooding on a regional scale. While designed for the San Antonio River
11
Page 12
Basin, this regional scale model may be used as a prototype for model applications in other areas of
the country.
Liu,et al.,2006, studied and applied three forecasting modelsto predict the ten-day
streamflow resulting from rainfallson the Kao-Ping river watershed. The three techniquesemployed in
establishing the models include: time series,Grey system theory, and the adaptive fuzzy neural
networkmodel. Through the research it is anticipated a moreappropriate hydrologic data series
forecasting model forthe Kao-Ping river watershed can be identified.
Tayfur,et al., 2006,presented presents the development of artificial neural network (ANN)
and fuzzy logic (FL) models for predicting event based R-R and tests these models against the
kinematic wave approximation (KWA). A three-layer feed-forward ANN was developed using the
sigmoid function and the back propagation algorithm. The FL model was developed employing the
triangular fuzzymembership functions for the input and output variables. The fuzzy rules were
inferred from the measured data. The measured event based R-R peak discharge data from laboratory
flume and experimental plots were satisfactorily predicted by the ANN, FL, and KWAmodels.
Similarly, all the three models satisfactorily simulated event-based rainfall-runoff hydrographs from
experimental plots withcomparable error measures. ANN and FL models also satisfactorily simulated
a measured hydrograph from a small watershed 8.44 km2in area. The results provide insights into the
adequacy of ANN and FL methods as well as their competitiveness against the KWA forsimulating
event-based R-R processes.
Li, et al.,2006, Proposed an intelligent forecasting method for medium-and-long term runoff
forecast , Based on the fuzzy optimum theory, neural network and genetic algorithm. Firstly, a fuzzy
optimum model is integrated with BP neural network to construct a new fuzzy neural network
describing the complicated relations between forecast factors and runoff. The network may fall into
local minimum during the training process. To overcome the shortcoming and improve training
efficiency, an improved genetic algorithm, RAGA, is introduced to optimize the network weights.
Finally, a case proves that the intelligent forecast methodology is efficient and has accuracy
forecasting results.
Lohani, et al., 2007, developed a technique. This technique had been applied to two gauging
sites in the Narmada basin inIndia. Performance of the conventional sediment rating curves, neural
networks and fuzzy rule-based models was evaluated using the coefficient of correlation, root mean
square error and pooled average relative (underestimation and overestimation) errors (PARE) of
sediment concentration. Comparison of results showed that the fuzzy rule-based model could be
successfully applied for sediment concentration prediction as it significantly improves the magnitude
of prediction accuracy.
Cheng, et al., 2007developed a new prior density and likelihood function model with BP
artificial neural network (ANN) to study the hydrologic uncertainty of short-term reservoir stage
forecasts based on the BFS framework. Markov chain Monte Carlo (MCMC) method is employed to
12
Page 13
solve the posterior distribution and statistics of reservoir stage. The results show that Bayesian
probabilistic forecasting model based on BP ANN not only increases forecasting precision greatly but
also offers more information for flood control, which makes it possible for decision makers consider
the uncertainty of hydrologic forecasting during decision-making and estimate risks of different
decisions quantitatively.
Jiang, et al.,2007, studied the classical algorithm of BP network model, itsconvergence rate is
slow and it may result in locallyoptimal solution. But on the condition of same arithmeticcomplicacy,
the Fletcher-Reeves algorithm can improvethe convergence rate and come to the least point along
theconjugate direction so as to improve the forecastingprecision of the BP network model. According
to thecheck results of the BP network model in Guanyingereservoir, it is proved that this model can
fulfill therequirement of forecasting precision and is valuable to beused for reference or be
generalized in real-time forecastof afflux runoff in other area under the same condition.
Jingbo, et al., 2008, recognized the problem of runoff forecasting is researched for water
supply reservoir group based on Phase Space ReconstructionTheory. The statistic method of BDS is
applied to prove its non-linearity and the largest Lyapunov exponent is computed, which manifests
thatthere is chaotic characteristics in the runoff sequence of reservoir group. Single-dimensional and
multi-dimensional runoff forecast models arebuilt and analyzed based on State Space Reconstruction
Theory, Artificial Neural Network and Genetic Algorithm. Their performances inpractice are
compared and analyzed, which manifests its validity and a broad prospect.
Aytek, et al., 2008, proposed an application of two techniques of artificial intelligence (AI)
for rainfall–runoff modeling: the artificial neural networks (ANN) and the evolutionary computation
(EC). Two differentANN techniques, the feed forward back propagation (FFBP) and generalized
regression neural network (GRNN) methods are compared with one EC method, Gene Expression
Programming (GEP) which is a new evolutionary algorithm that evolves computer programs. The
daily hydrometeorological data of three rainfall stations and one streamflow station for Juniata River
Basin in Pennsylvania State of USA are taken into consideration in the model development. Statistical
parameters such as average, standard deviation, coefficient of variation, skewness, minimum and
maximum values, as well as criteria such as mean square error (MSE) and determination coefficient
(R2) are used to measure the performance of the models. The results indicate that the proposed
genetic programming (GP) formulation performs quite well compared to results obtained by ANNs
and is quite practical for use. It is concluded from the results that GEP can be proposed as an
alternative to ANN models.
Li &Yuan, 2008, presented a runoff forecasting method basedon data mining. A runoff
forecasting model is built up bythe data mining tool ANN. Data mining is carried on byBP model and
the convergence speed is improved by themodification of the weight coefficients. The model istested
in a real project and compared with generalmodels. The test result and analysis illustrate its
goodprecision of forecasting and good value in theapplication.
13
Page 14
Vandeeiele& Win, 2009, had provided two types of monthly water balance models at basin
scale are used: PE models use precipitation and potential evapotranspiration (PET) as their observed
input data, whereas P models need only precipitation. Calibration proceeds by comparing model
runoff and observed runoff. Calibration is entirely automatic with the exclusion of subjective
elements. All models differ only by their actual evapotranspiration equations. PE models from
previous papers are generalized essentially by replacing the constant evapotranspiration parameter by
a periodic one, thus increasing the number of parameters by two (a “parameter” is an unknown
constant to be estimated, and which is a characteristic of the river basin to be described). P models use
a periodic “driving force”, which is intended to represent periodicity of hydrological phenomena,
normally originating in the (unavailable) PET time series. These eight PE models and three P models
are then applied to 55 river basins in 10 countries with widely diverging climates and soil conditions.
A marked improvement of model performance in about one third of the basins is due to the
introduction of the above mentioned periodic functions. Even when PET data are available it is
sometimes useful to consider P models. P models scarcely perform less well than PE models. An
engineer, wanting to try out as few models as possible on a given river basin, can restrict his attention
to the optimization of two or three models. The paper is an extension of a long effort towards monthly
water balance models, and is believed to give a solution in most circumstances.
Zhang, et al.,2009, applied to the annual runoffdata of the Baishan and Fengman hydrology
Stations; then theerror correction model is set up, which can predict the annualrunoff of Fengman
hydrology Station from 1989 to 1998. Theresults show that the model based on cointegration analysis
anderror correction is suitable in runoff forecasting.
Hung, et al., 2009, designed a new approach using an ANN technique to improve rainfall
forecast performance. The developed ANN model is being applied for real time rainfall forecasting
and flood managementforecasts by ANN model were compared to the convenient approach namely
simple persistent method. Results show that ANN forecasts have superiority over the ones obtained
by the persistent model.
Ren&Hao,2009, described a novel method to mid-longterm runoff prediction using moving
windows autoregressivequadratic model which combines autoregressive quadraticmodel and moving
windows method to improve predictioncapability of natural runoff. The parameters of the modelare
determined in light of the joints of half-sin function, selfadaptiveoptimization, smoothly moving
windows andgeneralized likelihood uncertainty estimation. Theapplication shows that the model can
not only improveprediction capability but keep robust, and shows that themodel has simpler structure
and less parameter thanartificial neural networks model, and avoids locally minimalpoint and excess
study, etc. Therefore, the moving windowsautoregressive quadratic model is a promising tool for
midlongterm runoff forecast.
Pradhan,et al., 2010, used remote sensing and GIS technology can be used to overcome the
problem of conventional method for estimating runoff caused due to rainfall. In this paper, modified
14
Page 15
Soil conservation System (SCS) CN model is used for rainfall runoff estimation that considers
parameter like slope, vegetation cover, area of watershed.
Güldal&Tongal, 2010, Compared of Recurrent Neural Network, Adaptive Neuro-Fuzzy
Inference System and Stochastic Models in Eğirdir Lake Level Forecasting.The performances of the
models are examined with the form of numerical and graphical comparisons in addition to some
statistic efficiency criteria. The results indicated that the RNN and ANFIS can be applied successfully
and provide high accuracy and reliability for lake-level changes than the AR and the ARMA models.
Also it was shown that these stochastic models can be used in the lake management policies with the
acceptable risk.
Xu, et ai.,2010, successfully appliedSupport Vector Machine (SVM) based rainfallrunoff
models to daily runoffmodeling in many basins. Most of them are however designed forsmall or
meso-scale basins rather than large-scale basins. One ofaims in the present work is therefore to
develop an SVM modelwith an optimized combination of input variables for dailystream flow
simulating. Another aim is to compare theperformance of SVM models with two different process-
basedhydrological models, namely TOPMODE and Xinanjiangmodel,in one day ahead stream flow
forecasting. Yingluoxia basin, witha drainage area of 10009 km2, is selected for testing them.
Theresults show that the precipitation, evaporation and antecedentobserved stream flow, are all
necessary as inputs to SVMmodeling for this basin. The optimized SVM model performsmuch better
than TOPMODE and Xinanjiang model both forcalibration period and the validation period in terms
of NashSutcliffeefficiency. The daily stream flows simulated by the SVMare in very good agreement
with the observed ones, while thosesimulated by Xinanjiang and TOPMODEL
significantlyunderestimate or overestimate the main peak-flows and aregreatly different from the
observed ones for low flow stages inboth calibration stage and validation period. SVM models
arepromising tools for short term daily runoff forecasting even if ina large-scale basin.
Wu&Chau, 2010, designed Accurately modeling of R-R transform (Wu and Chau, 2010)
remains a challenging task despite that a wide range of modeling techniques, either knowledge-driven
or data-driven, havebeen developed in the past several decades. Amongst data-driven models, ANN-
based R-R models have received great attentions in hydrologycommunity owing to their capability to
reproduce the highly nonlinear nature of therelationship between hydrological variables. However, a
lagged prediction effectoftenappears in the ANN modeling process.
Deshmukh&Ghatol,2010, applied The artificial neural networks (ANNs), to various
hydrologic problems recently. This researchdemonstrates a temporal approach by applying Jordan
andElman network for R-Rmodelling for the upper areaof Wardha River in India. The model is
developed byprocessing online data over time using recurrent connections.Methodologies and
techniques of the two models are presentedin this paper and a comparison of the short term runoff
prediction results between them is also conducted. Theprediction results of the Jordan network
indicate a satisfactoryperformance in the three hours ahead of time prediction. Theconclusions also
15
Page 16
indicate that the Jordan network is moreversatile than Elman model and can be considered as
analternate and practical tool for predicting short term floodflow.
Ma, et al.,2011, studied a combined model of chaos theory, wavelet and support vector
machine was built to overcome the limitations including challenges in determination of orders of
nonlinear models and low prediction accuracy which the simulated accuracy is high in runoff series
forecasting. Firstly, runoff series were decomposed into different frequency runoff components in
application of wavelet. Secondly, phase space was reconstructed in chaotic analysis. Thirdly, support
vector machine (SVM) was used to predict each component. Finally, all components were combined
into a model to predict runoff. In this study, annual and monthly runoff of two reservoirs located in
the Sha River and Li River of the Shaying River system within the Haihe River watershed were used
to examine the combined model. The results indicated that the simulated accuracy and predicted
accuracy were grade A and grade B, which met the requirements of the medium term accuracy and
long term accuracy and the combined model is applicable to medium term and long term prediction.
Hu, et al.,2011, developed a novel artificial intelligence-based method from statistical
learning theory, is adopted herein to establish R-R relationships model. The lags associated with the
input variables are determined by applying the hydrological concept of the response time, and a trial-
and-error with cross-validation was used to derive the support vector machine (SVM) model
parameters. The purpose of this study is to develop a parsimonious model used little observation gage
that accurately simulates semi-arid regions by using the SVM model. The R-R relations weretreated
as a non-linear input/output system to simulate the response of runoff to precipitation and applied the
model to the upstream of the Fenhe River, the branch of the Yellow River (China), a representative of
watershed in a semiarid area. The precipitation-runoff relationships on these regions were studied by
using SVM model. Moreover, the SVM model was compared with a previous Artifical neural
networks (ANN) model and it was found that the SVM model performed better. Results obtained
showed that runoff forecasts of daily time step were better in non-flood season than those made in
flood season and monthly runoff forecasts. It suggests that the SVM model and thedeveloped method
proposed are convenient and practical for semi-arid regions.
Brocca, et al.,2012,performed tworeal data and two synthetic experiments have been carried
outto assess the effects of assimilating soil moisture estimates into atwo-layer R-R model. By using
the ensemble Kalmanfilter, both the surface- and root-zone soil moisture (RZSM) products
derived by the Advanced SCATterometer (ASCAT) have been assimilated and the model
performance on flood estimation is analyzed. RZSM estimates are obtained through the application of
an exponential filter. Hourly rainfall–runoff observations for the period 1994–2010 collected in the
Niccone catchment (137 km2), Central Italy, are employed as case study. The ASCAT soil moisture
products are found to be in good agreement with the modelled soil moisture data for both the surface
layer (correlation coefficient (R) of 0.78) and the root zone (R = 0.94). In the real data experiment, the
assimilation of the RZSM product has a significantimpact on runoff simulation that provides a clear
16
Page 17
improvement in the discharge modeling performance. On the other hand, the assimilation of the
surface soil moisture product has a small effect. The same findings are also confirmed by the
synthetic twin experiments. Even though the obtained results are model dependent and site specific,
the possibility to efficiently employ coarse resolution satellite soil moisture products for improving
flood prediction is proven, mainly if RZSM data are assimilated into the hydrological model.
In recently, Mittal&Chowdhury, 2012proposed to develop a dual (combined and paral-leled)
artificial neural network (D-ANN), which aims to improve the models performance, especially in
terms of ex-treme values. The performance of the proposed dual-ANN model is compared with that of
feed forward ANN (FF-ANN) model, the later being the most common ANN model used in
hydrologic literature. The forecasting exercise is carried out for hourly river flow data of Kolar Basin,
India. The results of the comparison indicate that the D-ANN model per-forms better than the FF-
ANN model.
Patil,et al., 2012, neural network, fuzzy logic and genetic algorithms has become very
popular. It not only useful in IT sector, but also very useful in predicting or forecasting something
according to past information R-R modeling is very important and challenging area of research. The
issue becomes more crucial and difficult as population grows in particular city. The semiarid area of
western Maharashtra province is a important grain production base in India, in the area the nature
characteristics is small quantity and concentrate distribution in rainfall, and agriculture development
was restricted by drought and soil and water loss seriously. Surface runoff not only leads to rainfall
use efficiency decrease, it is also the important factor which causes soil erosion. The objective of this
paper is to review the different forecasting algorithm algorithms of R-R modeling. This paper find out
pros and cons of these algorithms and suggest framework of new algorithm for R-R modeling which
gives better water consumption.
Bell,et al.,2012, Described a Support Vector Machine based method for river runoff
forecasting. This method uses Smola/Scholkopf’s Sequential Minimal Optimization algorithm for
training a SupportVector Machine with a RBF kernel. The experimental results on predicting the full
natural flow of the American River at the Folsom Dam measurement station in California indicates
that, this method outperforms the current forecasting practices.
Chen, et al.,2013, a model for estimating runoff by using rainfall data from a river basin is
developed and a neural network technique is employed to recover missing data. For achieving the
objectives, hourly rainfall and flow data from Nanhe, Taiwu, and Laii rainfall stations and Sinpi flow
station in the Linbien basin are used. The data records were of 27 typhoons between the years 2005
and 2009. The feed forward back propagation network (FFBP) and conventional regression analysis
(CRA) were employed to study their performances. From the statistical evaluation, it has been found
that the performance of FFBP exceeded that of regression analysis as reflected by the determination
coefficients R2, which were 0.969 and 0.284 for FFBP and CRA, respectively.
17
Page 18
Gebregiorgis&Hossain, 2013,studied to characterize satellite rainfall errors and their impact
onhydrologic fluxes based on fundamental governing factors thatdictate the accuracy of passive
remote sensing of precipitation.These governing factors are land features—comprising
topography(elevation)—and climate type, representing the average ambientatmospheric conditions.
First, the study examines satelliterainfall errors of three major products, 3B42RT, Climate
predictioncenter MORHing technique (CMORPH), and PrecipitationEstimation from Remotely
Sensed Information using ArtificialNeural Networks (PERSIANN), by breaking the errors down
intoindependent components (hit, miss-rain, and false-rain biases) andinvestigating their contribution
to runoff and soil moisture errors.The uncertainties of three satellite rainfall products are exploredfor
five regions of the Mississippi River basin that are categorizedgrid cell by grid cell (at the native
spatial resolution of satelliteproducts) based on topography and climate. It is found that totaland hit
biases dictate the temporal trend of soil moisture andrunoff errors, respectively. Miss-rain and hit
biases are the leadingerrors in the 3B42RT and CMORPH products, respectively,whereas false-rain
bias is a pervasive problem of the PERSIANNproduct. For 3B42RT and CMORPH, about 50%–60%
of gridcells are influenced by the total bias during winter and 60%–70%of grid cells during summer.
For PERSIANN, about 70%–80% ofthe grid cells are marked by total bias during the summer
andwinter seasons. False-rain bias gradually increases from lowlandto highland regions universally
for all three satellite rainfall products.Overall, the study reveals that satellite rainfall uncertaintyis
dependent more on topography than the climate of the region.This study’s results indicate that it is
now worthwhile to assimilatethe static knowledge of topography in the satellite estimation
ofprecipitation to minimize the uncertainty in anticipation of theGlobal Precipitation Measurement
mission.
Recently, Robertson &Pokhrel, 2013,Improving statistical forecasts of seasonal stream flows
using hydrological model output. Statistical methods traditionally applied for seasonal stream flow
forecasting use predictors that represent the initial catchment condition and future climate influences
on future stream flows. Observations of antecedent stream flows or rainfall commonly used to
represent the initial catchment conditions are surrogates for the true source of predictability and can
potentially have limitations. This study investigates a hybrid seasonal forecasting system that uses the
simulations from a dynamic hydrological model as a predictor to represent the initial catchment
condition in a statistical seasonal forecasting method. We compare the skill and reliability of forecasts
made using the hybrid forecasting approach to those made using the existing operational practice of
the Australian Bureau of Meteorology for 21 catchments in eastern Australia. We investigate the
reasons for differences. In general, the hybrid forecasting system produces forecasts that are more
skilful than the existing operational practice and as reliable. The greatest increases in forecast skill
tend to be when the catchment is wetting up but antecedent stream flows have not responded to
antecedent rainfall, when the catchment is drying and the dominant source of antecedent stream flow
18
Page 19
is in transition between surface runoff and base flow, and when the initial catchment condition is near
saturation intermittently throughout the historical record.
These Conceptual and physically based models although try to account for all the physical
processes involved in the R–R process, their successful use is limited mainly because of the need of
catchment specific parameters and simplifications involved in the governing equations . On the other
hand the use of time series stochastic models (i.e., based on probability) is complicated due to non-
stationary behavior and nonlinearity in the data. These models often require experience and expertise
of the modeler .
Approximately, from the last two decades ANNs has emerged as a powerful computing
system for highly complex and nonlinear systems. ANN belongs to the black box time series models
and offers a relatively flexible and quick means of modeling. These models can treat the nonlinearity
of system to some extent due to their parallel architecture. A few studies reported poor performances
of ANN models in comparison with the conventional ones. For example compared feed forward
ANNs with a linear model and a conceptual model. They concluded that their conceptual model
outclassed the linear and ANN models. Some other studies show the successful applications of ANN
models in the simulation of future runoffs with high degree of accuracy.
From the above discussion it is evident to state that ANN models provide better predictions as
compared to the conventional models, however their application is as yet limited with the research
environment.
3. Noteworthy Contribution in the field of proposed work:
The R-R models are highly useful for water resources planning and development. The
rainfall–runoff model based on ANNs was developed and applied on a watershed in Pakistan by
Ghumman et al., 2012. The model was developed to suite the conditions in which the collected
dataset is short and the quality of dataset is questionable. The results of ANN models were compared
with a mathematical conceptual model. The cross validation approach was adopted for the
generalization of ANN models. The precipitation used data was collected from Meteorological
Department Karachi Pakistan. The results confirmed that ANN model is an important alternative to
conceptual models and it can be used when the range of collected dataset is short and data is of low
standard. Phuphong and Surussavadee2013, did a case study Khlong U-Tapao River Basin, Songkhla
Province, Thailand and applied ANN technology for Runoff Forecasting. He found that, good forecast
accuracy. Correlation coefficients between forecasted and observed water levels for Ban Takienphao
station are higher than 0.92 and rms errors are within 1.92% of the annual mean water level.
Correlation coefficients for Ban Muangkong station are higher than 0.86 and rms errors are within
6.67% of the annual mean water level.
It is concluded that, the ANN technology is sufficiently suitable in forecasting of runoff
however, proper design and training is remain tricky till date as well. No contributes have found in the
19
Page 20
literature have provided the accurate architecture of ANNs in this proposed area of research. Some
contributes also used fuzzy logic in their research and obtained good performance. The noteworthy
contributions of the proposed work are to
1. Provide appropriate design and development of BPN and FNN through identification of
their parameters and their training up to the level of global minima to overcome such a
great hydrological problem (i.e., simulation of R-R).
2. And the generalization of BPN and FNN for R-R modeling over Mahanadi basin.
3. Skill of BPN and FNN for R-R modeling.
4. Proposed Methodology during the tenure of research work:
As per the objectives of the research work the methodology is given in three phases are as
follows:
Phase-I
Step 1. Collection of meteorological data for Mahanadi basin, India.
Step 2. Pre-processing of data and splitting it for development and testing process.
Step 3. Identification of Input data (parameters).
Step 4. Design of BPN Model.
Step 5. Design of FNN Model.
Phase-II
Step 6. Development of BPN by using MATLAB R2010 (ANN toolbox)/Java.
Step 7. Development of FNN by using MATLAB R2010 (ANN toolbox)/Java.
Step 8. Apply the BPN and FNN for R-R for Mahanadi basin.
Phase-II
Step 9. Test the BPN and FNN.
Step 10. Evaluate the performance of BPN, FNN over the existing models.
Step 11. Finding the skill of BPN and FNN.
Step 12. Find the limitations of BPN, FNN in R-R forecasting and future work.
Step 13. Results & Discussions.
Step 14. Conclusions.
5. Expected outcome of the proposed work:
It is found that, the ANN models is sufficiently suitable to identify internal dynamics of high
dynamic system. It is extremely useful to obtain above described objectives of the research. As per
the methodology, the expected outcomes are as follows:
20
Page 21
Technical outcomes:
1. Design constraints of BPN and FNN parameters.
2. Application of BPN and FNN for R-R modeling over Mahanadi basin, India.
3. A MATLAB or Java based simulator of BPN and FNN.
4. Comparison results of BPN, FNN over the existing models.
5. Skill of BPN and FNN for R-R modeling.
6. Limitations of BPN and FNN for this specific application.
Academic outcomes:
1. Publications in National/International Journals/Conferences.
2. A book.
3. A copyright.
21
Page 22
6. Bibliography
Abulohom,M.S.,Shah,S.M.S.,&Ghumman,A.R.,(2001), Development of A Rainfall–Runoff Model,
its Calibration and Validation,Journal of Water Resources Management, 15, 149–163.
Agarwal,A., &Singh,R.D.,(2004), Runoff ModelingTthrough Back Propagation Artificial Neural
Network with Variable Rainfall–Runoff Data, Water Resources Management,18,285–300.
Alves,O., Wang,G., Zhong,A., Smith,N., Tzeitkin, F., Warren, G., Schiller, A., Godfrey, S.,
&Meyers,G.(2002), POAMA: Bureau of Meteorology Operational Coupled Model Seasonal
Forecast System, National Drought Forum.
Archer,D.R., &Fowler,H.J.,(2008), Using meteorological data to forecast seasonal runoff on the River
Jhelum, Pakistan, Journal of Hydrology, 361, 10-23.
Ashok,K.,Guan,Z.Y.,&Yamagata,T.,(2003), Influence of the Indian Ocean Dipole on the Australian
winter rainfall, Geophysical Research Letters,30(15),1-4.
Aytek,A., Asce,M., &Alp,M.,(2008), An application of artificial intelligence for rainfall–runoff
modeling, J. Earth Syst. Sci., 117(2),145-155.
Bach,H., Lampart,G., Strasser.G., &Mauser,W., (1999), First Results of an IntegratedFlood Forecast
System Based on Remote Sensing Data,IEEE Trans., 6(99), 864-866.
Bahremand,A., &Smedt,F.D. (2010), Predictive analysis and simulation uncertainty of a distributed
hydrological model, Water Resources Management, 24 (13), 2869–2880.
Baumgartner,M.F., Seidel,K., &Martinec,J., (1987), Toward Snowmelt Runoff Forecast Based on
Multisensor Remote-Sensing Information, IEEE Trans., 25(6), 746-750.
Bekele,E.G., &Knapp,H.V. (2010), Watershed modeling to assessing impacts of potential climate
change on water supply availability,Water Resources Management, 24 (13),3299–3320.
Bell,B.,Wallace,B., &Zhang,D.,(2012), Forecasting River Runoff through Support Vector
Machines,IEEE(Int. Conf. on Cognitive Informatics & Cognitive Computing), 58-64
Bhadra,A.,Bandyopadhyay,A.,Singh.,&Raghuwanshi,R.N.S.,(2010),Rainfall–Runoff Modeling:
Comparison of Two Wpproaches with Different Data Requirements, Journal of Water Resource
Management, (24),37–62.
Bierkens, M. F. P., &Beek, L. P. H.,(2009), Seasonal Predictability of European Discharge: NAO and
Hydrological Response Time, J. Hydrometeorol, 10, 953–968.
Box,G.E.,&Jenkins,G.M., (1976),Time Series Analysis: Forecasting and Control, Prentice Hall, San
Francisco, California.
Broersen,P.M.T.,(2007), Error Correction of Rainfall-Runoff Models With the ARMAsel Program,
IEEE Trans., 56(6), 2212-2219.
Cheng,C.T.,Chau,C.W.,&Li,X.Y.,(2007), Hydrologic Uncertainty for Bayesian Probabilistic
Forecasting Model Based on BP ANN, IEEE(Third International Conference on Natural
22
Page 23
Computation).
Cannas,B.,Fanni,A.,Pintusb,M., &Sechib,G.M.,(2002), Neural network models to forecast
hydrological risk, IEEE, 623-626.
Chang,F.J., Liang,J.M, &Chen,Y.C.,(2001), Flood Forecasting Using Radial Basis Function Neural
Networks, IEEE Trans., 31(4), 530-535.
Chang,N.B., &Guo,D.H.,(2006), Urban Flash Flood Monitoring, Mapping, and Forecasting via a
Tailored Sensor Network System, IEEE, 757-761.
Corani,G., &Guariso,G.,(2005), Coupling Fuzzy Modeling and Neural Networks for River Flood
Prediction, IEEE Trans., 35(3), 382-390.
Franchini,M., &Galeati,G., (1997), Comparing several genetic algorithm schemes for the calibration
of conceptual rainfall-runoffmodels, Bydrological Sciences-Journal-des Sciences
Hydrologiques, 42(3), 357-379.
Franchinia,M., Helmlinger,T.K.R., Foufoula-Georgioub.E., &Todini,E., (1996), Stochastic storm
transposition coupled with rainfall-runoff modeling for estimation of exceedance probabilities
of design floods, Journal of Hydrology, 175,511-532.
Gebregiorgis,A.S., &Hossain,F.,(2013), Understanding the Dependence of Satellite Rainfall
Uncertainty on Topography and Climate for Hydrologic Model Simulation,IEEE Trans.,
51(1), 704-718.
Gaume,E., &Gosset,R.,(2003), Over-parameterisation, a major obstacle to the use of artificial neural
networks in hydrology, Hydrol. Earth Syst. Sci., 7, 693-706.
Ghedira,H., Arevalo,J.C., Lakhankar,T., Azar.A., Khanbilvardi,R., &Romanov.P.,(2006), The Effect
of Vegetation Cover on Snow Cover Mapping from Passive Microwave Data, IEEE, 148-153.
Ghumman,A.R., Ghazaw,Y.M., Sohail.A.R., &Watanabe,K.,( 2012),Runoff forecasting by artificial
neural network and conventional model.,Alexandria Engineering Journal, 50(1),345 – 350.
Gmez-Landesa,E., &Rango,A.,(2000), Snow Mapping Technilque at Subpixel Level for Small
Basins, IEEE Trans., 3,1140-1142.
Guo,H., Dong,G.Z., &Chen,X.,(2008), WANN Model for Monthly Runoff Forecast , IEEE, 1087-
1089.
Guo,J., Xiong,W., &Chen,H.,(2009), Application of Rough Set Theory to Multi-factor Medium and
Long-period Runoff Prediction in Danjing Kou Reservoir,IEEE(Sixth
InternationalConference on Fuzzy Systems and Knowledge Discovery),177-182.
Guo,J.C.Y., &Urbonas.B.,(2000), Synthetic running capture and delivery curves for storm water
quality control designs, ASCE J. of Water Resources Planning and Management, 128(3), 1-
13.
G´omez-Landesa,E., &Rango,A.,(2002), Operational snowmelt runoff forecasting in the Spanish
23
Page 24
Pyrenees using the snowmelt runoff model, Hydrological Processes, 16, 1583–1591.
Ji,L., &Bende,W.,(2007), Parameters Selection for SVR based on the SCEM-UA Algorithmand Its
Application on Monthly Runoff Prediction, IEEE(International Conference on
Computational Intelligence and Security), 48-51.
Jiang,G., Shen,B., &Li.Y.,(2007), On the Application of Improved Back Propagation NeuralNetwork
in Real-Time Forecast. IEEE(Third International Conference on Natural Computation).
Jingbo,L., Zengchuan,D., Dezhi,W., &Shaohua.L.,(2008), Research on Runoff Forecast Model Based
on Phase Space Reconstruction, IEEE(7th World Congress on Intelligent Control and
Automation), 5339-5343.
Hossain,F., Anagnostou,E.N., &Dinku.T.,(2004),Sensitivity Analyses of Satellite Rainfall Retrieval
and Sampling Error on Flood Prediction Uncertainty, IEEE Trans., 42(1), 130-139.
Imrie,C.E.,&Durucan,S.,(2000), River flow prediction using artificial neural networks:
generalisationbeyond the calibration range, Hydrological Processes, 233(4), 138-153.
Ma,X., Ping,J.,Yang,L.,Yan,M., &Mu,H., (2011), Combined Model of Chaos Theory, Wavelet and
Support Vector Machine for Forecasting Runoff Series and its Application,IEEE (Natural
Science Foundation of Henan Province),842-845.
Mahabir,C,Hicks,F.E., &Fayek,A.R.,(2003), Application of Fuzzy Logic to Forecast Seasonal
Runoff,Hydrological Process,17,3749–3762.
Maria,C.M.,Wenceslao,G.M.,Manuel,G.M.,Jose.M.P.S.,&Roman,L.C. (2004),Modelling of the
Monthly and Daily Behaviour of the Runoff of the Xallas River Using Box-Jenkins and
Neural Networks Methods, Journal of Hydrology, 296, 38–58.
Mittal,P.,Chowdhury,S.,Roy,S.,Bhatia,N.,&Srivastav,R.,(2012), Dual Artificial Neural Network for
Rainfall-Runoff Forecasting,Journal of Water Resource and Protection,4,1024-1028.
Nash,J.E., &Sutcliffe,J.V.,(1970), River Flow Forecasting Through Conceptual Models, Part1 – A
Discussion of Principles, Journal of Hydrology, 10 , 282–290.
Nayak,P.C.,Sudheer,K.P,&Ramasastri,K.S.(2005),Fuzzy Computing Based Rainfall–Runoff Model
for Real Time Flood Forecasting, Hydrological Process,19, 955–968.
Patil,S.,Patil,S.,&Valunjkar,S.(2012),Study of Different Rainfall-Runoff Forecasting Algorithms for
Better Water Consumption,International Conference on Computational Techniques and
Artificial Intelligence,International Conference on Computational Techniques and Artificial
Intelligence,327-330.
Phuphong, S. &Surussavadee,C.,(2013), An Artificial Neural Network Based Runoff Forecasting
Model in the Absence ofPrecipitation Data: A Case Study of Khlong U-Tapao River Basin,
SongkhlaProvince, Thailand, IEEE(4th International Conference on Intelligent Systems,
Modelling and Simulation), 73-77.
Pradhan,R.,Pradhan,M.P.,Ghose.M.K.,Vivek,S.,Agarwal,V.S..&Agarwal,S.(2010),Estimation of
24
Page 25
RainfallRunoff using Remote Sensing and GIS in and Around Singtam, East
Sikkim,International Journal of Geomatics and Geosciences,1(3),466 – 476.
Ren,Z., &Hao,Z.C.,(2009), Application of Moving Windows Autoregressive Quadratic Model in
Runoff Forecast, IEEE(International Conference on Industrial Mechatronics and
Automation), 200-203.
Robertson,D.E., Pokhrel,P., &Wang,Q.J.( 2013),Improving Statistical Forecasts of Seasonal
Streamflows Using Hydrological Model Output, Hydrology and Earth System
Sciences,17,579–593.
Sajikumar,N., &Thandaveswara,B.S.A.,(1999), non-linear rainfall– runoff model using an artificial
neural network, Journal of Hydrology, 216,32–55.
Yonas, B. Dibike,&Solomatine,D.P.,(1999),River Flow Forecasting Using Artificial Neural
Network,EGS Journal of Physics and Chemistry of the Earth,1 – 22.
Shamseldin,A.Y., (1997), Application of A Neural Network Technique to Rainfall–Runoff Modeling,
Journal of Hydrology, 199, 272–294.
Sohail,,A.,Watanabe,K.,&Takeuchi,S.(2007), Runoff Analysis for A Small Watershed of Tono Area
Japan by Back Propagation Artificial Neural Network with Seasonal Data, Journal of Water
Resources Management Springer,22,1-22.
Tayfur,G.,Singh,V.P., &Asce,F.,(2006), ANN and Fuzzy Logic Models for Simulating Event-Based
Rainfall-Runoff,Journal of Hydraulic Engineering,12,1321 – 1330.
Tingsanchali,T.M.,&Gautam,M.R.,(2000), Application of Tank, NAM, ARMA and Neural Network
Models to Flood Forecasting, Hydrological Processes, 14, 2473–2487.
Toth.E.,Brath.A., &Montanari.A.,(2000), Comparison of Short-Term Rainfall Prediction Models for
Real-Time Flood Forecasting, Journal of Hydrology,239, 132 – 147.
Valença,M., Ludermir,T., &Valença,A.,(2005), Modeling of the rainfall-runoff relationship with
artificial neural network, IEEE.
Vandewiele,G.L., & WIN,N.L.,(2009),Monthly Water Balance Models for 55 Basins in 10
Countries,Hydrological Sciences Journal,43(5),687-699.
Vandewiele,G.L.,Xu,C.Y.,&Win,N.L.,(1992), Methodology and Comparative Study of Monthly
Water Balance Models in Belgium, China and Burma, Journal of Hydrology,134, 315–347.
Xia,J., O'Connor,K.M., Kachroo.R.K., &Liang,G.C.,(1997), A Non-Linear Perturbation Model
Considering Catchment wetness and Its Application in Fiver Flow Forecasting,Journal of
Hydrology,200,164 – 178.
Xu.J.,Wei.J., &Liu.Y.,(2010),Modeling Daily Runoff in a Large-scale Basin based on Support Vector
Machines, IEEE(International Conference on Computer and Communication Technologies in
Agriculture Engineering),601-604.
25