You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/290324877

AN OPTIMIZATION TECHNIQUE FOR ESTIMATION OF RATING CURVE


PARAMETERS

Conference Paper · December 2015

CITATION READS

1 1,087

3 authors, including:

Javed Alam Mohammad Zakwan


Aligarh Muslim University Aligarh Muslim University
81 PUBLICATIONS 569 CITATIONS 64 PUBLICATIONS 661 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Javed Alam on 13 January 2016.

The user has requested enhancement of the downloaded file.


Symposium on Hydrology, 22-23 December 2015, New Delhi, India

AN OPTIMIZATION TECHNIQUE FOR ESTIMATION OF RATING


CURVE PARAMETERS

MOHD MUZZAMMIL, JAVED ALAM, MOHAMMAD ZAKWAN

Department of Civil Engineering, Z.H. College of Engineering & Technology


A.M.U. Aligarh, INDIA

ABSTRACT
Streamflow measurements are of paramount importance in water resources engineering. It is the basic
data required for planning and design of various structures and water resource projects such as culverts,
bridges, operation of hydro power plants, flood protection works, flood plain zoning, flood warning
systems and assessment of available water resource potential. Accurate and continuous record of
streamflow data is, therefore, of utmost importance. However, continuous measurement of discharge is
very costly, time consuming, requires skilled labour and sometimes even impractical during floods. To
maintain a continuous record of stage in a river is rather easy and accurate as compared to streamflow.
The relationship between stage and discharge is single valued nonlinear relationship especially in
nonalluvial streams which is further utilised for estimation of discharge corresponding to observed stages.
The accuracy of estimated discharge depends on the accuracy in the estimation of parameters of stage
discharge relationship at a river section. In the present study, Parameters of stage discharge relationship
has been estimated using Generalized Reduced Gradient (GRG) solver, which is nonlinear optimization
code and considered as one of the most robust and reliable approaches to solve difficult nonlinear
programming problems. Development of rating curve and estimation of its parameter were performed on
excel spreadsheet. It has been found that GRG solver results in more accurate parameters of stage
discharge relationship as compared to those estimated from conventional approach.

Keyword : Stage, stream flow, rating curve, Generalized Reduced Gradient (GRG) solver
INTRODUCTION
Discharge data at various sections of a river are required for the design and construction of
hydraulic structures, flood management and optimum development of water resource. Accurate
stream flow measurement is of paramount importance for economic and safe design of water
resource project. However, continuous measurement of discharge at number of sections in a river
is very costly, sometimes impractical and a challenging job for field engineers. In majority of
streams, measurement of discharge is made by indirect method which depend on a relationship
between stage and discharge known as rating curve. Development of a rating curve eliminates
the need of continuous discharge measurement. Continuous measurement of stage is very easy
and cheap as compared to continuous measurement of discharge.
Discharge at a section in a river is not a function of stage alone. It depends on number of factors
like channel geometry, bed roughness and longitudinal roughness , but quantification of all these

234
Indian Association of Hydrologists (IAH), Roorkee
Symposium on Hydrology, 22-23 December 2015, New Delhi, India

factors is impractical. Hence, it becomes imperative to establish an accurate relationship between


stage and discharge which is considered epitome of all the characteristics of a reach of stream.
To establish the stage discharge relationship, historical stage discharge data of the gauging site is
used to plot the graph between the observed stage (G) and observed discharge (Q) at the gauging
site and fitting it with the best fit curve. Once the stage discharge relationship is established it is
then used to convert records of water level into discharge, thereby, eliminating time consuming,
costly and sometime impractical exercise of continuous discharge measurement.
Several attempts have been made by the researchers to establish rating curves using graphical
and numerical techniques. Stage discharge relationship was established based on polynomial
models by (Herschy,1995, 1999). (Tawfik et al. 1997) introduced the use of multilayer artificial
neural network (ANN) to model the stage discharge relationship. ANN was also employed by
(Jain and Chalisgaonkar 2000, Sudheer and Jain, 2003, Bisht 2010) to model the stage discharge
relationship. (Bhattacharya and Solomatine , 2005) established stage discharge relationship using
model tree M5 and ANN. Support vector machine (SVM) was used by (Sivapragasam and
Mutlie, 2005) and Gene expression programming (GEP) was used by (Guven and Aytek, 2009)
to model the rating curve. (Ghimire and Reddy, 2010) compared Genetic algorithm (GA) and
Model tree (M5) with GEP, MLR and classical stage discharge rating curve to establish the
superiority of GA and M5 over the other method to capture the stage discharge relationship more
accurately.
The reliability of discharge estimation from the rating curve is subject to the accuracy in stage
measurement and methods of the development of the rating curve. The conventional method of
regression analysis for the development of the rating curve often fails to establish stage discharge
relationship accurately. The main objective of the present study is, therefore, to implement a
simple and quick nonlinear optimization GRG solver to estimate the parameters of the stage-
discharge relationship and assess the performance of GRG solver with conventional method to
model stage discharge relationship.

METHODOLOGY
The stage discharge relationship is generally a single valued relation for majority of non-alluvial
streams and rivers (Subramanya 2008) and it is represented as
(1)
Q = stream discharge; G = stage height; a = constant representing the gauge reading
corresponding to zero discharge; K and n are the rating curve parameters.

-a) has to be prepared on double log scale with


discharge on abscissa and (G-
that results a straight line is considered as the correct va
For the conventional method the rating curve parameters K and n are obtained by the least square
error method. The logarithm form of equation (1) may be written as
log Q = n log (G-a) + log K (2)

235
Indian Association of Hydrologists (IAH), Roorkee
Symposium on Hydrology, 22-23 December 2015, New Delhi, India

Which may be expressed as


Y = AX + B (3)
Where
Y = log Q; A= n; X = log (G-a); B = log K.
Independent and dependent variables are regressed to get best fit straight line. The equation of
is
determined, the rating curve parameters can also be estimated by power law fitting of discharge
Q and (G-a) inbuilt in excel spreadsheet.
In this paper spreadsheet inbuilt regression method (RM 2) and GRG solver based on nonlinear
optimization method (NOM) has been used to obtain the stage discharge relationship. In order to
employ GRG solver the stage discharge relationship was modelled in on a spreadsheet using the
observed stage discharge data and assumed values of the rating curve parameters. Then the sum
of square of error was minimized by running the GRG solver. Over the years GRG solver have
been proved as one of the most robust and reliable approaches to solve difficult and highly
nonlinear programming problems. A brief description on this solver is provided herein.
Excel solver
Excel solver is an optimization tool embedded within Microsoft excel basically meant for
optimization of profit and loss in business problems. Excel solver is capable of optimizing linear
as well as nonlinear equations by changing specified parameters. It consist of linear
programming solver to optimize linear equations, GRG solver (Generalized Reduced Gradient)
and Evolutionary solver for optimizing nonlinear equations. Rating curve equations are basically
of nonlinear form therefore, GRG nonlinear solver is used to obtain the optimum values of rating
curve parameters.
GRG solver is a nonlinear optimization code which was developed by Leon Lasdon from
university of Texas at Austin and Allan Waren from Cleveland University. GRG and its specific
implementations have been proved in use over many years as one of the most robust and reliable
approaches to solve difficult and highly nonlinear programming problems (Lasdon and smith,
1992). GRG solver uses two techniques for determination of search direction. The default choice
is quasi-newton method, a gradient based technique and the second choice is the conjugate
gradient method. Depending on the available storage GRG solver may automatically switch
between the quasi-newton or conjugate gradient method (premium solver platform 2010).
Data
Stage- discharge data of river Yamuna at the proposed Lakhwar dam for the monsoon period of
year 1978 as reported by Mutreja has been used for the present study. Lakhwar dam (latitude
30º31'05'' and Longitude 77º56'56'') is a gravity dam under construction at Lakhwar town,
Dehradun, Uttarakhand in India. Data for the period June 01, 1978 to September 16, 1978 has
been used. The data set pertains to a period before the construction of the dam. Some of the
statistical parameters of the data set are shown in table 1.
Table 1: statistical parameters of the data set
Data Mean (µ) Maximum value Minimum value
Stage (m) 623.06 0.59 624.24 622.23
Discharge 170.89 114.02 451.87 30.56

236
Indian Association of Hydrologists (IAH), Roorkee
Symposium on Hydrology, 22-23 December 2015, New Delhi, India

(m3/s)
RESULTS AND DISCUSSION
The rating curve parameters a, K, and n were estimated using RM 1, RM 2 and NOM. The
rating curve parameters as obtained by Mutreja after regression have been reported for RM 1.
The parameters for RM 2 and NOM were estimated by the methodology as described earlier. The
values of these parameters obtained from the three approaches have been reported in table 2.

Table 2: Rating curve parameters for various method


Method a K n
RM 1 621.30 34.800 2.405
RM 2 621.30 37.768 2.382
NOM 621.83 113.840 1.523

The performance of various methods for the parameter estimation was assessed with respect to
the relation between the observed and the predicted discharges based on the following indices

Root mean square (rmse) = (4)

Nash-Sutcliffe criteria (e) = [1- ] (5)

Correlation coefficient (r) = (6)

Mean Absolute Relative Error (MARE) = (7)

Where Qo is observed discharge; Qp is predicted discharge and is mean of observed


discharge.The estimation method with lowest error and the highest correlation coefficient would
be considered as the best method. The values performance indices as calculated for the different
method have been provided in table 3 for the data set under considerations.

Table 3: Performance indices by different method


Method RMSE MARE Nash coefficient Correlation coefficient
RM 1 22.94 0.094 0.96 0.98
RM 2 22.41 0.082 0.97 0.98
NOM 17.56 0.070 0.98 0.99

237
Indian Association of Hydrologists (IAH), Roorkee
Symposium on Hydrology, 22-23 December 2015, New Delhi, India

Table 3 clearly demonstrate that the performance of nonlinear optimization method (NOM) is
better than that of the conventional methods of regression (RM 1 and RM 2). The reduction in
root mean square error (RMSE) using NOM was 23% and 21% in comparison to RM1 and RM2
respectively. A qualitative assessment of performance for the parameter estimation of the rating
curve was also made as shown in Figure 1and Figure 2. These figures also indicate that the
present approach based on nonlinear optimization is better than the conventional method.

624.5

624

623.5
Stage (m)

623
RM 1
RM 2
NOM
622.5 Observed data

622
0 100 200 300 400 500 600
Discharge (m3/s)

Figure 1. A comparison of rating curves obtained by different methods

238
Indian Association of Hydrologists (IAH), Roorkee
Symposium on Hydrology, 22-23 December 2015, New Delhi, India

480

430
Line of perfect agreement
380 RM 2
330 RM1
Observed discharge (m3/s)

280

230

180

130

80

30
30 80 130 180 230 280 330 380 430 480
Predicted discharge (m3/s)

Figure 2. A comparison of observed and predicted discharges for the Lakhwar site

CONCLUSION
The parameters of the rating curves were estimated using conventional method and nonlinear
optimization based GRG solver code for data set of Lakhwar dam site. The performance of both
the methods were assessed quantitatively as well as qualitatively. It has been found that GRG
solver is a promising tool for predicting rating curve parameters precisely. Moreover it
eliminates the need of time consuming and tedious process of trial and error for computation of
stage discharge relationships as in the case of the conventional method.
REFERENCES
ANN based river stage-discharge modelling
for Godavari -62.
Neural networks and M5 model trees in
modelling water level- -396.
stage-discharge rating curve in river
using genetic algorithms and model tree International Workshop Advances in Statistical
Hydrology. Taormina, Italy.
New approach for stage-discharge relationship: Gene
Expression Programming -492.
Streamflow measurement
Setting up stage-discharge relations using
Hydrologic Engineering, 5(4), 428-433.

239
Indian Association of Hydrologists (IAH), Roorkee
View publication stats

Symposium on Hydrology, 22-23 December 2015, New Delhi, India

nonlinear programs using GRG


Journal on Computing, 4 (1), 2-15.
Lasdon, L. S., Waren, A. D., Jain, A., and Ratner, M. (1978) Design and testing of a
generalized reduced gradient code
Mathematical Software, 4(1), 34-50.
Applied Hydrology. -Hill Limited, New Delhi, 110 002.
Reddy, M.J. and Ghim Use of Model tree and Gene expression programming
J. Intelligent Systems, 18(3), 211-227.
A textbook of hydrology
154-200.
Engineering Hydrology. -Hill, New
Delhi, 101-133.
Radial basis function Neural Network for modelling rating
curves -164.
Hysteresis sensitive Neural Network for
modelling rating curves -211.

240
Indian Association of Hydrologists (IAH), Roorkee

You might also like