Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

Performance of The Dispin Models with Automatic Parameter

Calibration on The Transformation of Rainfall to Runoff Data

Sulianto1, M Bisri2, L M Limantara2, D Sisinggih2

1Civil Engineering Department, University of Muhammadiyah Malang, 65144

Malang, Indonesia
2Water Resources Engineering Department, Universitas Brawijaya, 65145 Malang,

Indonesia

sulianto1967@email.com

Received 16-07-2019; revised 31-07-2019; accepted 20-08-2019

Abstract. This article presents a new model of the DISPRIN Model combination with two

different level optimization methods. The new model of DISPRIN Model combination and

Differential Evolution (DE) algorithm is called DISPRIN25-DE Models and its incorporation

with Monte Carlo Simulation method called DISPRIN25-MC Models. The case study is Lesti

Watershed (319.14 Km2) in East Java. The model test uses a 10-year daily data set, from

January 1, 2007 to December 31, 2016. Data series Year 2007 ~ 2013 as a set of training data

for calibration and data Year 2014 ~ 2016 as testing data set for model validation. Running

program DISPRIN25-DE Models with input parameter value C_min = 0, C_max = 1, H_min =

0, H_max = 600 mm obtained best fitness 0.044 m3/sec, NSE = 0.762 and PME = -0.059. The

DISPRIN25-MC Models analysis generates a minimum RMSE of 0.056 m3/sec, NSE = 0.779,

PME = -0.70. From the RMSE and NSE indicators it appears that both models can show an

equivalent level of performance, but in terms of the PME indicator and iteration time is

apparent The DISPRIN25-MC model has worse performance than the two DISPRIN25-DE

models.

Keywords: automatic, calibration, disprin model, rainfall-runoff

1. Introduction
A combination of metaheuristic methods with a conceptual hydrological model can produce a reliable

and practical new model applied to divert the rainfall data series into runoff data. In the application the

new model can perform automatic calibration by utilizing the rainfall data [P(t)], Evapotranspiration

[Ep(t)] and runoff [Q(t)] with limited series length [10].

Several new models resulted from a combination of conceptual models of hydrology and

metaheuristic methods have been successfully developed by previous researchers, among others; The

combination of GA with HBV Modified Model [15], CTSM Algorithm with HBV Model and NAM

Model [13], combination of GA with HBV Modelling Model [8], Shuffle Complex Evolution

Algorithm (SCE) with AFFDEF Model [4], Combination Dynamically Dimensioned Search (DDS)

Algorithm and SCE Algorithm with SWAT 2000 Model [21]. Xin'anjiang Models with SCE

Algorithms [1], GA and GA hybrid [20].

mailto:sulianto1967@email.com
mailto:sulianto1967@email.com

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

Metaheuristic methods for the automatic calibration of the Tank Model parameters have been

widely proposed by world researchers, including a combination of Particle Swam Optimization (PSO)

Algorithms with Tank Models [16], Tank Models combined with Marquard Algorithm [17], GA [13].

The combination of the Tank Models with the PSO Algorithm for flood discharge analysis with the

hourly period in urban areas in Taiwan has also performed very well [7]. Multi Tank Model 6 tanks

system combined with DDS Algorithm can show better results than the output of the Finite element

method (FEM) model [9]. The Tank Model 8 tank system combined with DDS Algorithm and GA can

perform well in predicting groundwater fluctuations in Japanese Yamagata. In this case the two

developed models can show nearly the same error rate, but the DDS Algorithm based optimization

method is more effective in terms of speed reaching convergent conditions [6, 11].

This article presents two new models result the combination of the DISPRIN Model (Dee

Investigation Simulation Program for Regulating Network) as described by D.G. Jamieson & J.C.

Wilkinson (1972) in Shaw, 1985, page 367 with parameter optimization method based on DE

Algorithm and Monte Carlo Simulation Method. The new model of the combination of DISPRIN 25

Model parameters with Algorithm DE called DISPRIN25-DE Model and its merger with Monte Carlo

Simulation Method named DISPRIN25-MC Model [18]. The application program system is compiled

using M-FILE MATLAB programming language with the reason of practicality in matrix operation

and easiness in data management in graphical form. The results of this study are expected to be an

alternative solution to solve the problem of limited river flow data which is often a classical constraint

in water resource development activities in developing countries.

2. Material and Methods

1.1 Hydrological Model

The DISPRIN model is included in the lumped model’s category which technically can be solved by

using the analogy of the Sugawara Tank Model simulation. In "the UK's Water Resources Board's

Model DISPRIN", this model was developed at the Dee River research program in the United States

[12]. The scheme developed as the basis of simulation is shown Figure 1.

Figure 1. Simulation scheme of DISPRIN Models.

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

In the application of the DISPRIN Model a watershed should be divided into 3 zones according to

their position and physical characteristics, i.e. up-land zone, hill-slope zone and bottom-slope zone.

The up-land zone is located in an upstream basin that is physically sloped by a steep slope, zone hill-

slope located in a central basin with a relatively moderate surface slope and a bottom-slope zone

located downstream of a watershed that tends to have a relatively flat surface slope. Each zone of the

watershed is presented by two vertically arrayed tanks. The first tank or upper tank represents the

combined surface and intermediate reservoir that contribute to the flow on the surface flow and the

intermediate flow. The second tank or bottom tank is the reservoir sub-base that contributes to the

flow of the sub-base flow. The tanks in each zone are interconnected with the principle of gravity

flow. The horizontal outflow flow from the up-land zone tank group will flow in the hill-slope zone

tank group, then the hill-slope zone tank group will drain the water in the bottom slope zone. In the

vertical upstream flow, the upper tank will fill the bottom tank when there is sufficient water reservoir

in the upper tank. However, if evapotranspiration is so dominant that it cannot be fulfilled by tank

water reserves, the water reserves in the lower tanks will be taken at the value of the deficit. This

process applies also to two tank groups in the hill-slope basin zone and bottom-slope zone.

The flow through a tank can be solved through the continuity equation. If Tank B1 is reviewed,

then the equation applies:

SB1(t) = SB1(t-1) + qA1(t) - qB1(t) - qB2(t) + Py(t) – Ey(t) (1)

with,

SB1(t) = Water level Tank B1 period t [mm].

SB1(t-1) = Water level Tank B1 period t-1 [mm].

qA1(t) = Horizontal flow Tank A1 period t [mm/day]

Py(t) = Precipitation on hill-slope zone period t [mm/day]

Ey(t) = Evapotranspiration on hill-slope zone period t [mm/day]

qB1(t) = Horizontal flow Tank B1period t [mm/day]

= cB1*[(SB1(t-1) + SB1(t)]/2) - hB1] (2)

qB2(t) = Vertical flow Tank B1 period t [mm/day]

= cB2*[SB1(t-1) + SB1(t)]/2] (3)

cB1 = Horizontal flow coefficient Tank B1

cB2 = Vertical flow coefficient Tank B1

The Tank D1 accommodates the channel flow factor in the attenuation effect component. The

water reservoir in this tank is not affected by the evapotranspiration process. The filling of water in the

Tank D1 is only influenced by the percolation flow of the three watershed zones. At the beginning of

the dry season the base flow in the river is caused by the intermediate flow and sub-base flow

components. However, at the end of the dry season when the water reserves in the intermediate zone

have been exhausted to meet evapotranspiration needs, the river flow is only supported by the Tank

D1. High water catches on Tank D1 stated:

SD1(t) = SD1(t-1 +qA4(t)+qB4(t)+qC4(t) (4)

In the DISPRIN Model stream flow is the result of translation effect of superposition of surface

flow, sub base flow and base flow. The translation effect factor is presented by Tank D2. The water

height in Tank D2 is calculated by the equation:

SD2(t) = SD2(t-1)+qC1t(t)+ qC3t(t)+qD1(t) (5)

with,

SD2(t) = Water lavel Tank D2 period t,

SD2(t-1) = Water level Tank D2 period (t-1),

qC1t(t) = (Ab/A_DAS) * qC1(t), and

qC3t(t) = (Ab/A_DAS) * qC3(t).

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

Q (t) is the river runoff period t at the control point of the watershed in units according to the

calculated analysis period. In data analysis with daily period input variables, then river runoff value in

m3/second unit can be converted by equation:

Q(t) = A_DAS * q(t)/(86.4) (6)

2.2 Calibration Model

2.2.1 Fitness function

The parameter calibration model is an analogy of solving the optimization problem to produce the

optimal value of DISPRIN Model parameters. The objective function of the optimization process is

the minimization of deviation curve of debit training data and the model simulation debit curve. In the

metaheuristic method the objective function is expressed as a fitness function. In this article the fitness

value is expressed as RMSE which is calculated by the equation [7, 21]:

𝐹 = 𝑅𝑀𝑆𝐸 = √
1

𝑁
∑ [𝑄𝑡𝑟𝑎𝑖𝑛,𝑡 − 𝑄𝑠𝑖𝑚,𝑡 ]

2𝑁
𝑡=1 (7)

with, F = fitness, QSim,t = discharge simulated period t, Qtraining, t = discharge training period t, N =

number of data points.

2.2.2 Parameter optimization based on Differential Evolution Algorithm (DE)

DE includes stochastic search methods and population based search. DE has similarities with other

evolutionary algorithms (EA), but differs in terms of distance and direction information from the

current population used to guide the search process for better solutions. In the field of hydrological

modelling the DE algorithm has been successfully applied for parameter optimization of SWAT

Model 16 parameters [21]. The DE algorithm contains 4 components, namely; 1) initialization, 2)

mutation, 3) recombination or crossover and 4) selection [2]. The process of calibrating the parameters

in the DISPRIN25-DE Model systematically can be explained as follows:

1) Input training data set : Evapotranspiration [Ep(t)], Rainfall [P(t)], Observation Debit
[Qtraining(t)] and area of up-land watershed zone [Au], hill-slope [Ah], bottom-slope [Ab].

2) Setting parameter DE : dimension (D), number of individual (N), upper limit (ub) and lower limit
(lb) parameter value, and maximum generation number (maximum iteration). The value of D is

corresponding to the number of optimized DISPRIN Model parameters. D = 25 for DISPRIN25-

DE Models.

3) Initialization: the generation of the initial value of the 0th generation vector, the variables to j and
vector ican be represented by the following notation.

𝑥𝑗,𝑖,0 = 𝑙𝑏𝑗 + 𝑟𝑎𝑛𝑑𝑗 (, 1)(𝑢𝑏𝑗 − 𝑙𝑏𝑗 ) (8)

The random number is generated by the rand function, where the resulting number lies between

(0,1). Index j denotes the variable to j. In case of minimization of function with 25 variables, then j

will be worth 1,2,3, .... 25.

4) Mutation, this process will produce population with size of N vector experiment. Mutation is done
by adding two vector differences to the third vector by the following notation :

𝑣𝑖,𝑔 = 𝑥𝑟0,𝑔 + 𝐹(𝑥𝑟1,𝑔 − 𝑥𝑟2,𝑔) (9)

It appears that two randomly selected vector differences need to be scaled before being added to

the third vector, xr0, g. Factor scale FЄ (0,1+) has real positive value to control population growth

rate. The base vector index r0 is determined by a random way that different from the index for the

target vector, i. Besides being different from each other and different from the index for the vector

base and the target vector, the vector index of the increments r1 and r2 can be chosen once per

mutant.

5) Crossover, at this stage DE crosses every vector (xi, g) with mutant vector (vi, g), to form the vector
of ui, g with the formula.

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

𝑢𝑖,𝑔 = 𝑢𝑗,𝑖,𝑔 = {
𝑣𝑗,𝑖,𝑔 → 𝑖𝑓 (𝑟𝑎𝑛𝑑(0,1) ≤ 𝐶𝑟𝑜𝑟𝑗 = 𝑗𝑟𝑎𝑛𝑑
𝑥𝑗,𝑖,𝑔 → 𝑖𝑓 (𝑟𝑎𝑛𝑑(0,1) > 𝐶𝑟𝑜𝑟𝑗 ≠ 𝑗𝑟𝑎𝑛𝑑

(10)

6) Selection, if trial vector ui,g has a goal function value smaller than the target destination function
xi,g, then ui,g will replace the position xi,g in the population in the next generation. If the opposite

happens then the target vector will remain in its position in the population.

The process of analysis of items 4), 5), 6) is done repeatedly from generation 0 to the maximum

generation specified (iter_max). Once the maximum generation is reached, then the optimal parameter

values and the model simulation debit from the calibration stage are generatedand the model

simulation debit from the calibration stage..

2.2.3 Parameter optimization base on Monte Carlo Simulation Method
In the field of hydrological modelling Monte Carlo simulations have been successfully used to

perform the uncertainty and sensitivity analysis of FORTRAN-HSPF model parameters [14]. The

Monte Carlo method can also be used to estimate the value of Tank Model parameters [3] and

successfully identify the value and structure of the HBV Model with great satisfaction [19]. The

process parameter optimization of DISPRIN25 Model is done through the following stages [14]:

1) Formulate a system of optimization model equations that will be simulated,
2) Input training data sets, namely: Au, Ah, Ab, P (i), Ep (i) and Qobs(i),
3) Input parameters Monte Carlo Method, ie: number of samples (N), and sample space limit (lb and

ub) of each variable analyzed, i.e. 25 parameters Model DISPRIN.

4) generates random numbers of uniform distributions or other probabilistic distributions of value
[0,1],

5) calculate the appropriate random variable for each model parameter under study based on the
number and space of the desired sample,

6) evaluate model performance by using random input parameter value result from step 5) according
to the equation developed in step 1),

7) steps 4) and 5) repeated as much as the number of samples given, and
8) analyzing and discussing the output of the models presented in graphical form and statistical

parameters.

2.3 Model validation
Model validation is done by reapplying the DISPRIN25 Model with input set of data testing and the

optimal parameter value of the result of the calibration process. The simulated discharge from the

model output will be compared with the discharge testing data, and the deviation test uses the RMSE,

Nash-Shutclife Efficiency (NSE) and Persistence Model Efficiency (PME) indicators. NSE and PME

are calculated by the formula [5].

𝑁𝑆𝐸 = 1 −
∑ (𝑞𝑡

𝑠𝑖𝑚−𝑞𝑡
𝑜𝑏𝑠)

2
𝑁
𝑡=1

∑ (𝑞𝑡
𝑜𝑏𝑠−𝑞𝑡

𝑚𝑒𝑎𝑛)
2𝑁

𝑡=1

(11)

𝑃𝑀𝐸 = 1 −
∑ (𝑞𝑡

𝑠𝑖𝑚−𝑞𝑡
𝑜𝑏𝑠)

2
𝑁
𝑡=1

∑ (𝑞𝑡
𝑜𝑏𝑠−𝑞𝑡−1

𝑜𝑏𝑠)
2𝑁

𝑡=1

(12)

2.4 Case Study
The position of Lesti Watershed in unity of Brantas watershed is shown in Figure 2. Lesti Watershed

has an area of 319.14 km2, divided into Up-land, hill-slope and bottom slope respectively of 67.02

Km2, 114.89 Km2 and 137.23 Km2.The data series of hydroclimatology in this study is data of

recording date from January 1, 2011 to December 31, 2013. Evapotranspiration data obtained from the

analysis using Penmann Method with data input wind speed, air temperature, air humidity and the

duration of solar irradiance monthly. Climatic parameters data obtained from the recording results

from Karangkates Station. There are 4 rain gauge stations covered in Lesti Watershed, namely; St.

Dampit, St. Turen, St. Wajak and St. Tirtoyudo. Rainfall data recorded in daily period. Average

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

regional precipitation is calculated by the Polygon Thiesen Method. Weighting factor of Polygon

Thiesen in the four rain stations was 0.380, 0.09, 0.19 and 0.34, respectively. River flow data from the

listing of St. AWLR Tawangrejeni is available in the hourly period. The transformation of the

discharge data into a daily mean period is calculated by means of an algebraic average. Furthermore,

the data series is divided into two groups. The first group as a set of data training for the calibration

process and the second group as set of data testing for the model validation process. Set of 7-year

training data, the record period of January 1, 2007 until December 31, 2013 and set of data testing

from the record results of the period January 1, 2014 to December 31, 2016. Hydroclimological data in

daily periods graphically shown Figure 3. Comparison of statistical parameters set of data training and

data testing sets are shown in Table 1. Average, minimum, maximum, and varied sets of data testing

tend to be larger than the set of data testing.

Figure 2. Case studi, Lesti Watershed

Table 1: Comparison of statistical parameters of training and testing data sets

Statistic of parameters
Discharge data (m3/sec) Rainfall data (mm/day)

Training Testing Training Testing

Mean 17.44 18.59 6.17 6.76

Minimum 5.91 5.99 0.00 0.00

Maximum 35.03 37.58 77.76 83.87

Deviation standard 6.02 6.87 9.62 10.82

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

Figure 3: daily training and testing data set

3. Result and discussion
The DISPRIN Models implementation reference is still very limited so the feasibility limit of the

parameters value becomes difficult to define. Referring to the application of the Sugawara’s Tank

Model of various references, the minimum (LB_C) and maximum (UB_C) parameters of the tank

outlet coefficient are "0" and "1" respectively. The initial high water catch and the position of the tank

outlet are positive numbers whose value varies depending on the watershed hydrological

characteristics being analyzed. In this article the minimum value of the initial parameter of the

container and the position of the tank outlet (LB_H) is set to "0", and the maximum value (UB_H) is

approached by trial and error. The results of the analysis by utilizing the application DISPRIN25-DE

Model generated the relevant UB_H value is 600 mm. Further analysis using input value LB_C = 0,

UB_C = 1, LB_H = 0, UB_H = 600 mm. The results of the DISPRIN25-DE Models analysis with the

parameters inputN = 350 and Iter_max = 250 and the results of the DISPRIN25-MCModels analysis

with the input sample number of 200,000 are shown in Table 2, Table 3 and Figure 4 to Figure 6.

Table 2: Comparison of model performance indicator values

Performance indicator Notation
Calibration stage Validation stage

DE MC DE MC

Root Mean Square Error RMSE 0.044 0.056 0.086 0.100

Nash-Sutcliffe Efficiency NSE 0.762 0.779 0.727 0.769

Persistence Model Efficiency PME -0.059 -0.700 -0.128 -0.505

Time of iteration t_iteration 126.67 172.59 - -

The best fitness value or minimum RMSE at the calibration stage of the DISPRIN25-DE Models is

obtained 0.044 m3/sec and the analysis of the DISPRIN25-MC Models with the sample input quantity

of 200,000 is obtained 0.056 m3/sec. The progress of achieving the minimum RMSE values of the two

models is shown in Figures 4 and 5. Based on the RMSE and NSE indicators both models can perform

as well, but on the PME indicator the both model gives a negative value which means it does not

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

perform well, The DISPRIN25-DE model has much better performance than the DISPRIN25-MC

Model. Model performance in the validation stage is generally decreased although not significant. This

is understandable because the data training sets and data testing sets have different statistical

parameters. In terms of the iteration the DISPRIN25-MC Model shows 30% longer than the

DISPRIN25-DE model. From the various considerations, it can be concluded that the process

optimization parameter DISPRIN25 Model based DE Algorithm is more effective than the analysis by

applying Monte Carlo Simulation Method.

Figure 4: Progress the best fitness value of DE Algorithm

Figure 5: RMSE value of Monte Carlo Simulation

results Sorted from large to small

Table 3: DISPRIN25 Models optimum parameters

Parameter Description
Constraint Optimum parameters

LBj UBj DE MC

hA1 height of surface outlet up-land zone 0.00 600.00 0.07 69.95

hA2 height of sub surface outlet up-land zone 0.00 600.00 600 433.8

CA2 infiltration coefficient up-land zone 0.00 1.00 0.348 0.502

CA3 sub surface coefficient up-land zone 0.00 1.00 0.499 0.0075

CA4 percolation coefficient up-land zone 0.00 1.00 0.18 0.926

SA1_0 initial storage of tank SA1 0.00 600.00 155.15 599.79

SA2_0 initial storage of tank SA2 0.00 600.00 0 434.53

hB1 height of surface outlet hill-slope zone 0.00 600.00 0 592.27

hB2 height of sub surface outlet hill-slope zone 0.00 600.00 500.31 281.28

CB2 infiltration coefficient hill-slope zone 0.00 1.00 0.672 0.954

CB3 sub surface coefficient hill-slope zone 0.00 1.00 0.906 0.898

CB4 percolation coefficient hill-slope zone 0.00 1.00 0.293 0.768

SB1_0 initial storage of tank SB1 0.00 600.00 67.72 213.99

SB2_0 initial storage of tank SB2 0.00 600.00 600 160.06

hC1 height of surface outlet bottom-slope zone 0.00 600.00 0 434.89

hC2 height of sub surface outlet bottom-slope zone 0.00 600.00 596.49 269.24

CC2 infiltration coefficient bottom-slope zone 0.00 1.00 0.846 0.684

CC3 sub surface coefficient bottom-slope zone 0.00 1.00 0.26 0.547

CC4 percolation coefficient bottom-slope zone 0.00 1.00 0.506 0.829

SC1_0 initial storage of tank SC1 0.00 600.00 164.48 221.5

SC2_0 initial storage of tank SC2 0.00 600.00 600 79.31

CD1 runoff coefficient 0.00 1.00 0 0.007

CD2 runoff coefficient 0.00 1.00 0.007 0.386

SC1_0 initial storage of tank (attenuation effect) 0.00 600.00 282.57 369.87

SC2_0 initial storage of tank (translation effect) 0.00 600.00 591.1 2.97

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

Figure 6: Comparison of the discharge observed data series and the model output

The comparison of the flow curve data and the curve from the model outcomes at the calibration

stage and validation stage is shown Figure 6. Results from the DISPRIN25-MC model outcomes

appear to be different at high flow conditions, where the results obtained tend to be under estimated.

The resulting flow curves of the both models at the calibration stage can generally follow the seasonal

trend of the training data flow curve. In low flow, normal flow and high flow conditions tend to place

themselves in a moderate position. Fluctuations in the sharp flows that occur due to the high rainfall

intensity in the daily period cannot be responded well. This condition is the cause of the low value of

the resulting PME indicator. Taking to the simulation scheme and the total equation of the Model

DISPRIN25 flow as shown in Figure 1, the existence of the translation effect factor (channel flow) is

likely to be the factor causing this condition. The existence of the translation effect factor presented by

a tank with a bottom outlet actually becomes an obstacle to model efforts in anticipating the

fluctuation of sharp flows. Rapid flow changes due to the high rainfall intensity occurring in the

accumulation of up-land zone tanks, hill-slope and bottom-slope are muted in translation effect tanks

and are delivered slowly. In the minimum RMSE condition the optimal value of the DISPRIN25

Model parameter is shown in Table 3 in column [5, 6]. The optimal value of DISPRIN25 Model

parameters from the output of the two models gives different results because the resulting performance

level is also significantly different.

4. Conclusions
Based on the RMSE, NSE, PME and iteration indicators the DE Algorithm is proven to work more

effectively than Monte Carlo Simulation Methods in solving the parameterization problem of the

DISPRIN25 Model. Testing Model DISPRIN25-DE on Lesti Watershed (319.14 Km2) with daily data

set can show good performance both calibration stage and validation stage. Generally generated flow

curves can follow the seasonal trend of the observation flow curve. In low flow conditions, the normal

flow as well as the high flow curve from the outline of the model tend to place themselves in a

moderate position. Fluctuations in the sharp flows that occur due to the high rainfall intensity of the

daily period cannot be responded well. This condition is the cause of the low value of the resulting

PME indicator. The translation effect factor presented by a tank with a basic outlet actually hampers

the efforts of the DISPRIN25 Model in anticipating the occurrence of sharp fluctuations. Rapid flow

changes due to the high rainfall intensity occurring in the accumulation of up-land zone tanks, hill-

slope and bottom-slope are muted in translation effect tanks and are delivered slowly. The optimal

0 500 1000 1500 2000 2500 3000 3500 4000
5

periode

D
is

c
h
a
rg

e
(m

3
/s

e
c
)

Comparation Q DISPRIN-DE, Q DISPRIN-MC & Q-obs

Q-obs

DISPRIN-DE

DISPRIN-MC

5 10 15 20 25 30 35 40
5

Q-DISPRIN DE [m3/sec]

Q
-o

b
s
[

m
3
/s

e
c
]

Plot Q-DISPRIN DE VS Q-Obs

5 10 15 20 25 30 35 40
5

Q-DISPRIN MC [m3/sec]

Q
-o

b
s
[

m
3
/s

e
c
]

Plot Q-DISPRIN MC VS Q-Obs

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

value of DISPRIN25 Model parameters from the output of the two models gives different results

because the resulting performance level is also different.

References

[1] Bao H. J, Wang L, Li Z. J, Zao L. N, Guo-ping ZHANG, 2008, Hydrological daily rainfall-

runoff simulation with BTOPMC model and comparison with Xin’anjiang model, Water

Science and Engineering, 2010, 3(2): 121-131, doi: 10.3882/j.issn.1674-2370.2010.02.001,

http://www.waterjournal.cn, e-mail: wse2008@vip.163.com

[2] Budi Santosa & Paul Willy, 2011, Metoda Metaheuristik Konsep dan implementasi, Cetakan

pertama, Guna Widya, Surabaya, Indonesia.

[3] Chen C, Shrestha D.L., Perez G.C., Solomatine D., 2006, Comparison of methods for

uncertainty analysis of hydrologic models, 7th International Conference on

Hydroinformatics HIC 2006, Nice, FRANCE.

[4] Darikandeh Dorsa, Akbarpour Abolfazl, Bilondi Mohsen Pourreza & Hashemi Seyyed Reza,

2014, Automatic calibration for Estimation of The Parameters of Rainfall - Runoff Model,

SCIJOUR, Journal of River Engineering, Volume 2, Issue 8 2014,

http://www.scijour.com/jre.

[5] Gupta, Sorooshian & Yapo, 1999, “Status of automatic calibration for hydrologic Models:

Comparation With Multi Level Expert Calibration”, Journal of Hydrologic Engineering, Vol.

4, No. 2, April, 1999. ASCE, ISSN:1084-0699/99/0002-0135–0143,

http://www.ascelibrary.org.

[6] Huang X. L. & Xiong J., 2010, Parameter Optimization of Multi-Tank Model with Modified

Dynamically Dimensioned Search Algorithm, Proceedings of the Third International

Symposium on Computer Science and Computational Technology (ISCSCT ’10), Jiaozuo, P.

PUBLISHER, AP-PROC-CS-10CN007.

[7] Hsu & Yeh, 2015, Study on Flood Para-Tank Model Parameters with Particle Swarm

Optimization, Journal of Information Hiding and Multimedia Signal Processing, Ubiquitous

International Volume 6, Number 5, September 2015, @ 2015 ISSN 2073-4212

[8] Jonsdottir Harpa, Madsen Henrik & Palsson Olafur Petur, 2005, Parameter estimation in

stochastic rainfall-runoff models, ELSEVIER, Journal of Hydrology 326 (2006) 379–393.

[9] Kenji T.，Yuzo O.，Xiong J. & Koyama T, 2008, Tank Model and its Application to
Predicting Groundwater Table in Slope, Chinese Journal of Rock Mechanics and

Engineering, Vol.27 No.12 Dec，2008, CLC number：P 642.22 Document code：A Article

ID：1000–6915(2008)12–2501–08.
[10] Kim Oong H., Paik Kyung R., Kim Hung S. & Lee, 2005 Dong R., A Conceptual Rainfall –

runoff Model Considering Seasonal Variation, ARTICLE in HYDROLOGICAL

PROCESSES · DECEMBER 2005Impact Factor: 2.68 · DOI:

10.1002/hyp.5984ADVANCES IN HYDRO-SCIENCE AND –ENGINEERING, VOLUME

VI, http://www.researchgate.net/publication/ 227599393

[11] Kuok King Kuok, Sobri Harun and Po-Chan Chiu, 2011, Comparison of Particle Swarm

Optimization and Shuffle Complex Evolution for Auto-Calibration of Hourly Tank Model’s

Parameters, Int. J. Advance. Soft Comput. Appl., Vol. 3, No. 3, November 2011, ISSN 2074-

[12] M. Shaw, Elizabeth, 1985, Hydrology in Practice, Van Nostrand Reinhold (UK) Co. Ltd.

[13] Ngoc T. A., Hiramatsu K. & Haramada M., 2012, Optimizing Parameters for Two Conceptual

Hydrological Models Using a Genetic Algorithm: A Case Study in the Dau Tieng River

Watershed, Vietnam, JARQ 47 (1), 85 – 96 (2013) http://www.jircas.affrc.go.jp.

[14] Ramires J.D, Camacho R, McAnally W, Martin J, 2012, Parameter uncertainty methods in

evaluating a lumped hydrological model, Obrasy Proyectos 12, 42-56.

http://www.waterjournal.cn/
http://www.waterjournal.cn/
http://www.scijour.com/jre
http://www.scijour.com/jre
http://www.researchgate.net/publication/%20227599393
http://www.researchgate.net/publication/%20227599393
http://www.jircas.affrc.go.jp/
http://www.jircas.affrc.go.jp/

Civil and Environmental Science Journal

Vol. II, No. 02, pp. 084-094, 2019

[15] Saibert Jan, 2000, Multi-criteria Calibration of Conceptual Runoff Model Using a Genetic

Algorithm, Hydrology and Earth System Sciences, 4(2), 215-224 (2000) EGS.

[16] Santos, 2011, Application of a particle swarm optimization to the Tank Model, Risk in Water

Resources Management (Proceedings of Symposium H03 held during IUGG2011 in

Melbourne, Australia, July 2011) (IAHS Publ. 347, 2011).

[17] Setiawan B., Fukuda T. & Nakano Y., 2003, Developing Procedures for Optimization of Tank

Model’s Parameters, Agricultural Engineering International: the CIGR Journal of Scientific

Research and Development.

[18] Tolson B. A. & Shoemaker C. A., 2007, Dynamically dimensioned search (DDS) algorithm for

computationally efficient watershed model calibration, ATER RESOURCES RESEARCH,

Geophysical Union. 0043-1397/07/2005WR004723$09.00

[19] Uhlenbrook S, Seibert J, Leibundgut C, Rodhe A, 1999, Prediction uncertainty of conceptual

rainfallrunoff models caused by problems in identifying model parameters and structure,

Hydrological Sciences-Joumal-des Sciences Hydrologiques, 44(5) October.

[20] Wang W. C., Cheng C. T., Chau K. W & Xu D. M., 2012, Calibration of Xinanjiang model

2012, Journal of Hydroinformatic, 14.3 – 2012, 785 W.-C. Wang et al. | Calibration of

Xinanjiang model parameters using HGA based FOM.

[21] Zhang, Srinivasan, Zhao & Liew (2008), Evaluation of global optimization algorithms for

parameter calibration of a computationally intensive hydrologic model, HYDROLOGICAL

PROCESSES, Published online in Wiley InterScience, www.interscience.wiley.com.