1th International Conference on Water, Environment and Sustainable Development, 27-29 September, 2016 University of Mohaghegh Ardabili, Ardabil,

Simulation Of Water Level In The River Using One- dimensional Hydrodynamic Model

Nasrin Badrzadeh 1, Jamal Mohammad Vali Samani2, Mehdi Mazaheri3, Bahman Fakori4 1,4- MSc. Student of Water Structures, University of Tarbiat Modares 2- Prof, Department of Water Structures, University of Tarbiat Modares 3- Assistant Prof, Department of Water Structures, University of Tarbiat Modares

[email protected]

Abstract Using of computer models is an adequate tool for simulation of flow and water level conditions in the rivers which can be the basis for planning and implementation of water engineering projects. A one- dimensional (1D) hydrodynamic model is used to simulate the water level with limited available data in the Aras River. In this study, an attempt has been made to extract the river cross-sections from GIS using HEC-GeoRAS and extracted cross-sections were used in the MIKE 11HD model for the simulation of water level at various sections of the Aras river. Manning’s roughness coefficient 0.033 was considered for Aras River and computational results were compared and evaluated with observational data. The model was calibrated and validated for the periods May 22 to December 21, 2006, and December 22, 2007 to May 21, 2008 respectively. The model-simulated water level are found to be in close agreement with the observed ones. The study demonstrates the usefulness of using the GIS to derive river cross- sections for use in hydrodynamic modelling studies. Keywords: Hydrodynamic model, MIKE 11, Aras river, River cross-sections.

1. INTRODUCTION

Administration of the water resources engineering projects of rivers basin requires detailed knowledge and theoric simulation of various phenomena that are involved in the process. Using of computer models is an adequate tool for simulation of flow and water level conditions in the rivers which can be the basis for planning and implementation of water engineering projects. One dimensional (1D) river flow modeling became easier after the development of fast computers. A good number of studies on various aspects of river hydraulics have been carried out for last two decades. The space–time variables of flow and water level in the rivers are computed using St. Venant equations [1]. The equations are applicable to shallow water flow conditions, which are governed by law of conservation of mass and momentum. Unsteady state flow conditions in the rivers are simulated with considerable accuracy making use of the above mentioned equations. The exact solutions of the equations are very difficult to attain. However, approximate solutions are obtained using numerical methods with some assumptions [1]. Various numerical methods used for solving St. Venant equations are finite difference method (FDM), finite element method (FEM) and finite volume method (FVM). In FDM, the partial differential equations are converted into equivalent finite difference equations and then the difference equations are solved using some computational schemes. During the past decade, many studies have been conducted on computational river hydraulics using different hydraulic models, which are based on the above-mentioned methods [6]. The numerical models with 1D approximation are mostly based on the finite difference method and the finite element method. The finite difference method is more popular due to comparatively less computational effort. Various softwares, such as HEC-RAS (Hydrologic Engineering Center River Analysis System) from the US Army Corps of Engineer’s HEC [13], MIKE 11 developed at the Danish Hydraulic Institute, Denmark [7], SOBEK-1D developed at the Delft Hydraulics, Delft [15], are some hydraulic models, which use the above numerical methods to compute flow and water level at different grid points along the rivers. One-dimensional hydraulic modeling is the simplest option which is best suited for representing flows within interconnected channel networks [2]. However Simulation of water levels at different sections of a river using hydrodynamic models is cumbersome, because it requires many types of data, such as hydrologic time series, river geometry, hydraulics of existing control structures and channel roughness coefficients.

1

1th International Conference on Water, Environment and Sustainable Development, 27-29 September, 2015 University of Mohaghegh Ardabili, Ardabil, Iran

Hydraulic models require accurate river geometry data, which may not be available at the desired locations in the rivers. River cross-sections are the prime input to any river hydraulic model for simulation of water level and discharge. Availability of measured river cross-sections is scanty in most of the developing countries, countries, thereby making it difficult to simulate the water level and discharge using hydraulic models. The measurement of the riverbed topography is not straightforward. Many different methods have been developed to estimate the riverbed topography. The integration of geographic information system (GIS) with the hydraulic models (HEC-GeoRAS, MIKE 11 GIS and WaterRIDE FLOOD) is the recent advance in computational river hydraulics [12]. Currently, several commercial software packages are available that connect GIS and river hydraulic modeling. Mathematical models are useful tool in analysis of river flow or hydraulic structures. Of the conducted studies in this field, studies of Thakur et al [14], Pramanik et al [11], Petro et al [10] can be pointed out that extracted rivers’ cross sections using geographical information system (GIS) DEM of the study area; subsequently they simulated river flow rate and its water level by utilizing the morphological information derived from hydrodynamic model MIKE11. In this study; MIKE 11, has been chosen because of it flexibility, speed, accuracy and popularity for the hydraulic studies. MIKE 11 is a software developed at the Danish Hydraulic Institute (DHI) for the simulation of water flow, sediment transport and water quality in estuaries, rivers, irrigation systems and similar water bodies. This model is one of the best models and one of characteristic of the model is a simulation under unsteady modeling. Studies on 1D and 2D modelling of river flow are limited in the Aras river basin due to unavailability of good quality of measured data. Availability of limited river cross-sections and scarcity of time series of water level and flow data are the main causes, limiting. A hydraulic model coupled with GIS use the HEC- GeoRAS to extract the river cross-sections, which define the geometry of the river. The extracted river cross- sections are then used in the MIKE 11HD model for the computation of water level at various sections of the Aras river.

2. STUDY AREA

This project focuses on the Aras river basin. Aras rive poses as the most important and fully watered river in the northern of Iran in the region flowing from the west of Julfa to Aslanduz Moghan (Ardabil Province) along the Iran-Azerbaijan border. This river springs from Mingun highlands in the south of Erzurum in Turkey and pours firstly river in Azerbaijan and then Caspian Sea by crossing from the three countries of Iran, Armenia and Azerbaijan and passing approximately 1072 km [3]. Maximum of flow rate average in the study area is 1100 m3/s located at and 2,600 m3/s at the Mill-Moghan diversion dam site. However, the values listed in dry years may be declined to 32 and 180 m3/s. Aras river water is chiefly used in terms of agricultural and animal husbandry usages [3]. Aras river comprises coarse-grained bed materials with relatively high slope, shallow and often high width, which is generally situated in the middle or mountainous region. This study aims at hydrodynamic simulation of the Aras river between the Mill-Moghan diversion dam to Parsabad city using hydrodynamic model MIKE 11. A Study area of this research includes Aras’s main branch located between longitude and latitude of 47◦22 to 47◦ 58 E and 39◦41 to 39◦26 N, respectively. Accordingly, two hydrometric stations of s1 (Mill-Moghan dam output), s2 (Parsabad city) were taken into account. Fig. 1 illustrates the spatial position of the study area and hydrometric stations over the Aras river.

3. DATA USED IN THE HYDRODYNAMIC MODEL

The input data requirement for hydrodynamic modelling generally consists of two types of data sets: (i) topographic data and (ii) hydraulic data. Topographic data include river course geometry and channel geometry, i.e. surveyed cross-sections of the river at different locations and longitudinal slope of the riverbed. Hydraulic data include an inflow hydrograph on the upstream end of the river together with a downstream water-level time series and appropriate channel roughness/friction coefficients. A MIKE 11HD model does not require such a large number of data sets in the calibration process. It only requires setting up of appropriate model parameters and input boundary conditions to represent the physical process of river flow. The river geometry was defined by inserting the river cross-sections at different locations of the river. 144 river cross-sections at different locations of the rivers were extracted from the GIS and were used in the MIKE 11 hydrodynamic (MIKE 11HD) model. Daily discharge data and daily water level data for a period of 365 days May 22, 2007– May 21, 2008, were collected from the Water Resources Management

2

1th International Conference on Water, Environment and Sustainable Development, 27-29 September, 2015 University of Mohaghegh Ardabili, Ardabil, Iran

Organization (WRM) Iran and were used as upstream and downstream boundary conditions in the hydrodynamic model.

Figure1. Location of Hydrometric stations (s1, s2) over Aras River

4. METHODOLOGY

4.1 EXTRACTION OF CROSS-SECTION DATA FROM HEC-GEORAS

Cross sections data are being employed to determine the carrying capacity of the river and surrounding floodplains. Amongst the diverse software, HEC-GeoRas is capable of extracting cross sections parameters from a digital elevation model (DEM) and introducing this information into the model so as to display and analyze in GIS. With the preprocessor tool of HECeGeoRAS, terrain data (TIN or Grid) are used to digitize the river centerline, banks, flow direction, and the start and end stations of a river reach. This is a quick and an easy approach for situations where the flow is predominantly in a single direction. The results however are likely to depend on the type, quality and resolution of the terrain data. In order to extract the cross-sections parameters, associated with studied reach, digital maps in 1: 25000 of mapping agency were used to construct a TIN as the basis of DEM data. The Cross-section data extraction procedure involves defining a TIN as a level and creating the digitalization of the cuts pertained to the cross-sections’ lines. Cut lines demonstrate the cross-sections conformity and alignment in the X, Y coordinates. After this step, ground height beneath each DEM surface vertex is interpolated. Therefore, the cross-sections, height and cut lines are saved as three-dimensional lines, or are extracted and transferred as a table. The following figure shows how to extract cross sections from a TIN.

3

1th International Conference on Water, Environment and Sustainable Development, 27-29 September, 2015 University of Mohaghegh Ardabili, Ardabil, Iran

Cross section cut line

TIN

Resultant cross section

Figure 2. Extraction of cross-section from TIN

Figure 3. Extracted river cross-section in MIKE 11

As well as, to extract cross-sections parameters and simulate study reach morphology, in ArcGIS using HEC-GEORAS, a series of points, lines and polygon layers associated with the development of geometric data were built to define the stream’s centerline, the centerline of flow path, the sides of the main channel to introduce the hydraulic model. According to the circumstances of the riverbed and sides, 144 cross sections along the 70 Km of the Aras river were considered to be representing river’s general condition.

4.2 MIKE11 HYDRODYNAMIC MODEL (MIKE 11HD)

MIKE 11HD model is a one-dimensional hydraulic modeling package, developed at Danish Hydraulic Institute in 1987. The model has been widely used to simulate water levels and flow in the river systems. MIKE 11 is an implicit finite difference mathematical model for computation of unsteady flow in rivers and floodplains. The governing flow equations of MIKE 11 are 1D and are of shallow water types that are the modifications of Saint Venant equations [9]. These equations are solved numerically using an implicit finite difference scheme known as the six-point Abbott scheme [8]. MIKE 11HD model solves fully dynamic shallow water flow equations (St. Venant Equations) using FDM technique. Equations 1 and 2 are the St. Venant Equations, which represent the law of conservation of mass and momentum respectively.

4

1th International Conference on Water, Environment and Sustainable Development, 27-29 September, 2015 University of Mohaghegh Ardabili, Ardabil, Iran

Where, Q = Discharge. A = Area of Cross Section. C = Chezy Coefficient. R = Hydraulic Radius. q = Lateral Discharge. α = Momentum Correction Factor. g = acceleration due to gravity. The MIKE 11HD model can be applied to a river of any size and gradient, i.e. from steep streams to tidal estuaries. The program is designed for irregular open channels and can account for varying flow resistance through different branches of the network. The flow and water level at the boundaries can be steady or can vary with time. The model has the capability to simulate flow over embankments as well as through weirs and bridges MIKE 11 is an implicit finite difference mathematical model for computation of unsteady flow in rivers and floodplains. In the present study, the boundary conditions of the model were defined by the time series of discharge at the upstream point (Mill-Moghan dam output) and time series of water level at the extreme down stream point (Parsabad city). The performance of the model was evaluated using the goodness-of-fit indices like R2, RMSE, E, and d.

5. RESULTS AND DISCUSSION

5.1 CALIBRATION AND VALIDATION OF MIKE 11 HD MODEL

The MIKE 11HD model setup was prepared by defining the model boundary conditions, inserting the river cross-sections, fixing the initial conditions of water level and discharge for hydrodynamic simulation of the river flow. One of the parameters that should be calibrated in mathematical models of hydrodynamic simulations is the resistance coefficient of river’s bed which defined as Manning roughness coefficient or Chezy coefficient. The MIKE 11 hydrodynamic model was calibrated and validated for the periods May 22 to December 21, 2006, and December 22, 2006 to May 21, 2007 respectively. The MIKE 11HD model was calibrated by changing the values of Manning’s roughness coefficient (n) in the river reach. During calibration, the simulated water level and the observed water level were compared for different values of n till the simulated and observed water level matched closely. To avoid model instability, appropriate computational time step and grid size were selected. In the model setup, the computational time step and grid size were assigned as 30 sec and 500 m, respectively. After primary simulation of the river, the calibration and verification tests of the model were performed for existing data (data obtained from hydrometry stations) and finally, the Manning roughness coefficient for MIKE11 model was selected as 0.03. The model validation was performed by comparing the simulated daily water level at Aras with the corresponding observed values. The results obtained from model validation showed an improvement in model prediction accuracies. Figure 4 a and b present a comparison between simulated and observed water levels for calibration and validation. The performance of the calibration and validation results of the hydrodynamic model is evaluated for all the years using different performance indices. The coefficient of determination (R2), modeling efficiency (E) and index of agreement (d) were found to be 0.944, 0.847 and 0.847 respectively for the Aras with the respect to water level simulation. From Figure 4 it is observed that the simulated water level matched well with the observed values. The performance of the model in simulating water level at all the locations was found to be satisfactory as revealed from their goodness-of-fit indices (Table 1). Also WL of observed data varies from 21.485 to 19.662 m (variation of 1.823 m) and simulated WL varies from 21.4 to 19.8 m (variation of 1.6 m), with difference in peak flood of 0.223 m. This difference can be attributed to the uncertainty in river cross-section data, which affects overall conveyance of river channel. The reason of not taking water level in the model’s calibration and verification as a comparison base into account is dated back to obtaining flow rates values in the model using the continuity equation. Coupled with, since roughness coefficient has no effect on the continuity equation, so flow rate changes will be negligible. In contrast, since

5

1th International Conference on Water, Environment and Sustainable Development, 27-29 September, 2015 University of Mohaghegh Ardabili, Ardabil, Iran the water level is obtained from momentum equation and considering that roughness coefficient affects the momentum equation, the slightest change in this factor will lead to greatly alter the values obtained for the stage.

Table 1- Performance indices obtained of the MIKE 11 HD model.

Performance indices R2 RMSE E d Calibration 0.875 0.140 0.85 0.906 Validation 0.944 0.122 0.847 0.961

21.5 (b) Observed WL Simulated WL 21

20.5

Water Level (m) Level Water 20

19.5 26-Nov 15-Jan 6-Mar 25-Apr 14-Jun Time (day)

Figure 4. Comparisons of observed and simulated water level during (a) calibration, (b) validation

Water surface profiles of the Aras river were simulated using the model. Figure 5 present the simulated water surface profile of the river Aras. Figure 5 is showing how to fluctuate the water level from the bed of river. Also Figure 6 shows a graphical representation of river’s transverse profile and water level in different points. As can be seen from Figure 6, as we come down to the downstream station the river width becomes wider. This indicates a reduction at height and flow rate in the river downstream.

6

1th International Conference on Water, Environment and Sustainable Development, 27-29 September, 2015 University of Mohaghegh Ardabili, Ardabil, Iran

Elevation (m)

Distance (m) Figure 5. Simulated water surface profile along the Aras river.

Figure 6. A Sample MIKE11 Cross-sectional profiles of Aras in chainage (a) 0, (b) 12.445, (c) 38.089, (d) 67.306

6. CONCLUSIONS

7

1th International Conference on Water, Environment and Sustainable Development, 27-29 September, 2015 University of Mohaghegh Ardabili, Ardabil, Iran

The cross-section elevations at various locations of the river course play a vital role in hydrodynamic modeling. Also, detailed field surveys for deriving such information on topography are often time-consuming and expensive. According to previously carried out studies, using GIS functionality and capabilities will result in desired outcomes in the simulation of river conditions of its adjacent lands [5]. Additionally, the use of GIS will enhance the efficiency and reduce the costs of research [4]. In the current study, riverbed geometry was also simulated by using the geographic information system (GIS) as well as the hydraulic behavior of the river by using Mike 11 model with high accuracy and short time. The results reveal that the above methodology is quite useful to extract cross-sections for the places with limited data. This study reveals that Manning’s roughness coefficient is one of the most important parameters for calibration and validation studies. The validation results of the MIKE 11HD model revealed a close agreement between the observed and simulated water levels as revealed from the values of goodness of fit criteria. The study demonstrates the usefulness of using the HEC-GEORAS to derive river cross-sections for use in hydrodynamic modelling studies. The results of the study are quite useful for the Aras basin, where as the methodology is quite broad based and useful as well for other river basins with inadequate quantum of measured river cross section data.

7. REFERENCES

1. Abbott MB (1979), Computational hydraulics. Pitman, London 2. Abbott MB, Refsgaard JC (eds). (1996), “ Distributed hydrological modelling”, Kluwer AcademicPress, Dordrecht, vol 321. 3. Bagirov. Z.A. and Bravarnik, S.E. (1985), “Water-management and power use of the Araks River”, Power Technology and Engineering (formerly Hydrotechnical Construction), 19.1: 35-40. 4. Barr. T. (2002), “Application of Tools for Hydraulic Power Point Presentation”, 105-Upper Gotvand Hydroelectric Power Project Feasibility Study, 1996.Reservior Operation Flood,14p. 5. Beavers, M., (1994), “Floodplain determination using HEC-2 and Geographical Information System”, Masters thesis.Department of Civil Engineering.University of Texas at Austin , Austin.110p. 6. Chowdhury S, Kjeld J. (2002), “ Simulation of coastal flooding with MIKE 11 and HEC-UNET”, Solutions to coastal disasters, pp. 205–214. 7. DHI, MIKE 11 reference an users manuals, Danish Hydraulic Institute (2000). 8. DHI. 1997. MIKE 11 GIS Reference and User Manual. Danish Hydraulic Institute: Horsholm, Denmark. 9. DHI. 2000. MIKE 21 User Guide. Danish Hydraulic Institute: Horsholm, 10. Patro, S., Chatterjee, C., Singh, R., & Raghuwanshi, N. S. (2009), “Hydrodynamic modelling of a large flood‐prone river system in India with limited data”, Hydrological Processes, 23.19: 2774-2791. 11. Pramanik, N., Panda, R. K., & Sen, D. (2010), “One dimensional hydrodynamic modeling of river flow using DEM extracted river cross-sections”, Water Resources Management 24.5: 835-852. 12. Renyi L, Nan L. (2002), “Flood area and damage estimation in Zhejiang, China”, Journal of environmental management 66.1 : 1-8. 13. Tate EC. (1999), “Floodplain mapping using HEC-RAS and ArcView GIS”. MSc thesis, Center for Research in Water Resources, University of Texas, Austin. 14. Thakur, P. K., Aggarwal, S., Aggarwal, S. P., & Jain, S. K. (2016), “One-dimensional hydrodynamic modeling of GLOF and impact on hydropower projects in Dhauliganga River using remote sensing and GIS applications”, Natural Hazards, 1-19. 15. Werner M. (2001), “Impact of grid size in GIS based flood extent mapping using a 1-D flow model”, Physics and Chemistry of the Earth, Part B: Hydrology, Oceans and Atmosphere 26: 517–522.

8