APPLICATION OF MIKE11 MODEL FOR THE SIMULATION OF SNOWMELT RUNOFF IN YUVACIK DAM BASIN, TURKEY

Transcription

1 APPLICATION OF MIKE11 MODEL FOR THE SIMULATION OF SNOWMELT RUNOFF IN YUVACIK DAM BASIN, TURKEY Fatih KESKİN State Hydraulic Works, Planning and Investigation Department, Yucetepe-Ankara/Turkey, Aynur ŞENSOY A. Arda ŞORMAN AnadoluUniversity, Department of Civil Engineering, Eskişehir-TURKEY A. Ünal ŞORMAN Middle East Technical University, Department of Civil Engineering, Ankara-TURKEY ABSTRACT Modeling and predicting daily discharges from snowmelt is important in mountainous basins where snow is the main component of the water balance and is especially effective in regions prone to seasonal flooding. In this study, Mike11 model is calibrated and validated to simulate the runoff from snowmelt in a semi-distributed manner. The model is applied to Yuvacık Basin to simulate the rain/snowmelt runoff for the period in between the years 2001 and The main objective of the reservoir operational strategy is to provide domestic water supply to the Izmit Great Municipality and to control the downstream flooding. Therefore the effective modeling of the snowmelt runoff is needed for the optimum operation of the reservoir to provide the demand of the city when the supply is insufficient. The basin is divided into three sub-basins where each sub-basin is represented by its own characteristics and the sub-basins are divided into elevation zones for snowmelt computations. The model is calibrated with snowmelt and rain on snow events and verified with the daily events which are not used in the calibration process. The resulting coefficient of determination (R 2 ) based on the average daily discharge values is bigger than 0.7 in most of the events. The modeling studies give promising results for the computation of runoff due mainly to snow during different seasons of a year. The results of this study would also be helpful for the TEFER project model calibration study which is the on going issue in DSI agenda. Keywords: Mike11, Snowmelt, Runoff, Yuvacık, Hydrological model

2 RIVER BASIN FLOOD MANAGEMENT 473 INTRODUCTION Snow cover is an important phenomenon in hydrology, therefore modeling the snow accumulation and melting becomes an important issue in places where snowmelt significantly contributes to runoff and have significant effect on water balance. Runoff processes are related with vegetation, soils and especially amount of soil moisture in various depths. Runoff generation from snowmelt is related not only to snowmelt mechanisms but also to hydrological processes. Snow-water equivalent (SWE) can be used to define the snowpack and snowmelt rate. A case study with MIKE11 model is used to illustrate formation of the runoff due to snowmelt from snow-water equivalent and rain on snow events. The model is applied to Yuvacık Basin to simulate the rain/snowmelt runoff for the period in between the years 2001 and The biggest inflow to the dam reservoir comes from the events of snowmelt and rain on snow. The snow starts to accumulate at the mid of December, and almost all of the the snow reserve melts at the end of March except from the locations at the higher elevations of the basin. The dam is planned to supply domestic water to Kocaeli city and to prevent the city from flooding. The dam is operated by Thames Water Türkiye (TWT), Turkish branch of Thames Water which is the biggest water company in United Kingdom. The optimum operation of the dam reservoir is required in terms of effective usage of water, the optimum operation methodology is especially important for the periods of high (flood) and low (drought) flow conditions as the low water level was observed in The modeling studies can be considered as a part of decision support system and forms an example for other future studies. STUDY AREA The study area is the Yuvacık Dam Basin which is in the eastern part of Marmara Region of Turkey at about º41 North and 29º29-30º08 East and located 12 km south of Izmit city. The drainage area of Yuvacık Dam basin is about km 2. In the north of the basin there exists a dam called Yuvacık or with its old name Kirazdere (DSI Report, 1983). Yuvacık dam was built in 1999 and subjected to one of the biggest earthquakes in Turkey. The main objective of the reservoir operational strategy is to provide domestic water supply to Izmit Great Municipality and to control the downstream flooding. Therefore, the effective modeling of the snowmelt runoff is needed for the optimum operation of the dam reservoir to provide the demand of the city when the supply is insufficient. Yuvacık dam basin is divided into four sub-basins named as Kirazdere, Kazandere, Serindere and Contributing catchment where Serindere is the largest and

3 474 INTERNATIONAL CONGRESS ON RIVER BASIN MANAGEMENT Kazandere is the smallest catchment. The location of the basin and its catchments is shown in Figure 1. The catchment that contains the lake is named as Contributing catchment, the others are Kirazdere, Kazandere and Serindere sub-basins from left to right. The discharge measurements are being done in stations FP1-FP4, where FP1, FP2 and FP3 flow plants provide streamflow measurements for Kirazdere, Kazandere and Serindere catchments, respectively. FPs and RGs in the legend of Figure 1 represents the flow plant stations and the raingauge stations. The M1-M3 stations are the newly installed mobile stations which will be removed after analyzing the spatial distribution of rainfall over the basin. The elevations of each station is given in Table 1. Figure 1: Yuvacık dam basin, sub-basins and the station network map AVAILABLE DATA The Digital Elevation Model (DEM) is produced from 1/25000 topographic maps. The boundaries of the sub-basins are formed by the ArcHydro Module of ArcGIS ver 9.1(ESRI, 2005) and the locations of the discharge measurement stations are used as sub-basin outlets.

4 RIVER BASIN FLOOD MANAGEMENT 475 There are 15 meteorological stations and 4 discharge measurement stations in the basin. Some of the stations are old stations and some of them are newly installed in the basin. The old meteorological station network was not adequate to represent the spatial distribution of the rainfall over the basin because the locations of these stations are very close to each other and the elevations of the available stations are limited to lower altitudes of the basin. Thus, new stations were recommended to be installed to measure temperature, rainfall and snow depth with new techniques and equipments. The elevations of old rainfall measurement stations vary from 173 meters to 520 meters, the new elevation range of the upgraded network changes between 76 meters to 1487 meters which indicates a better representation of the spatial distribution of both rainfall and snowfall over the whole basin. All of the available data from these stations were used in the modeling studies of the basin. The oldest meteorological station in the area is the Kocaeli station (in operation since 1930 s) which is operated by Turkish State Meteorological Service (DMI). The data that are being recorded in this station are temperature, rainfall and snow depth. State Hydraulic Works (DSI) operates a meteorological station in Hacıosman location (HO) which is in the basin and represents the basin. Rainfall and snow depth data have been recorded since 1980s at this station. The discharge measurements have been done since 2001 at different locations in the basin (FP1 to FP4) in 5 minutes intervals, one of which is just at the entrance site of the reservoir lake located for the measurement of the lake level, and the other three are at the outlets of three sub-basins. Rainfall data was recorded since 2001 from different stations by TWT (RG1 to RG6) in 5 minutes intervals. Since the stations(rg1 to RG6) are recording stations and recording the data at 5 minutes time intervals, these data are converted to daily cumulative values which are used in the modeling. Since the evaporation data were not available for the sites, the evaporation data of the nearby station were used with a proper elevation adjustment. Table 1: Hydrometeorological stations elevations Station Name FP1 FP2 FP3 FP4 RG1 RG2 RG3 RG4 RG5 RG6 Elevation (m) Station Name RG7 RG8 RG9 RG10 M1 M2 M3 HO KE Elevation (m) MIKE 11 (NAM) MODEL DESCRIPTION The MIKE 11 Rainfall-Runoff (RR) NAM model is a continuous precipitationrunoff model of the deterministic, lumped, conceptual type. It includes simulation of snowmelt in different altitude zones within the catchment. NAM is the abbreviation

5 476 INTERNATIONAL CONGRESS ON RIVER BASIN MANAGEMENT of the Danish ʺNedbør-Afstrømnings-Modelʺ, meaning precipitation-runoff-model. NAM represents various components of the rainfall-runoff process by continuously accounting for the water content in four different and mutually interrelated storages. These storages are Snow, Surface, Root Zone and Ground Water Storages. (DHI, 2004) Using rainfall, potential evaporation and temperature as input, the model simulates: Snow accumulation and melting, interception, evapotranspiration, overland flow, groundwater recharge and baseflow. PARAMETERS NAM Model-specific parameters involve: Surface, Root-zone, Snow melt data, Ground water data, Initial Conditions, Irrigated Area. The basic parameters and their short description is given in Table 2. Table 2: NAM parameters and their definitions Umax Lmax CQof CKIF TOF TIF TG CKBF CK1,2 Maximum contents of surface storage Maximum contents of root zone storage Overland flow coefficient Time constant for interflow Root zone threshold value for overland flow Root zone threshold value for inter flow Root zone threshold value for recharge Time constant for routing base flow Time constant for routing overland flow In case of using snowmelt component, NAM module needs additional snowmelt parameters which are given in Table 3. The model utilizes the temperature index model for the computation of daily snowmelt. The Snow melt module uses a temperature input time series, usually mean daily temperature. If one shoose the simple snowmelt, the Csnow parameter is provided as one value, but in case of extended snowmelt module, Csnow may provided to be monthly values. Table 3 : Extended Snow Parameters Csnow, Constant Degree-day coefficient (T0), Base Temperature snow/rain Radiation Coefficient if available Elevation Zones Reference level for temperature station Dry temperature lapse rate ( per 100 m) Wet temperature lapse rate( per 100 m) Reference level for precipitation station Correction of precipitation(per 100m)

6 RIVER BASIN FLOOD MANAGEMENT 477 The initial conditions required by the NAM model consist of the initial water contents in the surface and root zone storages together with initial values of overland flow, interflow and baseflow. If the snow module is included, the initial value of the snow storage should be specified. The initial condition statement is the most important part of the model. Parameters within the model depend on the initial conditions if the simulation period is less than a year (DHI, 2004). If a simulation commences at the end of a dry period, it is often sufficient to set all initial values to zero, except the water content in the root zone and the baseflow. Improved estimates of the initial conditions may be obtained from a previous simulation, covering several years, by noting the appropriate moisture contents of the root zone and baseflow at the same time of the year as the new simulation will start. CALIBRATION AND VERIFICATION OF THE MODEL All of the available hydrological and meteorological data (Rainfall, temparature, discharge, evaporation) of the basin were analyzed and used in the calibration and validation periods of the model. The model was simulated in daily time steps in a semi distributed manner. The hydrological and meteorological data were provided in daily time steps except that the evaporation data which were used as monthly values in the model. Simulation Periods The greatest peaks were recorded in winter and spring months due to snowmelt beside the rainfall, therefore the study focuses on the period starting from december upto may, where snowmelt is very effective in runoff generation chain. The simulation period is decided to be according to the available precipitation and water level data. There were data discontinuities for the flow plants of Kazandere and Serindere subbasins between the period because of the missing records due to the inconsistencies in water level measurements. Therefore, there are more storm events observed in Kirazdere than that of the other catchments. Rainfall Distribution Available rainfall data of all the stations are used to determine the daily areal mean rainfall of the sub-basins. The available rainfall data are recorded by TWT from the stations RG1-RG6 for the period of As it was mentioned before, the old meteorological station network (RG1-RG6) was not adequate to represent the spatial distribution of the rainfall over the basin because the locations of these stations are very close to each other and the elevations of the available stations are limited to lower altitudes of the basin.. HO (900 m) and KE (76m) stations are located at a relatively higher and lower elevations, respectively, therefore these two stations were

7 478 INTERNATIONAL CONGRESS ON RIVER BASIN MANAGEMENT used to represent the higher and lower altitudes of the basin for the period of Elevation Bands The catchments are divided into three elevation zones to represent the high elevation difference in the basin. The average altitude of each elevation zone is an important factor since some of the model variables are assigned to an average altitude of an elevation zone. The mean elevation can be used as a representative elevation of each zone, but this will not show the areal contribution of each elevation. The hypsometric mean altitude is more representative for an average elevation by considering the area factor in it, so the hypsometric mean elevation is used for the average altitude of each elevation zone. Therefore, the hypsometric curve of the each elevation zone is derived from the DEM of the basin and the elevation which corresponds to 50% area is selected as hypsometric mean elevation. The representation of each elevation band within each sub-basin is carried within the GIS environment and shown in Figure 3. Snow Water Equivalant(SWE) Figure 3 : Snow Elevation Zones of the Sub-Basins SWE is defined as the amount of water stored within the snowpack. It is a common snowpack measurement. The SWE determines the amount of water in the snowpack, and hence the amount available for runoff in the winter. Field studies is important to define the SWE in a snow cover area. In the field study, snow surveys are done at designated locations throughout the winter to determine snow depth and vertical density. These studies enable to identify the possible SWE in the snow so that it can be used in predicting runoff from snowmelt.

8 RIVER BASIN FLOOD MANAGEMENT 479 As mentioned before snowmelt is very effective in rainfall-runoff process. The snow water equivalant for each elevation band must be specified such that the effect and amount of snowmelt on runoff can be shown correctly. No measurement was done to analyse the relation between the rainfall and snow depth, nor for the relation between the temperature and snow depth. Thus, defining the snow depth and related snow water equivalant value for each elevation band was very cumbersome issue in the basin. The only snow depth measurement was done in HO station which is operated by DSI for the period of However, these measurements were not included the snow water equivalant values. Although, the snow water equivalant values were not available at the beginning of each event, the measured snow depth data were very valuable for the study. The snow data were very scarce for the entire basin for the period of HO is the only station where proper snow depth measurements were carried out in the basin between and this station data were used during the simulation studies. Manual snow density and automatic snow depth measurements have been carried out in the basin since november It is observed that the measured snow density changes from %14 to %45 from December to March, respectively, as well, values change by location and elevation. After the installation of the new stations, snow depth measurements are being done automatically by Thames Water Turkiye for the 2006 water year. Snow depth measurements at different elavtions of the basin were analyzed and compared with each other in Figure 2. Snow Depth( ) RG7 TEPECİK RG8 AYTEPE RG9 KARTEPE RG10 ÇİLEKLİ Hacıosman Snow Depth (m) Ara Ara Oca Oca Şub Şub Mar Mar Nis-06 Figure 2 : Snow depth values in the stations

9 480 INTERNATIONAL CONGRESS ON RIVER BASIN MANAGEMENT Lapse Rate Correction The lapse rate correction approach is a very simple but powerful way of adjustment in which the temperature and the precipitation are assumed to vary linearly with the altitude. The only input data required are the average altitude of the various zones, a reference altitude of the time series, and the lapse rate values. The temperature lapse rate values are known to be quite variable, ranging from high values under dry conditions to lower values under wet conditions. Hence, in some models (e.g.mike 11) it is possible to specify two different temperature lapse rate values that are to be used during dry and wet weather conditions, respectively. The model applies the wet lapse rate during days with precipitation larger than 10 mm and the dry lapse rate during rest of the time (DHI, 2004). Generally, wet lapse rate is less than the dry lapse rate, because, as water condenses it releases heat, so the temperature decrease is less. In literature the lapse rate is between -0.5 and -0.7 o C per 100 m elevation increase. The average of temperature lapse rate is calculated using the observed temperature values for the period of January 06, 2006 to April 24, An analysis showed that the lapse rate between the stations is negative with the elevation increase as expected expect for the two stations. The overall average of temperature lapse rate is calculated as 0.61 o C per 100 meter. The Dry Lapse rate is taken as o C/100 meter and the wet lapse rate is taken as -0.4 o C/100 meter for the basin. Figure 4 presents observed temperature data together with calculated temperature data using temperature lapse rate. Çilekli(-0.6 Lapse rate) Real Çilekli Temp. Value Comparision of Temperature of Çilekli Station with -0.6 Lapse Rate Ara Oca Oca Şub Mar Mar Nis May Figure 4 : Comparision of the real observed temperature and the transformed temperature with lapse rate in RG8 station

10 RIVER BASIN FLOOD MANAGEMENT 481 MODEL SIMULATION RESULTS Several tries during the model simulations showed that the most effective parameters are Lmax, Umax, CKOF, CKIF which also define the base flow in the basin. It was found that six parameters namely Lmax, Umax, CKOF,CKIF, TG, CKBF can be taken as constant allthroughout the calibration process. On the other hand, these values are differerent from sub-basin to sub-basin, as expected. These parameters are similar for Kazandere and Kirazdere, but they are different in Serindere. The other three parameters which are CK1,2, TOF, TIF vary from period to period, however, they do not have too much influence on the total volume in the simulations. The resulting parameters from the calibration processes are given in Table 4., Table 4 : The parameter values resulting from the simulation and default values Parameter Unit Default Values in the manual Recommended Values for Kirazdere Recommended Values for Kazandere Recommended Values for Serindere U max mm L max mm C KQF C KIF hours CK 1,2 hours T OF T IF T G C KBF hours Kirazdere - Kazandere - Serindere Catchments - Degree Day Coefficient Degree Day Coefficient (mm/day/ o C) December January February March April Months Figure 6 : Monthly Degree Day Coefficient

11 482 INTERNATIONAL CONGRESS ON RIVER BASIN MANAGEMENT Another important parameter is the degree-day coefficient which is defined for each month in a year. This variation reflects in a conceptual way the seasonal variation of the incoming short wave radiation and the variation in the albedo of the snow surface during the snow season. As it can be seen from the Figure 6, the degree-day coefficient start to increase at the end of the year and takes its maximum value in April. Validation of Results The calibrated parameters should be validated with another set of data which was not used in the calibration. For Kirazdere sub-basin, , , and water year periods were used for the calibration and water year was used for the validation. Temp (ºC) SWE (mm) HO (mm) MIKE SWE Figure 7 : Temperature and SWE graph KIRAZDERE, Observed RunOff [m^3/s] KIRAZDERE, Simulated RunOff [m^3/s] C:\Documents and Settings\engin\Desktop\Y-deneme\RRcalibration\KIRAZDERE dfs Figure 8 : Simulation Graph (Observed(Black) - Simulated(Red))

12 RIVER BASIN FLOOD MANAGEMENT 483 The SWE and temperature values are shown in Figure 7. As the catchments are divided into 3 elevation zones, the average value of SWE is found for the catchment by taking the average SWE of each zone with their areal weights. As seen from in Figure 7, simulated SWE values have the same trend with observed SWE of HO station. An example of the simulation result is shown in Figure 8. As seen from the graphs, the obtained results showed that the model can be used for simulation of rainfall-runoff and especially for the snowmelt-runoff in the basin. Model efficiency is tested with visual and statistical evaluations. Nash and Sutcliffe model efficiency (NSE) statistical criteria and coefficient of determination (R 2 ) which are widely used for the model performance evaluation in hydrology is used for the evaluation of the model. Moreover, the results were also analyzed in terms of the percent volume and peak error. Statistical analysis yields high goodness of fit for Kirazdere sub-basin the possible reason is the availability of appropriate modeling data. The model efficiencies are higher than 0.7 at least for half of the events for Kazandere and Serindere. The model efficiency reduces with low flows which is the main issue for Kazandere sub-basin. Either the percent peak and volume percent difference or model efficiency is in the acceptable ranges for almost all simulations. Parameters that are found in the calibration process are applicable in the validation process. So these values can be used in any hydrological simulation within the basin. CONCLUSION Snowpack accumulation and snowmelt are found to have a major importance on the inflow of Yuvacık dam reservoir. Air temperature is a critical climate factor to distinguish between the snowfall and the rainfall, and to determine the snowmelt rate. Mike 11 model incorporated the climate factors, land use and soil properties to estimate the streamflow, and they have represented the internal hydrologic dynamics and the impact of climate factors on them. This study uses Mike 11 to explore the hydrologic processes in situations where inflows and peaks are significantly contributed by snowmelt. The performance of the hydrological model shows that the simulated streamflow reflects the variation in temperature and precipitation as well as the moisture interaction between the surface, subsurface and the groundwater storages. Comparison of the results from simulation and observation indicates that the model can reproduce well the observed inflow starting time, peak and the time base. Statistical analysis revealed that the model quantitatively matches the historical data. The model in its present form provides a daily prediction of streamflow on a daily basis if

13 484 INTERNATIONAL CONGRESS ON RIVER BASIN MANAGEMENT rainfall and temperature forecast values are input to the model. If the model is used to make precise predictions of flood events for a short term, hourly variations in temperature and precipitation should be taken into account. This study is a very useful as an application that can be used for Decision Support tool for dam reservoir operation. The same model is being used in Flood Forecasting Center in DSI. The results of this study would also be helpful for the Turkey Emergency Flood and Earthquake Recovery (TEFER) project model calibration study in the different part of the country which is on going issue in DSI agenda. ACKNOWLEDGEMENTS The authors would like to thank TWT personnel, especially to Mr. Hasan Akdemir, Mr. Tolga Gezgin, Mr. Sinan Çelebci and Mr. Türker Akgün, who provided data used in this study and organized the site trips to Yuvacık Basin. Moreover, sincere thanks are extended to Government Organizations: General Directorate of State Hydraulic Works (DSI) for Hacıosman gage data and State Meteorological Service (DMI) for Kocaeli gage data. Also sincere thanks are extended to Danish Hydraulic Institute (DHI) from Denmark for supplying Mike11 model program to Middle East Technical University (METU) for model studies. REFERENCES DHI (2004): MIKE 11 User & Reference Manual, Danish Hydraulic Institute, Denmark. DSI Report (1983) Izmit-Kirazdere Projesi, Kirazdere Barajı Muhendislik Hidrolojisi Planlama Raporu, Bursa 1983 Zengin M., Hizal A., Karakas A., Serengil Y., Tugrul D.,Ercan M. (2005), Planning of the Renewable Natural Resources of the Izmit-Yuvacık Watershed for Water Production(in terms of quality, amount and regime), Poplar and Fast Growing Forest Trees Research Institute, Izmit, Turkey.