1. DEVELOPMENT OF A CELL BASED MODEL FOR STREAM FLOW PREDICTION IN UNGAUGED BASINS USING GIS DATA P B Hunukumbura & S B Weerakoon Department of Civil Engineering, University of Peradeniya, Peradeniya 10 th November 2005

2. Overview Introduction Objectives of the study Model development Case study Results Conclusions

3. Introduction Prediction in Ungauged Basins (PUB) “the prediction of the hydrologic response of ungauged or poorly gauged basins using all data sets other than past observations of the particular hydrologic response that is being predicted” (Sivapalan, et al. 2003) The International Association for Hydrological Sciences (IAHS) has announced a decade (from year 2003 to 2012) for research focused on PUB

4. PUB - Relevance to Sri Lanka Sri Lanka has 103 river basins and their sizes vary from 10 km 2 to 10450 km 2 Only 35 stream flow measuring stations are available within these basins WR development potentials with respect to water supply and hydropower exist mostly in ungauged river basins Many basins are totally ungauged

5. To develop a model which capable of predicting stream flow in ungauged basins using basin’s rainfall, topography, land use and stream network data. Objective of the study

6. Model development The basin was divided in to several grids to represent the basin spatial properties. The total time taken to flow water from each grid cell to the basin outlet was calculated. The S-curve for the basin was developed using the total travel time distribution of the basin and hence the unit hydrograph of the basin was obtained. The direct runoff hydrograph for a given effective rainfall time series was obtained using the unit hydrograph of the basin.

7. Slope grid Slope grid of a basin represents the maximum slope of each grid cell to its surrounding cells Flow direction grid Flow direction grid represents the maximum slope direction of each cell to its surrounding cells. Flow accumulation grid Number of cells that flow to the particular grid cell

8. Computer model - “CellBasin" A computer model with a Graphical User Interface (GUI) was developed using Visual Basic programming language Input file path File select dialog box

9. Case study The experimental Upper Kotmale basin The Upper Kotmale basin is the upper most basin of the Mahaweli River and this was setup as the experimental basin for PUB studies The model was applied to the experimental Upper Kotmale basin and predictions were compared with the observed stream flow data at basin outlet, Thalawakelle

10. The Upper Kotmale basin Basin area = 304 km2 Elevation= 1200m to 2500m Annual rainfall= 2200 mm to 2600 mm

11. Nuwara Eliya Horton Thalawakelle

12. GIS data used for the model 90 m SRTM - DEM 90 m Slope grid 90 m Flow accumulation grid 90 m Land use grid 90 m Flow direction grid

13. Results One hour unit hydrograph of the basin

14. Stream flow prediction Present version of the model is not facilitating an algorithm for base flow calculations. After analyzing the available stream flow data in the wet zone of the country, it is found that the flow having 95% probability of excedance is about 9 l/s/km 2 for the wet zone. Therefore, the model performance is compared with constant base flow of 9 l/s/km 2

16. ParameterValue Nash Coefficient0.50 Root Mean Squire Error4.48 Observed Mean7.84 Predicted Mean7.64 Observed Standard Deviation 6.34 Predicted Standard Deviation 7.29

17. Conclusions A cell based computer model, which is capable of predicting stream flows in ungauged basin, was developed using Visual Basic programming language. The model was applied to the mountainous Upper Kotmale basin and the predictions were satisfactory. The model, which requires only the basin rainfall, topographical and land use digital data is a useful tool for stream flow prediction in ungauged basins.

19. Overland flow Overland flow can be modeled using kinematic wave equation by assuming the flow is turbulent (Chow, 1988, Nurunnisa and Yilmaz, 2002). The kinematic wave equation is used to calculate the flow velocity through each grid cell assuming the steady uniform turbulent flow. The overland flow through a grid cell can be assumed as a flow through a wide canal and hence the value of the hydraulic radius ( R) is taken as the flow depth (Y) in calculating the flow velocity through a grid cell. As the final target is to get the unit hydrograph for the basin, the flow depth (Y) is calculated due to constant rainfall rate of 1mm/hr. Considering the continuity and kinematic wave equation, Y (m) can be written as Flow Velocity through a grid cell is ;

20. Land use type Manning’s roughness coefficient Water body0.01 Build up land0.011 Tea0.17 Forest0.8 Grass land0.24 Other crop land0.17 Source: USDA, 1986, Nurunnisa and Yilmaz, 2002

21. Canal flow Measured cross section details and the condition of the canals are used to estimate a reasonable value for the hydraulic radius and roughness coefficient of the particular section of the canal. In this regard, starting point and ending point of the canal are defined by the cell value of the FAccGrid. For the canal section, If Starting point FAccGrid value < cell value of FAccGrid < End point FAccGrid value, Where, MnReach and HRadius represent Manning’s roughness coefficient and hydraulic radius of the canal respectively

22. Flow accumulation value Hydraulic radius Manning’s roughness coefficient 0 - 1500Effective flow depth According to Table 1 1500 - 6000 0.32 0.04 6000 - 120000.71 0.05 >12000 0.90.06 Source: USDA, 1986, Connecticut, 2001

23. Direct runoff hydrograph of for the rainfall event 1

24. Direct runoff hydrograph of for the rainfall event 2