Sediment Yield Assessment and Mitigation Measures in Finchaa Watershed, Ethiopia

1. Comparisons of SWAT with other models

   For this study the program SWAT was selected due to its continuous time scale, computational efficiency, its ability to simulate long-term impacts, its applicability to large-scale catchment, distributed spatial handling of parameters and integration of multiple processes such as climate, hydrology, nutrient and pesticide, erosion, land cover, management practices, channel processes, and processes in water bodies has an important tool for watershed scale studies.

   The model was applied for LULCC impact assessment in different parts of the world and also in Ethiopia. [10], compared SWAT with several otherwatershed-scale models. In the study, they reported that the Dynamic Watershed Simulation Model (DWSM), Hydrologic Simulation Program-Fortran (HSPF) model. SWAT and other models have hydrology, sediment and chemical routines applicable to watershed-scale catchments and concluded that SWAT is a promising model for continuous simulations in predominantly agricultural watersheds [11]. They found that

   SWAT and HSPF could predict yearly flow volumes and pollutant losses, were adequate for monthly predictions except for months having extreme storm events and hydrologic conditions and were poor in simulating daily extreme flow events. In Contrast, DWSM reasonably predicted distributed flow hydrographs and concentration or discharge graphs of sediment and chemicals at small time intervals. [12] Found that the average daily flow, sediment loads, and nutrient loads simulated by SWAT were closer than HSPF to measured values collected at five sites during both the calibration and verification periods for the upper north Bosque River watersheds in Texas.

   [13], found that both SWAT and the MIKE-SHE model simulated the hydrology of Belgiums Jeker River basin in an acceptable way. However, MIKE-SHE predicted the overall variation of river flows slightly better. [14], found that SWAT estimated flow more accurately than the Soil Moisture Distribution and Routing (SMDR) model that SWAT was also more accurate on a seasonal basis.

2. Benefits of SWAT Model Approach

   • Watersheds with no monitoring data (e.g., stream gage or water quality data) can be modeled.

     • The relative impact of alternative input data (e.g. changes in management practices, climate, vegetation, or land use) on water quality or another variable of interest can be quantified.

     • The model uses readily available inputs.

     • SWAT is computationally efficient. The Simulation of very large basins management strategies can be performed without excessive investment. The model enables users to study long-term impacts.

     • SWAT explicitly incorporates elevation or orographic effects on precipitation and temperature.

     • It is a continuous time or long term yield model able to simulate long term impacts of land use, land management practices and build-up of pollutants.

     • SWAT has daily data weather simulation model that generates for rainfall, solar radiation, relative humidity, wind speed and temperature from the average monthly variables for the data provides a useful tool to fill in gaps in daily data in the observed records.

     • SWAT is designed to use either observed meteorological data or statistically generated meteorology, facilitating the development of long term analysis.

     • SWAT was developed for and has been widely applied to simulation of watersheds in arid regions.

     • SWAT explicitly incorporates routines for agricultural diversions and irrigation.

     • The other advantage of the SWAT model is the ability to build different scenarios.

     • SWAT includes routines designed to address the impacts on flow and pollutant loading of multiple small (or large) farm ponds within a basin.

3. SWAT-CUP

   SWAT-CUP is an interface that was developed for SWAT. SWAT-CUP is designed to integrate various sensitivity analysis, calibration, validation and uncertainty programs for SWAT using different interface. The main function of an interface is to provide a link between the input/output of a calibration program and the model.Using this generic interface; any calibration, validation/uncertainty or sensitivity program can easily be linked to SWAT.

   The recently developed SWAT-CUP interfaced program for calibration and uncertainty analysis procedures [15] also made the SWAT model more attractive for this study. SWAT-CUP is linked to five different algorithms such as: Sequential Uncertainty FItting (SUFI-2)[15] Generalized Likelihood Uncertainty Estimation (GLUE) [16], Parameter Solution (ParaSol) [17], Particle swarm optimization (PSO) [18], and Markov Chain Monte Carlo (MCMC) [19], procedures to SWAT.

   • SUFI2 [15]: Sequential Uncertainty Fitting Ver. 2, the parameter uncertainty in driving variables (e.g., rainfall), conceptual model, parameters, and measured data.

   • GLUE [16]: Generalized Likelihood Uncertainty Estimation is based on the estimation of the weights or probabilities associated with different parameter sets, based on the use of a subjective likelihood measure to derive a posterior probability function, which is subsequently used to derive the predictive probability of the output variables.

   • Parasol [17]: Parameter Solution method aggregates objective functions into a global optimization criterion and then minimizes these objective functions or a global optimization criterion using the SCE-UA (Shuffled Complex Evolutionalgorithm, which is a global search algorism for minimization of a single function, were utilized in the calibration process.

   • MCMC: Markov Chain Monte Carlo generates samples from a random walk which adapts to the posterior distribution [19]. This simple technique from this class is the Metropolis Hasting algorithm [20].

   Various SWAT parameters for estimation discharge were estimated using the SUFI-2 program [15]. In SUFI-2, parameter uncertainty accounts for all sources of uncertainties such as uncertainty in driving variables (e.g., rainfall), conceptual model, parameters, and measured data. Uncertainty is defined as discrepancy between observed and simulated variables in SUFI-2 where it is counted by variation between them.

   The SUFI-2 was the most suitable way to find the SWAT Uncertainty under the condition that the parameter range. The Goodness of fit in SUFI-2 is expressed by the 95PPU band, it cannot be compared with observation signals using the traditional indices such as R2, Nash-Sutcliffe (NS). For this reason two measures referred to as the P-factor and the R-factor [15], the P-factor is the percentage of the measured data bracketed by the 95PPU. The R-factor, on the other hand, is a measure of the quality of calibration and indicates the thickness of the 95PPU.As all forms of uncertainties are reflected in the measurements (e.g., discharge), the

   parameter uncertainties generating the 95PPUaccount for all uncertainties.

4. Sensitivity analysis

   Sensitivity analysis determines the sensitivity of the input parameters by comparing the output variance due to the input variability. The sensitivity analysis was carried out to identify the ensitive parameters of the SWAT model. The sensitivity analysis is done by varying parameters value and checking how the model reacts. If small change on a given parameter value results on a remarkable change on the model output, the parameter is said to be sensitive to the model.

5. Model calibration and validation

   In hydrologic simulation there are two main exercises that must be successfully achieved before using a model. These are calibration and validation of the models.

   1. Calibration

      Calibration is an intensive exercise used to establish the most suitable parameter in modeling studies and an iterative process that compares simulated and observed data of interest (typically streamflow data) through parameter evaluation. The exercise is vital because reliable values for some parameters can only be found by calibration [21].

   2. Validation

Model validation is the process of representing that a given site specific model is capable of making sufficiently accurate simulation. The degree of accuracy of parameter estimates was assessed by applying the model to different data set that was not used for calibration. This implies the application of the model without changing the parameter values that were set during calibration [22].The model is validated if its accuracy and predictive capability in the validation period have been proven to lie within acceptable limits [22].

3.8. Assessment of model performance

The performance of SWAT is evaluated using statistical measures to determine the quality and reliability of predictions when compared to observed values.During calibration and validation of a hydrological model it is necessary to assess the performance of the model. This is done by statistically comparing the model output and observed values using various statistical measures. These measures include the coefficient of determination (R2) and Nash-Sutcliffe Efficiency (NS). The range of values for R2 is

1.0 (best) to 0.0(poor). The R2 coefficient measures the fraction of the variation in the measured data that is replicated in the simulated model results. Nash-Sutcliffe simulation efficiency (NS) indicates the degree of fitness of the observed and simulated plots with the 1:1 line. The statistical index of modelling Nash-Sutcliffe Efficiency (NS) values range from 1.0(best) to negative infinity. NS is a strictertest of performance than R2 and is never larger than R2.

1. RESULTS AND DISCUSSIONS

   1. Evaluation of sediment yield due to land use/ land cover changes to hydropower reservoir

      Evaluation of the impacts of LULCC on sediment yield was one of the most significant parts of this study. The major land use changes that affect stream flow and sediment yield in this study catchments were changes to farmland, grassland and other Agriculture land use types. One of the most important things of the study was to evaluate the impacts of LULCCs on sediment yield to hydropower reservoir. The evaluation was done in terms of the impacts of LULCCs on the variation of LULC types and the variations caused depending on the LULC types on the major components of sediment yield including surface runoff and groundwater flow.

      To assess the changes in the contribution of the components of the stream flow due to the LULCC, analysis were made on the surface runoff and sediment yield, ground water flow. The surface runoff and sediment yield has increased and the ground water flow has decreased during the expansion of agricultural land over the forest, the surface runoff and sediment yield decreased and the ground water flow has increased during the expansion of forests. Conversion to Agriculture has the largest impact on the yearly surface run off and sediment yield while has the smallest. The expansion of agricultural land use type results the increasing of sediment yield and reduction of water infiltrating in to the ground. Generally, the Hydrological investigation with respect to the LULCCs within Finchaa watershed showed that the flow characteristics/ water balance components have changed through different LULCCs of the study.

   2. Sediment yield calibration and validation

      Initially one year data was taken as the warming period and the rest of the period was used for the model calibration.The coefficient of determination R2 and the Nash- Sutcliffe equation has been applied for model testing between simulated and observed flows and calculated on monthly basis was 0.74 and 0.70 respectively. The amount to which all uncertainties are accounted for is quantified by a measure of the P-factor, which is known as the percentage of measured data bracketed by the 95% prediction uncertainty. Validation proves the performance of the model for simulated flows in periods different from the calibration periods, but without any further adjustment in the calibrated parameters.The correlation coefficient (R2 = 0.76) and the Nash-Sutcliffe (NS=0.74) shows a good agreement between the observed and simulated values.

      Table 1. Percent changes in annual average water balance components for the Finchaa watershed land use/ land cover change scenarios.

      	Conversion to agriculture	Conversion to forest
      Sur_Q (mm)	0.664	-15.979
      Lat_Q (mm)	-26.265	-25.369
      Gw (shalAq)_Q (mm)	0.973	5.412
      Gw (Deep Aq)_Q (mm)	0.976	5.418
      Total WYLD (mm)	-0.036	-0.160
      Perc (mm)	0.977	5.389
      Total SEDYLD (T/Ha)	0.949	-16.207

   3. Adaptation options to mitigate the adverse impacts of the land use/ land cover changes on the reservoir

   The adaptation of LULC patterns are an essential aspect of minimizing the expected impact of LULCCs at the regional and local scales. LULCC can play an important role in reducing the amount of sediment yield to the reservoir through LULC and forestry activities that can occur through avoiding deforestation. Improved management of grassland over the agricultural land use type was also one type of mitigation measure to improve the LULCC. The adaptation of watershed land use patterns used to mitigate the impact of LULCC on the regions hydrology.

   The LULC based mitigation measures include afforestation of the areas. The afforestation/ reforestation has a function to reduce over land flow and rainfall erosivity. Appropriate soil conservation measures based on suitable afforestation techniques can be highly influential in risk mitigation of soil erosion. Therefore, the adaptation options to be taken to mitigate the adverse impacts of the LULCCs on reservoir would be, if possible to cover the land use type byforest or increase the forest type or grass land type of the watershed area in order to decrease the amount of sediment yield to the reservoir. figure 2 and 3 shows the afforestation areas have reduced the intensity of soil loss rate in the Finchaa watershed by reducing the amount of sediment yield generated and the annual surface runoff to the reservoir.

   Figure 2.Sediment yield of the watershed before and after applying the mitigation measures.

   Figure 3. Surface flow of the watershed before and after applying the mitigation measures.

2. CONCLUSIONS

Land use/ land cover change in Finchaa watershed should be controlled to reduce deforestation, which increases the frequencies and concentrations of sediment in the reservoir. Re- afforestation must be introduced within the catchment area of Finchaa watershed which tends to increase filtration of rainfall water and reduce surface runoff which subsequently reduces erosion within the catchment. In addition, solutions to the problems of LLCC should include improving the productivity of the agricultural sector through technical intervention.Reservoir sediment management, especially in sediment rich areas of Ethiopia and other countries around the world is becoming more and more a major problem for the hydro projects.

In this study SUFI-2 was used for model calibration and validation, it could perform uncertainty analysis and calibrate the model for more number of parameters. The model simulations considered only LULCC scenarios assuming one variable at a time and keeping other values unchanged. To analyze the impacts of LULCC on sediment yield to hydropower reservoir sedimentation some changes of land use land cover/options were developed. The options were developed simply to show the potential change of stream flow and sediment yield from the corresponding land use/ land cover change to the hydropower reservoir. The LULCC scenarios developed shows in LULC variables are likely to have significant impacts on the flow volume into reservoirs and the hydrological components were changed. Therefore, decision makers and all concerned stakeholders should plan and implement an integrated watershed development program in advance to alleviate the problem.

REFERENCES

1. Brandon, R. (1998). Mapping Ruaral Land Use and Land Cover Change in Carroll Country, Arkansas: Utilizing Multi- Tempora Landsat Thematic Mapper Satellite Imagery. UniversitynOf Arkansas.

2. Calder, I. R. (1992). Hydrologic effects of land use change, Ed. in Chief D. R. Maidment. Handbook of Hydrology. 13.1-13.50.

3. Sidle, R. C., Ziegler, A. D., and Negishi, J. N. (2006). Erosion processes in steep terrain- truths, myths and uncertainties related to forest management in southeast Asia. Forest Ecology and Management 224(1-2). 199 – 225.

4. Bakker, M., Govers, G., Van Doorn, A., Quetier, F., Chouvardas, D., and Rounsevell, M. (2008). The response of soil erosion and sediment export to land-use change in four areas of a Europe: the importance of landscape pattern. Geomorphology 98. 213 – 226.

5. Tamene, L., Park, S., Dikau, R., and Vlek, P. (2005). Analysis of factors determining sediment yield variability in the highlands of Northern Ethiopia. Geomorphology, 76. 76-91.

6. Arnold, J. G., Allen, P. M., and Bernhardt, G. (1993). A comprehensive surface – groundwater flow model. Journal of Hydrology 142, 47 – 69.

7. Jha, M. K., Gassman, P. W., and Arnold, J. G. (2007). Water quality modelling for the raccoon river watershed using SWAT. American Society of Agricultural and Biological Engineers, 50. 479 – 493.

8. Tripathi, M. P., Panda, R. K., and S., R. N. (2003). Identification and prioritization of critical sub-watersheds for soil conservation management using SWAT model. Bio systems Engineering, Vol.85.

9. Tibebe, D., and Bewket, W. (2010). Surface rRunoff and Soil Erosin Estimation Using the SWAT Model in the Keeta Watershed, Ethiopia. Land Degradation and Developement.

10. Borah, D. K., and Bera., M. (2003). Watershedscale hydrologic and nonpointsource pollution models: Review of mathematical bases. Trans. ASAE 46(6). 15531566.

11. Bicknell, B. R., Imhoff, J. C., Donigian, A. S., and Johanson., R. C. (1997). Hydrological simulation program – FORTRAN (HSPF): User's manual for release 11. EPA600/R97/080. Athens, Ga.: U.S. Environmental Protection Agency.

12. Saleh, A., and Du., B. (2004). Evaluation of SWAT and HSPF within BASINS program for the upper North Bosque River watershed in central Texas. Trans. ASAE 47(4). 10391049.

13. ElNasr, Arnold, A. J., Feyen, J., and Berlamont., J. (2005). Modelling the hydrology of a catchment using a distributed and a semidistributed model. Hydrol. Process. 19(3), 573587.

14. Srinivasan, M. S., Gerard-Marchant, P., Veith, T. L., Gburek, W. J., and Steenhuis., T. S. (2005). Watershed scale modeling of critical source areas of runoff generation and phosphorus transport. J. Am. Water Resour. Assoc. 41, 361-375.

15. Abbaspour, K., M.Vejdani, and Haghighat, S. (2007). SWAT-CUP calibration and uncertainty programs for SWAT. In proc. Int. Congress on Modelling and Simulation (MODSIM'07). 1603 -1609.

16. Beven, K., and Binley, A. (1992). The Future of Distributed Model- Model Calibration and Uncertainty Prediction. Hydrological Processes, 6(3), 279 – 298.

17. Van Griensven, A., and Meixner, T. (2006). Methods to quantify and identify the sources of uncertainty for river basin water quality models. Water science and Technology, 53(1), 51 – 59.

18. Eberhart, R., and Kennedy, J. (1995). A new optimizer using particle swarm theory. Proceedings of the sixth International symposium on Micro Machine and Human Science, Nagoya, Japan, 39 – 43. Piscataway, NJ: IEEE Service Center.

19. Kuczera, G., and Parent, E. (1998). Monte Carlo assessment of parameter uncertainty in conceptual catchment models: the Metropolis algorithm. Journal of Hydrology 211(1-4), 69 – 85.

20. Gelman, S., Carlin, J. B., Stren, H. S., and Rubin, D. B. (1995). Bayesian Data Analysis, Chapman and Hall, New York, USA.

21. Beven, K. J. (1989). Changing ideas in Hydrology. The Case of Physically – Based Models. Journal of Hydrology 105: 157 – 172. Catchments, Tanzania. Proceedings of the 3rd International SWAT conference, Zurich, 2005.

22. Refsgaard, K., and Storm, B. (1996). Construction, calibration and validation of hydrological models, Distributed Hydrological Modelling(eds. M. B. Abott and J. C. Refsgaard), Kluwer Academic Publishers.

23. Ankit Chakravarti, Nitin Joshi and Himanshu Panjiar (2015) Rainfall runoff modeling using artificial neural networks. Indian Journal of Science and Technology, Vol. 8(14), pp 1-7, ISSN 0974-5645.

24. Ankit Chakravarti and M. K. Jain (2014) Experimental investigation and modeling of rainfall runoff process. Indian Journal of Science and Technology,Vol. 7(12), pp 2096-2106, ISSN 0974-5645.