Abstract
The impact of climate change on the water resource potential of the Gibe III watershed, Omo-Gibe Basin, Ethiopia, was investigated using the Soil and Water Assessment Tool (SWAT) model and selected climate models of Coupled Model Intercomparison Project Phase 5 (CMIP5) for future projection. Because the Omo-Gibe Basin in general and the Gibe III watershed in particular was the source of hydropower generation, more work toward updating knowledge of climate change impact on it is required so as to manage the sustained use of the water resource and prevent sedimentation of the reservoir. High-resolution (0.25° × 0.25°) datasets of some general circulation models (GCMs) such as GFDL-ESM2M, MPI-ESM-MR, CSIRO-MK3-6-0, NorESM1-M, and MIROC5 were downloaded for six stations. After calibrating and validating the Soil and Water Assessment Tool (SWAT) model, the impact of climate change was simulated. Accordingly, the annual precipitation was expected to increase by 8.4 and 21.1% during 2050 and 2080, respectively; mean temperature was projected to increase by 1.85 and 2.8 °C in 2050 and 2080, respectively; the stream flow was expected to increase by 55.5 and 81% by 2050 and 2080, respectively, from the base period (1990–2017). The scenario of mean annual sediment yield would increase by 64.5 and 138% by 2050 and 2080, respectively. Therefore, actions toward reducing excess runoff production in the catchment and timely removal of sediment from the reservoir are required.
HIGHLIGHTS
The impact of climate change on the water resource potential of the Gibe III watershed was evaluated. Accordingly, the following results were projected for 2050 and 2080, respectively.
Annual precipitation projected to increase by 8.4 and 21.1%.
The mean temperature is projected to increase by 1.85 and 2.8 °C.
The stream flow will increase by 55.5 and 81%.
The sediment yield is expected to increase by 64.5 and 138%.
INTRODUCTION
Worldwide, climate change causes considerable economic and environmental risks. Increase in concentration of greenhouse gases could nearly double by the year 2100 (IPCC 2007) owing to the historic increase in atmospheric carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), F-gases (fluorinated gases) (HFCs), among others (IPCC 2014). General Circulation Models (GCMs) are the vital resource used to perform climate change experiments globally, regionally, and at very fine scale-ups to identify climate patterns (Trzaska & Schnar 2014). While the resolution of GCM is quite coarse relative to the scale of exposure units in most impact assessments, several physical processes around the globe occur at smaller scales and cannot be properly modeled. Nowadays, GCMs take more and more components into account including sophisticated models of the sea ice, the carbon cycle, ice sheet dynamics, and even atmospheric chemistry (Goosse et al. 2013). Even though climate models are powerful tools for describing regional and local-scale climate conditions, climate variables either for present or future periods obtained directly from simulation of the model outputs are of limited value for impact study as they are coarse in nature (Jakob et al. 2011).
Evaluating the expected impact of climate change is a standard practice to develop adaptation mechanisms for its adverse effect on water resource availability and on agriculture in general, and stream flow and sediment yield of a given catchment in particular (Stocker 2013). With climate change threatening and uncertainty rising to its potential impacts, research is needed to examine how possible changes might affect water resources. The sustainable economic growth and development of Ethiopia is dependent on the proper utilization of agricultural land as well as on hydropower generation. The potential impact of climate change on the hydropower projects, such as flow condition and sedimentation problems, requires up-to-date information. Surface runoff was dominantly increased in the future scenarios and periods from the water balance components. A monthly analysis of precipitation indicated the reduction in April and May with drastic increase in wet seasons, peaking in October rather than in August. Belg season rainfall (March, April, and May) will likely reduce under all future scenarios. Higher precipitation is expected in November, indicating that the wet season will be long, and the dry season rainfall will be reduced or at risk. The future prediction of annual sediment yield showed an increase of above 50 and 100% during the 2050s and 2080s, respectively, for both scenarios. Therefore, integrated watershed management activities that could reduce excess runoff production in the catchment and facilitate infiltration to ground water and stream flow as well as timely removal of sediment from the reservoir are required. As this study was done with station-based climate model selection, further study is recommended applying each of the climate model datasets in all stations around the watershed. Among physically based, spatially distributed hydrological models, the Soil and Water Assessment Tool (SWAT) model has been applied widely (Fontaine et al. 2001; Ficklin et al. 2009; Githui et al. 2009; Setegn 2010; Li et al. 2011; Mengistu & Sorteberg 2012; Shrestha et al. 2012) to evaluate the potential impact of climate change on stream flow and sedimentation. Those who applied the SWAT model in this study area did not give emphasis to analyzing the sediment yield component, which was the main focus here in this study, and therefore the application of SWAT under future climate change scenarios (IPCC 2013) to get the stream flow and sediment yield at Gibe III watershed is very important. Therefore, the objective of this study was to evaluate the impacts of climate change on the stream flow, sediment yield, and water resource potential at the Gibe III watershed using climate models with NASA NEX-GDDP datasets and the SWAT hydrologic model. This article is structured as follows: Introduction; Materials and Methods; Results and Discussion; Conclusion and Recommendations, and finally, a list of References.
MATERIALS AND METHODS
Description of the study area
Location
The catchment area is 34,154 km2 and the estimated long-term average discharge is 438.2 m3/s (EEPCO 2009). According to the Water Works Design and Supervision Enterprise (WWDSE), in association with the South Design and Construction Supervision Enterprise (SDCSE) (2013), the monthly flow ranges from 60 m3/s in March to 1,500 m3/s in August. The Gibe III hydropower station includes a 243-m-high dam, which forms a reservoir with an area of approximately 200 km2, having live storage of approximately 11,750 mm3. Administratively, the reservoir spans 5 zones and 12 woredas, extending from the southern Omo zone downstream to Lake Turkana (UNEP 2013). Climates vary from hot and dry climates in the south to tropical humid highlands including the extreme north and northwest. The rainfall ranges from 1,200 to 1,900 mm and is uni-modal in the northern and central parts of the basin and bimodal in the southern part. The calculated average annual rainfall for the entire Gibe III Basin where the dam is located is 1,426 mm, with 75–80% of the annual precipitation falling between May and September. The mean annual temperatures in the basin vary from 16 °C in the northern highlands to more than 29 °C in the southern lowlands (EEPCO 2009).
The topography of the Omo-Gibe Basin is characterized by its physical variation. The northern two-thirds of the basin has mountainous to hilly terrain cut by deeply incised gorges of the Omo, Gojeb, and Gilgel Gibe River, while the southern one-third of the basin is a flat alluvial plain punctuated by hilly areas. The study watershed has an altitude range of 681–3,466 m.a.s.l (EEPCO 2009). The major land-use types in the study area are forest lands, urban areas, rangeland, agricultural lands, water bodies, and other built-up areas (Takala et al. 2016). In very broad terms, most of the northern catchments of the watershed are under extensive cultivation with increased pressure of cultivated areas into marginal lands at the expense of woodlands. The flatter and poorly drained bottomlands of the northern catchments are usually not cultivated but are used for dry season grazing and eucalyptus tree plantations. The eastern part of the area has some of the most densely populated and intensively farmed areas. The natural vegetation at the south of the area is sparse owing to deforestation, which has been increasing at an alarming rate (EEPCO 2009).
Method of analysis
NEX-GDDP NASA dataset and climate change scenarios
Climate change evaluation was based on model results collected as part of the Coupled Model Intercomparison Project Phase 5 (CMIP5) World Climate Research Program's (WCRP's) multi-model dataset (Meehl et al. 2007; Flato et al. 2013). Among different GCM models, GFDL-ESM2M, MPI-ESM-MR, CSIRO-MK3-6-0, NorESM1-M, and MIROC5 for the Woliso, Jimma, Shebe, Hosaina, and Sodo & Sekoru stations, respectively, were used as data source models (Table 1) (refer to Supplementary material Figures 1-6). The data of each of the models were spatially interpolated to represent the watershed area. The model results were available for the Representative Concentration Pathway (RCP) 4.5 (medium to low emission) and RCP 8.5 (high to medium emissions) scenarios (Table 2). These models were selected based on the evaluation of station-based data with observations of respective station data obtained from the National Meteorological Agency (NMA) of Ethiopia (Abiy et al. 2022). The models provide daily precipitation and temperature data, which were bias-corrected statistical downscaled future data downloaded for each scenario, from the NEX-GDDP NASA dataset.
Number . | Model . | Country and institution . |
---|---|---|
01 | MPI-ESM-MR | Max Planck Institute for Meteorology, Germany |
02 | CSIRO-Mk3-6-0 | Commonwealth Scientific and Industrial Research Organization Queensland Climate Change Centre of Excellence, Australia |
03 | MIROC5 | Atmosphere and Ocean Research Institute, Japan |
04 | GFDL-ESM2M | Geophysical Fluid Dynamics Laboratory, America |
05 | NorESM1-M | Norway Consumer Council, Norway |
Number . | Model . | Country and institution . |
---|---|---|
01 | MPI-ESM-MR | Max Planck Institute for Meteorology, Germany |
02 | CSIRO-Mk3-6-0 | Commonwealth Scientific and Industrial Research Organization Queensland Climate Change Centre of Excellence, Australia |
03 | MIROC5 | Atmosphere and Ocean Research Institute, Japan |
04 | GFDL-ESM2M | Geophysical Fluid Dynamics Laboratory, America |
05 | NorESM1-M | Norway Consumer Council, Norway |
Scenarios . | Definition . | Country . |
---|---|---|
RCP 4.5 | Becomes peak in total radiative forcing at +4.5 W/m2 before 2100 and decline | United States by the Joint Global Change Research Institute (JGCRI) |
RCP 8.5 | Raising radiative forcing pathway leading to +8.5 W/m2 in 2100 | Austria at the International Institute for Applied System Analysis (IIASA) |
Scenarios . | Definition . | Country . |
---|---|---|
RCP 4.5 | Becomes peak in total radiative forcing at +4.5 W/m2 before 2100 and decline | United States by the Joint Global Change Research Institute (JGCRI) |
RCP 8.5 | Raising radiative forcing pathway leading to +8.5 W/m2 in 2100 | Austria at the International Institute for Applied System Analysis (IIASA) |
SWAT model set up
For setting up of the SWAT model, meteorological data, hydrological data, digital elevation model (DEM) data, and land-use, land-cover (LULC) data of 2018 were used as input data. Accordingly, routine steps including watershed delineation, hydrological response unit (HRU) analysis, weather data input preparation, running SWAT, parameter sensitivity analysis using SWATCUP SUFI2 algorithm, calibrations, and model performance evaluation were applied procedurally. The SWAT default multiple HRU definition criteria with land, soil, and slope ratio of 20:10:20 (settings for land-use threshold (20%); soil threshold (10%); and slope threshold (20%) of individual sub-basin area) was selected in most SWAT applications (Winchell et al. 2012). The SWAT description is available at the SWAT link: https://swat.tamu.edu/docs/
Study of climate change impact on stream flow and sediment yield
Relative to the base period of 1990–2017, the effect of future climate change on stream flow and sediment yield was studied for the short term (2040–2069) and long term (2071 2099) using statistically downscaled datasets for climate variables (precipitation, maximum and minimum temperatures). The NEX-GDDP dataset for the selected five models output at each station (GFDL-ESM2M model for Woliso station, MPI-ESM-MR for Jimma station, CSIRO-MK3-6-0 for Shebe station, NorESM1-M for Hosaina station, and MIROC5 model for both Sodo and Sekoru stations) were used as input for the SWAT model to generate the future flow and sediment yield. The NEX-GDDP dataset of NASA, used in the aforementioned five models for emission scenario RCPs 4.5 and 8.5, was downloaded and used for climate change projection. After the routine steps of parameter selection, sensitivity analysis, calibration, and validation, the calibrated SWAT model was used for estimation of stream flow and sediment yield under future scenarios for 2050 and 2080.
Uncertainty analysis
In general, the sources of uncertainties of climate scenarios are multiple. The climate system itself is too complex to be represented in a numerical model and contains a number of assumptions and parameterizations that each climate modeling center approaches differently. Uncertainties in climate scenarios and GCM outputs may be much larger (Abdo et al. 2009). Uncertainties in projected changes in a hydrological system arise from internal variability of the climate; uncertainty in future greenhouse gas and aerosol emissions; the translation of these emissions into climate change by global climate models; hydrological model uncertainty; uncertainty caused by sufficiency of field data at all scales; and uncertainty of downscaling techniques (Bates et al. 2008).
The uncertainties emanating from SWAT parameters, input data, and model structure are handled through the sequential and uncertainty fitting algorithm (SUFI-2) tool of the SWATCUP (Soil and Water Assessment Tool–Calibration and Uncertainty Procedure) software (Abbaspour et al. 2007). The program links the Generalized Likelihood Uncertainty Estimation (GLUE; Beven and Binley 1992). The paper also included r-factors and p-factors, which also indicate the result of the uncertainty analysis.
RESULTS AND DISCUSSIONS
Evaluation of climate data for selected models
The results of the statistical evaluation for mean monthly and mean annual rainfall for the base scenario (1976–2005) at the selected stations for the representative models are provided in Table 3.
Station . | Selected model . | Evaluation statistics for monthly mean . | Annual mean (mm) . | |||||||
---|---|---|---|---|---|---|---|---|---|---|
RMSE . | . | NSE . | Observed rainfall (mm) . | Model rainfall (mm) . | Observed . | Model . | Difference . | Difference (%) . | ||
Hosaina | NorESM | 19.75 | 0.89 | 0.88 | 102.62 | 101.4 | 1,231.47 | 1,216.9 | −14.57 | −1.2 |
Jimma | MPI-ESM-MR | 11.97 | 0.98 | 0.97 | 123.49 | 123.2 | 1,481.92 | 1,478 | −3.92 | −0.3 |
Shebe | CSIRO-Mk3-6-0 | 16.03 | 0.97 | 0.94 | 127.29 | 126.4 | 1,527.49 | 1,516.7 | −10.79 | −0.7 |
Sodo | MIROC5 | 19.43 | 0.97 | 0.91 | 108.60 | 100.8 | 1,302.90 | 1,209.7 | −93.2 | −7.2 |
Sekoru | MIROC5 | 17.95 | 0.94 | 0.96 | 108.91 | 111.8 | 1,306.88 | 1,342.1 | 35.22 | 2.7 |
Woliso | GFDL-ESM2M | 26.68 | 0.92 | 0.91 | 103.04 | 109.0 | 1,236.51 | 1,308.1 | 71.59 | 5.8 |
Overall Mean | 18.64 | 0.95 | 0.93 | 112.33 | 112.1 | 1,347.86 | 1,345.3 | −2.6 | −0.9 |
Station . | Selected model . | Evaluation statistics for monthly mean . | Annual mean (mm) . | |||||||
---|---|---|---|---|---|---|---|---|---|---|
RMSE . | . | NSE . | Observed rainfall (mm) . | Model rainfall (mm) . | Observed . | Model . | Difference . | Difference (%) . | ||
Hosaina | NorESM | 19.75 | 0.89 | 0.88 | 102.62 | 101.4 | 1,231.47 | 1,216.9 | −14.57 | −1.2 |
Jimma | MPI-ESM-MR | 11.97 | 0.98 | 0.97 | 123.49 | 123.2 | 1,481.92 | 1,478 | −3.92 | −0.3 |
Shebe | CSIRO-Mk3-6-0 | 16.03 | 0.97 | 0.94 | 127.29 | 126.4 | 1,527.49 | 1,516.7 | −10.79 | −0.7 |
Sodo | MIROC5 | 19.43 | 0.97 | 0.91 | 108.60 | 100.8 | 1,302.90 | 1,209.7 | −93.2 | −7.2 |
Sekoru | MIROC5 | 17.95 | 0.94 | 0.96 | 108.91 | 111.8 | 1,306.88 | 1,342.1 | 35.22 | 2.7 |
Woliso | GFDL-ESM2M | 26.68 | 0.92 | 0.91 | 103.04 | 109.0 | 1,236.51 | 1,308.1 | 71.59 | 5.8 |
Overall Mean | 18.64 | 0.95 | 0.93 | 112.33 | 112.1 | 1,347.86 | 1,345.3 | −2.6 | −0.9 |
The results indicated that each model best represents the station. The maximum deviation of the respective model's mean annual rainfall from the observation was obtained at Sodo station, with 7.2%. However, the minimum difference was observed as 0.3% at Jimma station indicating the best representation of observation by the model rainfall data. Higher and Nash Sutcliffe efficiency (NSE) values that approach 1 indicate the model efficiency to represent the respective observation data in the area. Accordingly, the relation between model and observed data resulted in and NSE of greater than 0.7 for the selected models of the station. According to Table 3, the models selected were applicable in representing the observation in each station based on statistical test values such as RMSE, RSQ, and NSE. Moreover, the overall variation of the mean precipitation between the model and observation was about 0.9%, which indicates the acceptance of models for future prediction of precipitation (refer to Supplementary data Tables 3–4).
Future climate change projections
The future climate changes in the short term (2040–2069) and the long term (2071–2099) were analyzed by comparing the future downscaled climate parameter with the base period (1990–2017). The climate change estimation was made using the RCPs 4.5 and 8.5 from the NEX-GDDP dataset from NASA.
Mean annual rainfall and temperature changes from the base scenario
During the 2050s and 2080s, the projected rainfall result indicates that mean annual rainfall is expected to increase by 3.2 and 11.0% in the RCP 4.5 scenario, while the increments were 9.7 and 29.5% in the case of the RCP 8.5 scenario, respectively (Table 4) (refer to Supplementary data Figures 7–10). The result is consistent with the report of the United Nations Environment Programme (UNEP) (2013) undertaken for the Gibe III watershed. It summarized that the annual rainfall would increase in the range of −3 to +25% by the 2080s. According to McSweeney et al. (2012), by the year 2100, an increase of more than 20% in precipitation is expected. A similar result was obtained by using the prediction model ensemble mean compared with the base line period, with the rainfall expected to increase by 36.57% (Ashenafi 2017).
. | Data period . | Tmax (°C) . | ΔTmax (°C) . | Tmin (°C) . | ΔTmin (°C) . | Annual rainfall (mm) . | Δ Annual rainfall (%) . |
---|---|---|---|---|---|---|---|
Base scenario | 2000s | 25.8 | – | 12.8 | – | 1,354.7 | – |
RCP 4.5 | 2050s | 26.6 | 0.8 | 12.2 | −0.6 | 1,398.9 | 3.2 |
2080s | 26.96 | 1.16 | 12.7 | −0.1 | 1,503.7 | 11.0 | |
RCP 8.5 | 2050s | 27.2 | 1.4 | 12.8 | 0 | 1,486.6 | 9.7 |
2080s | 28.6 | 2.8 | 14.3 | 1.5 | 1,754.6 | 29.5 |
. | Data period . | Tmax (°C) . | ΔTmax (°C) . | Tmin (°C) . | ΔTmin (°C) . | Annual rainfall (mm) . | Δ Annual rainfall (%) . |
---|---|---|---|---|---|---|---|
Base scenario | 2000s | 25.8 | – | 12.8 | – | 1,354.7 | – |
RCP 4.5 | 2050s | 26.6 | 0.8 | 12.2 | −0.6 | 1,398.9 | 3.2 |
2080s | 26.96 | 1.16 | 12.7 | −0.1 | 1,503.7 | 11.0 | |
RCP 8.5 | 2050s | 27.2 | 1.4 | 12.8 | 0 | 1,486.6 | 9.7 |
2080s | 28.6 | 2.8 | 14.3 | 1.5 | 1,754.6 | 29.5 |
Note: Change in temperature and rainfall is relative to 2000s (1990–2017, base period).
Similarly, the mean annual maximum and minimum temperature changes are provided in Table 4, indicating that, during 2050s and 2080s, maximum temperature increase of about 0.8 and 1.2 °C in the case of RCP 4.5 is expected, while it is expected to increase by 1.4 and 2.8 °C in the RCP 8.5 scenario. The finding of the report by UNEP (2013) at the same watershed shows that the change in mean temperature during the 2080s would range between 1.8 and 4.3 °C. Field et al. (2014) reported that as a result of the current and past anthropogenic greenhouse gas emissions, the global surface temperature is likely to increase by 3–6 °C for the RCPs 4.5 and 8.5 by 2100.
According to the work of different scholars such as Rowel et al. (2015), Souverijns et al. (2016), Ongoma and Chen (2017), Muhati et al. (2018), and Shiferaw (2018), the projected increase of surface temperatures and rainfall for the East African region by GCM dataset was expected. Their results conclude that annual rainfall and temperature will increase in the 21st century relative to the 20th century. In addition to the increased annual temperature by 0.38, 0.96, and 0.77 °C by the end of 2035, 2050, and 2065, respectively, the report of similar studies at this watershed showed increasing annual precipitation by 5.31, 3.07, and 0.31% in the short term, mid-term, and long term, respectively (Melkamu et al. 2017; Mango et al. 2021).
Impact of climate change on water resource potential
SWAT model parameter sensitivity analysis
The parameter selection and sensitivity analysis were performed with repeated and procedural steps in a SWATCUP setup; the most sensitive parameters were obtained for Abelti gauge station for stream flow and sediment yield as provided in Table 5.
Flow parameter sensitivity . | . | Sediment parameter sensitivity . | . | ||||
---|---|---|---|---|---|---|---|
Parameter . | t-Stat . | p-value . | Rank . | Parameter . | t-Stat . | p-value . | Rank . |
1:R__CN2 | −13.6 | 0.00 | 1 | 6:V__USLE_C | −27.26 | 0.00 | 1 |
6:V__LAT_TTIME | 11.98 | 0.00 | 2 | 13:R__CN2 | −21.11 | 0.00 | 2 |
13:V__GW_REVAP | 2.25 | 0.03 | 3 | 5:V__SPEXP | 11.08 | 0.00 | 3 |
12:V__HRU_SLP | −1.56 | 0.12 | 4 | 9:V__USLE_P | −8.47 | 0.00 | 4 |
3:R__SOL_BD | 1.52 | 0.13 | 5 | 14:V__CH_N2 | −6.37 | 0.00 | 5 |
5:V__OV_N | 1.25 | 0.22 | 6 | 1:V__HRU_SLP | −4.13 | 0.00 | 6 |
22:V__CH_K2 | −1.16 | 0.25 | 7 | 4:V__SPCON | 2.68 | 0.01 | 7 |
20:V__SLSUBBSN | 0.91 | 0.37 | 8 | 3:V__CH_K2 | 2.51 | 0.01 | 8 |
9:R__SOL_AWC | −0.80 | 0.42 | 9 | 8:V__BIOMIX | −1.85 | 0.07 | 9 |
17:R__SOL_K | 0.71 | 0.48 | 10 | 2:V__SLSUBBSN | 0.84 | 0.40 | 10 |
2:V__ALPHA_BF | −0.52 | 0.61 | 11 | 12:V__CH_COV2 | 0.75 | 0.45 | 11 |
18:V__RCHRG_DP | 0.50 | 0.62 | 12 | 10:V__RSDIN | −0.42 | 0.67 | 12 |
10:V__REVAPMN | −0.47 | 0.64 | 13 | 7:R__USLE_K | −0.24 | 0.81 | 13 |
16:V__GWQMN | −0.39 | 0.69 | 14 | 11:V__CH_COV1 | 0.22 | 0.83 | 14 |
11:V__ESCO | 0.39 | 0.69 | 15 | – | – | – | |
7:V__ALPHA_BNK | −0.39 | 0.70 | 16 | – | – | – | |
4:V__GW_DELAY | −0.34 | 0.73 | 17 | – | – | – | |
8:V__CANMX | 0.29 | 0.77 | 18 | – | – | – | |
15:V__CH_S2 | 0.23 | 0.82 | 19 | – | – | – | |
21:V__SURLAG | −0.21 | 0.83 | 20 | – | – | – | |
14:V__CH_N2 | 0.05 | 0.96 | 21 | – | – | – | |
19:V__EPCO | −0.04 | 0.97 | 22 | – | – | – |
Flow parameter sensitivity . | . | Sediment parameter sensitivity . | . | ||||
---|---|---|---|---|---|---|---|
Parameter . | t-Stat . | p-value . | Rank . | Parameter . | t-Stat . | p-value . | Rank . |
1:R__CN2 | −13.6 | 0.00 | 1 | 6:V__USLE_C | −27.26 | 0.00 | 1 |
6:V__LAT_TTIME | 11.98 | 0.00 | 2 | 13:R__CN2 | −21.11 | 0.00 | 2 |
13:V__GW_REVAP | 2.25 | 0.03 | 3 | 5:V__SPEXP | 11.08 | 0.00 | 3 |
12:V__HRU_SLP | −1.56 | 0.12 | 4 | 9:V__USLE_P | −8.47 | 0.00 | 4 |
3:R__SOL_BD | 1.52 | 0.13 | 5 | 14:V__CH_N2 | −6.37 | 0.00 | 5 |
5:V__OV_N | 1.25 | 0.22 | 6 | 1:V__HRU_SLP | −4.13 | 0.00 | 6 |
22:V__CH_K2 | −1.16 | 0.25 | 7 | 4:V__SPCON | 2.68 | 0.01 | 7 |
20:V__SLSUBBSN | 0.91 | 0.37 | 8 | 3:V__CH_K2 | 2.51 | 0.01 | 8 |
9:R__SOL_AWC | −0.80 | 0.42 | 9 | 8:V__BIOMIX | −1.85 | 0.07 | 9 |
17:R__SOL_K | 0.71 | 0.48 | 10 | 2:V__SLSUBBSN | 0.84 | 0.40 | 10 |
2:V__ALPHA_BF | −0.52 | 0.61 | 11 | 12:V__CH_COV2 | 0.75 | 0.45 | 11 |
18:V__RCHRG_DP | 0.50 | 0.62 | 12 | 10:V__RSDIN | −0.42 | 0.67 | 12 |
10:V__REVAPMN | −0.47 | 0.64 | 13 | 7:R__USLE_K | −0.24 | 0.81 | 13 |
16:V__GWQMN | −0.39 | 0.69 | 14 | 11:V__CH_COV1 | 0.22 | 0.83 | 14 |
11:V__ESCO | 0.39 | 0.69 | 15 | – | – | – | |
7:V__ALPHA_BNK | −0.39 | 0.70 | 16 | – | – | – | |
4:V__GW_DELAY | −0.34 | 0.73 | 17 | – | – | – | |
8:V__CANMX | 0.29 | 0.77 | 18 | – | – | – | |
15:V__CH_S2 | 0.23 | 0.82 | 19 | – | – | – | |
21:V__SURLAG | −0.21 | 0.83 | 20 | – | – | – | |
14:V__CH_N2 | 0.05 | 0.96 | 21 | – | – | – | |
19:V__EPCO | −0.04 | 0.97 | 22 | – | – | – |
Among selected parameters, the SCS_CN for moisture condition II (Cn2), lateral flow travel time (LAT_TTIME), groundwater ‘revap’ coefficient (GW_REVAP.gw), average slope steepness (HRU_SLP), soil bulk density (SOL_BD.sol), Manning's ‘n’ value for overland flow (OV_N), effective hydraulic conductivity in main channel alluvium (CH_K2, average slope length (SLSUBBSN), soil conductivity (SOL_K), and soil available water capacity (Sol_AWC) were found to be sensitive parameters that have important roles in the hydrological process in the study watershed among all the tested 22 parameters. In addition, about eight of the most sensitive parameters were obtained to have high effect on the simulated sediment yield values. Therefore, parameters such as USLE cover or management factor (USLE_C), the SCS_CN for moisture condition II (Cn2), exponential factor for channel sediment routing (SPEXP), USLE support practice factor (USLE_ P), Manning's ‘n’ value for the main channel (CH_N2), average slope steepness (HRU_SLP), linear factor for channel sediment routing (SPCON), and effective hydraulic conductivity in main channel alluvium (CH_K2) were highly sensitive parameters for sediment yield simulation in the watershed. Considering the limitations of parameterization with respect to physical meaning of the parameter in the area (parameters expected to affect stream flow and sediment yield but not top sensitive), additional parameters were considered for calibration and validation of the SWAT model for both variables (flow and sediment).
Stream flow and sediment yield calibration and validation
For flow calibration, about 11 parameters from flow parameters and 7 parameters (not including common parameters of flow and sediment) from sediment parameters were selected. Attention was given to prevent over-parameterization, loss of meaning of the parameters, and missing the reasonability of the output. Table 6 shows the selected sensitive parameters, ranges of parameter values, and fitted value results.
Parameters of flow . | Parameters of sediment . | ||||||
---|---|---|---|---|---|---|---|
Parameter . | Value fitted . | Min . | Max . | Parameter . | Fitted value . | Min . | Max . |
CN2 * | −0.06 | 35 | 98 | SPCON | 0.0009 | 0.0001 | 0.01 |
ALPHA_BF | 0.30 | 0 | 1 | SPEXP | 1.14 | 1.11 | 1.5 |
SOL_BD* | 0.14 | 0.9 | 2.5 | USLE_C | 0.28 | 0.001 | 0.5 |
OV_N | 26.58 | 0.01 | 30 | USLE_K* | 0.09 | 0 | 0.65 |
LAT_TTIME | 25.78 | 0 | 180 | USLE_P | 0.57 | 0 | 1 |
SOL_AWC* | 0.09 | 0 | 1 | RSDIN | 8,954 | 0 | 10,000 |
HRU_SLP | 0.36 | 0 | 1 | CH_COV2 | 0.47 | −0.001 | 1 |
GW_REVAP | 0.10 | 0.02 | 0.2 | ||||
SLSUBBSN | 37.02 | 10 | 150 | ||||
CH_K2 | 16.26 | –0.01 | 500 | ||||
SOL_K* | −0.25 | 0 | 2,000 |
Parameters of flow . | Parameters of sediment . | ||||||
---|---|---|---|---|---|---|---|
Parameter . | Value fitted . | Min . | Max . | Parameter . | Fitted value . | Min . | Max . |
CN2 * | −0.06 | 35 | 98 | SPCON | 0.0009 | 0.0001 | 0.01 |
ALPHA_BF | 0.30 | 0 | 1 | SPEXP | 1.14 | 1.11 | 1.5 |
SOL_BD* | 0.14 | 0.9 | 2.5 | USLE_C | 0.28 | 0.001 | 0.5 |
OV_N | 26.58 | 0.01 | 30 | USLE_K* | 0.09 | 0 | 0.65 |
LAT_TTIME | 25.78 | 0 | 180 | USLE_P | 0.57 | 0 | 1 |
SOL_AWC* | 0.09 | 0 | 1 | RSDIN | 8,954 | 0 | 10,000 |
HRU_SLP | 0.36 | 0 | 1 | CH_COV2 | 0.47 | −0.001 | 1 |
GW_REVAP | 0.10 | 0.02 | 0.2 | ||||
SLSUBBSN | 37.02 | 10 | 150 | ||||
CH_K2 | 16.26 | –0.01 | 500 | ||||
SOL_K* | −0.25 | 0 | 2,000 |
Note: Parameters with asterisk were treated as relative change so that the original parameter value in SWAT is multiplied by (1 + fitted value).
Flow and sediment yield calibration and validation result
Based on the streamflow and sediment yield calibration and validation results, the SWAT model's efficiency was assessed using the criteria provided in Table 7 which is supported by Moriasi et al. (2007), Nash & Sutcliffe (1970), and Santhi et al. (2001).
Rating . | . | . | PBIAS (%) . | |
---|---|---|---|---|
. | NSE . | Stream flow . | Sediment . | |
Very good | 0.75 − 1.0 | 0.75 − 1.00 | Less than ±10 | Less than ±15 |
Good | 0.65 − 0.75 | 0.65 − 0.75 | ±10 to ±15 | ±15 to ±30 |
Satisfactory | 0.60 − 0.65 | 0.50 − 0.65 | ±15 to ±25 | ±30 to ±55 |
Unsatisfactory | <0.6 | <0.5 | More than ±25 | More than ±55 |
Rating . | . | . | PBIAS (%) . | |
---|---|---|---|---|
. | NSE . | Stream flow . | Sediment . | |
Very good | 0.75 − 1.0 | 0.75 − 1.00 | Less than ±10 | Less than ±15 |
Good | 0.65 − 0.75 | 0.65 − 0.75 | ±10 to ±15 | ±15 to ±30 |
Satisfactory | 0.60 − 0.65 | 0.50 − 0.65 | ±15 to ±25 | ±30 to ±55 |
Unsatisfactory | <0.6 | <0.5 | More than ±25 | More than ±55 |
Note: PBIAS means percent bias in %.
Variable . | . | Mean obs.a . | Mean sim.a . | R2 . | NSE . | PBIAS . | p-factor . | r-factor . |
---|---|---|---|---|---|---|---|---|
Flow | Calibration | 218.75 | 231.39 | 0.9 | 0.87 | −5.80 | 0.85 | 1.05 |
Validation | 226.37 | 192.92 | 0.82 | 0.77 | 14.8 | 0.83 | 1.23 | |
Sediment | Calibration | 105 × 104t | 111 × 104t | 0.87 | 0.86 | −6.0 | 0.78 | 0.85 |
Validation | 101 × 104t | 90 × 104t | 0.75 | 0.73 | 10.9 | 0.67 | 0.67 |
Variable . | . | Mean obs.a . | Mean sim.a . | R2 . | NSE . | PBIAS . | p-factor . | r-factor . |
---|---|---|---|---|---|---|---|---|
Flow | Calibration | 218.75 | 231.39 | 0.9 | 0.87 | −5.80 | 0.85 | 1.05 |
Validation | 226.37 | 192.92 | 0.82 | 0.77 | 14.8 | 0.83 | 1.23 | |
Sediment | Calibration | 105 × 104t | 111 × 104t | 0.87 | 0.86 | −6.0 | 0.78 | 0.85 |
Validation | 101 × 104t | 90 × 104t | 0.75 | 0.73 | 10.9 | 0.67 | 0.67 |
aMonthly (m3/s) for flow and tons for sediment load.
Uncertainty analysis for the Abelti sub-watershed
The degree to which all uncertainties are accounted for is quantified by a measure of the p-factor, which is the percentage of the measured data bracketed by 95% prediction uncertainty (95PPU) (Abbaspour et al. 2007). Similarly, r-factor is used as the measure quantifying the strength of a calibration or uncertainty analysis, which is the average thickness of the 95PPU band divided by the standard deviation of the measured data (Abbaspour 2013). Theoretically, a p-factor of 1 (all observations bracketed by the prediction uncertainty) and r-factor of 0 (achievement of rather small uncertainty band) indicate that the simulation exactly corresponds to the measured data but p-factor ranges between 0 and 100%, while that of r-factor ranges between 0 and infinity. The degree to which they are away from these numbers can be used to judge the strength of our calibration. When acceptable values of r-factor and p-factor are reached, then the parameter uncertainties are the desired parameter ranges (Moriasi et al. 2007).
According to Table 8, the result of flow calibration and validation at Abelti gave p-factors of 0.85 and 0.83 and r-factor 1.05 and 1.23, respectively. Calibration and validation of sediment yield indicated p-factors of 0.78 and 0.67 and r-factors of 0.85 and 0.67, respectively. The result indicates that 85% of the measured data were bracketed by the 95PPU range of uncertainty in the output for flow calibration, and 83% during validation, which is in the range of acceptance. During the calibration and validation, the thickness of the uncertainty band was obtained as 95PPU of the r-factors 1.05 and 1.23 for flow and 0.85 and 0.67 for sediment yield, respectively. Setegn et al. (2009) justified that low p-factors and large r-factors in uncertainty simulation could be due to the error in the rainfall and temperature input data, which may be the reason for the results in this study as well, for the validation periods indicating a 38% r-factor. The area has complex topography and a highly converting land-use land-cover condition that could also increase uncertainty more than the model structure.
Stream flow and sediment yield simulation for future periods
The climate data of 2050 and 2080 for both RCPs 4.5 and 8.5 for the selected models were used as input for the calibrated SWAT model, and the flow and sediment responses were analyzed as follows. The result in Table 9 shows that climate change will have a prominent effect on stream flow and sediment load. It indicates an increasing amount of stream flow and sediment yield by 2050 and 2080 in both the RCPs 4.5 and 8.5 scenarios. The relative increase of flow from the current situation was 54 and 74% for RCP 4.5 for 2050 and 2080, while it was 56 and 117% for RCP 8.5 for 2050 and 2080, respectively. The response was directly linked to increasing future rainfall (Dao & Tadashi 2013). Similarly, sediment load was expected to increase in future climate change conditions under the RCP 4.5 and 8.5 scenarios (Table 9) from the 3.78Mton at the current condition to the 10.15Mton (169% increase) at RCP 8.5 in 2080.
Variable . | Scenario . | RCP 4.5 . | RCP 8.5 . | ||
---|---|---|---|---|---|
Current . | 2050 . | 2080 . | 2050 . | 2080 . | |
Mean monthly stream flow (m3/s) | 499.9 | 770.8 | 873.4 | 783.4 | 1,087.1 |
Percent change relative to current (%) | – | 54 | 74 | 56 | 117 |
Mean monthly sediment load (Mton) | 3.78 | 6.02 | 7.83 | 6.44 | 10.15 |
Percent change relative to current (%) | – | 59 | 107 | 70 | 169 |
Variable . | Scenario . | RCP 4.5 . | RCP 8.5 . | ||
---|---|---|---|---|---|
Current . | 2050 . | 2080 . | 2050 . | 2080 . | |
Mean monthly stream flow (m3/s) | 499.9 | 770.8 | 873.4 | 783.4 | 1,087.1 |
Percent change relative to current (%) | – | 54 | 74 | 56 | 117 |
Mean monthly sediment load (Mton) | 3.78 | 6.02 | 7.83 | 6.44 | 10.15 |
Percent change relative to current (%) | – | 59 | 107 | 70 | 169 |
The projected effect of climate change in monthly mean stream flow and sediment yield in each month of the future periods for the RCPs 4.5 and 8.5 scenarios are indicated in Table 9. The result shows a significant increase in stream flow and sediment yield for both periods and scenarios. At the same watershed, Ashenafi (2017) also reported the projections with increasing stream flow in both RCP 4.5 and RCP 8.5 emission scenarios.
As it is shown in Figure 10, the high flow months (June to October) were expected to increase in all scenarios relative to the dry months (November to May) in the current situation. The result shows significantly higher mean monthly stream flow and sediment yield under the RCP 8.5 emission scenario during 2080 than others in the rainy season. The peaks of sediment yield exactly follow similar patterns of stream flow peaks as also reported by Kumar & Mishra (2014), justifying that the sediment yields have a direct link and trend with simulated runoff peaks.
Evaluation of water balance components
The main water balance components of the basin include: the total amount of precipitation, actual evapotranspiration, and the net amount of water that contributes to stream flow in the reach (water yield). The water yield includes surface runoff contributing to stream flow, lateral flow contributing to stream flow, groundwater contributing to stream flow minus the losses. The period from 1990 to 2017 was used as the current period (base period) to analyze the impact of climate change on stream flow and sediment yield. The daily precipitation and minimum and maximum temperatures used for the future prediction purpose were from the period 2040–2069 for the 2050s and 2070–2099 for the 2080s, which was input for calibrating the SWAT model. The result in Table 10 shows that the variation of discharge is strongly linked with the increasing trend of precipitation in the future.
Hydrological component . | Current period . | 2050s . | 2080s . | Δ(2050s) % . | Δ(2080s) % . | ||||
---|---|---|---|---|---|---|---|---|---|
RCP 4.5 . | RCP 8.5 . | RCP 4.5 . | RCP 8.5 . | RCP 4.5 . | RCP 8.5 . | RCP 4.5 . | RCP 8.5 . | ||
PCP (mm) | 1,286.1 | 1,442.1 | 1,509.9 | 1,551.6 | 1,800.6 | 12.1 | 17.4 | 20.6 | 40.0 |
Sur_Q (mm) | 149.91 | 260.5 | 316.8 | 321.0 | 446.9 | 73.7 | 111.3 | 114.1 | 198.1 |
Lat_Q (mm) | 52.12 | 48.3 | 43.7 | 48.2 | 48.3 | −7.3 | −16.2 | −7.5 | −7.3 |
Qgw (mm) | 282.47 | 438.6 | 398.9 | 474.3 | 554.0 | 55.3 | 41.2 | 67.9 | 96.1 |
WYD (mm) | 484.5 | 747.4 | 759.4 | 843.5 | 1,049.2 | 54.2 | 56.7 | 74 | 116.5 |
ET (mm) | 646.1 | 567.7 | 528.2 | 577.7 | 614.6 | −12.1 | −18.2 | −10.6 | −4.9 |
Sediment load (t/ha) | 60.6 | 90.8 | 80.0 | 145.2 | 122.0 | 49.7 | 32.0 | 139.5 | 101.1 |
Hydrological component . | Current period . | 2050s . | 2080s . | Δ(2050s) % . | Δ(2080s) % . | ||||
---|---|---|---|---|---|---|---|---|---|
RCP 4.5 . | RCP 8.5 . | RCP 4.5 . | RCP 8.5 . | RCP 4.5 . | RCP 8.5 . | RCP 4.5 . | RCP 8.5 . | ||
PCP (mm) | 1,286.1 | 1,442.1 | 1,509.9 | 1,551.6 | 1,800.6 | 12.1 | 17.4 | 20.6 | 40.0 |
Sur_Q (mm) | 149.91 | 260.5 | 316.8 | 321.0 | 446.9 | 73.7 | 111.3 | 114.1 | 198.1 |
Lat_Q (mm) | 52.12 | 48.3 | 43.7 | 48.2 | 48.3 | −7.3 | −16.2 | −7.5 | −7.3 |
Qgw (mm) | 282.47 | 438.6 | 398.9 | 474.3 | 554.0 | 55.3 | 41.2 | 67.9 | 96.1 |
WYD (mm) | 484.5 | 747.4 | 759.4 | 843.5 | 1,049.2 | 54.2 | 56.7 | 74 | 116.5 |
ET (mm) | 646.1 | 567.7 | 528.2 | 577.7 | 614.6 | −12.1 | −18.2 | −10.6 | −4.9 |
Sediment load (t/ha) | 60.6 | 90.8 | 80.0 | 145.2 | 122.0 | 49.7 | 32.0 | 139.5 | 101.1 |
PCP, precipitation; Sur_Q, surface runoff; Lat_Q, lateral flow into stream; WYD, total water yield; ET, evapotranspiration; GW, groundwater recharge (percolation below root zone).
It was observed for RCP 4.5 that the mean annual sediment yield was expected to increase by 49.7 and 139.5% in the 2050s and 2080s, respectively. Under the RCP 8.5 scenario, it was also expected to increase by 32 and 101.1% during the 2050s and 2080s, respectively. According to the results, precipitation and sediment yield share a strong relationship, which confirms the previous findings of Mullan et al. (2012), Dao & Tadashi (2013), Cousino et al. (2015), and Azim et al. (2016). The Gibe III reservoir's typical maximum capacity since it began filling in mid-2015 is approximately 8.3 km3. At periods of minimum water levels, the reservoir's capacity is approximately 4.1 km3. According to SOGREAH 2010, the anticipated capacity of the Gibe III reservoir was expected to be 14,700 mm3 at its maximum level, with 2,950 Mm3 of dead volume and 11,750 Mm3 of active volume. It controls about 50% of the Omo catchment area, and about 70% of the total water runoff. Based on the estimated volume in the Gibe III reservoir at an annual maximum water level, the reservoir is approximately 6 km3 below its planned capacity (Sogreah 2010). This indicates that there may be a potential for additional water abstraction behind the dam in the future.
CONCLUSION AND RECOMMENDATION
The study on the response of water resource potential to possible future climate change at the Gibe III watershed, Omo-Gibe Basin, is a very important and current issue. The impact of climate change on stream flow, sediment yield and water resource potential in the Gibe III watershed was evaluated. The dataset output downloaded from five GCM climate models (GFDL-ESM2M, MPI-ESM-MR, CSIRO-MK3-6-0, NorESM1-M, and MIROC5), spatially covering the watershed area for different climate stations, were used to project the future climate condition and predict future stream flow, sediment yield, and water resource potential. These climate models used the NASA NEX-GDDP dataset centered at two different time slices (2050s and 2080s) under two emission scenarios (RCPs 4.5 and 8.5). This study was expected to show the effect of climate change on the water balance component so that planners of the watershed could develop preventive action plans for sustaining water resource potential. Similarly, due to the expanding agricultural land in the watershed, the risk of sediment flow to the reservoir was currently considered as the issue in the area that should be predicted and known under different climate change scenarios and future periods. With regard to monthly rainfall variation in the area, which could affect monthly stream flow, modeling the future projection was very important to implement preventive action as a solution to an expected event. The finding showed an increasing trend for annual precipitation of 6.2 and 11.8% under RCP 4.5 during the 2050s and 2080s as well as 10.5 and 30.4% under RCp 8.5 during the 2050s and 2080s, respectively. The maximum temperature was expected to increase by 1.9 and 2.80 °C during the 2050s and 2080s, while the minimum temperature was also projected to increase by 1.8 and 2.81 °C during the 2050s and 2080s, respectively. The result showed in the future surface runoff amount would likely dominate, increasing relative to other water balance components. The annual sediment yield would increase by about 50% during the 2050s under the RCP 4.5 emission scenario and by more than 100% during the 2080s under both emission scenarios. The monthly sediment yield will change by more than 50 and 100% during the 2050s and 2080s under both emission scenarios, which requires more attention with respect to saving reservoir life year. The belg season (from February to May) will likely be dry and the kiremt season (from June to September) is projected to be long, shifting the peak month from August to October by the 2050s.
The SWAT model output shows that the stream flow was expected to increase by 55 and 95.5% during the 2050s and 2080s, respectively, relative to base period (1990–2017). The response was directly related to increasing future rainfall and surface runoff. The annual sediment yield at the watershed outlet was projected to increase by 64.5 and 138% during the 2050s and 2080s, respectively. Except for the months of April and May, all other months showed increasing sediment loads. Under RCP 4.5, the average annual basin sediment load was 91 and 145.2 t/ha (during the 2050s and 2080s, respectively), while under RCP 8.5 it was 80 and 122.0 t/ha (during the 2050s and 2080s, respectively) so that the mean of 118 t/ha at the RCP 4.5 and 101 t/ha at the RCP 8.5 emission scenario was predicted. The slight reduction of annual basin mean sediment load at the high emission scenario may be due to the combined effect of higher temperature, rainfall, and CO2 concentration. In general, it is likely that the climate change may increase the runoff and sediment yield at the watershed much more if the watershed is not properly managed to reduce surface flow. The monthly analysis shows that the wet months may bring flooding to the river, and thus, environmental conservation measures need to be emphasized to avoid the failure of structures and ecological disturbance. It is also better to focus on harvesting excess water from the wet season to minimize the effects of droughts that could follow the high flow season. The sedimentation of the reservoir could be reduced by upstream watershed management interventions and frequent removal of sediment from the reservoir. Overall, the results highlight the need for development and cooperation in the region to encourage robust management strategies for climate resilience and to combat rapid climate change in the basin. The overall output of this study gave us critical issues that we should focus on, such as reducing excess surface runoff, preventing sedimentation in the reservoir, and preparing flood control structures at downstream sites because increasing precipitation especially during the wet season is expected. Generally, the findings show the requirement of additional dam construction downstream in the future. Moreover, a further similar study is recommended using the NASA NEX-GDDP dataset on other climate models of CMIP5 as well as using the dataset of one single climate model downloaded for all climate stations spatially covering the watershed. Studies with other hydrological models are also suggested so as to build confidence by reducing possible uncertainty in the decision-making in hydrological design and water resources management of the watershed.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the National Meteorological Agency of Ethiopia for provision of observed data. Similarly, the authors acknowledge the Climate Analytics Group and NASA Ames Research Center for preparing the official website (https://dataserver.nccs.nasa.gov/thredds/catalog/bypass/NEX-GDDP/bcsd/catalog.html) and making it possible to download the NASA NEX-GDDP dataset distributed by the NASA Center for Climate Simulation (NCCS).
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.