*3.1. Results of the Model Parameters Sensitivity Analyses*

The parameter sensitivity analyses were carried out together with the calibration process, as it is necessary to include the flows estimated by SWAT and the monitored flows in the sub-basins. In general, a higher "t-stat" and a lower p-value indicate that the parameter is sensitive [55]. Based on the sensitivity scale developed by Lenhart et al. (2002), shown in Table 2, the following parameters were identified as highly sensitive in the Ketar sub-basin: EPCO, RCHRG\_DP, SOL\_K, GW\_DELAY, CN2, REVAPMIN, and SURLAG. Similarly, ESCO, REVAPMIN, GWQMN, HRU\_SLP, and GW-DEALY were very highly sensitive parameters in the Meki sub-basin, and ESCO, CH\_K2, SOL\_K, and GWQMN were very highly sensitive in the Shalla sub-basin. The description of the parameters is presented in Table 3. The differences in the sensitivity of the hydrological parameters in the sub-basins indicate that the sub-basins are heterogeneous, although they refer to a single, closed, lakes region. The differences are mainly due to land use, soil, hydrogeologic, and anthropogenic variations. The t-stat values of each of the selected parameters for each sub-basin are indicated in Table 6. The parameter description and their adjusting values are indicated in Tables 3 and 4.

#### *3.2. Results of the Calibration and Validation of the Model*

The calibration results indicate good agreement between the simulated and observed discharges in the sub-basins. The results for simulated and observed discharges in the sub-basins were evaluated against *R*2, *NSE*, and *PBIAS* during calibration and validation. The values in the Ketar sub-basin are in good agreement with *R*<sup>2</sup> > 0.6, *NSE* > 0.5, and *PBIAS* ≤ "±"25, (Figure 4a,b). Similarly, the results showed that the simulated and observed monthly discharges were in a good agreement during calibration and validation for the Meki and Shalla sub-basins (Table 7).


**Table 6.** Sensitivity or mean of index I values of the selected parameters for the sub-basins, according to their "t-stat" results as per the scale indicated in Table 2.

Note: NI\* = not identified, \*\* Parameter description is presented in Table 3.


**Table 7.** Model performance statistics for the Ketar, Meki, and Shalla sub-basins.

Overall model performance statistics (*R*2, *NSE,* and *PBIAS*) for the Ketar, Meki, and Shalla sub-basins are presented in Table 7.

#### *3.3. Climate Scenario Analyses Results and Discussion*

The results of the impacts of climate change on the major components of the water balance such as surface runoff (Q), water yield (WY), and evapotranspiration (ET) were evaluated in terms of their annual, seasonal, and monthly variations. The Q, WY, and ET were identified as the most sensitive elements of the water balance components in the CRVB. The simulated impacts of the climate scenarios on the water balance components are substantial. The percentage change in the Q, WY, and ET from their baseline simulated outputs for each sub-basin are presented in Table 8, together with the indication of the baseline annual rainfall data (averaged for years 1984–2010).

**Table 8.** The simulated mean annual changes, as a percentage, from the annual average values of the baseline outputs for the major components of the water balance in the sub-basins.



**Table 8.** *Cont.*

Note: % of Δ = Percentage of change of the component from its baseline output.

#### 3.3.1. Ketar Sub-Basin

The resulting simulated ET, WY, and Q mean monthly values for the Ketar sub-basin are graphically displayed in Figure 5a. Changes in the Q pattern over the seasons in the Ketar sub-basin can be observed in Figure 5a. The highest Q season has shifted both in the near and long term of RCP4.5 to the months from March to May while it used to be between mid-June to the end of September in the baseline data outputs. The simulated annual variations from the base data are between −65.2% (LT-RCP8.5) and 22.9% (LT-RCP4.5). RCP 2.6 and RCP 8.5 analyses indicate that the expected runoff will decrease both in the near term and in the long term in relation to the baseline data simulation outputs. In all the seasons, for all RCPs, the runoff condition in the long term (LT) is higher than the runoff in the near-term (NT) period. However, the general trend indicates that the runoff is decreasing in this sub-basin in relation to the historical (baseline) period, but the rate of its reduction differs from one RCP to another and from one period to another.

In similar analyses, the WY in the Ketar sub-basin decreases for all RCPs, in both the NT and LT periods, except in the long-term periods of RCP4.5 for the months from April to June (Figure 5a). Generally, the impact is expected to reduce the WY in all projected scenarios, especially for the periods from July to October. However, the rate of reduction varies from RCP to RCP and varies from season to season. Nevertheless, the annual WY generation capacity of the Ketar sub-basin is higher than in the Meki and Shalla sub-basins, corresponding to the annual precipitation that is supplied. Almost half of the rainfall, 50% on an average, goes to the WY in all the scenarios, while the proportion is about 40% in the Meki sub-basin and about 44% in the Shalla sub-basin. The simulated WY in the RCPs follows a similar pattern to the observed base year simulations. It means that the seasonal change in WY is not disturbed in pattern but in quantity.

The ET in the sub-basin has bi-annual peaks between March and mid-May, and between July and September (Figure 5a). The ET is relatively low between mid-May and June. The rate of ET decreases between March and May in all the scenarios in relation to the observed data simulations except between June and September. ET will be higher in the Ketar sub-basin for RCP2.6 and RCP8.5, between June and September, than outputs from the base data. The significant change in ET mainly reflects the increase in temperature. Therefore, according to the RCP2.6 and RCP8.5 climate projections, the increase in ET will be higher than the RCP4.5 projections for ET. This is in line with the works of Musie et al. (2020) and Gadissa et al. (2019) in the Lake Ziway and CRV basins in Ethiopia, respectively [21,33]. Musie et al. (2020) used the SWAT model to evaluate the impacts of regional climate variabilities and land use change on the water resources in the Lake Ziway basin. They found an increase in surface runoff and water yield due to the climate scenarios from the year 2000 to 2017. Gadisa et al. (2019) used projected climate scenarios to evaluate stream flows for the medium-term (2040 to 2070) periods for the RCP4.5 and RCP8.5 scenarios.

**Figure 5.** The simulated monthly distributions of Q, WY, and ET in the Ketar, Meki, and Shalla sub-basins for the applied climate scenarios. (**a**) Ketar, (**b**) Meki, (**c**) Shalla.

The results reported in both studies, and in Getnet et al. (2014), in the CRVB indicated that the hydrologic variations in water balance due to climate variability were highly significant [20,27,32]. However, in contrast to the study by Musie et al. (2020) [20], the hydroclimate in our study was more predominant in WY than ET in the Ketar sub-basin. Another study conducted in the CRVB in 2007 on climate change impacts on water availability with a SWAT model indicating an increase of averaged annual rainfall from 2001 to 2099 can also be found [56]. However, Gadissa et al. (2019) projected a reduction in precipitation by 7.97% and 2.55% under RCP4.5 and RCP8.5 respectively for the future period from 2040 to 2070 [32]. Reduction in precipitation has strong correlation with reduction in water yield and surface runoff. Our study is thus in line with the findings of Gadissa et al. (2019) [32] with minimal differences in the periods of occurrences. There are seasonal shifts in the pattern of occurrences of the components of the water balance when compared with the baseline data sets. These shifts are mainly from the changes in precipitation, temperature, and humidity patterns caused by greenhouse gases and other emissions.

#### 3.3.2. Meki Sub-Basin

The Meki sub-basin is characterized by greater annual amounts of ET than in the Shalla and Ketar sub-basins. The annual surface runoff rises in all the RCP scenarios. There will be a seasonal shift of the peak runoff period from the usual July-to-September period to April-to-June in the sub-basin (Figure 5b). In the long-term periods of RCP2.6 and RCP8.5, the runoff will increase greatly in relation to the baseline data simulation outputs. However, RCP4.5 will create a moderate range of changes in relation to RCP2.6 and RCP8.5. The change in annual average runoff varies from 6% to 85% in reference to the baseline outputs. The projected monthly distribution shows that this water balance component varies significantly over the months in both the NT and LT period.

The change in averaged annual WY ranges from −1.1% to +23.9% in relation to the baseline data simulated. The scenario analysis also showed a remarkable increment in the WY amount between May and October for all RCP outputs. ET is the major water balance component of the sub-basin (Figure 5b). About 56% of the rainfall on average turns into ET. This indicates that the sub-basin water balance is highly sensitive to changes in temperature. Even though WY is good in the rainy seasons, most of it will be lost via ETs. Thus, for the Meki sub-basin, the impacts were more predominant in ET than in WY. This indicates the high seasonal weather variabilities in the sub-basin and its low hydroclimatic impact resilience. Similar findings were reported by Gadissa et al. (2019) and Musie et al. (2021) for this sub-basin. They used modeling approaches of RCM projections to assess the conditions of the Q, ET, and stream flows using the SWAT and WEAP models, respectively. In addition, Molla, (2014) has used physical assessment methods to indicate the sub-basin climate conditions [16,17,36]. These studies reported that the Meki sub-basin is the most hydroclimate-sensitive region. The strong weather variabilities in the sub-basin have resulted in wide ranges of changes in water resources similar to the findings of another study conducted by Getnet et al. (2014) in the CRVB [16,26,27]. The annual variations in this study are also relatively large for the sub-basin (Table 8). The modeling results in this study for the sub-basin are thus inconsistent with the above study findings.
