Climate change impacts on the water and groundwater resources of the Lake Tana Basin, Ethiopia

Climate change impacts on the water cycle can severely affect regions that rely on groundwater to meet their water demands in the midto long-term. In the Lake Tana basin, Ethiopia, discharge regimes are dominated by groundwater. We assess the impacts of climate change on the groundwater contribution to streamflow (GWQ) and other major water balance components in two tributary catchments of Lake Tana. Based on an ensemble of 35 bias-corrected regional climate models and a hydrologic catchment model, likely changes under two representative concentration pathways (RCP4.5 and 8.5) are assessed. No or only slight changes in rainfall depth are expected, but the number of rainy days is expected to decrease. Compared to the baseline average, GWQ is projected to decrease whereas surface runoff is projected to increase. Hence, rainfall trends alone are not revealing future water availability and may even be misleading, if regions rely heavily on groundwater.


INTRODUCTION
Climate change causes strong impacts on the natural systems of all continents (Field et al. ). Likewise, the hydrological cycle and its components are highly affected by climate change. According to Marx et al. (), half of the rivers in Europe will experience a decrease of low flows (7-12%) under 1.5 K warming for the period 2047- To minimize uncertainties associated with future water management plans and to secure the future water needs of the global community, many efforts have been made to investigate the impact of climate change on water resources across the planet. However, the majority of the studies focus more on surface water bodies than on groundwater (Meixner et al. ; Saha et al. ), and African countries are underrepresented in these efforts (Field et al. ). It is well known that the social-ecological systems of developing countries are severely impacted by climate change due to the lowest capacity to adapt (Dile et al. ). This indicates that more efforts are required to investigate the effect of climate change in developing countries like Ethiopia.
Compared to surface water, groundwater is a reliable and cost-effective resource especially in many African countries and other parts of the world where availability of surface water is limited (Bovolo et al. ). One-third of the global freshwater is extracted from the groundwater source (Taylor et al. ). On Ethiopia's national level, as well as in the Lake Tana basin, groundwater provides 80% of the water demand (Kebede ).
In the Lake Tana basin, being a home for more than three million people and a headwater source of the Blue Nile River, several studies were carried out to investigate the impact of climate change on its hydrology. Nevertheless, the vast majority of the climate change studies were focused on streamflow or runoff analyses and there are disagree- (CMIP3). Their results indicated that the annual streamflow tends to decline significantly for most of the GCMs considered for the study, while the AET is expected to increase. A likely decrease in groundwater flow and soil moisture was reported. However, the magnitudes of changes in the groundwater flow as well as the spatial patterns of changes were not considered. Woldesenbet et al. () also studied the impacts of future climate change on the hydrological components of the Lake Tana Basin under RCP6.0. Their results showed that the groundwater contribution to the streamflow, percolation, and AET would increase on the annual time scale, but decreases are expected during the small rainy season compared to their reference period . Although this research is comprehensive in addressing the major water balance components, it does not address the impact of climate change on the major hydrological components during the mid-term and long-term of the 21st century as the authors focused only on a short-term period (2016)(2017)(2018)(2019)(2020)(2021)(2022)(2023)(2024)(2025)(2026)(2027)(2028)(2029)(2030). In addition, the high-level representative concentration pathway (RCP8.5) was not considered.
Thus, the main purpose of this study is to enhance our understanding of how projected changes in rainfall and temperature will affect the groundwater contribution to the streamflow and other major water balance components in Gilgelabay and Gumara catchments, which are the two major tributaries of Lake Tana during the mid-term (2031-2060) and long-term (2065-2094). The specific objectives are (i) to assess if groundwater contribution to the streamflow and other major water balance components has significantly changed in response to climate change and (ii) to determine if changes in the groundwater contribution to the streamflow show a distinct spatial pattern in the study area.

Study area
The Lake Tana basin is entirely located in the Amhara regional state in the north-western highlands of Ethiopia ( Figure 1). It is one of the sub-basins of the Blue Nile River in which the largest fresh water lake in the country is found. The total catchment area of the Lake Tana basin is 15,321 km 2 . More than 40 streams flow into Lake Tana (Alemayehu et al. ). Gilgelabay (catchment area ∼5,004 km 2 ) and Gumara (catchment area ∼1,394 km 2 ) contribute about 70% of the inflow to the lake. The Blue Nile (locally referred to as Abay) is the only surface outflow from the lake with an average annual flow volume, as calculated from raw data , of 3.9 billion m 3 (123 m 3 /s) measured at the lake outlet.
Similar to its hydrologic variability, the Lake Tana basin has a heterogeneous hydrogeology. According to Kebede et al. (), areas surrounding Lake Tana are covered by quaternary basalts and alluvial sediments. Being driven by the diversified parent geology, the soil structure, lateral and vertical extents, and hydraulic conductivity, the basin is highly diversified. The soils vary from hydrologic group B to group D, which represent infiltration rates from moderate to very slow. Additional information can be found in The input component to the hydrologic balance equation is rainfall and is partitioned into AET, water entering to the vadose zone, and surface runoff.
In this study, calibrated SWAT models for Gilgelabay and Gumara catchments (Tigabu et al. ) were used. As the catchments are characterized by diversified soil, topographic, and hydrogeological features, model parameters that are related to the groundwater flow system were adjusted through intensive sensitivity analysis tests in SWAT-CUP (Abbaspour et al. ). Considering the availability and continuity of the streamflow data of the catchments, we split our modeling period into a warmup period (1980)(1981)(1982)(1983)(1984), calibration period (1985( -19951988-1996, and validation period (1996( -20141997-2011. The model parameter settings were defined using the Latin hypercube sampling algorithm implementation in hydroGOF (Zambrano-Bigiarini ). Six thousand model simulations that include different combinations of 10 sensitive parameter values were tested, and then the best model run was chosen based on the optimized Nash-Sutcliffe efficiency (NSE) from 6,000 simulations. Moreover, other objective functions, such as Kling-Gupta efficiency (KGE), percent bias (PBIAS), and standardized root mean square error (RSR) were also used to test the model performance.
In addition to statistical indices, the fitness of the models in capturing the monthly streamflow data was also evaluated by comparing the measured and simulated streamflow hydrographs and flow duration curves. The flow duration curves showed very good agreement for the middle and CORDEX coordinates RCMs to improve regional climate downscaling models and techniques and to produce coordinated sets of regional downscaled projections worldwide (Giorgi & Gutowski Jr     . Those RCMs that show a bimodal pattern of rainfall, which is not observed for the study area (e.g. CLM4) are not used for further analysis.  | Exemplary plots presenting seasonal patterns and agreement of the bias-corrected mean monthly rainfall generated from regional climate model  with the monthly mean of observed rainfall of Gumara catchment. The red lines represent the monthly mean rainfall of different climate models before bias correction was applied and the other colours represent the rainfall after different bias correction methods were applied. Suitability of bias correction methods differs from one regional climate model to the other.
Here, we presented the rainfall data generated from different regional climates that showed good agreement with observed rainfall at Debretabor station only as an example.
between the baseline and mid-term, baseline and long-term,  reported that there was no significant change in the ensemble median rainfall from 18 GCMs. While the direction of projected rainfall changes varies across the GCM-RCM outputs (Figure 6), all models show a rising trend in both minimum and maximum temperature. Setegn et al. () also concluded that there is no consensus on the direction of rainfall changes of different GCMs. In our study, we noticed that each and every 30 years moving average maximum and minimum temperatures (for a data series from 1985 to 2100) is expected to be higher than the preceding one (Figure 7). The absolute temperature change varies between 1.3 to 2.7 C and 2.0 to 3.8 C for the mid-and long-term of the century, respectively. The mid-term projection deviates from the baseline average for    The numbers represent the absolute change (first number) and p-value (second number), where numbers in bold indicate significant changes.

Impacts on groundwater
During the baseline period, groundwater accounts for more than 50% of the streamflow for each catchment (Tigabu et al. ). Under RCP4.5, the groundwater contribution to the streamflow (GWQ) is decreased. The decreases of GWQ vary between 3 and 57%. These decreases are projected even for positive changes in the projected rainfall.
For instance, in the Gumara catchment, statistically significant decreases in GWQ (high confidence, p-values ¼ 0.001) for both time periods are likely (Table 3)  periods for Gumara ( Figure 10(b), Table 3). In summary, changes related to GWQ are mainly negative for both catchments under the two RCPs, and these changes are being driven by the increasing level of rainfall intensity that could increase SURQ. Moreover, the annual percolation will decrease under the given RCPs (Table 3)

Surface runoff
Under the assumption that agricultural management stays the same, our results indicate that SURQ in Gumara is more considerable than in the Gilgelabay catchment ( Figure 11)

Spatial patterns of changes in GWQ
Besides the expected temporal changes of GWQ in the study catchments, changes are also projected in the spatial patterns. Influences of the projected rainfall and temperature vary from the highland to the lowland portion of the catchments. The projected changes are higher in the lowlands than in the highlands for both catchments. For Gumara catchment, the expected changes are negative and vary from À3 to À32% for both RCPs (Figure 12). Although we expect differences in the magnitude of changes between RCP4.5 and RCP8.5 for both mid-and long-term periods, our results reveal that spatial differences between RCPs are small ( Figure 12). On the contrary, in Gilgelabay catchment, the degree of influence of the two RCPs on the spatial variability differs more considerably. Under RCP8.5, projected changes will mostly be negative while positive changes are expected under RCP4.5 ( Figure 13). Compared to Gumara, the percentage changes in Gilgelabay show a wider range that varies from 26 to À58%. This wide range of changes in Gilgelabay catchment can be explained by the variability in rainfall. Four rainfall stations were used for Gigelabay, while only one station was used for Gumara (Tigabu et al. ).

CONCLUSION
In this study, we investigated the future (2031-2060 (midterm), 2065-2094 (long-term)) temporal changes of groundwater contribution to streamflow and of major water balance components in two catchments (Gilgelabay and Gumara) of the Lake Tana basin, in north-western Ethiopia under RCP4.5 and RCP8.5 using CORDEX datasets (CMIP5). Impacts on water resources were assessed using the hydrological model SWAT under the assumption that the current agricultural management practices and land cover conditions would not change.
Distinct spatial patterns are expected in the groundwater contribution to streamflow for the two catchments.
The declines of groundwater contribution to streamflow will be higher in lowland portions of the two catchments than in the highlands.
The ensemble mean rainfall is not expected to show significant change for both the mid-and long-term periods.
However, the rainfall intensity is expected to be higher than during the baseline period. Consequently, the anticipated surface runoff is expected to increase, whereas groundwater contribution to streamflow is projected to decline. Therefore, we recommend to analyse changes in rainfall intensities alongside changes in rainfall amounts prior to a hydrologic impact assessment.
The higher projected surface runoff that results from the increases in rainfall intensity will lead to an increase in soil erosion in both catchments. Moreover, the flood plain area of Gumara catchment may experience a higher flood risk in the future. Hence, to mitigate erosion and flood risks that are anticipated as a result of increasing surface runoff, the construction of small scale reservoirs to store more surface water for domestic water supply and small scale irrigation may be advisable. Furthermore, agricultural management practices that enhance infiltration can be recommended to mitigate climate change impacts.