Assessment of the Impacts of Climate Change on Hydrological Characteristics of the Mbarali River Sub Catchment Using High Resolution Climate Simulations from CORDEX Regional Climate Models

This study assesses the impacts of climate change on water resources over Mbarali River sub-catchment using high resolution climate simulations from the Coordinated Regional Climate Downscaling Experiment Regional Climate Models (CORDEX_RCMs). Daily rainfall, minimum and maximum temperatures for historical climate (1971-2000) and for the future climate projection (2011-2100) under two Representative Concentration Pathways RCP 8.5 and RCP 4.5 were used as input into the Soil and Water Assessment Tool (SWAT) hydrological model to simulate stream flows and water balance components for the Mbarali River sub-catchment. The impacts of climate change on hydrological conditions over Mbarali river catchment were assessed by comparing the mean values of stream flows and water balance components during the present (2011-2040), mid (2041-2070) and end (2071-2100) centuries with their respective mean values in the baseline (1971-2000) climate condition. The results of the study indicate that, in the future, under both RCP 4.5 and RCP 8.5 emission scenarios, the four main components that determine change in catchment water balance (rainfall, ground water recharge, evaporation and surface runoff) over Mbarali river catchment are projected to increase. While the stream flows are projected to decline in the future by 13.33% under RCP 4.5 and 13.67% under RCP 8.5 emission scenarios, it is important to note that simulated surface runoff under RCP8.5 emission scenario is higher than that which is obtained under the RCP4.5 emission scenario.


Introduction
According to the Intergovernmental Panel on Climate Change (IPCC) assessment reports, climate change is projected to substantially reduce water availability in the watershed (IPCC, 2007(IPCC, , 2013(IPCC, and 2014)).In developing countries, Tanzania inclusive, by the year 2020, between 75 to 250 million of people are predicted to be exposed to increased water stress due to climate change (IPCC, 2007).Indeed, the only strategy to reduce impacts of climate change on water resources is to invest in the development of adaptation strategies.However, the development of water resources adaptation strategies to overcome impacts of climate change on hydrological systems is challenging (Muerth et al., 2014;Piani et al., 2010).The challenges on one hand are attributed by a lack of scientific evidence which shows the projection of how the future climate change will impact the hydrological systems.On the other hand, there are high uncertainties associated with the projections.
In Tanzania several studies that address climate change impacts on water resources have been done (e.g.Mwandosya et al., 1998).These studies have used climate simulations derived directly from the General Circulation Models (GCMs) to evaluate the impacts of climate change on water resource in Tanzania.However, the GCMs have coarse spatial resolutions (500 or 1000 km) and are designed to simulate global or continent climate characteristics like global or continent temperature or rainfall amount.The coarse spatial resolutions of The RCMs and the driving GCMs used in this study are listed in Table 1.These models were chosen based on their ability to simulate the historical  climate condition over the southern regions of Tanzania with relatively minimum error (Luhunga et al., 2017).For a detailed description of the regional climate models and their driving general circulation models used in the CORDEX program a reader may consult Nikulin et al. (2012).

Hydrological Model and Model Input Files
To assess future water resource availability for the Mbarali River Catchment the Soil and Water Assessment Tool (SWAT) models were used.SWAT is a semi distributed hydrological model that can quantitatively explain the processes and mechanisms that influence the behavior of the watershed (Arnold et al., 1998).This model has been used intensively by different researchers, consultants, hydrologists, policy as well as decision makers worldwide (Abbaspour et al., 2007).In this study, SWAT v10.4 was used to simulate hydrological conditions over the mbarali River sub catchment.For a detailed description of SWAT v10.4 the reader may consult Abbaspour (2009).
SWAT requires DEM, soil data, and weather information to simulate hydrological conditions.Weather information required to run SWAT includes daily values of incoming solar radiation (MJ/m 2 -day), maximum and minimum daily air temperature (°C) and daily rainfall (mm).The model input files were created using several levels in SWAT.For a detailed description on how to create the model input files using SWAT reference can be made to Abbaspour(2009).

Model Calibration and Validation
The Soil and Water Assessment Tool (SWAT)-hydrological model is incorporated within Arc GIS version 10.4 to simulate hydrological condition as influenced by catchment characteristics.The calibration and validation of SWAT-Hydrological model to obtain reasonable estimates of model evaluation performance was performed by comparing simulated and observed stream flows data from 1KA11 Igawa maji gauge station.Daily rainfall, minimum and maximum temperatures from four CORDEX regional climate models were used to force SWAT to simulate hydrological parameters.

Assessment of the Impacts of Climate Change on Water Resource
The assessment of the impacts of climate change on water resources was carried out by comparing hydrological parameters simulated in historical climate  against future climate (2011-2040, 2041-2070and 2071-2100 under two emission scenarios (RCP 4.5 and RCP 8.5).Since different CORDEX regional climate models simulate climate variables at specific location differently, this may contribute to large uncertainties in the simulated hydrological parameters.To address the issue of uncertainties introduced from the climate models, the ensemble average of four CORDEX regional climate models driven by three different GCMs was constructed.
Outputs from the constructed ensemble average for RCP 4.5 and RCP 8.5 emission scenarios were used to force SWAT model to simulate the stream flows and water balance components during historical , and future (211-2100) climate.

Results
The results are presented in two sub sections.The first sub section presents the climate variables over the Mbarali river sub catchment as simulated by the climate models.The second sub section presents the hydrological characteristics over Mbarali river sub catchment as simulated by SWAT hydrological model.

Climate Variables over Mbarali River Sub-catchment
In the Mbarali river sub catchment, the RCMs from CORDEX simulate historical  climate variables (rainfall, minimum and maximum temperatures) differently .All the RCMs reproduce the unimodal rainfall pattern over Mbarali river sub catchment (Figure 3).However, all RCMs simulate phase change and the magnitude of rainfall in the historical   In order to reduce uncertainties associated by the individual RCMs and GCMs, the multi-model approach or ensemble average of simulated climate variables (rainfall, minimum and maximum temperatures) was created and the results can be compared with those from individual models (see Figures 3-5).Taking the ensemble average as a reference, it is found that the absolute error (biases) from the RCMs and driving GCMs contribute almost equally in simulating both rainfall and temperatures over Mbarali river sub catchment.Moreover, during the historical climate  the RCMs simulated the patterns and magnitude of minimum and maximum temperatures better than those of rainfall.

Impact of Climate Change on Stream Flows
The long term means annual cycles of the simulated stream in baseline period  and future climate projections (2011-2100) under two emission scenarios (RCP 4.5 and RCP 8.5) for the Mbarali river sub-catchment are presented in Figure 9.This figure indicates two peaks of stream flows; the maximum peak is found in March and the minimum peak is found in December.The figure further shows a decline of stream flows during present, mid and end centuries from the month of March to July and from October to January.More decline of stream flows is projected to occur in the end century under both RCP 4.5 and RCP 8.5 emission scenarios.The decline of stream flows particularly during the month of October to January, affects the beginning of the hydrological year of the Mbarali river sub-catchment, which starts in November and lasts in April.The presented results agree with those obtained in the study carried by Rajabu et al. (2007) in the Upper Great Ruaha River Catchment (UGRRC).They found a decline in the annual stream flow, but with large variations along the year.
Table 2 shows the simulated annual streamflow during the baseline period  is 36.74m3s-1.These flows are projected to decrease during the present (2011-2040), mid (2041-2070) and end (2071-2100) centuries under both RCP 4.5 and RCP 8.5 emission scenarios.More decrease of 15.91% is projected to occur during the mid and end centuries under RCP 4.5.In the present century, more decrease of stream flows of 13.99% is projected to occur over Mbarali river sub-catchment under RCP 8.5 emission scenario., the highest surface runoff of 13.95 mm is simulated by SWAT forced with HIRHAM-ICHEC and the lowest surface runoff of 0.15 mm is simulated by SWAT forced with RACMO22T-ICHEC.The uncertainties associated by different CORDEX-RCMs can be analyzed, when SWAT is forced by different RCMs, but similar simulates water balance components differently.For instance, in the historical climate, SWAT simulates water yield of 719.72 mm and 614.99 mm when forced by HIRHAM-ICHEC and RACMO22T-ICHEC respectively.Moreover, SWAT forced by similar RCM, but different GCMs simulates water yield differently.For instance water yield of 825.42 mm and 1030.86 mm was simulated by SWAT driven by RCA4-CNRM and RCA4-MPI respectively.The uncertainties from the climate model may be associated with different formulation and parameterization schemes in those models.To account for uncertainties that arise from RCM-GCM combinations, the ensemble averages of four climate model members were constructed and used to force SWAT.
Table 7 shows annual average for the water balance components simulated by SWAT forced with the ensemble averages.These water balance components differ greatly from those of individual models.This is an indication of large uncertainties involved in climate change impact studies for the Mbarali river sub-catchment.However, the results from the ensemble averages that take into account the uncertainties from individual RCMs and driving GCMs are the best estimate of future climate change in the sub basin.Table 7 indicates that the ground water and water yield component are projected to decrease under RCP 8.5 during the present, mid and end centuries, while the surface runoff and evapotranspirations are projected to increase under two emission scenarios during the present, mid and end centuries.

Discussion
The uncertainties in the previous study done by Wambura (2015) on the impacts of climate change on stream flows using climate change simulations derived by GCMs offer a room for more research to be done on the hydrological responses due to climate change.In this study, we have used high resolution climate information from four Regional Climate Models (RCMs) driven by three General Circulation Models (GCMs) and Soil and water Assessment tool embedded in Arc GIS to simulate stream flows over the Mbarali River sub catchment during historical, and future periods under two emission scenarios (RCP 4.5 and RCP 8.5).
This work aimed at assessing, how climate change will affect hydrological characteristics of the Mbarali River Sub catchment in the Upper Great Ruaha sub basin of Rufiji Basin.This River sub-catchment is an agricultural watershed which produces food and cash crop in Tanzania country (Milder et al., 2013).As a result, much assessment on how climate change will impact future water availability within Mbarali River sub catchment is of great important.The study revealed that temperatures are projected to increase throughout the sub-catchment.Different RCM-GCM combinations project increase in temperatures differently.Yet, the ensemble average of the RCMs approves that the sub-catchment will have high change in temperatures during the end century under RCP 8.5.
The ensemble average project that minimum temperature will increase by 14% (1.9 0 C) under RCP 8.5 and maximum temperature will increase by 7.68% (1.8 o C) under the same scenario RCP 8.5.These increases in temperature will raise the evaporation within the sub-catchment and hence contributing to the decrease in the stream flows, particularly in the warmer low altitude areas.
The ensemble average of the River Sub-catchment value and climate variable are presented on Table 7.The hydrological model SWAT simulates differently when two different RCMs are forced by similar GCM.The differences are related with the design of the RCMs themselves.For the case, RACMO22T and HIRHAM5 enforced by similar GCM(ICHEC-EARTH) simulate different amount of stream flows , surface runoff , water yield and ground water yield in historical climate , present climate (20110-2040), mid (2041-2070) and end (2071-2100) centuries.
Likewise, different amount of water balance components are simulated when are forced by the same RCM derived by different GCMs.For instance, SWAT simulates different amount of stream flows and surface runoff, and water yields even when is forced by the same RCM (RCA4) driven by different GCMs (CNRM, and MPI).
In this case, the difference comes from the variations of formulation of the driving GCMs.The annual rainfall projections show a significant increase in the averages of the three projected periods relative to the baseline period for both RCP4.5 and RCP8.5 scenarios.These projections converge with the projections of temperatures in the River sub catchment.Moreover, as with the temperatures, the increase under the RCP8.5 scenario is higher than that obtained under the RCP4.5 scenario.
The projected decrease of the stream flows in the Mbarali River Sub-catchment may result in the decrease in the amount of water availability.This may have negative impacts on the aquatic biota, water quality and agriculture.The projected increase in the surface run-off may cause serious environmental problems such as soil erosion and floods.
It is important to note that the uncertainties related to the current assessment, which is based on the downscaling of the high resolution regional climate model driven by three members of the GCM (ICHEC, MPI and CNRM) simulations of the RCPs emission scenario were considered.The included simulations are the most and the least climate sensitivity member of the GCM runs, and it is believed that more environmentally concerned and more business-as-usual scenarios have been somewhat considered in these simulations.As a result, some insights in the possible future changes to the river flow under the future climate change have been provided in this work.Despite uncertainties of the projections, the results showed some level of agreement with other works, and therefore they contain some level confidence to the hydrological impacts in the Mbarali River Sub-catchment, which makes the information a useful guidance for local adaptation measures.

Conclusion
In this study we assessed the impact of climate change on stream flows and hydrological condition of the Mbarali River sub catchment which is located 1530 km 2 in the Upper Great Ruaha of the Rufiji Basin, Tanzania.SWAT, a distributed hydrological model was employed to simulate the stream flows and hydrological condition of the River sub-catchment.The climate variable graded from the cordex project for the present period  was compared with the three future periods (2011-2040, 2041-270 and 2071-2100) with their respective RCP 4.5 and RCP8.5 as input into a distributed hydrological model, SWAT.Calibration and validation of SWAT model show that the tool can be a reliable for hydrology cycle simulation in Mbarali River sub-catchment.
The future annual rainfall and temperature are projected to increase while the annual stream flows are projected to decrease.A significant decrease of projected monthly stream flow during the year might cause drying up of the Mbarali river and hence decreasing the environmental condition of the aquatic organism in the future.Additionally, the decreases in monthly stream flow during March and April would lead to water shortage problems that have been occurring in previous years.The climate of the study area is becoming hotter in both wet and dry seasons.To couple with all this condition, the basin water users and other stakeholder should apply integrated water resources management at all levels.

Figures 6 -
Figures 6-8 present the projected annual cycles of climate variables (rainfall, minimum and maximum temperatures) in the Mbarali river sub catchment as simulated by the CORDEX RCMs under two Representative Concentration Pathways (RCP8.5 and RCP 4.5).It can be seen that all the models can reproduce the annual cycles of the historical climate in the future climatic condition.However, rainfall is expected to increase in the present century (2011-2040) under both RCP 8.5 and RCP 4.5 emission scenarios.Moreover, minimum and maximum temperatures are projected to increase in the present century (2011-2040) under both RCP 8.5 and RCP 4.5 emission scenarios.

Figure 3 .Figure 6 .
Figure 3. Simulated annual cycle of rainfall during the historical climate

Figure 8 .
Figure 8. Simulated annual cycles of climate variables (rainfall, minimum and maximum temperatures) in the end century (2071-2100): the upper panel is for RCP 8.5 and the bottom panel is for RCP 4.5 emission scenarios Mean monthly stream flows of RCM ensemble under RCP 4.5 and RCP 8.5 for the three future period and baseline period 3.2.2The Impact of Climate Change on Water Balance ComponentsThe average annual values of water balance components for the Mbarali river sub catchment simulated by SWAT hydrological model when is forced by climate variables from different RCM-GCM combinations during the baseline and future (2011-2100) climate conditions are presented in tables 3-7.During historical climate Average annual basin values as simulated by SWAT Model fed with climate data from HIRHAM -ICHEC for two scenarios RCP 4.5 and RCP 8

Table 1 .
CordexRcms and the Driving Gcms

Table 2 .
Mean annual stream flows for the baseline and future period as simulated by SWAT forced by ensemble RCMs under RCP 8.5 and RCP 4.5 scenario (All values are in m3/sec)

Table 4 .
Average annual basin values as simulated by SWAT Model fed with climate data from RACMO22T-ICHEC for two scenarios RCP 4.5 and RCP 8.5.Note that figures in brackets are % change

Table 5 .
Average annual basin values as simulated by SWAT Model fed with climate data from RCA -CNRM for two scenarios RCP 4.5 and RCP 8.5: Note that figures in brackets are % change

Table 6 .
Average annual basin values as simulated by SWAT Model fed with climate data from RCA -MPI for two scenarios RCP 4.5 and RCP 8.5: Note that figures in brackets are % change

Table 7 .
Average annual basin values as simulated by SWAT Model fed with climate data from RCM Ensemble average for two scenarios RCP 4.5 and RCP 8.5: Note that figures in brackets are % change