The Songwe sub-basin (Figure 1) covers an area of roughly 10,800 km² and is situated in the eastern portion of the Lake Rukwa Basin in southern-western Tanzania. It is found between latitudes 07˚40'S and 09˚20'S and longitudes 33˚00'E and 33˚50'E. All river catchments that pour into Lake Rukwa have one outlet, describing an internal drainage system. There are various hills in the area that cuts through fertile valleys, and their elevations range from 600 to 2400 masl. The catchment is distinguished by generally healthy, fertile soils, consistent, and copious amounts of precipitation, mild temperatures, and excellent agricultural potential. According to the 2012 National Census, the sub-basin has a population of roughly 843,278 people, 40% of whom reside in Mbeya City and the adjacent peri-urban areas. By 2035, the sub-basin population is expected to increase to around 1,643,629 people. One of Tanzania’s most productive regions is the Songwe watershed. About 80% of the population works in agriculture is the primary industry. Mbeya City, a significant regional economic and industrial hub, is also located within the watershed.

Due to topographical height, differences in rainfall distribution based on time

and place, and related land use activities that occurred in the sub-basin that influenced river flow, soil erosion, and sediment yield to rivers and Lake Rukwa, the Songwe sub-basin has a diverse bioclimate. The sub-basin primarily has a tropical climate and has one extended rainy season (October to May). From about June to September, there is a dry period. The lowlands receive 650 mm of annual precipitation, while the highlands receive 2600 mm. The altitude-dependent mean temperature ranges from about 16˚C in the highlands to roughly 30˚C in the lowlands. Three significant rivers drain the northern, central, and southern portions of the Songwe Sub-basin: the River Lupa, the River Songwe, and the River Zira, which originate in the Poroto Mountains and drain the southern portion. The rivers travel through the gold mine area, crossing wide flat lowlands, combining, and then emptying into Lake Rukwa. Strong seasonality is present in all sub-basin rivers, with high flows during the rainy season and low flows during the rest of the year. The rivers see their highest flows between February and March and their lowest flows between July and November, with September being the driest month. Climate change, which has led to significant fluctuations in temperature and precipitation and has serious effects on the management and use of the sub-basin water resources, has made the issue worse.

The stream network and hydrologic response units are easier to define in the SWAT model simulation due to the streamflow data, topographic features, and spatial distribution of land cover/use and soil types. The Ministry of Water Tanzania’s Lake Rukwa Water Basin office provided the daily streamflow information for the Songwe sub-basin at Galula station. One of the essential inputs used by SWAT to divide the sub-basin into smaller sub-basins and to examine the sub-drainage basin’s pattern, slope, stream length, and channel width is the Digital Elevation Model (DEM). In this investigation, a DEM with a 30 m spatial resolution was collected from the Advanced Land Observing Satellite Phased Array L-band Synthetic Aperture Radar (ALOS PALSAR) at the Alaska Satellite Facility, which may be found online at http://www.asf.alaska.edu. Images of land use and land cover (LULC) for the year 2020 used in this study were downloaded for free from the United States Geological Survey (USGS) website at https://earthexplorer.usgs.gov/, and categorized in line with the input specifications of the SWAT model. From the FAO-UNESCO Soil Map of the World, Volume VI at https://swat.tamu.edu/data/ online, a soil map was constructed. The water balance model in SWAT depends on data on daily precipitation, maximum and minimum air temperatures, relative humidity, wind speed, and solar radiation. Data on the daily rainfall, were gathered from the Ministry of Water’s Lake Rukwa Water Basin office. While data on the minimum and maximum temperatures, solar radiation, wind speed, relative humidity, and wind direction were obtained from the updated Global Weather Database for SWAT. The Lake Rukwa Water Basin office provided daily baseline data for the Lupa tinga tinga and Galula stations. The baseline scenario was examined and calibrated using data from the baseline era, which ran from 1981 to 1992.

The CORDEX-Africa runs of the Climate Model Intercomparison Project Phase 5 (CMIP5) were downscaled from GCMs utilized in this study [8] . The Consortium for Small-scale Modeling (COSMO), Climate Limited-Area Model (CCLM), SMHI Ross by Centre Regional Climate Model (RCA4), Max Planck Institute Regional Model (REMO), KNMI-Regional Atmospheric Climate Model (RACMO22T), and High-Resolution Limited Area Model (HIRHAM5) are some of the most well-known CORDEX RCMs used in Africa [7] . For this investigation, the four RCMs produced through the CORDEX-Africa Data Search (HIRHAM5, CCLM4, RACMO22T, and RCA4) were employed (Table 1). RCMs used in this work have a 0.44 degree grid spacing (50 km of resolution). The reference period is changed to reflect the period of observations that are available (1981-1992). Using the spatial resolution of 0.44˚ and the RCP4.5 and RCP8.5 emission scenario paths, an ensemble mean from the regional climate models CORDEX Africa [15] was taken at the model boundaries. The Earth System Grid Federation-Lawrence Livermore National Laboratory website, https://esgf-node.llnl.gov/projects/esgf-llnl/, is where the CORDEX data were obtained. In the context of the Coordinated Regional Climate Downscaling

Experiment [12] , the ensemble means data are available over Africa at 0.44˚ resolution, and they have already been used over Africa. The studies of the data concentrated on the future eras and the reference period (1981-1992) (2011-2100). Climate model data for hydrologic modeling (CMhyd) for four RCMs outputs and give climate data for the SWAT model were bias-corrected by the linear scaling method because precipitation and temperature are the main drivers of the hydrological regime of climate change. For the bias correction of the four RCMs’ output of climate data, data on precipitation measured at gauging stations in the sub-basin and the on minimum and maximum temperature gathered from SWAT Global Weather Data were used.

In vast catchments with changing soils, land use/land cover, and management conditions, the SWAT model uses a physically-based, semi-distributed approach to anticipate the effects of land management methods on water, sediment, and nutrients [16] . Comprehensive data on the weather, soil characteristics, and terrain, vegetation, and land management techniques are required for the SWAT model [17] . The SWAT model was used to simulate the water balance for the chosen period and under changing climatic conditions [18] . [16] provides a thorough explanation of the SWAT model. The DEM-derived drainage patterns that the model uses to construct small sub-basins that are then divided into smaller sub-basins by a threshold that specifies the required amount of drainage to create a stream [19] . Hydrologic response units (HRU) with unique LULC classes, soil types, and slope classes are created from these small sub-basins [20] . Each hydrologic response unit runoff is anticipated independently and routed to determine the sub-overall basin’s runoff. In general, the SWAT model integrates climatic station data at the sub-basin while solving the water balance equation for each HRU and adding the HRU calculations for each sub-basin [16] . For each HRU, the hydrological balance is simulated using the water balance equation [16] . The Songwe sub-basin, which has an extent of around 10,800 km², is one of the catchment that the SWAT model can be applied to. The application addresses the bias-correction of RCMs temperature and precipitation, as well as the development, calibration, and validation of the SWAT model for the Songwe sub-basin and the forced simulation of the model using a set of bias-corrected RCM outputs to evaluate future climatic and river flow change for the years 2011-2100. Galula station streamflow data from 1981 to 1992 were used to assess the model’s calibration and performance. Due to its ability to physically foundation and continually simulate hydrological processes, the SWAT model was chosen for this study. These characteristics are crucial for modeling the effects of climate change on streamflow. The model can calculate the basin’s hydrologic water cycle by integrating various spatial data, observed data, and anticipated climatic data (Arnold et al., 2012), which makes it valuable for catchment management. The SWAT is a hydrological model with a physical foundation that integrates with QGIS. For the model setup in this study, QSWAT 1.7 version, compatible with QGIS 2.6.1 interface, was employed. Spatial, hydrological, meteorological, and projected climate data were among the inputs gathered to create the model. Figure 2 shows satellite pictures of land use and land cover,

Figure 3 shows a digital elevation model with a 30 m resolution, and Figure 4 shows a soil map of the Songwe sub-basin. Figure 5 depicts the placement of the metrological and hydrological stations. Table 2 in Figure 2 displays the land use/cover in the Songwe sub-basin. The DEM and river channels of the study area served as the foundation for the model setup.

Water Balance Equation;

S W t = S W o + ∑ i = 1 t ( R d a y − Q s u r f − E a − W s w e e p − W g w ) t (1)

S W t is the final soil water content,

S W o is the initial soil water content on day i in mm,

R d a y is the amount of precipitation on day in mm,

Q s u r f is the amount of surface runoff on day i in mm,

E a is the amount of evapotranspiration on day i in mm,

W s w e e p is the amount of water entering the vadose zone from the soil profile on day i in mm,

W g w is the amount of return flow on day i in mm,

t is the time in da.

The land phase of the hydrological cycle and the water phase of the hydrological cycle are the two stages of the SWAT. The process of water transportation through the stream network was chosen for this study’s focus area: the water

phase. Additionally, the SWAT model employed the Penman-Monteith model to estimate potential evapotranspiration [14] . The total volume of surface runoff was calculated using The Curve Number (CN), which requires daily precipitation data. The Songwe sub-basin was divided into several smaller sub-basins for the SWAT model, and these smaller sub-basins were then separated into units with distinctive soil and land use characteristics called hydrological response units (HRUs). These HRUs are described as homogeneous spatial units with comparable hydrological and geomorphological characteristics [1] . The sub-basins were then divided into a total of 345 hydrologic response units based on land use, soil type, and slope (HRUs). The model was first filled with the reclassified land use, land cover, and soil data before the HRUs were specified. The climatic data were entered into the model and written using the SWAT Model’s Write SWAT Input Tables interface. The Write SWAT Input Tables interface of the SWAT Model was used to load and write the model’s climate data, which included precipitation (pcp), temperature (tmp), relative humidity (rh), solar radiation, and wind speed. The SWAT model was run once the model’s settings were modified and tables for observed weather data had been produced.

The goal of model calibration is to optimize the model’s agreement with a collection of experimental data by adjusting a set of parameters. It involves changing model parameters based on comparing outcomes to observations in

order to maintain the same response over time. From the standpoint of the model’s intended usage, validation is the act of assessing how accurately a model represents the real world. Using observed streamflow data from Galula station for the years 1981 to 1992, the model was calibrated and validated. The SWAT-Calibration and Uncertainty Program’s Sequential Uncertainty Fitting version 2 technique (SUFI-2) was used to identify and calibrate the SWAT model’s sensitive parameters (SWAT-CUP). The SWAT model had to simulate flows with bias-corrected temperature and rainfall over the future periods and in the baseline period after being calibrated and validated using data on observed stream flows. In accordance with four regional climate models, the simulated stream flows for the 2040s, 2070s, and 20100s correspond to two scenarios.

The Galula gauging station in the Songwe sub-basin served as the calibration and validation site for the semi-distributed SWAT hydrologic model. With goodness-of-fit values of NSE 0.45, R² 0.59, and RSR 0.73 for the calibration period and NSE 0.59, R² 0.59, and RSR 0.64 for the validation period, the model was able to mimic stream flows. Table 3 shows that stream flows from the Songwe sub-basin could be simulated by the model.

The parameters utilized for the SWAT model’s calibration were subjected to a global sensitivity analysis using the SWAT-CUP. The findings demonstrated that the stream flow calibration’s sensitive parameters included CN2, SOL AWC, GWQWN, GW DELAY, and GW REVAP (Table 4).