Uncertainity reduction in climate and hydrological models predictions at catchment scale in the upper great Ruaha river sub-basin, Tanzania

Abstract

Water resources have become scarce in most tropical areas of Tanzania due to climate change. Any changes to the hydrological cycle may have significant effects on the water resources in the river basins of Tanzania. The impact of climate change on water resources in Tanzania have been studied using General Circulation Models (GCM) which run at low spatial resolutions of 100-300 km. The resolution is too coarse to provide useful information about climate change impact in small catchments as many physical processes which control local climate e.g.; vegetation, hydrology, topography is not fully parameterized and hence results on uncertainty in model prediction. The main aim of this research was to quantify the uncertainty in model predictions for the Mbarali River Sub-catchment of the Upper Great Ruaha River Sub-basin in the Rufiji River Basin, Tanzania. Three research objectives were analyzed; the first objective was to evaluate the performance of the Coordinated Regional Downscaling Experiment Regional Climate Model (CORDEX, Regional Climate Models) in simulating rainfall characteristics of the Mbarali River Sub catchment. The area weighted average method was used to calculate the average rainfall from the CORDEX RCMs and from ERA-Interim reanalysis over the entire Mbarali River sub-catchment. Comparison between rainfall data from CORDEX RCMs and ERA-Interim reanalysis was done to test the ability of the CORDEX RCMs to reproduce the annual cycles, interannual variability, annual total and trends of rainfall as presented by the ERA-Interim reanalysis. The second objective assessed the impact of climate change on hydrological characteristics using the Soil and Water Assessment Tool (SWAT) model. The ability of the SWAT model to simulate catchment processes was assessed through a calibration and validation process, which was a key factor in reducing uncertainty and increasing user confidence in its predictive abilities. The SWAT model was driven by high resolution climate simulations for historical climate condition (1971-2000) as well as future climate projections (2011-2040, 2041-2070 and 2071-2100) for two Representative concentration Pathways (RCPs): RCP 4.5 and RCP 8.5. Furthermore, Ensemble of RCMs was applied into SWAT to simulate water resources availability and the results were compared with individual models (HIRHAM5, CCLM4, RACMO22T, RCA4). The Rainfall and Temperature data were obtained from the selected four CORDEX RCMs driven by three different General Circulation Models (GCMs). Inverse Distance Weight Average (IDWA) was used to interpolate model gridded climate simulation to the location of weather station. The third objective assessed the impacts of land use and land cover change on the hydrology using integration of remote sensing data, QGIS and SWAT model. The land use and land cover (LULC) maps for three window period snapshots, 1990, 2006 and 2017 were created from Landsat TM and OLI_TIRS. Supervised classification was used to generate LULC maps using the Maximum Likelihood Algorithm and Kappa statistics for assessment of accuracy. The findings of the first objective are that CORDEX RCMs were able to capture well the seasonal and annual cycles of rainfall. However, they underestimated the amount of rainfall in March, April and May (MAM) and overestimated in October, November and December (OND) respectively. CORDEX RCMs reproduce interannual variation of rainfall. The source of uncertainties was revealed when the same RCMs driven by different GCMs and when different RCMs driven by the same GCM in simulating rainfall. It was found that the error and biases from RCMs and driving GCMs contribute roughly equally. Overall, the evaluation found reasonable (although variable) model capability in representing the mean climate, interannual variability and rainfall trends. The results suggest that CORDEX RCM is suitable in simulating rainfall, maximum temperature and minimum temperature. The findings of the second objective showed that SWAT model simulated stream flow and water balance components differently when two different RCMs were forced by the same GCMs as well as when the same RCMs were forced by different GCMs. The differences are related to the formulation of the RCMs themselves. For example, RACMO22T and HIRHAM5 driven with the same GCM (ICHEC-EARTH) simulate different amount of stream flows, surface runoff, water yield and groundwater yield in historical (1971–2000) as well as in present century (2011-2040), mid-century (2041- 2070) and end century (2071-2100). Ensemble RCMs projected decrease in stream flows by 13.67% under RCP 8.5. However annual rainfall was shown to increase in averages by 1.62% under RCP 4.5 and by 1.96% for RCP 8.5 relative to the 1177.1mm of the baseline period (1971-2000). The results also showed that, temperature will slightly increase relative to the baseline during present century (2011-2040) for RCP 4.5 and RCP 8.5. The ensemble average project that the minimum temperature will increase by 14% (1.9 0 C) under RCP 8.5 and maximum temperature by 7.68% (1.8 o C) under RCP 4.5 The findings of the third objective showed that there were significant changes in land use and cover for the three-time periods (1990, 2006 and 2017). The cultivated land and built up area increased from 25.69% in 1990 to 31.53% in 2006 and 43.57% in 2017 compared to other land classes. Increase of cultivated land and built up area led to decrease in forest cover. Forests occupied 7.54% in 1990, but decreased to 5.51% in 2006 and 5.23% in 2017. This decrease in forest cover has resulted in increased surface runoff for the same periods (2006-2017). The increase in surface runoff in the study area could be attributed to deforestation and poor land husbandry, where during land preparation much of the vegetation is cleared, hence decreasing canopy interception and allowing water to drain off. Also, poor farming practices including cultivation on hillslopes without soil conservation, reducing soil compaction, hence allowing more water to drain as surface runoff. The calibrated SWAT model using the three different land use and land cover change of 1990, 2006 and 2017 indicate that during the wet season, the mean monthly flow increased by 1.48% relative to the 28.09 m 3 /s of the baseline 1990 while during the dry season, the mean monthly flow decreased by 16.7% relative to the 0.20 m 3 /s baseline flow. Assessment of the impacts of land use and land cover changes on catchment water balance component revealed that surface runoff increased by 3.9% in 2006 and 9.01% in 2017 while groundwater contribution to stream flow decreased by 6.3% and 12.86% in 2006 and 2017, respectively. The decrease in stream flow could also be attributed to abstraction of water for irrigation activities upstream of the Igawa gauge station. The findings of the study may help basin water officers, planners in water sector and agriculture sector in addressing uncertainty in policy and decision-making specifically when preparing strategies and adaptations plans for river catchment. The science used in this study can be applicable to another river basin in Tanzanian in a climate change impact study.