This dissertation employs hydrologic modeling to assess probable impacts of changes in vegetation cover and in the channel network on streamflow and floodplain groundwater levels in the Baviaanskloof catchment, South Africa. The Baviaanskloof serves as a case study of a semi-arid, mountainous, meso-scale catchment that has been subject to agricultural land use and is regionally important for water supply. In this catchment livestock grazing has resulted in a loss of subtropical thicket cover on hillslopes and the channel network in the central valley has become increasingly connected and incised.
In order to build an appropriate model of the Baviaanskloof, streamflow, groundwater, surface runoff, and soil moisture data were analyzed for diagnostic patterns that revealed information about hydrologic connectivity at different spatial and temporal scales. Critical results of these analyses were that: a) the central valley alluvial aquifer is recharged by subsurface flows from surrounding mountain areas following two major pathways, a likely interflow contribution following large rainfall events and a more temporally consistent contribution from the bedrock aquifer, and b) the dominant direction of exchange of water between the alluvial aquifer and the main floodplain channel regularly fluctuates between losing and gaining. To capture the observed patterns, the numeric model structure consisted of a coarse-scale sub-model of the mountain tributary subcatchments surrounding the central valley alluvial fill, and a higher resolution, coupled hydraulic-hydrologic sub-model of the central valley alluvial fans and floodplain. This model was calibrated in a multi-criteria calibration process using various observational datasets. It was found that including multiple streamflow-based criteria and including criteria based on additional data types improved model performance and better constrained the model parameter space.
Alternative scenarios of further degradation and of restoration of the hillslope vegetation, alluvial fan surfaces, and floodplain channel were modeled individually and in combination. Models were run using 38 years of local climate data and differences between model predictions for the current catchment state and each alternative scenario were assessed. Outputs suggested that, of the individual restoration intervention scenarios considered, hillslope thicket restoration would have the most significant impact on streamflow, driven by large reductions in storm event runoff and floodplain channel restoration would have the largest impact on the floodplain water table, driven by decreased drainage into the channel and increased recharge due to overbank flooding. Results indicated that restoring hillslopes could reduce flood peaks by 56-60% and annual average yield by 22-27%. Greater modeled water retention and evapotranspiration on vegetated hillslopes reduced runoff to the floodplain, resulting in a deepened water table and decreased baseflow in the model. Restoring alluvial fans was predicted to reduce flood peaks by 11-17% compared to the current scenario, but modeled impacts on average yield, baseflow, and floodplain aquifer levels were not statistically detectable. Comparing the alluvial fan restoration scenario to a more extreme channelized case did show small, but detectable, increases in baseflow and floodplain groundwater levels. Reducing floodplain channel incision was predicted to reduce peak flows by 14-20%. Modeled impacts on average yield and baseflow were not statistically detectable. Groundwater levels were predicted to rise with channel restoration, with average depth decreasing 17-21%. Simultaneous restoration at all three positions was predicted to reduce flood peaks more substantially than any individual intervention (69-71%) while also decreasing the average depth of the floodplain water table (8-11%). Average annual yield was predicted to decrease by 32-37% as was baseflow, with a 20-40% decrease in average annual minimum monthly flow.
These results highlight various potential tradeoffs that would need to be considered in restoration planning and catchment management. Predictions made here would need further integration into reservoir and water supply system models, as well as sediment transport models to consider reservoir sedimentation, in order to better understand implications for local and downstream water supply availability. Nevertheless a net decrease in annual average available supply appears likely.
The catchment and climatic contexts of these changes were shown to be important in determining the magnitude and direction of the predicted impacts, with dispersive flow paths through central valley alluvium dampening impacts of hillslope vegetation changes and with the frequency and magnitude of storm events determining how often different thresholds of flow path connectivity were reached. This highlighted the importance of including best available understanding of a landscape's hydrologic connectivity in modeling, even when focusing on a local scale change, and of modeling impacts over a long time period to include long-term weather patterns.
