The further development, application and evaluation of a sediment yield model (WQSED) for catchment management in African catchments
- Authors: Gwapedza, David
- Date: 2021-04
- Subjects: Sedimentation and deposition -- South Africa , Sedimentation and deposition -- Zimbabwe , Watersheds -- South Africa , Watersheds -- Zimbabwe , Watershed management -- Africa , Water quality -- South Africa , Water quality -- Zimbabwe , Modified Universal Soil Loss Equation (MUSLE) , Water Quality and Sediment Model (WQSED) , Soil and Water Assessment Tool (SWAT)
- Language: English
- Type: thesis , text , Doctoral , PhD
- Identifier: http://hdl.handle.net/10962/178376 , vital:42934 , 10.21504/10962/178376
- Description: Erosion and sediment transport are natural catchment processes that play an essential role in ecosystem functioning by providing habitat for aquatic organisms and contributing to the health of wetlands. However, excessive erosion and sedimentation, mostly driven by anthropogenic activity, lead to ecosystem degradation, loss of agricultural land, water quality problems, reduced reservoir storage capacity and damage to physical infrastructure. It is reported that up to 25% of dams in South Africa have lost approximately 30% of their initial storage capacity to sedimentation. Therefore, excessive sedimentation transcends from an ecological problem to a health, livelihood and water security issue. Erosion and sedimentation occur at variable temporal and spatial scales; therefore, monitoring of these processes can be difficult and expensive. Regardless of all these prohibiting factors, information on erosion and sediment remains an urgent requirement for the sustainable management of catchments. Models have evolved as tools to replicate and simulate complex natural processes to understand and manage these systems. Several models have been developed globally to simulate erosion and sediment transport. However, these models are not always applicable in Africa because 1) the conditions under which they were developed are not as relevant for African catchments 2) they have high data requirements and cannot be applied with ease in our data-scarce African catchments 3) they are sometimes complicated, and there are little training available or potential users simply have no time to dedicate towards learning these models. To respond to the problems of erosion, sedimentation, water quality and unavailability of applicable models, the current research further develops, applies and evaluates an erosion and sediment transport model, the Water Quality and Sediment Model (WQSED), for integration within the existing water resources framework in South Africa and application for practical catchment management. The WQSED was developed to simulate daily suspended sediment loads that are vital for water quality and quantity assessments. The WQSED was developed based on the Modified Universal Soil Loss Equation (MUSLE), and the Pitman model is a primary hydrological model providing forcing data, although flow data from independent sources may be used to drive the WQSED model. The MUSLE was developed in the United States of America, and this research attempts to improve the applicability of the MUSLE by identifying key issues that may impede its performance. Assessments conducted within the current research can be divided into scale assessment and application and evaluation assessment. The scale assessment involved evaluating spatial and temporal scale issues associated with the MUSLE. Spatial scale assessments were conducted using analytical and mathematical assessments on a hypothetical catchment. Temporal scale issues were assessed in terms of the vegetation cover (C) factor within the Tsitsa River catchment in South Africa. Model application and evaluation involved applying and calibrating the model to simulate daily time-series sediment yield. The model was applied to calibrated and validated (split-sample validation) in two catchments in South Africa, two catchments in Zimbabwe and three catchments were selected from the USA and associated territories for further testing as continuous daily time-series observed sediment data could not be readily accessed for catchments in the Southern African region. The catchments where the model was calibrated and validated range in size from 50 km2 to 20 000 km2. Additionally, the model was applied to thirteen ‘ungauged’ catchments selected from across South Africa, where only long-term reservoir sedimentation rates were available to compare with long term model simulations converted to sediment yield rates. The additional thirteen catchments were selected from areas of different climatic, vegetation and soils conditions characterising South Africa and range in size from 30 km2 to 2 500 km2. The current research results are split into a) MUSLE scale dependency and b) WQSED testing and evaluation. Scale dependency testing showed that the MUSLE could be spatially scale-dependent, particularly when a lumped approach is used, resulting in simulations of up to 30% more sediment. Spatial scale dependence in the MUSLE was found to be related to the runoff and topographic factors used and how they are calculated. The current study resorted to adopting a reference grid in applying the MUSLE, followed by scaling up the outputs to the total catchment area. Using a reference grid resulted in a general avoidance of the problem of spatial scale. The adoption of a seasonal vegetation cover factor was shown to significantly account for temporal changes of vegetation cover within a year and reduce over-estimations in sediment output. The temporal scale evaluation demonstrated the uncertainties associated with using a fixed vegetation cover factor in a catchment with variable rainfall and runoff pattern. The WQSED model evaluation showed that the model could be calibrated and validated to provide consistent results. Satisfactory model evaluation statistics were obtained for most catchments to which the model was applied, based on general model evaluation guidelines (Nash Sutcliffe Efficiency and R2 > 0.5). The model also performed generally well compared to established models that had been previously applied in some of the study catchments. The highest sediment yields recorded per country were 153 t km-2 year-1 (Tsitsa River; South Africa), 90 t km-2 year-1 (Odzi River; Zimbabwe) and 340 t km-2 year-1 (Rio Tanama; Puerto Rico). The results also displayed consistent underestimations of peak sediment yield events, partly attributed to sediment emanating from gullies that are not explicitly accounted for in the WQSED model structure. Furthermore, the calibration process revealed that the WQSED storage model is generally challenging to calibrate. An alternative simpler version of the storage model was easier to calibrate, but the model may still be challenging to apply to catchments where calibration data are not available. The additional evaluation of the WQSED simulated sediment yield rates against observed reservoir sediment rates showed a broad range of differences between the simulated and observed sediment yield rates. Differences between WQSED simulated sediment and observed reservoir sediment ranges from a low of 30% to a high of > 40 times. The large differences were partly attributed to WQSED being limited to simulating suspended sediment from sheet and rill processes, whereas reservoir sediment is generated from more sources that include bedload, channel and gully processes. Nevertheless, the model simulations replicated some of the regional sediment yield patterns and are assumed to represent sheet and rill contributions to reservoir sediment in selected catchments. The outcome of this study is an improved WQSED model that has successfully undergone preliminary testing and evaluation. Therefore, the model is sufficiently complete to be used by independent researchers and water resources managers to simulate erosion and sediment transport. However, the model is best applicable to areas where some observed data or regional information are available to calibrate the storage components and constrain model outputs. The report on potential MUSLE scale dependencies is relevant globally to all studies applying the MUSLE model and, therefore, can improve MUSLE application in future studies. The WQSED model offers a relatively simple, effective and applicable tool that is set to provide information to enhance catchment, land and water resources management in catchments of Africa. , Thesis (PhD) -- Faculty of Science, Institute for Water Research, 2021
- Full Text:
- Date Issued: 2021-04
- Authors: Gwapedza, David
- Date: 2021-04
- Subjects: Sedimentation and deposition -- South Africa , Sedimentation and deposition -- Zimbabwe , Watersheds -- South Africa , Watersheds -- Zimbabwe , Watershed management -- Africa , Water quality -- South Africa , Water quality -- Zimbabwe , Modified Universal Soil Loss Equation (MUSLE) , Water Quality and Sediment Model (WQSED) , Soil and Water Assessment Tool (SWAT)
- Language: English
- Type: thesis , text , Doctoral , PhD
- Identifier: http://hdl.handle.net/10962/178376 , vital:42934 , 10.21504/10962/178376
- Description: Erosion and sediment transport are natural catchment processes that play an essential role in ecosystem functioning by providing habitat for aquatic organisms and contributing to the health of wetlands. However, excessive erosion and sedimentation, mostly driven by anthropogenic activity, lead to ecosystem degradation, loss of agricultural land, water quality problems, reduced reservoir storage capacity and damage to physical infrastructure. It is reported that up to 25% of dams in South Africa have lost approximately 30% of their initial storage capacity to sedimentation. Therefore, excessive sedimentation transcends from an ecological problem to a health, livelihood and water security issue. Erosion and sedimentation occur at variable temporal and spatial scales; therefore, monitoring of these processes can be difficult and expensive. Regardless of all these prohibiting factors, information on erosion and sediment remains an urgent requirement for the sustainable management of catchments. Models have evolved as tools to replicate and simulate complex natural processes to understand and manage these systems. Several models have been developed globally to simulate erosion and sediment transport. However, these models are not always applicable in Africa because 1) the conditions under which they were developed are not as relevant for African catchments 2) they have high data requirements and cannot be applied with ease in our data-scarce African catchments 3) they are sometimes complicated, and there are little training available or potential users simply have no time to dedicate towards learning these models. To respond to the problems of erosion, sedimentation, water quality and unavailability of applicable models, the current research further develops, applies and evaluates an erosion and sediment transport model, the Water Quality and Sediment Model (WQSED), for integration within the existing water resources framework in South Africa and application for practical catchment management. The WQSED was developed to simulate daily suspended sediment loads that are vital for water quality and quantity assessments. The WQSED was developed based on the Modified Universal Soil Loss Equation (MUSLE), and the Pitman model is a primary hydrological model providing forcing data, although flow data from independent sources may be used to drive the WQSED model. The MUSLE was developed in the United States of America, and this research attempts to improve the applicability of the MUSLE by identifying key issues that may impede its performance. Assessments conducted within the current research can be divided into scale assessment and application and evaluation assessment. The scale assessment involved evaluating spatial and temporal scale issues associated with the MUSLE. Spatial scale assessments were conducted using analytical and mathematical assessments on a hypothetical catchment. Temporal scale issues were assessed in terms of the vegetation cover (C) factor within the Tsitsa River catchment in South Africa. Model application and evaluation involved applying and calibrating the model to simulate daily time-series sediment yield. The model was applied to calibrated and validated (split-sample validation) in two catchments in South Africa, two catchments in Zimbabwe and three catchments were selected from the USA and associated territories for further testing as continuous daily time-series observed sediment data could not be readily accessed for catchments in the Southern African region. The catchments where the model was calibrated and validated range in size from 50 km2 to 20 000 km2. Additionally, the model was applied to thirteen ‘ungauged’ catchments selected from across South Africa, where only long-term reservoir sedimentation rates were available to compare with long term model simulations converted to sediment yield rates. The additional thirteen catchments were selected from areas of different climatic, vegetation and soils conditions characterising South Africa and range in size from 30 km2 to 2 500 km2. The current research results are split into a) MUSLE scale dependency and b) WQSED testing and evaluation. Scale dependency testing showed that the MUSLE could be spatially scale-dependent, particularly when a lumped approach is used, resulting in simulations of up to 30% more sediment. Spatial scale dependence in the MUSLE was found to be related to the runoff and topographic factors used and how they are calculated. The current study resorted to adopting a reference grid in applying the MUSLE, followed by scaling up the outputs to the total catchment area. Using a reference grid resulted in a general avoidance of the problem of spatial scale. The adoption of a seasonal vegetation cover factor was shown to significantly account for temporal changes of vegetation cover within a year and reduce over-estimations in sediment output. The temporal scale evaluation demonstrated the uncertainties associated with using a fixed vegetation cover factor in a catchment with variable rainfall and runoff pattern. The WQSED model evaluation showed that the model could be calibrated and validated to provide consistent results. Satisfactory model evaluation statistics were obtained for most catchments to which the model was applied, based on general model evaluation guidelines (Nash Sutcliffe Efficiency and R2 > 0.5). The model also performed generally well compared to established models that had been previously applied in some of the study catchments. The highest sediment yields recorded per country were 153 t km-2 year-1 (Tsitsa River; South Africa), 90 t km-2 year-1 (Odzi River; Zimbabwe) and 340 t km-2 year-1 (Rio Tanama; Puerto Rico). The results also displayed consistent underestimations of peak sediment yield events, partly attributed to sediment emanating from gullies that are not explicitly accounted for in the WQSED model structure. Furthermore, the calibration process revealed that the WQSED storage model is generally challenging to calibrate. An alternative simpler version of the storage model was easier to calibrate, but the model may still be challenging to apply to catchments where calibration data are not available. The additional evaluation of the WQSED simulated sediment yield rates against observed reservoir sediment rates showed a broad range of differences between the simulated and observed sediment yield rates. Differences between WQSED simulated sediment and observed reservoir sediment ranges from a low of 30% to a high of > 40 times. The large differences were partly attributed to WQSED being limited to simulating suspended sediment from sheet and rill processes, whereas reservoir sediment is generated from more sources that include bedload, channel and gully processes. Nevertheless, the model simulations replicated some of the regional sediment yield patterns and are assumed to represent sheet and rill contributions to reservoir sediment in selected catchments. The outcome of this study is an improved WQSED model that has successfully undergone preliminary testing and evaluation. Therefore, the model is sufficiently complete to be used by independent researchers and water resources managers to simulate erosion and sediment transport. However, the model is best applicable to areas where some observed data or regional information are available to calibrate the storage components and constrain model outputs. The report on potential MUSLE scale dependencies is relevant globally to all studies applying the MUSLE model and, therefore, can improve MUSLE application in future studies. The WQSED model offers a relatively simple, effective and applicable tool that is set to provide information to enhance catchment, land and water resources management in catchments of Africa. , Thesis (PhD) -- Faculty of Science, Institute for Water Research, 2021
- Full Text:
- Date Issued: 2021-04
A combined modelling approach for simulating channel–wetland exchanges in large African river basins
- Authors: Makungu, Eunice J
- Date: 2020
- Subjects: Watersheds -- Africa , Watershed management -- Africa , Water resources development -- Africa -- International cooperation , Floodplain management -- Africa , Wetland ecology -- Simulation methods -- Africa , Wetland management -- Africa
- Language: English
- Type: text , Thesis , Doctoral , PhD
- Identifier: http://hdl.handle.net/10962/123288 , vital:35424
- Description: In Africa, many large and extensive wetlands are hydrologically connected to rivers, and their environmental integrity, as well as their influence on downstream flow regimes, depends on the prevailing channel–wetland exchange processes. These processes are inherently complex and vary spatially and temporally. Understanding channel–wetland exchanges is therefore, indispensable for the effective management of wetlands and the associated river basins. However, this information is limited in most of the river basins containing large wetlands in Africa. Furthermore, it is important to understand the links between upstream and downstream flow regimes and the wetland dynamics themselves, specifically where there are water resource developments that may affect these links (upstream developments), or be affected by them (downstream developments). Hydrological modelling of the entire basin using basin-scale models that include wetland components in their structures can be used to provide the information required to manage water resources in such basins. However, the level of detail of wetland processes included in many basin-scale models is typically very low and the lack of understanding of the wetland dynamics makes it difficult to quantify the relevant parameters. Detailed hydraulic models represent the channel-wetland exchanges in a much more explicit manner, but require relatively more data and time resources to establish than coarser scale hydrological models. The main objective of this study was, therefore, to investigate the use of a detailed hydraulic wetland model to provide a better understanding of channel–wetland exchanges and wetland dynamics, and to use the results to improve the parameterisation of a basin-scale model. The study focused on improving the water resource assessments modelling of three data-scarce African river basins that contain large wetlands: the floodplains of the Luangwa and Upper Zambezi River basins and the Usangu wetland in the Upper Great Ruaha River basin. The overall objective was achieved through a combined modelling approach that uses a detailed high-resolution LISFLOOD-FP hydraulic model to inform the structure and parameters of the GW Pitman monthly hydrological model. The results from the LISFLOOD-FP were used to improve the understanding of the channel–wetland exchange dynamics and to establish the wetland parameters required in the GW Pitman model. While some wetland parameters were directly quantified from the LISFLOOD-FP model results, others, which are highly empirical, were estimated by manually calibrating the GW Pitman wetland sub-model implemented in excel spreadsheets containing the LISFLOOD-FP model results. Finally, the GW Pitman model with the inclusion of the estimated wetland parameters was applied for each basin and the results compared to the available downstream observed flow data. The two models have been successfully applied in southern Africa, with the GW Pitman model being one of the most widely applied hydrological models in this region. To address the issue of data scarcity, during setup of these models, the study mainly relied on the global datasets which clearly adds to the overall uncertainty of the modelling approach. However, this is a typical situation for most of the data scarce regions of the continent. A number of challenges were, however, faced during the setup of the LISFLOOD-FP, mainly due to the limitations of the data inputs. Some of the LISFLOOD-FP data inputs include boundary conditions (upstream and downstream), channel cross-sections and wetland topography. In the absence of observed daily flows to quantify the wetland upstream boundary conditions, monthly flow volumes simulated using the GW Pitman monthly model (without including the wetland sub-model) were disaggregated into daily flows using a disaggregation sub-model. The simulated wetland inflows were evaluated using the observed flow data for downstream gauging stations that include the wetland effects. The results highlighted that it is important to understand the possible impacts of each wetland on the downstream flow regime during the evaluations of the model simulation results. Although the disaggregation approach cannot be validated due to a lack of observed data, it at least enables the simulated monthly flows to be used in the daily time step hydraulic model. One of the recommendations is that improvements are required in gauging station networks to provide more observed information for the main river and the larger tributary inflows into these large and important wetland systems. Even a limited amount of newly observed data would be helpful to reduce some of the uncertainties in the combined modelling approach. The SRTM 90 m DEM (used to represent wetland topography) was filtered to reduce local variations and noise effects (mainly vegetation bias), but there were some pixels that falsely affect the inundation results, and the recently released vegetation-corrected DEMs are suggested to improve the simulation results. Channel cross-section values derived from global datasets should be examined because some widths estimated from the Andreadis et al. (2013) dataset were found to be over-generalised and did not reflect widths measured using high-resolution Google Earth in many places. There is an indication that channel cross-sections digitised from Google Earth images can be successfully used in the model setup except in densely vegetated swamps where the values are difficult to estimate, and in such situations, field measured cross-section data are required. Small channels such as those found in the Usangu wetland could play major role in the exchange dynamics, but digitising them all was not straightforward and only key ones were included in the model setup. Clearly, this inevitably introduced uncertainties in the simulated results, and future studies should consider applying methods that simplify extractions of most of these channels from high-resolution images to improve the simulated results. The study demonstrated that the wetland and channel physical characteristics, as well as the seasonal flow magnitude, largely influence the channel–wetland exchanges and wetland dynamics. The inundation results indicated that the area–storage and storage–inflow relationships form hysteretic curves, but the shape of these curves vary with flood magnitude and wetland type. Anticlockwise hysteresis curves were observed in both relationships for the floodplains (Luangwa and Barotse), whereas there appears to be no dominant curve type for the Usangu wetlands. The lack of well-defined hysteretic relationships in the Usangu could be related to some of the difficulties (and resulting uncertainties) that were experienced in setting up the model for this wetland. The storage–inflow relationships in all wetlands have quite complex rising limbs due to multiple flow peaks during the main wet season. The largest inundation area and storage volume for the Barotse and Usangu wetlands occurred after the peak discharge of the wet season, a result that is clearly related to the degree of connectivity between the main channel and those areas of the wetlands that are furthest away from the channel. Hysteresis effects were found to increase with an increase in flood magnitudes and temporal variations in the wetland inflows. Overall, hysteresis behaviour is common in large wetlands and it is recommended that hysteresis curves should be reflected in basin-scale modelling of large river basins with substantial wetland areas. At a daily time scale, inflow–outflow relationships showed a significant peak reduction and a delayed time to peak of several weeks in the Barotse and Usangu wetlands, whereas the attenuation effects of the Luangwa floodplain are minimal. To a large extent, the LISFLOOD-FP results provided useful information to establish wetland parameters and assess the structure of Pitman wetland sub-model. The simple spreadsheet used to estimate wetland parameters did not account for the wetland water transfers from the upstream to the next section downstream (the condition that is included in the LISFLOOD-FP model) for the case when the wetlands were distributed across more than one sub-basin. It is recommended that a method that allows for the upstream wetland inflows and the channel inflows should be included in the spreadsheet. The same is true to the Pitman model structure, and a downstream transfer of water can be modelled through return flows to the channel. The structure of the wetland sub-model was modified to allow an option for the return flows to occur at any time during the simulation period to provide for types of wetlands (e.g. the Luangwa) where spills from the channel and drainage back to the channel occur simultaneously. The setup of the GW Pitman model with the inclusion of wetland parameters improved the simulation results. However, the results for the Usangu wetlands were not very satisfactory and the collection of additional field data related to exchange dynamics is recommended to achieve improvements. The impacts of the Luangwa floodplain on the flow regime of the Luangwa River are very small at the monthly time scale, whereas the Barotse floodplain system and the Usangu wetlands extensively regulate flows of the Zambezi River and the Great Ruaha River, respectively. The results highlighted the possibilities of regionalising some wetland parameters using an understanding of wetland physical characteristics and their water exchange dynamics. However, some parameters remain difficult to quantify in the absence of site-specific information about the water exchange dynamics. The overall conclusion is that the approach implemented in this study presents an important step towards the improvements of water resource assessments modelling for research and practical purposes in data-scarce river basins. This approach is not restricted to the two used models, as it can be applied using different model combinations to achieve similar study purpose.
- Full Text:
- Date Issued: 2020
A combined modelling approach for simulating channel–wetland exchanges in large African river basins
- Authors: Makungu, Eunice J
- Date: 2020
- Subjects: Watersheds -- Africa , Watershed management -- Africa , Water resources development -- Africa -- International cooperation , Floodplain management -- Africa , Wetland ecology -- Simulation methods -- Africa , Wetland management -- Africa
- Language: English
- Type: text , Thesis , Doctoral , PhD
- Identifier: http://hdl.handle.net/10962/123288 , vital:35424
- Description: In Africa, many large and extensive wetlands are hydrologically connected to rivers, and their environmental integrity, as well as their influence on downstream flow regimes, depends on the prevailing channel–wetland exchange processes. These processes are inherently complex and vary spatially and temporally. Understanding channel–wetland exchanges is therefore, indispensable for the effective management of wetlands and the associated river basins. However, this information is limited in most of the river basins containing large wetlands in Africa. Furthermore, it is important to understand the links between upstream and downstream flow regimes and the wetland dynamics themselves, specifically where there are water resource developments that may affect these links (upstream developments), or be affected by them (downstream developments). Hydrological modelling of the entire basin using basin-scale models that include wetland components in their structures can be used to provide the information required to manage water resources in such basins. However, the level of detail of wetland processes included in many basin-scale models is typically very low and the lack of understanding of the wetland dynamics makes it difficult to quantify the relevant parameters. Detailed hydraulic models represent the channel-wetland exchanges in a much more explicit manner, but require relatively more data and time resources to establish than coarser scale hydrological models. The main objective of this study was, therefore, to investigate the use of a detailed hydraulic wetland model to provide a better understanding of channel–wetland exchanges and wetland dynamics, and to use the results to improve the parameterisation of a basin-scale model. The study focused on improving the water resource assessments modelling of three data-scarce African river basins that contain large wetlands: the floodplains of the Luangwa and Upper Zambezi River basins and the Usangu wetland in the Upper Great Ruaha River basin. The overall objective was achieved through a combined modelling approach that uses a detailed high-resolution LISFLOOD-FP hydraulic model to inform the structure and parameters of the GW Pitman monthly hydrological model. The results from the LISFLOOD-FP were used to improve the understanding of the channel–wetland exchange dynamics and to establish the wetland parameters required in the GW Pitman model. While some wetland parameters were directly quantified from the LISFLOOD-FP model results, others, which are highly empirical, were estimated by manually calibrating the GW Pitman wetland sub-model implemented in excel spreadsheets containing the LISFLOOD-FP model results. Finally, the GW Pitman model with the inclusion of the estimated wetland parameters was applied for each basin and the results compared to the available downstream observed flow data. The two models have been successfully applied in southern Africa, with the GW Pitman model being one of the most widely applied hydrological models in this region. To address the issue of data scarcity, during setup of these models, the study mainly relied on the global datasets which clearly adds to the overall uncertainty of the modelling approach. However, this is a typical situation for most of the data scarce regions of the continent. A number of challenges were, however, faced during the setup of the LISFLOOD-FP, mainly due to the limitations of the data inputs. Some of the LISFLOOD-FP data inputs include boundary conditions (upstream and downstream), channel cross-sections and wetland topography. In the absence of observed daily flows to quantify the wetland upstream boundary conditions, monthly flow volumes simulated using the GW Pitman monthly model (without including the wetland sub-model) were disaggregated into daily flows using a disaggregation sub-model. The simulated wetland inflows were evaluated using the observed flow data for downstream gauging stations that include the wetland effects. The results highlighted that it is important to understand the possible impacts of each wetland on the downstream flow regime during the evaluations of the model simulation results. Although the disaggregation approach cannot be validated due to a lack of observed data, it at least enables the simulated monthly flows to be used in the daily time step hydraulic model. One of the recommendations is that improvements are required in gauging station networks to provide more observed information for the main river and the larger tributary inflows into these large and important wetland systems. Even a limited amount of newly observed data would be helpful to reduce some of the uncertainties in the combined modelling approach. The SRTM 90 m DEM (used to represent wetland topography) was filtered to reduce local variations and noise effects (mainly vegetation bias), but there were some pixels that falsely affect the inundation results, and the recently released vegetation-corrected DEMs are suggested to improve the simulation results. Channel cross-section values derived from global datasets should be examined because some widths estimated from the Andreadis et al. (2013) dataset were found to be over-generalised and did not reflect widths measured using high-resolution Google Earth in many places. There is an indication that channel cross-sections digitised from Google Earth images can be successfully used in the model setup except in densely vegetated swamps where the values are difficult to estimate, and in such situations, field measured cross-section data are required. Small channels such as those found in the Usangu wetland could play major role in the exchange dynamics, but digitising them all was not straightforward and only key ones were included in the model setup. Clearly, this inevitably introduced uncertainties in the simulated results, and future studies should consider applying methods that simplify extractions of most of these channels from high-resolution images to improve the simulated results. The study demonstrated that the wetland and channel physical characteristics, as well as the seasonal flow magnitude, largely influence the channel–wetland exchanges and wetland dynamics. The inundation results indicated that the area–storage and storage–inflow relationships form hysteretic curves, but the shape of these curves vary with flood magnitude and wetland type. Anticlockwise hysteresis curves were observed in both relationships for the floodplains (Luangwa and Barotse), whereas there appears to be no dominant curve type for the Usangu wetlands. The lack of well-defined hysteretic relationships in the Usangu could be related to some of the difficulties (and resulting uncertainties) that were experienced in setting up the model for this wetland. The storage–inflow relationships in all wetlands have quite complex rising limbs due to multiple flow peaks during the main wet season. The largest inundation area and storage volume for the Barotse and Usangu wetlands occurred after the peak discharge of the wet season, a result that is clearly related to the degree of connectivity between the main channel and those areas of the wetlands that are furthest away from the channel. Hysteresis effects were found to increase with an increase in flood magnitudes and temporal variations in the wetland inflows. Overall, hysteresis behaviour is common in large wetlands and it is recommended that hysteresis curves should be reflected in basin-scale modelling of large river basins with substantial wetland areas. At a daily time scale, inflow–outflow relationships showed a significant peak reduction and a delayed time to peak of several weeks in the Barotse and Usangu wetlands, whereas the attenuation effects of the Luangwa floodplain are minimal. To a large extent, the LISFLOOD-FP results provided useful information to establish wetland parameters and assess the structure of Pitman wetland sub-model. The simple spreadsheet used to estimate wetland parameters did not account for the wetland water transfers from the upstream to the next section downstream (the condition that is included in the LISFLOOD-FP model) for the case when the wetlands were distributed across more than one sub-basin. It is recommended that a method that allows for the upstream wetland inflows and the channel inflows should be included in the spreadsheet. The same is true to the Pitman model structure, and a downstream transfer of water can be modelled through return flows to the channel. The structure of the wetland sub-model was modified to allow an option for the return flows to occur at any time during the simulation period to provide for types of wetlands (e.g. the Luangwa) where spills from the channel and drainage back to the channel occur simultaneously. The setup of the GW Pitman model with the inclusion of wetland parameters improved the simulation results. However, the results for the Usangu wetlands were not very satisfactory and the collection of additional field data related to exchange dynamics is recommended to achieve improvements. The impacts of the Luangwa floodplain on the flow regime of the Luangwa River are very small at the monthly time scale, whereas the Barotse floodplain system and the Usangu wetlands extensively regulate flows of the Zambezi River and the Great Ruaha River, respectively. The results highlighted the possibilities of regionalising some wetland parameters using an understanding of wetland physical characteristics and their water exchange dynamics. However, some parameters remain difficult to quantify in the absence of site-specific information about the water exchange dynamics. The overall conclusion is that the approach implemented in this study presents an important step towards the improvements of water resource assessments modelling for research and practical purposes in data-scarce river basins. This approach is not restricted to the two used models, as it can be applied using different model combinations to achieve similar study purpose.
- Full Text:
- Date Issued: 2020
- «
- ‹
- 1
- ›
- »