Prompt location of sources and sinks of sediment within a catchment would allow more effective So... more Prompt location of sources and sinks of sediment within a catchment would allow more effective Soil and Water Conservation (SWC) planning. Distributed erosion models are valuable tools for watershed planning, but the quality of spatially distributed model predictions is seriously hampered by the natural complexity and spatial heterogeneity of the landscape system, coupled with limited spatio-temporal datasets of sufficient accuracy. This study aimed at developing a semi-empirical, spatially distributed erosion model to locate sources of sediment within a catchment in data scarce environments. In the experimental catchment of Kwalei, in the West Usambara Mountains of Tanzania, the spatial distribution of erosion and erosion factors was observed during two rainy seasons. In the catchment, overland flow was of dynamic Hortonian type: it was triggered by short and intense showers, but as it moved downward, it quickly reinfiltrated. These observations and measurements at the catchment outlet were used to build a hydrologic model to predict event-based overland flow depth that accounted for rainfall characteristics, land use, field topology, and reinfiltration length, i.e. the average travel distance of overland flow. The hydrologic model was coupled with the sediment phase of the Morgan, Morgan and Finney model to estimate field erosion rates. The best model simulations predicted correctly around 75 % of erosion pattern, but the uncertainty of model prediction due to sediment transport parameterisation was high: 10 % of fields were either classified as subject to severe or slight erosion depending on the sediment transport parameters. Analysis of the spatial patterns of erosion and erosion factors showed that in the Kwalei catchment the location of severely eroded areas was correlated to crust and vegetation cover, but the spatial extent of erosion depended upon the overland flow travel distance. Moreover, the spatial scale of the distribution of some farmers¿ indicators of erosion, i.e. signs that farmers use to assess erosion in their fields, was very close to that of eroded areas and overland flow distribution. Farmers¿ indicators of erosion were used to build a classification tree to predict the distribution of erosion. The resulting Farmers¿ Indicator Tree was the best among several erosion models tested in the area in predicting the spatial pattern of erosion. These findings open up possibilities to integrate more effectively farmers' knowledge into distributed modelling of hydrology and erosion.
ABSTRACT [1] Assessment of constituent loads in rivers is essential to evaluate water quality of ... more ABSTRACT [1] Assessment of constituent loads in rivers is essential to evaluate water quality of streams and estuaries; however, uncertainty in load estimation may be large and must be considered and communicated together with estimates. In this comparative study, the usefulness of two existing methods (bootstrap and Bayesian inference) to assess uncertainty in constituent loads estimated with an improved eight-parameter rating curve is demonstrated. Bootstrap prediction intervals and Bayesian credible intervals were estimated for daily and monthly loads obtained with a rating curve applied to routine monitoring sampling data sets of nitrate (NO3-N), reactive phosphorus (RP), and total phosphorus (TP) of the Duck River, in Tasmania (Australia). Predicted loads and prediction intervals were compared to benchmark loads obtained by an independent, high frequency monitoring program. The eight-parameter rating curve resulted in better prediction of NO3-N and TP than RP loads. Both inference methods successfully generated prediction intervals. The bracketing frequency (i.e., the fraction of prediction intervals that comprised benchmark loads) of bootstrap prediction intervals was 50–65% of daily or monthly benchmark loads. Bracketing frequency of Bayesian credible intervals was consistently higher, and included 74–85% of benchmark daily loads and 80% or more of benchmark monthly loads. Both methods proved to be robust to the presence of an artificial outlier. Prediction intervals were affected by the distribution of the regression error, hence they reflected uncertainty in the regression data set and limitations in the rating curve formulation. They did not account for other sources of uncertainty, i.e., they were still conservative predictions of load uncertainty.
A parametrically parsimonious, data-based model was built on observations at hillslope and catchm... more A parametrically parsimonious, data-based model was built on observations at hillslope and catchment scale to simulate the distribution of overland flow within a small East African Highlands catchment (Kwalei, Tanzania). A rainfall-flow Data Based Mechanistic model identified catchment effective rainfall and separated the discharge quick flow, interpreted as the combination of overland flow plus reinfiltration along the slopes, and the slow flow, interpreted as ground water displacement. Observations of overland flow occurrence along the slopes were used to derive probability distribution functions (pdfs) of overland flow in relation to effective rainfall for two pre-defined hydrologic response units (HRUs): perennial (HRU_1) versus other crops (HRU_2). At low effective rainfall, overland flow was more frequent in HRU_2, while at high effective rainfall overland flow in the two HRUs was similar. The pdfs were employed to disaggregate the quick flow into HRU overland flow depth. Reinfiltration was accounted for by assuming that only the overland flow generated in the lower part of the field would drain downslope. Effective reinfiltration length was about 4 m. Comparison of model simulations versus Gerlach trough measurements indicated that rainfall intensity was not accounted for sufficiently. The use of smaller time steps or, alternatively, of a rainfall intensity threshold could improve model performance. However, given the high variability of overland flow observed along the slopes and the limited dataset, model simulations were considered satisfactory. Though the model needs further testing on other datasets, the disaggregating approach represents an inductive alternative to the use of infiltration equations to model overland flow within a catchment. q
Sediment monitoring, tracing and modelling are widely used to identify suspended sediment sources... more Sediment monitoring, tracing and modelling are widely used to identify suspended sediment sources. Although each method has inherent limitations and uncertainties, their integration provides opportunities to form collective knowledge and encourages robust management strategies. This paper presents a Weight-of-Evidence approach to integrate multiple Lines-of-Evidence for identifying suspended sediment sources. Three sources of evidence were used: i) stream flow and suspended sediment monitoring at river gauges; ii) geochemical sediment tracing at river junctions; and iii) catchment-scale suspended sediment modelling of hillslope, gully, streambank and unsealed road erosion. We applied this approach on two data-poor catchments in Australia. Some reaches were consistently identified as major sources of sediment from all Lines-of-Evidence. However, inconsistencies between the types of evidence in other areas highlighted the high uncertainty in identifying suspended sediment sources in these areas and the need for further investigation. The integration framework maximised the use of scarce information, enabled explicit consideration of uncertainties for catchment management and identified where future monitoring and research should be targeted.
This study assessed the ability of several models to locate areas affected by severe erosion and ... more This study assessed the ability of several models to locate areas affected by severe erosion and identified the factors controlling the distribution of erosion in a catchment characterized by a dynamic Hortonian hydrologic regime. The spatial patterns of severely eroded areas predicted by five erosion models were compared with the pattern of erosion observed during an extensive field survey conducted in the Kwalei catchment, north-eastern Tanzania. The actual erosion pattern was also compared with the spatial distribution of some erosion factors: overland flow (whose distribution was simulated with a hydrologic model that took overland flow reinfiltration into account), slope, crust, canopy cover and ground cover. The patterns of severely eroded areas varied markedly among the models. The best predictions were those of (i) a classification tree based on farmers' indicators of erosion (Pearson's Phi correlation coefficient q = 0.72, n = 334, a b 0.01); (ii) a semi-empirical model that accounted for overland flow reinfiltration (q = 0.43); and (iii) a logit regression model based on slope and ground cover (q = 0.34). The erosion factor that most correlated with eroded areas was crust cover (q = 0.52). Lacunarity analysis of the spatial patterns showed that the erosion models could not characterize the spatial scale of eroded areas correctly. Instead, the spatial scale of erosion distribution in the catchment did coincide with the overland flow distribution at short reinfiltration length (0.5-5 m), even though severely eroded areas were not spatially correlated to areas of high overland flow depth (q = 0.12, a N 0.05). In the dynamic Hortonian regime of the Kwalei catchment, the travel distance of overland flow determined the spatial scale of severely eroded areas. Spatially distributed erosion model predictions could improve if the configuration of sources and sinks of overland flow in the landscape is taken into account. D
Under increasing population pressure, soil erosion has become a threat in the East African Highla... more Under increasing population pressure, soil erosion has become a threat in the East African Highlands, and erosion modelling can be useful to quantify this threat. To test its applicability for this region, the LISEM soil erosion model was applied to two small catchments, one in the Usumbara Mountains, Tanzania, and the other on the slopes of Mount Kenya. Input data for the model were collected in both catchments, as were data on runoff and erosion that were used for calibration and validation of the model. LISEM was first calibrated on catchment outlet data, and afterwards simulated spatial patterns of erosion were compared to available erosion data. The results showed that LISEM can, after calibration, give good discharge predictions for some events, but not for all. However, LISEM generally overpredicted soil loss from the catchments. Comparison with observed erosion patterns did not show overprediction, but according to the model, erosion was more widespread than was observed. There are several reasons for these discrepancies. First, it is difficult to obtain enough accurate data to run the model, such as accurate maps, rainfall data and soil and plant characteristics. Second, it is also difficult to obtain accurate data to evaluate the performance of the model, either for the catchment outlet or spatially, therefore observed erosion rates are also uncertain. Third, the model could not deal correctly with complex events, i.e. those having double rainfall peaks, and might also have difficulties with catchment characteristics such as soil type and the complexity of land use. Finally, LISEM could not deal with events in which throughflow or baseflow played a role, which was to be expected since those processes are not simulated by LISEM. Nevertheless, LISEM could be calibrated to give good discharge predictions for some events, and also gave reasonable results when compared to data obtained from erosion plots. Furthermore, only complex, distributed, storm-based models such as LISEM can give spatial predictions for single storms. Therefore, it is concluded that if the aim is spatial prediction on an event basis, there is no alternative to complex erosion models such as LISEM, but if the aim is to predict average annual erosion, the data-demanding, physically based LISEM erosion model may not be the most appropriate model.
The Soil and Water Assessment Tool (SWAT) is used worldwide for water quality assessment and plan... more The Soil and Water Assessment Tool (SWAT) is used worldwide for water quality assessment and planning. This paper aimed to assess and adapt SWAT hillslope sediment yield model (Modified Universal Soil Loss Equation, MUSLE) for applications in large basins, i.e. when spatial data is coarse and model units are large; and to develop a robust sediment calibration method for large regions. The Upper Danube Basin (132,000km(2)) was used as case study representative of large European Basins. The MUSLE was modified to reduce sensitivity of sediment yields to the Hydrologic Response Unit (HRU) size, and to identify appropriate algorithms for estimating hillslope length (L) and slope-length factor (LS). HRUs gross erosion was broadly calibrated against plot data and soil erosion map estimates. Next, mean annual SWAT suspended sediment concentrations (SSC, mg/L) were calibrated and validated against SSC data at 55 gauging stations (622 station-years). SWAT annual specific sediment yields in su...
Australian agriculture is under pressure to reduce sediment exports and improve stream water qual... more Australian agriculture is under pressure to reduce sediment exports and improve stream water quality. However data on soil losses of different land management is lacking, thus potential benefits of adopting Best Management Practices cannot be quantified. In this study we assessed field soil losses of farming systems of North Central Victoria by using the HowLeaky2008 model. HowLeaky2008 is a one-dimensional farm system model that simulates the effect of soil and land management on daily water, sediment and nutrient exports. HowLeaky2008 was applied to current and alternative land management options in the Avon Richardson catchment, which is typical of the Wimmera-Mallee region of Victoria. Simulated long term average soil losses were similar to erosion rates previously assessed with 137 Cs techniques. Modelling suggests that changing from minimum to zero tillage can reduce sediment exports by 40-75% in cropping land; and switching from annual to perennial pastures or lucerne can red...
Ambitious nutrient reduction targets have been set for the Gippsland Lakes, Victoria but at what ... more Ambitious nutrient reduction targets have been set for the Gippsland Lakes, Victoria but at what cost to productive agriculture? An interdisciplinary approach is addressing this question for the Moe River catchment, a dairy-dominated catchment that is a major source of pollutants to the Gippsland Lakes. Off-farm nitrogen exports are being estimated by biophysical modellers, and economists are quantifying the impact
Prompt location of sources and sinks of sediment within a catchment would allow more effective So... more Prompt location of sources and sinks of sediment within a catchment would allow more effective Soil and Water Conservation (SWC) planning. Distributed erosion models are valuable tools for watershed planning, but the quality of spatially distributed model predictions is seriously hampered by the natural complexity and spatial heterogeneity of the landscape system, coupled with limited spatio-temporal datasets of sufficient accuracy. This study aimed at developing a semi-empirical, spatially distributed erosion model to locate sources of sediment within a catchment in data scarce environments. In the experimental catchment of Kwalei, in the West Usambara Mountains of Tanzania, the spatial distribution of erosion and erosion factors was observed during two rainy seasons. In the catchment, overland flow was of dynamic Hortonian type: it was triggered by short and intense showers, but as it moved downward, it quickly reinfiltrated. These observations and measurements at the catchment outlet were used to build a hydrologic model to predict event-based overland flow depth that accounted for rainfall characteristics, land use, field topology, and reinfiltration length, i.e. the average travel distance of overland flow. The hydrologic model was coupled with the sediment phase of the Morgan, Morgan and Finney model to estimate field erosion rates. The best model simulations predicted correctly around 75 % of erosion pattern, but the uncertainty of model prediction due to sediment transport parameterisation was high: 10 % of fields were either classified as subject to severe or slight erosion depending on the sediment transport parameters. Analysis of the spatial patterns of erosion and erosion factors showed that in the Kwalei catchment the location of severely eroded areas was correlated to crust and vegetation cover, but the spatial extent of erosion depended upon the overland flow travel distance. Moreover, the spatial scale of the distribution of some farmers¿ indicators of erosion, i.e. signs that farmers use to assess erosion in their fields, was very close to that of eroded areas and overland flow distribution. Farmers¿ indicators of erosion were used to build a classification tree to predict the distribution of erosion. The resulting Farmers¿ Indicator Tree was the best among several erosion models tested in the area in predicting the spatial pattern of erosion. These findings open up possibilities to integrate more effectively farmers' knowledge into distributed modelling of hydrology and erosion.
ABSTRACT [1] Assessment of constituent loads in rivers is essential to evaluate water quality of ... more ABSTRACT [1] Assessment of constituent loads in rivers is essential to evaluate water quality of streams and estuaries; however, uncertainty in load estimation may be large and must be considered and communicated together with estimates. In this comparative study, the usefulness of two existing methods (bootstrap and Bayesian inference) to assess uncertainty in constituent loads estimated with an improved eight-parameter rating curve is demonstrated. Bootstrap prediction intervals and Bayesian credible intervals were estimated for daily and monthly loads obtained with a rating curve applied to routine monitoring sampling data sets of nitrate (NO3-N), reactive phosphorus (RP), and total phosphorus (TP) of the Duck River, in Tasmania (Australia). Predicted loads and prediction intervals were compared to benchmark loads obtained by an independent, high frequency monitoring program. The eight-parameter rating curve resulted in better prediction of NO3-N and TP than RP loads. Both inference methods successfully generated prediction intervals. The bracketing frequency (i.e., the fraction of prediction intervals that comprised benchmark loads) of bootstrap prediction intervals was 50–65% of daily or monthly benchmark loads. Bracketing frequency of Bayesian credible intervals was consistently higher, and included 74–85% of benchmark daily loads and 80% or more of benchmark monthly loads. Both methods proved to be robust to the presence of an artificial outlier. Prediction intervals were affected by the distribution of the regression error, hence they reflected uncertainty in the regression data set and limitations in the rating curve formulation. They did not account for other sources of uncertainty, i.e., they were still conservative predictions of load uncertainty.
A parametrically parsimonious, data-based model was built on observations at hillslope and catchm... more A parametrically parsimonious, data-based model was built on observations at hillslope and catchment scale to simulate the distribution of overland flow within a small East African Highlands catchment (Kwalei, Tanzania). A rainfall-flow Data Based Mechanistic model identified catchment effective rainfall and separated the discharge quick flow, interpreted as the combination of overland flow plus reinfiltration along the slopes, and the slow flow, interpreted as ground water displacement. Observations of overland flow occurrence along the slopes were used to derive probability distribution functions (pdfs) of overland flow in relation to effective rainfall for two pre-defined hydrologic response units (HRUs): perennial (HRU_1) versus other crops (HRU_2). At low effective rainfall, overland flow was more frequent in HRU_2, while at high effective rainfall overland flow in the two HRUs was similar. The pdfs were employed to disaggregate the quick flow into HRU overland flow depth. Reinfiltration was accounted for by assuming that only the overland flow generated in the lower part of the field would drain downslope. Effective reinfiltration length was about 4 m. Comparison of model simulations versus Gerlach trough measurements indicated that rainfall intensity was not accounted for sufficiently. The use of smaller time steps or, alternatively, of a rainfall intensity threshold could improve model performance. However, given the high variability of overland flow observed along the slopes and the limited dataset, model simulations were considered satisfactory. Though the model needs further testing on other datasets, the disaggregating approach represents an inductive alternative to the use of infiltration equations to model overland flow within a catchment. q
Sediment monitoring, tracing and modelling are widely used to identify suspended sediment sources... more Sediment monitoring, tracing and modelling are widely used to identify suspended sediment sources. Although each method has inherent limitations and uncertainties, their integration provides opportunities to form collective knowledge and encourages robust management strategies. This paper presents a Weight-of-Evidence approach to integrate multiple Lines-of-Evidence for identifying suspended sediment sources. Three sources of evidence were used: i) stream flow and suspended sediment monitoring at river gauges; ii) geochemical sediment tracing at river junctions; and iii) catchment-scale suspended sediment modelling of hillslope, gully, streambank and unsealed road erosion. We applied this approach on two data-poor catchments in Australia. Some reaches were consistently identified as major sources of sediment from all Lines-of-Evidence. However, inconsistencies between the types of evidence in other areas highlighted the high uncertainty in identifying suspended sediment sources in these areas and the need for further investigation. The integration framework maximised the use of scarce information, enabled explicit consideration of uncertainties for catchment management and identified where future monitoring and research should be targeted.
This study assessed the ability of several models to locate areas affected by severe erosion and ... more This study assessed the ability of several models to locate areas affected by severe erosion and identified the factors controlling the distribution of erosion in a catchment characterized by a dynamic Hortonian hydrologic regime. The spatial patterns of severely eroded areas predicted by five erosion models were compared with the pattern of erosion observed during an extensive field survey conducted in the Kwalei catchment, north-eastern Tanzania. The actual erosion pattern was also compared with the spatial distribution of some erosion factors: overland flow (whose distribution was simulated with a hydrologic model that took overland flow reinfiltration into account), slope, crust, canopy cover and ground cover. The patterns of severely eroded areas varied markedly among the models. The best predictions were those of (i) a classification tree based on farmers' indicators of erosion (Pearson's Phi correlation coefficient q = 0.72, n = 334, a b 0.01); (ii) a semi-empirical model that accounted for overland flow reinfiltration (q = 0.43); and (iii) a logit regression model based on slope and ground cover (q = 0.34). The erosion factor that most correlated with eroded areas was crust cover (q = 0.52). Lacunarity analysis of the spatial patterns showed that the erosion models could not characterize the spatial scale of eroded areas correctly. Instead, the spatial scale of erosion distribution in the catchment did coincide with the overland flow distribution at short reinfiltration length (0.5-5 m), even though severely eroded areas were not spatially correlated to areas of high overland flow depth (q = 0.12, a N 0.05). In the dynamic Hortonian regime of the Kwalei catchment, the travel distance of overland flow determined the spatial scale of severely eroded areas. Spatially distributed erosion model predictions could improve if the configuration of sources and sinks of overland flow in the landscape is taken into account. D
Under increasing population pressure, soil erosion has become a threat in the East African Highla... more Under increasing population pressure, soil erosion has become a threat in the East African Highlands, and erosion modelling can be useful to quantify this threat. To test its applicability for this region, the LISEM soil erosion model was applied to two small catchments, one in the Usumbara Mountains, Tanzania, and the other on the slopes of Mount Kenya. Input data for the model were collected in both catchments, as were data on runoff and erosion that were used for calibration and validation of the model. LISEM was first calibrated on catchment outlet data, and afterwards simulated spatial patterns of erosion were compared to available erosion data. The results showed that LISEM can, after calibration, give good discharge predictions for some events, but not for all. However, LISEM generally overpredicted soil loss from the catchments. Comparison with observed erosion patterns did not show overprediction, but according to the model, erosion was more widespread than was observed. There are several reasons for these discrepancies. First, it is difficult to obtain enough accurate data to run the model, such as accurate maps, rainfall data and soil and plant characteristics. Second, it is also difficult to obtain accurate data to evaluate the performance of the model, either for the catchment outlet or spatially, therefore observed erosion rates are also uncertain. Third, the model could not deal correctly with complex events, i.e. those having double rainfall peaks, and might also have difficulties with catchment characteristics such as soil type and the complexity of land use. Finally, LISEM could not deal with events in which throughflow or baseflow played a role, which was to be expected since those processes are not simulated by LISEM. Nevertheless, LISEM could be calibrated to give good discharge predictions for some events, and also gave reasonable results when compared to data obtained from erosion plots. Furthermore, only complex, distributed, storm-based models such as LISEM can give spatial predictions for single storms. Therefore, it is concluded that if the aim is spatial prediction on an event basis, there is no alternative to complex erosion models such as LISEM, but if the aim is to predict average annual erosion, the data-demanding, physically based LISEM erosion model may not be the most appropriate model.
The Soil and Water Assessment Tool (SWAT) is used worldwide for water quality assessment and plan... more The Soil and Water Assessment Tool (SWAT) is used worldwide for water quality assessment and planning. This paper aimed to assess and adapt SWAT hillslope sediment yield model (Modified Universal Soil Loss Equation, MUSLE) for applications in large basins, i.e. when spatial data is coarse and model units are large; and to develop a robust sediment calibration method for large regions. The Upper Danube Basin (132,000km(2)) was used as case study representative of large European Basins. The MUSLE was modified to reduce sensitivity of sediment yields to the Hydrologic Response Unit (HRU) size, and to identify appropriate algorithms for estimating hillslope length (L) and slope-length factor (LS). HRUs gross erosion was broadly calibrated against plot data and soil erosion map estimates. Next, mean annual SWAT suspended sediment concentrations (SSC, mg/L) were calibrated and validated against SSC data at 55 gauging stations (622 station-years). SWAT annual specific sediment yields in su...
Australian agriculture is under pressure to reduce sediment exports and improve stream water qual... more Australian agriculture is under pressure to reduce sediment exports and improve stream water quality. However data on soil losses of different land management is lacking, thus potential benefits of adopting Best Management Practices cannot be quantified. In this study we assessed field soil losses of farming systems of North Central Victoria by using the HowLeaky2008 model. HowLeaky2008 is a one-dimensional farm system model that simulates the effect of soil and land management on daily water, sediment and nutrient exports. HowLeaky2008 was applied to current and alternative land management options in the Avon Richardson catchment, which is typical of the Wimmera-Mallee region of Victoria. Simulated long term average soil losses were similar to erosion rates previously assessed with 137 Cs techniques. Modelling suggests that changing from minimum to zero tillage can reduce sediment exports by 40-75% in cropping land; and switching from annual to perennial pastures or lucerne can red...
Ambitious nutrient reduction targets have been set for the Gippsland Lakes, Victoria but at what ... more Ambitious nutrient reduction targets have been set for the Gippsland Lakes, Victoria but at what cost to productive agriculture? An interdisciplinary approach is addressing this question for the Moe River catchment, a dairy-dominated catchment that is a major source of pollutants to the Gippsland Lakes. Off-farm nitrogen exports are being estimated by biophysical modellers, and economists are quantifying the impact
Uploads
Papers by Olga Vigiak