The effect of input data complexity on the uncertainty in simulated streamflow in a humid, mountainous watershed

Mapping Intimacies ◽

10.5194/hess-2018-46 ◽

2018 ◽

Author(s):

Linh Hoang ◽

Rajith Mukundan ◽

Karen E. B. Moore ◽

Emmet M. Owens ◽

Tammo S. Steenhuis

Keyword(s):

New York ◽

Land Use ◽

Parameter Uncertainty ◽

Input Data ◽

Assessment Tool ◽

Model Performance ◽

Data Complexity ◽

Delaware River ◽

Output Uncertainty ◽

Uncertainty Model

Abstract. Uncertainty in hydrological and water quality modelling is of significant concern due to its effects on prediction and subsequent application in watershed management. Similar to other distributed hydrological models, model uncertainty is an issue in applying the Soil and Water Assessment Tool (SWAT). Previous research has shown how SWAT predictions are affected by uncertainty in parameter estimation and input data resolution. Nevertheless, little information is available on how parameter uncertainty and output uncertainty are affected by input data of varying complexity. In this study, SWAT-Hillslope (SWAT-HS), a modified version of SWAT capable of predicting saturation excess runoff was applied to assess the effects of input data with varying degrees of complexity on parameter uncertainty and output uncertainty. Four digital elevation model (DEM) resolutions (1, 3, 10 and 30 m) were tested for their ability to predict streamflow and saturated areas. In a second analysis, three soil maps and three land use maps were used to build nine SWAT-HS setups from simple to complex (fewer to more soil types/ land use classes), which were then compared to study the effect of input data complexity on model prediction/output uncertainty. The case study was the Town Brook watershed in the upper reaches of the West Branch Delaware River in the Catskill Region, New York, USA. Results show that DEM resolution did not impact parameter uncertainty or affect the simulation of streamflow at the watershed outlet but significantly affected the spatial pattern of saturated areas, with 10 m being the most appropriate grid size to use for our application. The comparison of nine model setups revealed that input data complexity did not affect parameter uncertainty. Model setups using intermediate soil/land use specifications were slightly better than the ones using simple information, while the most complex setup did not show any improvement from the intermediate ones. We conclude that increasing spatial input details may not necessarily improve model performance or reduce parameter and output uncertainty, but using multiple temporal and spatial observations can aid in finding the appropriate parameter sets and in reducing prediction/output uncertainty.

Download Full-text

The effect of input data resolution and complexity on the uncertainty of hydrological predictions in a humid vegetated watershed

Hydrology and Earth System Sciences ◽

10.5194/hess-22-5947-2018 ◽

2018 ◽

Vol 22 (11) ◽

pp. 5947-5965 ◽

Cited By ~ 1

Author(s):

Linh Hoang ◽

Rajith Mukundan ◽

Karen E. B. Moore ◽

Emmet M. Owens ◽

Tammo S. Steenhuis

Keyword(s):

Land Use ◽

Parameter Uncertainty ◽

Input Data ◽

Hydrological Modeling ◽

Assessment Tool ◽

Model Performance ◽

Data Complexity ◽

Delaware River ◽

Output Uncertainty ◽

Data Resolution

Abstract. Uncertainty in hydrological modeling is of significant concern due to its effects on prediction and subsequent application in watershed management. Similar to other distributed hydrological models, model uncertainty is an issue in applying the Soil and Water Assessment Tool (SWAT). Previous research has shown how SWAT predictions are affected by uncertainty in parameter estimation and input data resolution. Nevertheless, little information is available on how parameter uncertainty and output uncertainty are affected by input data of varying complexity. In this study, SWAT-Hillslope (SWAT-HS), a modified version of SWAT capable of predicting saturation-excess runoff, was applied to assess the effects of input data with varying degrees of complexity on parameter uncertainty and output uncertainty. Four digital elevation model (DEM) resolutions (1, 3, 10 and 30 m) were tested for their ability to predict streamflow and saturated areas. In a second analysis, three soil maps and three land use maps were used to build nine SWAT-HS setups from simple to complex (fewer to more soil types/land use classes), which were then compared to study the effect of input data complexity on model prediction/output uncertainty. The case study was the Town Brook watershed in the upper reaches of the West Branch Delaware River in the Catskill region, New York, USA. Results show that DEM resolution did not impact parameter uncertainty or affect the simulation of streamflow at the watershed outlet but significantly affected the spatial pattern of saturated areas, with 10m being the most appropriate grid size to use for our application. The comparison of nine model setups revealed that input data complexity did not affect parameter uncertainty. Model setups using intermediate soil/land use specifications were slightly better than the ones using simple information, while the most complex setup did not show any improvement from the intermediate ones. We conclude that improving input resolution and complexity may not necessarily improve model performance or reduce parameter and output uncertainty, but using multiple temporal and spatial observations can aid in finding the appropriate parameter sets and in reducing prediction/output uncertainty.

Download Full-text

The impact of local surface changes in Borneo on atmospheric composition at wider spatial scales: coastal processes, land-use change and air quality

Philosophical Transactions of the Royal Society B Biological Sciences ◽

10.1098/rstb.2011.0060 ◽

2011 ◽

Vol 366 (1582) ◽

pp. 3210-3224 ◽

Cited By ~ 22

Author(s):

J. A. Pyle ◽

N. J. Warwick ◽

N. R. P. Harris ◽

Mohd Radzi Abas ◽

A. T. Archibald ◽

...

Keyword(s):

Land Use ◽

Air Quality ◽

Land Use Change ◽

Spatial Scales ◽

Model Performance ◽

Atmospheric Composition ◽

Model Studies ◽

Uncertainty Model ◽

Emission Changes ◽

The Impact

We present results from the OP3 campaign in Sabah during 2008 that allow us to study the impact of local emission changes over Borneo on atmospheric composition at the regional and wider scale. OP3 constituent data provide an important constraint on model performance. Treatment of boundary layer processes is highlighted as an important area of model uncertainty. Model studies of land-use change confirm earlier work, indicating that further changes to intensive oil palm agriculture in South East Asia, and the tropics in general, could have important impacts on air quality, with the biggest factor being the concomitant changes in NO x emissions. With the model scenarios used here, local increases in ozone of around 50 per cent could occur. We also report measurements of short-lived brominated compounds around Sabah suggesting that oceanic (and, especially, coastal) emission sources dominate locally. The concentration of bromine in short-lived halocarbons measured at the surface during OP3 amounted to about 7 ppt, setting an upper limit on the amount of these species that can reach the lower stratosphere.

Download Full-text

A pre-calibration approach to select optimum inputs for hydrological models in data-scarce regions

10.5194/hess-2016-242 ◽

2016 ◽

Cited By ~ 1

Author(s):

Esraa Tarawneh ◽

Jonathan Bridge ◽

Neil Macdonald

Keyword(s):

Input Data ◽

Assessment Tool ◽

Swat Model ◽

Model Performance ◽

Global Scale ◽

Reanalysis Data ◽

Local Soil ◽

Sediment Data ◽

Uncalibrated Model ◽

Selection Of

Abstract. This study uses the Soil and Water Assessment Tool (SWAT) model to quantitatively compare available input datasets in a data-poor dryland environment (Wala catchment, Jordan; 1743 km2). Eighteen scenarios combining best available land-use, soil and weather datasets (1979–2002) are considered to construct SWAT models. Data include local observations and global reanalysis data products. Uncalibrated model outputs assess the variability in model performance derived from input data sources only. Model performance against discharge and sediment load data are compared using r2, Nash–Sutcliffe Efficiency (NSE), RSR and PBIAS. NSE statistic varies from 0.56 to −12 and 0.79 to −85 for best and poorest-performing scenarios against observed discharge and sediment data respectively. Global weather inputs yield considerable improvements on discontinuous local datasets, whilst local soil inputs perform considerably better than global-scale mapping. The methodology provides a rapid, transparent and transferable approach to aid selection of the most robust suite of input data.

Download Full-text

Impact of spatial data resolution on simulated catchment water balances and model performance of the multi-scale TOPLATS model

Hydrology and Earth System Sciences Discussions ◽

10.5194/hessd-2-2183-2005 ◽

2005 ◽

Vol 2 (5) ◽

pp. 2183-2217 ◽

Cited By ~ 1

Author(s):

H. Bormann

Keyword(s):

Land Use ◽

Stream Flow ◽

Input Data ◽

Model Performance ◽

Multi Scale ◽

Data Resolution ◽

Land Use Scenarios ◽

Simulation Results ◽

Water Balances ◽

Flow Components

Abstract. This paper analyses the effect of spatial input data resolution on the simulated water balances and flow components using the multi-scale hydrological model TOPLATS. A data set of 25m resolution of the central German Dill catchment (693 km2 is used for investigation. After an aggregation of digital elevation model, soil map and land use classification to 50 m, 75 m, 100 m, 150 m, 200 m, 300 m, 500 m, 1000 m and 2000 m, water balances and water flow components are calculated for the entire Dill catchment as well as for 3 subcatchments without any recalibration. The study shows that both model performance measures as well as simulated water balances almost remain constant for most of the aggregation steps for all investigated catchments. Slight differences occur for single catchments at the resolution of 50–500 m (e.g. 0–3% for annual stream flow), significant differences at the resolution of 1000 m and 2000 m (e.g. 2–12% for annual stream flow). These differences can be explained by the fact that the statistics of certain input data (land use data in particular as well as soil physical characteristics) changes significantly at these spatial resolutions, too. The impact of smoothing the relief by aggregation occurs continuously but is not reflected by the simulation results. To study the effect of aggregation of land use data in detail, three different land use scenarios are aggregated which were generated aiming on economic optimisation at different field sizes (0.5 ha, 1.5 ha and 5.0 ha). The changes induced by aggregation of these land use scenarios are comparable with respect to catchment water balances compared to the current land use. A correlation analysis only in some cases reveals high correlation between changes in both input data and in simulation results for all catchments and land use scenarios combinations (e.g. evapotranspiration is correlated to land use, runoff generation is correlated to soil properties). Predominantly the correlation between catchment properties (e.g. topographic index, transmissivity, land use) and simulated water flows varies from catchment to catchment. This study indicates that an aggregation of input data for the calculation of regional water balances using TOPLATS type models leads to significant errors from a resolution exceeding 500 m. A meaningful aggregation of data should in the first instance aim on preserving the areal fractions of land use classes.

Download Full-text

A pre-calibration approach to select optimum inputs for hydrological models in data-scarce regions

Hydrology and Earth System Sciences ◽

10.5194/hess-20-4391-2016 ◽

2016 ◽

Vol 20 (10) ◽

pp. 4391-4407 ◽

Cited By ~ 7

Author(s):

Esraa Tarawneh ◽

Jonathan Bridge ◽

Neil Macdonald

Keyword(s):

Input Data ◽

Assessment Tool ◽

Swat Model ◽

Model Performance ◽

Global Scale ◽

Reanalysis Data ◽

Local Soil ◽

Error Standard Deviation ◽

Sediment Data ◽

Uncalibrated Model

Abstract. This study uses the Soil and Water Assessment Tool (SWAT) model to quantitatively compare available input datasets in a data-poor dryland environment (Wala catchment, Jordan; 1743 km2). Eighteen scenarios combining best available land-use, soil and weather datasets (1979–2002) are considered to construct SWAT models. Data include local observations and global reanalysis data products. Uncalibrated model outputs assess the variability in model performance derived from input data sources only. Model performance against discharge and sediment load data are compared using r2, Nash–Sutcliffe efficiency (NSE), root mean square error standard deviation ratio (RSR) and percent bias (PBIAS). NSE statistic varies from 0.56 to −12 and 0.79 to −85 for best- and poorest-performing scenarios against observed discharge and sediment data respectively. Global weather inputs yield considerable improvements on discontinuous local datasets, whilst local soil inputs perform considerably better than global-scale mapping. The methodology provides a rapid, transparent and transferable approach to aid selection of the most robust suite of input data.

Download Full-text

Modeling Hydrologic Response to Land Use and Climate Change in the Krueng Jreu Sub Watershed of Aceh Besar

Aceh International Journal of Science and Technology ◽

10.13170/aijst.5.3.5762 ◽

2016 ◽

Vol 5 (3) ◽

Author(s):

Teuku Ferijal ◽

Mustafril Bachtiar ◽

Dewi Sri Jayanti ◽

Dahlan Jafaruddin

Keyword(s):

Climate Change ◽

Land Use ◽

Agricultural Land ◽

Assessment Tool ◽

Irrigation System ◽

Swat Model ◽

Model Performance ◽

Primary Source ◽

Surface Flow ◽

Water Yield

Soil and Water Assessment Tool (SWAT) model was used to simulate impact of landuse and climate change on water resources in Krueng Jreu subwatershed located in Aceh Province – Indonesia. The subwatershed is a primary source of water to irrigated 233.52 km2 paddy field area through a surface irrigation system. The model performance was considerably good in predicting streamflow. The coefficients of determination varied between 0.58 and 0.72, while the Nash-Sutcliffe coefficients (ENS) ranged between 0.65-0.72 and the percentage bias were in the range of -0.36 to 2.30. Scenarios were applied to the best fit model to evaluate watershed responses to land use and climate changes. The model predicted increases in both runoff and water yield by 1% and 0.1% respectively as the result of increasing 15% settlement area. When all agricultural land within subwatershed converted to forest, water yield would increase by 1% during dry period and runoff contribution would decrease by 5%. Increases in surface flow by 23.6% and water yield by 15.1% were found under scenario of increasing 10% of daily precipitation. Increasing in evapotranspiration caused by an increase of 1.5⁰C in daily air temperature would decrease surface flow and water yield by 0.8% and 1.3%, respectively. Combination scenarios of changes in daily temperature and precipitation would increase evapotranspiration rate, annual water yield and runoff contribution.

Download Full-text

Application of the SWAT model to the Xiangjiang river watershed in subtropical central China

Water Science & Technology ◽

10.2166/wst.2013.100 ◽

2013 ◽

Vol 67 (9) ◽

pp. 2110-2116 ◽

Cited By ~ 9

Author(s):

Qiao Luo ◽

Yong Li ◽

Kelin Wang ◽

Jinshui Wu

Keyword(s):

Land Use ◽

Land Use Planning ◽

Stream Flow ◽

Assessment Tool ◽

Swat Model ◽

Model Performance ◽

Central China ◽

Xiangjiang River ◽

Stream Flows ◽

River Watershed

The Soil and Water Assessment Tool (SWAT) model was applied to simulate the water balance in the Xiangjiang river watershed for current and planning scenarios of land uses. The model was first calibrated for the period from 1998 to 2002 and then validated for the period from 2003 to 2007 using the observed stream flow data from four monitoring gages within the watershed. The determination coefficient of linear regression of the observed and simulated monthly stream flows (R2) and their Nash–Sutcliffe Index (NSI) was used to evaluate model performance. All values of R2 and NSI were above 0.8 and ranged from 0.82 to 0.92, which indicates that the SWAT model was capable of simulating the stream flow in the Xiangjiang river watershed. The calibrated and validated SWAT model was then applied to study the hydrological response of three land use change scenarios. Runoff was reduced by increasing the areas of forest and grassland while simultaneously decreasing the areas of agricultural and urban land. In the recent and future land use planning for the Xiangjiang river watershed, the hydrological effect should be considered in regional water management and erosion control.

Download Full-text

Impact of spatial data resolution on simulated catchment water balances and model performance of the multi-scale TOPLATS model

Hydrology and Earth System Sciences ◽

10.5194/hess-10-165-2006 ◽

2006 ◽

Vol 10 (2) ◽

pp. 165-179 ◽

Cited By ~ 41

Author(s):

H. Bormann

Keyword(s):

Land Use ◽

Stream Flow ◽

Input Data ◽

Model Performance ◽

Multi Scale ◽

Data Resolution ◽

Land Use Scenarios ◽

Simulation Results ◽

Water Balances ◽

Flow Components

Abstract. This paper analyses the effect of spatial input data resolution on the simulated water balances and flow components using the multi-scale hydrological model TOPLATS. A data set of 25m resolution of the central German Dill catchment (693 km2) is used for investigation. After an aggregation of digital elevation model, soil map and land use classification to 50 m, 75 m, 100 m, 150 m, 200 m, 300 m, 500 m, 1000 m and 2000 m, water balances and water flow components are calculated for the entire Dill catchment as well as for 3 subcatchments without any recalibration. The study shows that model performance measures and simulated water balances almost remain constant for most of the aggregation steps for all investigated catchments. Slight differences in the simulated water balances and statistical quality measures occur for single catchments at the resolution of 50 m to 500 m (e.g. 0–3% for annual stream flow), significant differences at the resolution of 1000 m and 2000 m (e.g. 2–12% for annual stream flow). These differences can be explained by the fact that the statistics of certain input data (land use data in particular as well as soil physical characteristics) changes significantly at these spatial resolutions. The impact of smoothing the relief by aggregation occurs continuously but is barely reflected by the simulation results. To study the effect of aggregation of land use data in detail, in addition to current land use the effect of aggregation on the water balance calculations based on three different land use scenarios is investigated. Land use scenarios were available aiming on economic optimisation of agricultural and forestry practices at different field sizes (0.5 ha, 1.5 ha and 5.0 ha). The changes in water balance terms, induced by aggregation of the land use scenarios, are comparable with respect to catchment water balances compared to the current land use. A correlation analysis between statistics of input data and simulated annual water fluxes only in some cases reveals systematically high correlation coefficients for all investigated catchments and data sets (e.g. actual evapotranspiration is correlated to land use, surface runoff generation is correlated to soil properties). Predominantly the correlation between catchment properties (e.g. topographic index, transmissivity, land use) and simulated water flows varies from catchment to catchment. Catchment specific properties determine correlations between properties and fluxes, but do not influence the effect of data aggregation. This study indicates that an aggregation of input data for the calculation of regional water balances using TOPLATS type models leads to significant errors from a resolution exceeding 500 m. Correlating statistics of input data and simulation results indicates that a meaningful aggregation of data should in the first instance aim on preserving the areal fractions of land use classes.

Download Full-text