Elsevier

Journal of Hydrology

Volume 602, November 2021, 126760
Journal of Hydrology

Variance decomposition of forecasted sediment transport in a lowland watershed using global climate model ensembles

https://doi.org/10.1016/j.jhydrol.2021.126760Get rights and content

Highlights

  • Hydrologic model parameterization produces greatest uncertainty for sediment transport forecasts via GCMs.

  • Water and sediment forecasts are under-predicted 49% and 35% when only precipitation and temperature are included.

  • Streamflow and sediment yield forecasts increase by 19% and 14% for the shear-limited landscape and transport-limited stream.

Abstract

Forecasting change in sediment transport using global climate model ensembles is under-developed in the hydrology community due to a lack of knowledge regarding the variance structure of predictions. We investigate the uncertainty of forecast sediment transport from sources, including global climate model realizations, global climate model ensemble design, forecasted hydrologic inputs, and hydrologic modeling parameterizations. We then forecast sediment transport for the gently rolling watershed in Kentucky, USA. Contrary to past research forecasting hydrology with global climate models, hydrologic model parameterization was the most significant source of variance impacting forecasted sediment transport variables. Forecast of sediment transport responses shows the propagation of uncertainty from the hydrologic model parameterization was over two times greater than the uncertainty from the selected global climate model realizations. This result emphasizes researchers focused on forecasting sediment transport with global climate models may need to give as much, or more, consideration to their water–sediment linkages as climate-water–sediment linkages. Hydrologic inputs from climate change, including forecast precipitation, temperature, relative humidity, solar radiation, and wind speed, impacted sediment transport. Considering changes in precipitation and temperature alone under-predicts streamflow and sediment transport by 49% and 35%, respectively, compared to including all meteorological variables inputs. Variance introduced across the three different global climate model ensembles was a relatively small source of variance impacting forecasted streamflow or sediment yield. The results suggest a quantitative effort by the researcher to design the global climate model ensemble by considering representativeness, historical performance, and independency will lead to robust results. Ensemble average forecasts forecast streamflow and sediment yield to increase by 19% and 14%, respectively, for the lowland watershed. The sediment transport forecasts reflect the shear-limited landscape of the watershed and the transport-limited conditions in the stream channel.

Introduction

Forecasting sediment transport requires projecting the change in future climate. Global climate model (GCM) ensembles provide an approach to input climate variables to hydrologic modeling for forecasting sediment transport. GCM forecasts physically account for changing climate variables and these variables are useful in sediment transport models for predicting sediment loadings that could disrupt ecosystems or fill reservoirs. Robust forecasts will allow watershed managers to avert problems before they occur. With these applications in mind, we advance sediment transport research by investigating GCM projections through water and sediment models, the variance structure associated with the simulations, and sediment processes and their connectivity in a basin.

Recently, a number of hydrology research studies couple GCM projections with sediment transport modeling (see Table 1). Bussi et al. (2016) used climate scenarios by changing precipitation and temperature to predict sediment loads in the River Thames catchment, United Kingdom. Gould et al. (2016) used climate scenarios and their precipitation and temperature results to study runoff erosion in northern Rocky Mountains, United States. Shrestha et al., 2016, Shrestha et al., 2018 used precipitation and temperature from multiple GCMs, multiple emission scenarios, and sets of model parameters to investigate sediment load predictions and reservoir sedimentation in the Mekong River Basin, Laos. Wallace et al. (2017) used multiple GCMs results of precipitation, temperature, and solar radiation to project sediment transport rates at three spatial scales in Indiana, United States. Yu et al. (2017) used climate scenarios from three GCMs to project sediment discharge due to changing precipitation and temperature in the Upper Yellow River Basin, China. Giri et al. (2019) applied multipole GCMs to project impacts of climate and land use change on the sediment production in Neshanic River Watershed, New Jersey, USA. Loiselle et al. (2020) used two climate scenarios to study the change in sediment and total organic carbon in Elbow River Watershed in Alberta, Canada. Mukundan et al. (2020) applied multiple GCM projections to study the impact of climate on nutrient and sediment loadings in the New York City water supply watershed, United States. Wang et al. (2020) applied GCM projections to study climate change impacts on the riverbed, wetland topography, and ecology in Yarlung Zangbo-Brahmaputra River, China.

The recent studies predict sediment flux in a changing climate throughout the world and advance the technology of model coupling. However, sources of variance impacting projected sediment transport has been understudied despite the fact that the variance structure can be conceptualized (see Fig. 1). Table 1 reviews sources of variance when projecting sediment transport with GCMs, and these sources of variance include: ‘GCM model realizations’, ‘GCM ensemble selection’, ‘hydrology inputs’, and ‘hydrologic model parameterization’. Only GCM model realization has been widely considered in previous studies, and other variance sources were rarely investigated.

All previous studies consider the source of variance termed ‘GCM model realizations’—meaning multiple GCMs with one or more downscaling methods and emission scenarios were input to show sets of sediment modeling results. The realization of GCM and its climate modeling factors (e.g., GCM type and model version, emission scenario, downscaling method, and bias correction technique) are well reported to impact hydrologic predictions due to underlying structure and assumptions of individual GCMs as well as dynamic or statistical methods when post-processing GCM results (Tu, 2009, Sheshukov et al., 2011, Harding et al., 2012, Chen et al., 2013, Al-Mukhtar et al., 2014, Neupane and Kumar, 2015, Al Aamery et al., 2016, Al Aamery et al., 2018). For this reason, ‘GCM model realization’ is considered a prominent source of variance impacting sediment transport predictions.

The set of GCM realizations chosen to include in GCM ensembles is the source of variance termed ‘GCM ensemble selection’. Many choices of GCMs are available, but GCMs may be highly correlated with one another or poorly suited for the basin studied. GCM ensemble selection procedures have been developed by climate scientists for choosing a set of GCMs as a function of model historical performance, future climate uncertainty, and model independency (Mendlik and Gobiet, 2016, Lee and Kim, 2017, Pechlivanidis et al., 2018). ‘GCM ensemble selection’ has not been investigated previously, to our knowledge, as a source of variance in sediment transport forecasts (see column 4, Table 1).

Hydrologic variables that change in a future climate that are output from GCMs and input to hydrology models is the source of variance termed ‘hydrology inputs’. GCM projections suggest precipitation, temperature, relative humidity, net radiation, and wind speed all may change substantially in the future (McVicar et al., 2012, Wild, 2009, Willett et al., 2008, Al Aamery et al., 2018). However, many previous studies only consider precipitation and temperature inputs to sediment transport models because these variables are assumed the main drivers for changing future hydrology (Al Aamery et al., 2016, Mejia et al., 2012, Chien et al., 2013, Brekke et al., 2013). Changes in relative humidity, solar radiation, and wind speed, might also shift future sediment transport, and these sources of variance have not been considered previously, to our knowledge (see column five, Table 1).

The set of parameters chosen or calibrated in hydrologic models is the source of variance termed ‘hydrologic model parameterization’. Parameterization of a hydrology model is well recognized as a source of variance impacting streamflow and sediment transport (e.g., Moriasi et al., 2007). However, the importance of hydrologic model parameterization has been marginalized to some degree when forecasting streamflow with GCMs (Niraula et al., 2015, Al Aamery et al., 2018). Only a few studies consider this source of variance previously (see column six, Table 1).

This paper provides methods and results for decomposing the variance from the four mentioned sources to gain understanding of their relative importance on sediment transport predictions. Knowledge of the most sensitive linkages in projections can provide insight to where river managers should invest their resources when making forecasts. Knowledge of the variance structure can also provide the most important areas of future research collaboration between climate researchers and hydrologic researchers.

Another contribution of this paper is GCM projection through a sediment transport model formulated with sediment connectivity theory. Seven previous studies used the semi-distributed modified universal soil loss equation (MUSLE) embedded in the soil water assessment tool (SWAT) to predict sediment transport with GCMs, and three previous studies used distributed sediment transport models (see column seven, Tab 1). No previous studies, to our knowledge, used sediment connectivity-based models with GCMs. Sediment connectivity is defined as an emergent system property reflecting the continuity and magnitude of sediment linkages within geomorphologic compartments during a given time period (Heckmann et al., 2018). Given the increase in availability of high-resolution geospatial data, the ability to simulate explicit sediment pathways with sediment connectivity approaches has become increasingly plausible and has been shown to improve understanding and simulation of sediment processes (e.g., Vigiak et al., 2012; Heckmann et al., 2018; Mahoney et al., 2020a, Mahoney et al., 2020b, Wu et al., 2020). An attractive feature of the sediment connectivity approach is explicitly representing these pathways, which is underdeveloped in semi-distributed and fully distributed sediment models (Mahoney et al., 2020a).

Several features of sediment connectivity modeling make it a potentially advantageous approach for coupling with GCMs. First, Mahoney et al., (2018) found sediment connectivity modeling overcomes spatial limitations of semi-distributed conceptual models such as MUSLE by grounding the model in physical equations and estimating sediment transport at the raster cell scale. Second, connectivity simulations first predict the spatial domain most likely contributing sediment to the river, and in turn provide reduced computational complexity of fully distributed sediment prediction (Mahoney et al., 2018). Third, several researchers have shown sediment connectivity models can highlight hotspots of sediment production, which ultimately improves decision making when analyzing sediment forecasts (e.g., Vigiak et al., 2012; Heckmann et al., 2018; Mahoney et al., 2018, Husic et al., 2020). Fourth, sediment connectivity captures the morphometry of the basin because parameters associated with watersheds shape, size, and gradient are spatially explicit in connectivity formula (Dikpal et al., 2017, Mahajan and Sivakumar, 2018).

Sediment connectivity modeling allows prediction of the sediment processes occurring in a basin under changing climate conditions. The potential controlling sediment transport process are many (see Fig. 2), including soil erosion and transport across the landscape, erosion of bed and bank sediment in the stream corridor, redistribution, and sediment deposition throughout the basin, and sediment flux leaving the watershed. The relative importance of these processes reflects the system's net behavior as shear, transport, or source limited and the aggrading or degrading condition of the river network. Questions remain as to a system’s net behavior and in turn response under future climates. For example, GCM projections suggest wet, temperate regions of the world will experience higher rain- and snowfall, higher humidity, more cloud cover, and lower wind speed over the next century (Melillo et al., 2014, McVicar et al., 2012, Wild, 2009, Willett et al., 2008, Al Aamery et al., 2016). These climate changes are suggested to in turn shift the hydrologic cycle to produce higher runoff and higher streamflow in wet temperate watersheds (Al Aamery et al., 2018). However, this net impact on a system’s behavior as shear, transport, or source limited remains under-discussed in the hydrology community.

The objectives of this study were to: (i) investigate the uncertainty structure of sediment transport fluxes impacted by numerous sources of variance when projecting with GCMs; (ii) investigate forecasted changes in specific soil erosion and sediment transport processes found from sediment connectivity modeling; and (iii) provide results of a net climate change impact on a system’s behavior as shear, transport, or source limited for a basin located in a wet, temperate region.

Section snippets

Study site and materials

The Upper South Elkhorn Watershed (Fig. 3), located in the Kentucky river basin, Kentucky USA, has been selected for this study. The watershed area is 61.7 km2, the average slope is 5.3%, and the elevation ranges from 255 m to 324 m asl. Urban lands dominate the watershed land use with 47%, followed by agricultural lands with 38.4% and forest with 14.6%. The watershed is characterized as mild sloped with gently rolling hills; the soil is mainly silt loam; the floodplains are long and flat; and

Methods

An overview of the method is shown in Fig. 4. Our approach was to permeate the model selection criterions to select representative subsets of GCMs. Then we used SWAT to simulate the hydrologic conditions of the watershed, where we extracted the results of streamflow, surface runoff depth, and daily curve number. SWAT extracted results have been used in our process-based-connectivity and instream sediment transport models. Surface runoff and daily curve number results are inputs to the upland

Select ensembles of GCMs

Results of the model selection method provided three GCM ensemble designs with different sets of GCMs across statistical downscaling and dynamic downscaling projects (see Table 2 and Fig. 5; and Tables A, B, C, and D and Figures A and B in Supplementary material). Design #2 showed the best choice for a GCM ensemble due to its highest value of the maximum entropy measure while still considering model representativeness, historical performance, and independency.

Results shown in Table 2 tend to

Conclusion

The conclusion of this paper is as follows:

  • 1.

    Model evaluation results show the GCM model selection procedure based on model representativeness, historical performance, independency and lack of additional clustering for the downscaling approach provided the best ensemble design. Model evaluation of bias correction approaches showed quantile delta mapping outperformed scaled distribution mapping and quantile mapping in this study. Hydrologic and sediment model performance was found adequate for the

CRediT authorship contribution statement

Nabil Al Aamery: Methodology, Software, Validation, Formal analysis, Writing - original draft, Visualization. James F. Fox: Conceptualization, Methodology, Investigation, Resources, Writing - original draft, Writing - review & editing, Supervision, Project administration, Funding acquisition. Tyler Mahoney: Conceptualization, Methodology.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

We gratefully acknowledge the financial support of the this research under National Science Foundation (NSF) Award 163288. We thank the CMIP3, CMIP5, NARCCAP and CORDEX climate projects for making their data publically accessible.

References (79)

  • S. Giri et al.

    Assessing the potential impacts of climate and land use change on water fluxes and sediment transport in a loosely coupled system

    J. Hydrol.

    (2019)
  • G.K. Gould et al.

    The effects of climate change and extreme wildfire events on runoff erosion over a mountain watershed

    J. Hydrol.

    (2016)
  • D. Loiselle et al.

    Projecting impacts of wildfire and climate change on streamflow, sediment, and organic carbon yields in a forested watershed

    J. Hydrol.

    (2020)
  • D.T. Mahoney et al.

    Watershed erosion modeling using the probability of sediment connectivity in a gently rolling system

    J. Hydrol.

    (2018)
  • D.T. Mahoney et al.

    Integrating connectivity theory within watershed modelling part II: application and evaluating structural and functional connectivity

    Sci. Total Environ.

    (2020)
  • D.T. Mahoney et al.

    Integrating connectivity theory within watershed modelling part I: model formulation and investigating the timing of sediment connectivity

    Sci. Total Environ.

    (2020)
  • T.R. McVicar et al.

    Global review and synthesis of trends in observed terrestrial near-surface wind speeds: Implications for evaporation

    J. Hydrol.

    (2012)
  • R. Mukundan et al.

    Climate change impact on nutrient loading in a water supply watershed

    J. Hydrol.

    (2020)
  • S.D. Nerantzaki et al.

    Modeling suspended sediment transport and assessing the impacts of climate change in a Karstic Mediterranean watershed

    Sci. Total Environ.

    (2015)
  • R.P. Neupane et al.

    Estimating the effects of potential climate and land use changes on hydrologic processes of a large agriculture dominated watershed

    J. Hydrol.

    (2015)
  • R. Niraula et al.

    Determining the importance of model calibration for forecasting absolute/relative changes in streamflow from LULC and climate changes

    J. Hydrol.

    (2015)
  • J. Russo et al.

    The role of the surface fine-grained laminae in low-gradient streams: a model approach

    Geomorphology

    (2012)
  • B. Shrestha et al.

    Uncertainty in flow and sediment projections due to future climate scenarios for the 3S Rivers in the Mekong Basin

    J. Hydrol.

    (2016)
  • C. Teutschbein et al.

    Bias correction of regional climate model simulations for hydrological climate-change impact studies: review and evaluation of different methods

    J. Hydrol.

    (2012)
  • J.R. Thompson et al.

    Assessment of uncertainty in river flow projections for the Mekong River using multiple GCMs and hydrological models

    J. Hydrol.

    (2013)
  • J. Tu

    Combined impact of climate and land use changes on streamflow and water quality in eastern Massachusetts, USA

    J. Hydrol.

    (2009)
  • O. Vigiak et al.

    Comparison of conceptual landscape metrics to define hillslope-scale sediment delivery ratio

    Geomorphology

    (2012)
  • J. Wang et al.

    Climate change impacts on the topography and ecological environment of the wetlands in the middle reaches of the Yarlung Zangbo-Brahmaputra River

    J. Hydrol.

    (2020)
  • C.K. Abbaspour

    SWAT Calibrating and Uncertainty Programs

    (2008)
  • M. Al-Mukhtar et al.

    Assessing the impacts of climate change on hydrology of the upper reach of the spree river: Germany

    Water Resour. Manage.

    (2014)
  • J.G. Arnold et al.

    Large area hydrologic modeling and assessment part I: model development 1

    JAWRA J. Am. Water Resour. Assoc.

    (1998)
  • Bagnold, R.A., 1966. An approach to the sediment transport problem from general physics. US government printing...
  • C.H. Bishop et al.

    Climate model dependence and the replicate Earth paradigm

    Clim. Dyn.

    (2013)
  • J. Boé et al.

    Statistical and dynamical downscaling of the Seine basin climate for hydro-meteorological studies

    Int. J. Climatol.: J. R. Meteorol. Society

    (2007)
  • Brekke, L., Thrasher, B.L., Maurer, E.P. and Pruitt, T., 2013. Downscaled CMIP3 and CMIP5 climate projections: release...
  • A.J. Cannon et al.

    Bias correction of GCM precipitation by quantile mapping: how well do methods preserve changes in quantiles and extremes?

    J. Clim.

    (2015)
  • V. Chirivella Osma et al.

    Dynamical versus statistical downscaling for the generation of regional climate change scenarios at a Western Mediterranean basin: the Jucar River District

    J. Water Clim. Change

    (2016)
  • D.L. Davies et al.

    A cluster separation measure

    IEEE Trans. Pattern Anal. Mach. Intell.

    (1979)
  • R.L. Dikpal et al.

    Evaluation of morphometric parameters derived from Cartosat-1 DEM using remote sensing and GIS techniques for Budigere Amanikere watershed, Dakshina Pinakini Basin, Karnataka, India

    Appl. Water Sci.

    (2017)
  • Cited by (0)

    View full text