TY - JOUR
T1 - Sampling variability in estimates of flow characteristics in coarse-bed channels
T2 - Effects of sample size
AU - Cienciala, Piotr
AU - Hassan, Marwan A.
N1 - Funding Information:
The data presented in this paper are available upon request from the authors. This research was funded by NSERC Discovery Grant awarded to Marwan A. Hassan. Piotr Cienciala was supported by a University of British Columbia Four Year Doctoral Fellowship. Tim Reid and Laurent Roberge provided assistance with hydraulic field data collection. The manuscript benefited from conversations with a number of people, including Michael Church, Dan Moore, John Richardson, Brett Eaton, Hal Voepel, and Rick White. Comments from Michael Church and three anonymous reviewers have greatly helped improve an earlier version of this manuscript.
PY - 2016/3/1
Y1 - 2016/3/1
N2 - Adequate description of hydraulic variables based on a sample of field measurements is challenging in coarse-bed streams, a consequence of high spatial heterogeneity in flow properties that arises due to the complexity of channel boundary. By applying a resampling procedure based on bootstrapping to an extensive field data set, we have estimated sampling variability and its relationship with sample size in relation to two common methods of representing flow characteristics, spatially averaged velocity profiles and fitted probability distributions. The coefficient of variation in bed shear stress and roughness length estimated from spatially averaged velocity profiles and in shape and scale parameters of gamma distribution fitted to local values of bed shear stress, velocity, and depth was high, reaching 15-20% of the parameter value even at the sample size of 100 (sampling density 1 m-2). We illustrated implications of these findings with two examples. First, sensitivity analysis of a 2-D hydrodynamic model to changes in roughness length parameter showed that the sampling variability range observed in our resampling procedure resulted in substantially different frequency distributions and spatial patterns of modeled hydraulic variables. Second, using a bedload formula, we showed that propagation of uncertainty in the parameters of a gamma distribution used to model bed shear stress led to the coefficient of variation in predicted transport rates exceeding 50%. Overall, our findings underscore the importance of reporting the precision of estimated hydraulic parameters. When such estimates serve as input into models, uncertainty propagation should be explicitly accounted for by running ensemble simulations.
AB - Adequate description of hydraulic variables based on a sample of field measurements is challenging in coarse-bed streams, a consequence of high spatial heterogeneity in flow properties that arises due to the complexity of channel boundary. By applying a resampling procedure based on bootstrapping to an extensive field data set, we have estimated sampling variability and its relationship with sample size in relation to two common methods of representing flow characteristics, spatially averaged velocity profiles and fitted probability distributions. The coefficient of variation in bed shear stress and roughness length estimated from spatially averaged velocity profiles and in shape and scale parameters of gamma distribution fitted to local values of bed shear stress, velocity, and depth was high, reaching 15-20% of the parameter value even at the sample size of 100 (sampling density 1 m-2). We illustrated implications of these findings with two examples. First, sensitivity analysis of a 2-D hydrodynamic model to changes in roughness length parameter showed that the sampling variability range observed in our resampling procedure resulted in substantially different frequency distributions and spatial patterns of modeled hydraulic variables. Second, using a bedload formula, we showed that propagation of uncertainty in the parameters of a gamma distribution used to model bed shear stress led to the coefficient of variation in predicted transport rates exceeding 50%. Overall, our findings underscore the importance of reporting the precision of estimated hydraulic parameters. When such estimates serve as input into models, uncertainty propagation should be explicitly accounted for by running ensemble simulations.
KW - bedload modeling
KW - channel hydraulics
KW - coarse-bed streams
KW - hydrodynamic modeling
KW - sampling variability
KW - spatial heterogeneity
UR - http://www.scopus.com/inward/record.url?scp=84960511195&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84960511195&partnerID=8YFLogxK
U2 - 10.1002/2015WR017259
DO - 10.1002/2015WR017259
M3 - Article
AN - SCOPUS:84960511195
VL - 52
SP - 1899
EP - 1922
JO - Water Resources Research
JF - Water Resources Research
SN - 0043-1397
IS - 3
ER -