Modelisation of suspended sediment transport in rivers Master thesis Véronique Briguet 2011 Alain Recking, Oldrich Navratil, Nicolle Mathys

Contents 2 1. Introduction 2. Data set 3. Data analysis 4. Modelling 5. Perspectives 6. Conclusion

3 Suspension versus bedload Data set

4 Fraser River at Yale Lochsa River 2 mm d50 du lit GBRSBR Sand Bed River versus Gravel Bed River

5 Connexion with hillslopes processes: Head Water Streams HWS versus Lowland Rivers LR Erosion on the Draix Catchment Data set

0.03 mg/L < C < 29 g/L; 0.01% < S < 18%; <Q< 3770 m 3 /s ; 0.08 < A < km². MeasurementNbr of value Rivers Source 1Instantaneous Bedload + susp reaches, 76 rivers, 15 SBR, 62 GBR USGS, USDA … 2Instantaneous susp SBR Brownlie Annual load139 1 SBR and 8 GBR, 139 years ORE Draix, McLean, Church et al 1999 … 4Event load213 5 HWS ORE Draix

Log(A)Log(C)Log(d50)Log(H)Log(P)Log(q)Log(qb)Log(qs)Log(S)Log(U)Log(W) Log(C) Log(qs) Correlation coefficient R A: Watersherd area D50: median diameter H: depth P: power  QS U: Velocity C: Suspended load concentration q s : Suspended load / unit width q b : bedload / unit width q: Q/W W: Width S: Slope q s =Cq => autocorrelation Correlation between q b and C Data Analysis 7

8 Dispersion du ratio instantané Qs / QT Suspension versus bedload in the total load Data Analysis Instantaneous measurements

9 Event scale Data Analysis Bedload (t) Suspension (t)

10 Data Analysis Annual load Bedload (t) Suspension (t)

11 Suspension – bedload interactions: some hypothesis  1 : bedload  2 >  1 : progressive suspension concentration SBR  1 : weak suspension  2 >  : Sharp suspension concentration with bedload GBR Data Analysis HWSInconsistencies in Qs/Qt between event and volumes Possible bias in instantaneous measurements with bedload absent for the flood conditions considered

Modelling Deterministics : Bagnold (1966), Einstein (1950), Celik et Rodi (1991)… Empirical models : Lefort (1990), Abrahams (2001)… 12 Equations of fluid mechanics Constants calibration with experimental data Deterministic Model Identification of representative variables Fit equations with experimental data Empirical model SUSPENSION Qs OR TOTAL TRANSPORT QT

13 Limits of deterministics and empirical models: Most of them calibrated in flume With uniform materials Fine sands Modelling Correlation analyses with field data Fit equations « black box » models Statistical models qs=f(q): Turowski & Rickenmann (2010) Use generally limited to the river used to built the data set (Prosser & Rustomji 2000)

14 Confluence Galabre - Bès Haute Bléone à Prads Bès à Sivan Modelling Width? Bed diameter? Transported diameter? Fall velocity?...

20/06/ Modelling Discrepancy ratio r = q s calculated q s measured Scores = % of r values obtained in a given interval Tested in the interval [0.1 – 10] Ex: a scores of 30% significates that 30% of the predictions are correct within plus or minus one order of magnitude

20/06/ Modelling

20/06/ Modèle de Bagnold (1966) Modèle de Celik et Rodi (1991) Modelling Auto correlation ! q s =Cq

20/06/ Modelling q s =Cq with C randomly choosen

20/06/ => An accurate concentration model is required Modèle de Celik et Rodi (1991) Modelling

20/06/ Perspectives ?

Tested with a new bedload model specifically developed for gravel beds (Recking 2010) 21 C calculated with q b measured C calculated with q b computed Perspectives ?

22 Perspectives ?

Conclusion 23 No tool is really efficient in the field, especially in gravel bed rivers Strong correlation between suspended load and bedload, especially in gravel bed rivers Necessity to develop a new concentration model

20/06/ Thank you for your attention