...
Div | |||||||||
---|---|---|---|---|---|---|---|---|---|
| |||||||||
|
Div | |||||||||
---|---|---|---|---|---|---|---|---|---|
| |||||||||
|
...
Div | |||||||||
---|---|---|---|---|---|---|---|---|---|
| |||||||||
|
Proposed short list
Div | ||||||
---|---|---|---|---|---|---|
| ||||||
Code Block | ||||||
| ||||||
deltagamma #Fraction Breakerof parameterwave inheight Baldockto oradd Roelvinkto formulation facAswater depth BRfac # Calibration factor surface slope C # Calibration factor time averaged flows due to wave asymmetry fw ## BedChezy frictioncoefficient factorCFL beta #Maximum Courant-Friedrichs-Lewy number Tbfac Breaker slope coefficient in roller model alpha# Calibration factor for#bore Waveinterval dissipationTbore: coefficientTbore in= RoelvinkTbfac*Tbore formulationTsmin wetslp # Critical avalanching slope under water (dz/dx and dz/dy) facSk# Minimum adaptation time scale in advection diffusion equation sediment
alpha# Calibration factor time averaged flows due to# Wave dissipation coefficient in Roelvink formulation bed wave skewness gammax # Maximum ratio# waveCalibration heightfactor tofor waterbed depth cftransports [0..1]
beta # Breaker slopeFriction coefficientin roller model betad # Dissipation parameter long wave breaking turbulence break # Type of breaker formulation (1=roelvink, 2=baldock, 3=roelvink adapted, 4=roelvink on/off breaking) breakerdelay # Turn on (1) or off (0) breaker delay model bulk # Option to compute bedload and suspended load seperately; 0 = seperately, 1 = bulk (as in previous versions) cats # Current averaging time scale for wci, in terms of mean wave periods cf # Friction coefficient flow cmax # Maximum allowed sediment concentration correctHm0 # Turn off or on Hm0 correction delta # Fraction of wave height to add to water depth dispc # Coefficient in front of the vertical pressure gradient, Default = 1. dryslp # Critical avalanching slope above water (dz/dx and dz/dy) dtbc # Timestep used to describe time series of wave energy and long wave flux at offshore boundary (not affected by morfac) dwetlayer # Thickness of the top soil layer interacting more freely with the surface water dzmax # Maximum bedlevel change due to avalanching eps # Threshold water depth above which cells are considered wet eps_sd # Threshold velocity difference to determine conservation of energy head vs momentum epsi # Ratio of mean current to time varying current through offshore boundary facAs # Calibration factor time averaged flows due to wave asymmetry facDc # Option to control sediment diffusion coefficient [0..1] facSk # Calibration factor time averaged flows due to wave skewness facsd # fraction of the local wave length to use for shoaling delay depth facsl # Factor bedslope effect facua # Calibration factor time averaged flows due to wave skewness and asymmetry fcutoff # Low-freq cutoff frequency for instat = 4,5,6 boundary conditions form # Equilibrium sed. conc. formulation: 1 = Soulsby van Rijn, 1997, 2 = Van Rijn 2008 with modifications by Van Thiel fw # Bed friction factor fwcutoff # Depth greater than which the bed friction factor is NOT applied gamma # Breaker parameter in Baldock or Roelvink formulation gamma2 # End of breaking parameter in break = 4 formulation gammax # Maximum ratio wave height to water depth hmin # Threshold water depth above which Stokes drift is included hswitch # Water depth at which is switched from wetslp to dryslp hwci # Minimum depth until which wave-current interaction is used jetfac # Option to mimic turbulence production near revetments [0..1] kdmin # Minimum value of kd ( pi/dx > minkd ) lws # 1 = long wave stirring, 0 = no long wave stirring lwt # Switch 0/1 long wave turbulence n # Power in Roelvink dissipation model nc # Smoothing distance (defined as nr of cells) for estimating umean nmax # maximum ratio of cg/c fro computing long wave boundary conditions nuh # Horizontal background viscosity nuhfac # Viscosity switch for roller induced turbulent horizontal viscosity nuhv # Longshore viscosity enhancement factor, following Svendsen (?) order # Switch for order of wave steering, 1 = first order wave steering (short wave energy only), 2 = second oder wave steering (bound long wave corresponding to short wave forcing is added) por # Porosity reformsteep # Wave steepness criterium to reform after breaking rfb # If rfb = 1 then maximum wave surface slope is feeded back in roller energy balance; else rfb = par%Beta roller # Turn on (1) or off(0) roller model rugdepth # Minimum depth for determination of last wet point in runup gauge scheme # Use first-order upwind (upwind_1), second order upwind (upwind_2) or Lax-Wendroff (lax_wendroff) secbrsteep # Secondary maximum wave steepness criterium secorder # Use second order corrections to advection/non-linear terms based on MacCormack scheme sedcal # Sediment transport calibration coefficient per grain type setbathyfile # Name of prescribed bed update file shoaldelay # Turn on (1) or off (0) shoaling delay sigfac # dsig scales with log(sigfac). Default = 1.3 smag # Switch for smagorinsky subgrid model for viscocity smax # Being tested: maximum Shields parameter for ceq Diane Foster snells # Turn on (1) or off (0) Snell's law for wave refraction solver_acc # accuracy with respect to the right-hand side used solver_urelax # Underrelaxation parameter sourcesink # In suspended transport use source-sink terms to calculate bed level change (1) or sus transport gradients (0) sprdthr # Threshold ratio to maxval of S above which spec dens are read in (default 0.08*maxval) sus # Calibration factor for suspensions transports [0..1] sws # 1 = short wave & roller stirring and undertow, 0 = no short wave & roller stirring and undertow trepfac # Compute mean wave period over energy band: par%trepfac*maxval(Sf) for instat 4,5,6; converges to Tm01 for trepfac = 0.0 and tsfac # Coefficient determining Ts = tsfac * h/ws in sediment source term turb # Switch to include short wave turbulence: turbadv # Switch 0/1 to activate turbulence advection model for short and or long wave turbulence ucrcal # Critical velocity calibration coefficient per grain type umin # Threshold velocity for upwind velocity detection and for vmag2 in eq. sediment concentration waveform # Option for waveshape model: 1 = Ruessink & Van Rijn, 2 = Van Thiel de Vries, 2009 wci # Turns on (1) or off (0) wave-current interaction wetslp # Critical avalanching slope under water (dz/dx and dz/dy) z0 # Zero flow velocity level in Soulsby van Rijn (1997) sed.conc. expression |
Expert panel
- Dano Roelvink
- Ad Reniers
- Jaap van Thiel de Vries
- Robert McCall
- Ap van Dongeren
- Gerben Ruessink
Invitation
—
Dear XBeach expert,
Deltares is reconsidering the default settings of XBeach for the use of XBeach as advanced assessment model for the Dutch coast. As scientific developer or advanced XBeach user we would like your opinion on this subject.
XBeach provides about 250 parameters to be set by the user. Many of those are related to case-specific input or are related to processes and functionalities that are not relevant for the application as assessment model for the Dutch coast. Discarding all these parameters, still a long list of over 80 parameters will be available to the user.
An assessment model is supposed to be suitable for application by ordinary users. Therefore we would like to minimise the fraction of these 80+ parameters exposed to these users, preferably by a 100% so that only case-specific data is to be provided by the user. In order to do so, we need to offer a default set of parameter settings that are applicable for dune assessment computations along the Dutch coast. We intend to derive such default set of parameters by systematically varying the parameter settings for a selected set of validation cases that are representative for the Dutch coast. Ultimately we will pick the set of parameters that resembles the validation measurements best.
Systematic variation of a set of 80+ parameters is, however, unfeasible. Only if we would pick 2 values per parameter and a single validation case it would take a zillion years to run the necessary computations. Therefore the first step is to eliminate the long list of 80+ parameters to a short list of less than 10 parameters based on expert judgement. This is where your help is appreciated.
Attached to this e-mail you will find a list of the 80+ relevant XBeach parameters. We would like you to assign 25 points to the parameters that should, according to you, be considered in the systematic variation. Optionally, you may indicate a value range for each parameter. We will divide about 10.000 XBeach simulations over the parameters that received the most points. The number of parameters is not set beforehand. We will consider including more parameters (dimensions) over including more variations per parameter (resolution).
As a guidance, we also attached two plots that provide some insight in the (marginal) sensitivity of XBeach for the 80+ parameters found in the list. The marginal sensitivity is determined by running a 1D XBeach simulation using a schematised profile representative for the Dutch coast and normative storm conditions representative for the location Hoek van Holland. During each simulation only a single parameter is varied within it's applicable range. Some parameters depend on others (hwci, jetfac, waveform, facsd, z0, betad, BRfac, bed). In these cases both parameters are set, but only the parameter of interest is varied within it's range.
The first plot shows the sensitivity of the erosion volume above SSL for each parameter over it's valid range of values. The black dot depicts the default value. The red line is a linear fit excluding outliers. The second plot shows the range of erosion volumes found when varying each parameter, again excluding outliers. The second plot is ordered based on the variation of erosion volumes found. Outliers are excluded since they often correspond to valid, but unrealistic values (e.g. dzmax=0).
Please use these plots as guidance only. The plots are generated using a single model set-up only and not representative for all cases that should be taken into consideration. Also some parameters with large influence are not good candidates for systematic variation. For example, the parameter "sus" has a large influence, but it is probably not a good idea to calibrate XBeach turning suspended sediment transport on or off. On the other hand, it might not be so useful to include a parameter to the short list that has no influence at all.
A few remarks on the continuation of this calibration effort might be important to your input:
- We only consider 1D cases (transects)
- We only consider erosion volumes above SSL as performance indicator (we will look at profiles separately, later on)
- We will scale the validation cases to prototype scale and run XBeach at that scale (we will look into scaling issues separately, later on)
- We do not include the depthscale parameter in this study (it is always 1), but we will include the parameters set by the depthscale parameter (we will look into scaling issues seperately, later on)
- We exclude the following processes / functionalities from calibration beforehand:
- vegetation
- wind
- ships
- non-hydrostatic
- drifters
- mpi
- morfac
- discharge
- groundwater flow
- beach wizard
- sediment fractions
- stationary wave solver
Your input is highly appreciated. You can provide your point assignments by a reply to this e-mail. Any list in any regular file format will do.
If you do not wish to participate, or if you will not be able to provide your input before May, 15th 2014, please let is know.
Best regards,
Bas Hoonhout
Pieter van Geerhttp://publicwiki.deltares.nl/display/XBEACH/Default+settings
—
Short list
Variation matrix
Resolution vs. dimensions
We can choose either to vary a large number of parameters (dimensions) using a limited resolution or vary a small number of parameters using a high resolution.
Cases
- Deltaflume ?
- Deltaflume 2006
- Boers ?
- 1953 storm surge
Result
Discussion
Several topics within this workflow are still being discussed. These discussions are briefly mentioned below.
Performance indicators
XBeach allows a variety of performance indicators to be used to compare the performance of one set of settings against another. For example, a profile comparison with measured data may be used by computing a BSS score. The BSS score itself, however, is subject to discussion for this purpose. Moreover, comparison data is often obtained from flume experiments on scale. These experiments use scaling rules that are derived by comparing erosion profiles rather than profile shape. The profile shape for these tests can thus not be used. For now, we use the erosion volume above SSL as performance indicator. Another advantage of this indicator is that it relates well to other calibration studies for dune assessment models in the Netherlands.
Scaling
Comparison with data is often done using data obtained from flume experiments. These experiments are performed on scale. XBeach allows us to simulate the experiment at the scale of the experiment rather than on a 1:1 scale. As for now, this is not done for two reasons. First, the scaling within XBeach is still subject to discussion. Second, other calibration studies for dune assessment models also first scaled the experiment results and then performed the calibration.
Depthscale
The depthscale parameter is subject to discussion and should preferably be omitted. It is therefore not part of the long list. A separate study will investigate the scaling performance in XBeach.
Processes / functionalities
What processes / functionalities do we include? Probably not the following:
- vegetation
- wind
- ships
- non-hydrostatic
- drifters
- mpi
- morfac
- discharge
- groundwater flow
- beach wizard
- sediment fractions
- stationary wave solver
But what about:
...
flow
|
Expert panel
- Dano Roelvink
- Ad Reniers
- Jaap van Thiel de Vries
- Robert McCall
- Ap van Dongeren
- Gerben Ruessink
- Kees den Heijer
- Arnold van Rooijen
- Joost den Bieman
- ...
Invitation
—
Dear XBeach expert,
Deltares is reconsidering the default settings of XBeach for the use of XBeach as advanced assessment model for the Dutch coast. As scientific developer or advanced XBeach user we would like your opinion on this subject.
XBeach provides about 250 parameters to be set by the user. Many of those are related to case-specific input or are related to processes and functionalities that are not relevant for the application as assessment model for the Dutch coast. Discarding all these parameters, still a long list of over 80 parameters will be available to the user.
An assessment model is supposed to be suitable for application by ordinary users. Therefore we would like to minimise the fraction of these 80+ parameters exposed to these users, preferably by a 100% so that only case-specific data is to be provided by the user. In order to do so, we need to offer a default set of parameter settings that are applicable for dune assessment computations along the Dutch coast. We intend to derive such default set of parameters by systematically varying the parameter settings for a selected set of validation cases that are representative for the Dutch coast. Ultimately we will pick the set of parameters that resembles the validation measurements best.
Systematic variation of a set of 80+ parameters is, however, unfeasible. Only if we would pick 2 values per parameter and a single validation case it would take a zillion years to run the necessary computations. Therefore the first step is to eliminate the long list of 80+ parameters to a short list of less than 10 parameters based on expert judgement. This is where your help is appreciated.
Attached to this e-mail you will find a list of the 80+ relevant XBeach parameters. We would like you to assign 25 points to the parameters that should, according to you, be considered in the systematic variation. Optionally, you may indicate a value range for each parameter. We will divide about 10.000 XBeach simulations over the parameters that received the most points. The number of parameters is not set beforehand. We will consider including more parameters (dimensions) over including more variations per parameter (resolution). Parameters that are not included will be set to their current default value.
As a guidance, we also attached two plots that provide some insight in the (marginal) sensitivity of XBeach for the 80+ parameters found in the list. The marginal sensitivity is determined by running a 1D XBeach simulation using a schematised profile representative for the Dutch coast and normative storm conditions representative for the location Hoek van Holland. During each simulation only a single parameter is varied within it's applicable range. Some parameters depend on others (hwci, jetfac, waveform, facsd, z0, betad, BRfac, bed). In these cases both parameters are set, but only the parameter of interest is varied within it's range.
The first plot shows the sensitivity of the erosion volume above SSL for each parameter over it's valid range of values. The black dot depicts the default value. The red line is a linear fit excluding outliers. The second plot shows the range of erosion volumes found when varying each parameter, again excluding outliers. The second plot is ordered based on the variation of erosion volumes found. Outliers are excluded since they often correspond to valid, but unrealistic values (e.g. dzmax=0).
Please use these plots as guidance only. The plots are generated using a single model set-up only and not representative for all cases that should be taken into consideration. Also some parameters with large influence are not good candidates for systematic variation. For example, the parameter "sus" has a large influence, but it is probably not a good idea to calibrate XBeach turning suspended sediment transport on or off. On the other hand, it might not be so useful to include a parameter to the short list that has no influence at all.
A few remarks on the continuation of this calibration effort might be important to your input:
- We only consider 1D cases (transects)
- We only consider erosion volumes above SSL as performance indicator (we will look at profiles separately, later on)
- We will scale the validation cases to prototype scale and run XBeach at that scale (we will look into scaling issues separately, later on)
- We do not include the depthscale parameter in this study (it is always 1), but we will include the parameters set by the depthscale parameter (we will look into scaling issues seperately, later on)
- We exclude the following processes / functionalities from calibration beforehand:
- vegetation
- wind
- ships
- non-hydrostatic
- drifters
- mpi
- morfac
- discharge
- groundwater flow
- beach wizard
- sediment fractions
- stationary wave solver
Your input is highly appreciated. You can provide your point assignments by a reply to this e-mail. Any list in any regular file format will do.
If you do not wish to participate, or if you will not be able to provide your input before May, 15th 2014, please let is know.
Best regards,
Bas Hoonhout
Pieter van Geer
http://publicwiki.deltares.nl/display/XBEACH/Default+settings
—
Short list
Variation matrix
Resolution vs. dimensions
We can choose either to vary a large number of parameters (dimensions) using a limited resolution or vary a small number of parameters using a high resolution.
Cases
- Deltaflume ?
- Deltaflume 2006
- Boers ?
- 1953 storm surge
Result
Discussion
Several topics within this workflow are still being discussed. These discussions are briefly mentioned below.
Performance indicators
XBeach allows a variety of performance indicators to be used to compare the performance of one set of settings against another. For example, a profile comparison with measured data may be used by computing a BSS score. The BSS score itself, however, is subject to discussion for this purpose. Moreover, comparison data is often obtained from flume experiments on scale. These experiments use scaling rules that are derived by comparing erosion profiles rather than profile shape. The profile shape for these tests can thus not be used. For now, we use the erosion volume above SSL as performance indicator. Another advantage of this indicator is that it relates well to other calibration studies for dune assessment models in the Netherlands.
Scaling
Comparison with data is often done using data obtained from flume experiments. These experiments are performed on scale. XBeach allows us to simulate the experiment at the scale of the experiment rather than on a 1:1 scale. As for now, this is not done for two reasons. First, the scaling within XBeach is still subject to discussion. Second, other calibration studies for dune assessment models also first scaled the experiment results and then performed the calibration.
Depthscale
The depthscale parameter is subject to discussion and should preferably be omitted. It is therefore not part of the long list. A separate study will investigate the scaling performance in XBeach.
Processes / functionalities
What processes / functionalities do we include? Probably not the following:
- vegetation
- wind
- ships
- non-hydrostatic
- drifters
- mpi
- morfac
- discharge
- groundwater flow
- beach wizard
- sediment fractions
- stationary wave solver
But what about:
- structures
- short-wave runup
- shoaling delay
- breaker delay
- wave-current interaction
- long-wave stirring
- long-wave turbulence
Python toolbox
In order to efficiently perform the workflow described above, some Python tools have been developed. The toolbox is available through the OpenEarthTools repository: https://svn.oss.deltares.nl/repos/openearthtools/trunk/python/applications/xbeach/default_settings. A IPython notebook is available at the same location that illustrates the simple workings of the toolbox.
The main functions are as follows:
Code Block | ||
---|---|---|
| ||
import default_settings
# read and write an overview of available XBeach parameters
default_settings.filesys.write_overview('all_params.txt')
ignore_params = default_settings.filesys.read_overview('ignore_params.txt')
# include dependencies and exclude parameters read from ignore list
dependencies = {'hwci' : {'wci' : 1},
'jetfac' : {'swrunup' : 1},
'waveform' : {'turb' : 'wave_averaged'},
'facsd' : {'shoaldelay' : 1},
'z0' : {'form' : 'soulsby_vanrijn'},
'betad' : {'lwt' : 1},
'BRfac' : {'rfb' : 1},
'bed' : {'bulk' : 0}}
params = default_settings.xbeach.get_parameters(dependencies=dependencies)
params_longlist = {k:v for k,v in params.iteritems() if k not in params_ignored}
# write models for marginal sensitivity analysis
default_settings.filesys.write_models(params_longlist, overwrite=False)
# run models for marginal sensitivity analysis
default_settings.filesys.run_models(overwrite=False)
# monitor model progress
default_settings.filesys.stat_models()
# read erosion volumes above SSL
V = default_settings.filesys.read_volumes()
# plot sensitivity
fig, axs = default_settings.plot.plot_volumes(V, params=params)
fig, axs = default_settings.plot.plot_sensitivity(V, params=params)
|
...