Building a model¶

Introduction¶

Preparing the input maps for a distributed model is not always trivial. wflow comes with two scripts that help in this process. The scripts are made with the assumption that the base DEM you have is a higher resolution as the DEM you want to use for the final model. When upscaling the scripts try to maintain as much information from the high resolution DEM as possible. The procedure described here can be used for all wflow models (wflow_sbm or wflow_hbv).

Using the scripts¶

The scripts assume you have a DEM, landuse and soil map available in pcraster format. If you do not have a soil or landuse map the you can generate a uniform map. The resolution and domain of these maps does not need to be the same, the scripts will take care of resampling. The process is devided in two scripts, wflow_prepare_step1.py and wflow_prepare_step2.py. In order to run the scripts the following maps/files need to be prepared.

• a DEM in pcraster format
• a land use map in pcraster format. If the resolution is different from the DEM the scripts will resample this map to match the DEM (or the DEM cutout). If no landuse map is found a uniform map will be created.
• a soil map in pcraster format. If no soil map is found a unifrom map will be created.
• a configuration file for the prepare scripts that defines how they operate (.ini format) file (see below)
• an optional shape file with a river network
• an optional catchment mask file

The scripts work in two steps, each script need to be given at least one command-line option, the configuration file. The first script performs the following tasks:

• wflow_prepare_step1.py

  1. Performs an initial upscaling of the DEM if required (set in the configuration file). This initial upscaling may be needed if the processing steps (such as determining the drainage network) take a very long time or if the amount of available memory is not sufficient. The latter may be the case on 32bit systems. For example a 90x90 m grid for the Rhine/Meuse catchment could not be handled on a 32 bit system.

  2. Create the local drainage network. If the ldd is already present if will use the existing ldd. Use the force option to overwrite an existing ldd.

  3. Optionally use a shape file with a river network to “burn-in” this network and force the ldd to follow the river. In flat areas wher the river can be higher than the surrounding area having a river shape is crucial.

     Tip

     Another option is to prepare a “pseudo dem” from a shape file with already defined catchment boundaries and outlets. Here all non boundary points would get a value of 1, all boundaries a value of 2 and all outlets a value of -10. This helps in generating a ldd for polder areas or other areas where the topography is not the major factor in determining the drainage network.

  4. Determine various statistics and also the largest catchment present in the DEM. This area will be used later on to make sure the catchments derived in the second step will match the catchment derived from the high resolution DEM

• wflow_prepare_step1.py

  1. Create a map with the extend and resolution defined in the configuration file and resample all maps from the first step to this resolution

  2. Create a new LDD using the following approach:

     • Construct a new dem to derive the ldd from suing the minimum dem from the first step for all the pixels that are located on a river and the maximum dem from the first step for all other pixels.
     • In addition raise all cells outside of the largest catchment defined in the first step with 1000 meter divided by the distance of each cell to the largest catchment.
     • Derive the ldd and determine the catchments

     

     Steps in creating the wflow model input

Once the script is finished successfully the following maps should have been created, the data type is shown between brackets:

• wflow_catchment.map (ordinal)
• wflow_dem.map (scalar)
• wflow_demmax.map (scalar)
• wflow_demmin.map (scalar)
• wflow-dem*percentile* - (10,25,33,50,66,75,90) (scalar)
• wflow_gauges.map (ordinal)
• wflow_landuse.map (nominal)
• wflow_soil.map (nominal)
• wflow_ldd.map (ldd)
• wflow_outlet.map (scalar)
• wflow_riverburnin.map (boolean)
• wflow_riverlength_fact.map (scalar)
• wflow_river.map (ordinal)
• wflow_streamorder.map (ordinal)
• wflow_subcatch.map (ordinal)

The maps are created in the data processing directory. To use the maps in the model copy them to the staticmaps directory of the case you have created.

Command line parameters¶

Both scripts take the same command-line parameters:

wflow_prepare_step1 -I inifile [-W workdir][-f][-h]

    -f force recreation of ldd if it already exists
    -h show this information
    -W set the working directory, default is current dir
    -I name of the ini file with settings

contents of the configuration file¶

An example can be found here.

[directories]
# all paths are relative to the workdir set on the command line
# The directories in which the scripts store the output:
step1dir = step1
step2dir = step2

[files]
# Name of the DEM to use
masterdem=srtm_58_14.map
# name of the lad-use map to use
landuse=globcover_javabali.map
soil=soil.map
# Shape file with river/drain network. Use to "burn in" into the dem.
river=river.shp
riverattr=river
# The riverattr above should be the shapefile-name without the .shp extension

[settings]
# Nr to reduce the initial map with in step 1. This means that all work is done
# on an upscaled version of the initial DEM. May be usefull for very
# large maps. If set to 1 (default) no scaling is taking place
initialscale=1

# Set lddmethod to dem (other methods are not working at the moment)
lddmethod=dem

# If set to 1 the gauge points are moved to the neares river point on a river
# with a strahler order higher of identical as defined in this ini file
snapgaugestoriver=1

# The strahler order above (and including) a pixel is defined as a river cell
riverorder=4

# X and y cooordinates of gauges (subcatchments). Please note the the locations
# are based on the river network of the DEM used in step2 (the lower resuolution
# DEM). This may need some experimenting... is most case the snap function
# will work by ymmv. To set multiple gauges use [x_gauge_1, x_gauge_2]

gauges_y = [-6.1037]
gauges_x = [107.4357]


# settings for subgrid to create. This also determines how the
# original dem is (up)scaled. If the cellsize is the same
# as the original dem no scaling is performed. This grid will
# be the grid the final model runs on
Yul = -6.07
Xul = 106.9
Ylr = -7.30271
Xlr = 107.992
cellsize = 0.009166666663

# tweak ldd creation. Default should be fine
lddoutflowdepth=1E35
lddglobaloption=lddout

Problems¶

In many cases the scripts will not produce the maps the way you want them in the first try. The most common problems are:

1. The gauges do not coincide with a river and thus the subcatchment is not correct

   • Move the gauges to a location on the rivers as determiend by the scripts. The best way to do this is to load the wflow_subcatch.map in qgis and use the cursor to find the nearest river cell fro a gauge.

2. The delimited catchment is not correct even if the gauges is at the rigth location

   • Get a better DEM or fix the current DEM.
   • Use a river shape file to fix the river location
   • Use a catchment mask to force the catchment delineated to use that. Or just clip the DEM with the catchment mask.

If you still run into problems you can adjust the scripts yourself to get better results.

wflow_prepare_step1¶

wflow data preparation script. Data preparation can be done by hand or using the two scripts. This script does the first step. The second script does the resampling. This scripts need the pcraster and gdal executables to be available in you search path.

Usage:

wflow_prepare_step1 [-W workdir][-f][-h] -I inifile 

-f force recreation of ldd if it already exists
-h show this information
-W set the working directory, default is current dir
-I name of the ini file with settings

$Id: $

wflow_prepare_step1.OpenConf(fn)¶

wflow_prepare_step1.configget(config, section, var, default)¶

gets parameter from config file and returns a default value if the parameter is not found

wflow_prepare_step1.main()¶

Variables:	masterdem – digital elevation model dem – digital elevation model river – optional river map

wflow_prepare_step1.usage(*args)¶

wflow_prepare_step2¶

wflow data preparation script. Data preparation can be done by hand or using the two scripts. This script does the resampling. This scripts need the pcraster and gdal executables to be available in you search path.

Usage:

wflow_prepare_step2 [-W workdir][-f][-h] -I inifile 

-f force recreation of ldd if it already exists
-h show this information
-W set the working directory, default is current dir
-I name of the ini file with settings

$Id: $

wflow_prepare_step2.OpenConf(fn)¶

wflow_prepare_step2.configget(config, section, var, default)¶

wflow_prepare_step2.main()¶

wflow_prepare_step2.resamplemaps(step1dir, step2dir)¶

Resample the maps from step1 and rename them in the process

wflow_prepare_step2.usage(*args)¶

Introduction¶

As with all hydrological models calibration is needed for optimal performance. Currently we are working on getting the link with the OpenDA calibration environment running (not tested yet). We have calibrated the Rhine/Meuse models using simple shell scripts and the XX and XX command-line parameters to multiply selected model parameters and evaluate the results later.

Parameters¶

M

Once the depth of the soil has been set for the different land-use types the M parameter is the most important variable in calibrating the model. The decay of the conductivity with depth controls the baseflow resession and part of the stormflow curve.

N

The Manning N parameter controls the shape of the hydrograph (the peak parts). In general it is advised to set N to realistic values for the rivers, for the land phase higher values are usually needed.

Ksat

Increasing the Ksat will lower the hydrograph (baseflow) and flatten the peaks. The latter also depend on the shape of the catchment.

FirstZoneCapacity

Increasing the storage capacity of the soil will decrease the outflow

RunoffGeneratingGWPerc

Default is 0.1. Determines the (upper) part of the groudwater that can generate runoff in a cell. This is only used of the RunoffGenSigmaFunction option is set to 1. In general generating more runoff before a cell is completely saturated (which is the case if RunoffGenSigmaFunction is set to 0) will lead to more baseflow and flattening of the peaks.

Changes in hydrographs for different values of parameters¶