Skip to content

Methodology

This chapter describes in more detail the steps FM2PROF takes to go from a 2D representation of reality to a 1D representation of reality. Conceptually, FM2PROF works in three distinct steps: (1) initialisation, (2) building cross-sections and (3) finalisation.

Initialisation

Image title
Image caption

The initialisation step involves parsing (i.e. reading and making available for further analysis) the 2D data. In this step control volumes and sections are defined as well. This step may take some time to complete, but this preprocessing greatly reduces the computation times in the next step.

Import region file

The Region file is specified in the configuration file.

Warning

For any reasonably sized river model, it is currently not advised to supply a polygon. Instead, a NetCDF file should be supplied. See issue_region_polygon for more information.

Import section file

The Section file is specified in the configuration file.

Warning

For any reasonably sized river model, it is currently not advised to supply a polygon. Instead, a NetCDF file should be supplied. See issue_region_polygon for more information.

Import 2D data

Parsing 2D data

Dflow2d uses a staggered grid to solve the (hydrostatic) flow equations. Because of this staggered approach, there is not a single 2D point that has all information. Flow information (flow velocity, discharge) is stored in [flow links]{.title-ref}, while geometry (bed level) is stored in cell faces. needs both information from the faces, as from the links.

Image title
The dflow2d staggered grid.

Below is a table that lists all variables read by from dflowd map output.

FM2PROF variable Variable in dflow2d output

x (at face) mesh2d_face_x

y (at face) mesh2d_face_y

area (at face) mesh2d_flowelem_ba

bedlevel (at face) mesh2d_flowelem_bl

x (at flow link) mesh2d_edge_x

y (at flow link) mesh2d_edge_y

edge_faces (at flow link) mesh2d_edge_faces

edge_nodes (at flow link) mesh2d_edge_nodes

waterdepth mesh2d_waterdepth

waterlevel mesh2d_s1

chezy_mean mesh2d_czs

chezy_edge mesh2d_czu

velocity_x mesh2d_ucx

velocity_y mesh2d_ucy

velocity_edge mesh2d_u1

: Overview of variables from the 2D model that are used by FM2PROF

Classification of volumes

Control volumes are used to define which 2D datepoints are linked to which 1D cross-section. This is done in the following steps:

  • Each node, link and face is assigned a Region{.interpreted-text role="term"}. This is currently done through DeltaShell (see issue_region_polygon)
  • For each region seperately, a scikit-learn KNearestNeighbour classifier is trained.
  • The classifier is used to uniquely identify each 2D link and each face to a 1D cross-section

Classification of sections

Sections <Section> are used to output a different roughness function for the main channel and the floodplains. The purpose of the classification is to determine whether a 2D point belongs to the main channel section, or to the floodplain section (see warning below).

Two methods are implemented:

  • Variance based classification<section_classification_variance>{.interpreted-text role="ref"}
  • Polygon-based classification using DeltaShell (see issue_region_polygon)

Warning

If classification is done using DeltaShell, it is possible to define more than two sections. However, this functionality is not tested and may not work properly.

Build Cross-Section

Image title
The dflow2d staggered grid.

Once initialisation is complete, will loop over each cross-section location<Cross-section location>{.interpreted-text role="term"}. In each iteration, the program takes two steps: (1) building the geometry and (2) building the roughness tables.

Warning

No cross-section will be generated for locations that have no 2D data assigned or have less than 10 faces assigned. This may happen if a location lies outside the 2D grid, or if there are many cross-section closely together. If this happens, an error is raised by FM2PROF. The user should check the cross-section location input file to resolve the problem.

Build Geometry

In each loop, a number of steps is taken based on the 2D data that is uniquely assigned to that cross-section:

  • Lakes are identified using the identify_lakes
  • Flow volume and Storage volume are seperated using the distinguish_storage
  • The water level dependent geometry<Water level (in)dependent geometry>{.interpreted-text role="term"} is computed using wl_dependent_css{.interpreted-text role="ref"}
  • The water level independent geometry<Water level (in)dependent geometry>{.interpreted-text role="term"} is computed using wl_independent_css
  • The parameters for Summerdikes are defined using the sd_optimisation
  • Finally, the cross-section is simplified using the Visvalingam-Whyatt method of poly-line vertex reduction simplify_css

Build roughness

At each cross-section point, a roughness look-up table is constructed that relates water level (in m + NAP) to a Chézy roughness coefficient. This is done in three steps:

  • For each section, a roughness table is constructed by averaging the 2D points

Finalisation

Image title
The dflow2d staggered grid.