Module Contents



An unstructured reader. Data is gridded irregularily.

class opendrift.readers.basereader.unstructured.UnstructuredReader[source]

Bases: opendrift.readers.basereader.variables.Variables

An unstructured reader. Data is gridded irregularily.

The initial type of grid that this class supports are triangular prisms. Unstructured in xy-coordinates, x and y is constant in z. z might be non-cartesian (e.g. sigma-levels).

Initialize self. See help(type(self)) for accurate signature.

abstract get_variables(self, variables, time=None, x=None, y=None, z=None)[source]

Obtain and return values of the requested variables at all positions (x, y, z) closest to given time.


Dictionary with arrays of length len(time) with values at exact positions x, y and z.

_get_variables_interpolated_(self, variables, profiles, profiles_depth, time, reader_x, reader_y, z)[source]

This method _must_ be implemented by every reader. Usually by subclassing one of the reader types (e.g. structured.StructuredReader).

Arguments are in _native projection_ of reader.

_build_boundary_polygon_(self, x, y)[source]

Build a polygon of the boundary of the mesh.

param x

Array of node x position, lenght N

param y

Array of node y position, length N


A shapely.prepareped.prep shapely.Polygon.

The boundary of the mesh, ideally including holes in the mesh.



Try this alogrithm:

Boundary edges (line between two nodes) are only referenced by a single triangle.

  1. Find a starting edge segment: [v_start, v_next] (v is vertex or node)

  2. Find another _unvisited_ edge segment [v_i, v_j] that has either v_i = v_next or v_j = v_next and add the one not equal to v_next to the polygon.

  3. Reset v_next to the newly added point. Mark edge as visited.

  4. Continue untill we reach v_start.

The polygon has a rotation, but this should not matter for our purpose of checking the bounds.

Note: In order to find holes in the polygon all points must be scanned.

Approximate using the convex hull:

An alternative simple approximation is to use the convex hull of the points, but this will miss points along the boundary which form a wedge in the boundary (as well as holes in the mesh).

Holes in the mesh will often be covered by the landmask anyway, so they will usually not be a problem.

covers_positions(self, x, y, z=0)[source]

Check which points are within boundary of mesh.

_build_rtree_(self, x, y)[source]

Builds an R-tree of x, y

_build_ckdtree_(self, x, y)[source]
__nearest_ckdtree__(self, idx, x, y)[source]

Return index of nearest point in cKDTree

static __nearest_rtree__(idx, x, y)[source]

Take array of points and get nearest point in rtree index.

_nearest_node_(self, x, y)[source]

Return nearest node (id) for x and y

_nearest_face_(self, xc, yc)[source]

Return nearest element or face (id) for xc and yc