This section introduces the different entities that can be created and stored in the geoh5 file format.



Groups are effectively containers for other entities, such as Objects (Points, Curve, Surface, etc.) and other Groups. Groups are used to establish parent-child relationships and to store information about a collection of entities.


By default, the parent of any new Entity is the workspace RootGroup. It is the only entity in the Workspace without a parent. Users rarely have to interect with the Root group as it is mainly used to maintain the overall project hierarchy.



A ContainerGroup can easily be added to the workspace and can be assigned a name and description.

from geoh5py.groups import ContainerGroup
from geoh5py.workspace import Workspace

# Create a blank project
workspace = Workspace("my_project.geoh5")

# Add a group
group = ContainerGroup.create(workspace, name='myGroup')

At creation, "myGroup" is written to the project geoh5 file and visible in the Analyst project tree.


Any entity can be accessed by its name or uid (unique identifier):

print(workspace.get_entity("myGroup")[0] == workspace.get_entity(group.uid)[0])


The geoh5 format enables storing a wide variety of Object entities that can be displayed in 3D. This section describes the collection of Objects entities currently supported by geoh5py.



The Points object consists of a list of vertices that define the location of actual data in 3D space. As for all other Objects, it can be created from an array of 3D coordinates and added to any group as follow:

from geoh5py.workspace import Workspace
from geoh5py.objects import Points
import numpy as np

# Create a blank project
workspace = Workspace("my_project.geoh5")

# Generate a numpy array of xyz locations
n = 100
radius, theta = np.arange(n), np.linspace(0, np.pi*8, n)

x, y = radius * np.cos(theta), radius * np.sin(theta)
z = (x**2. + y**2.)**0.5
xyz = np.c_[x.ravel(), y.ravel(), z.ravel()] # Form a 2D array

# Create the Point object
points = Points.create(
    workspace,       # The target Workspace
    vertices=xyz     # Set vertices



The Curve object, also known as a polyline, is often used to define contours, survey lines or geological contacts. It is a sub-class of the Points object with the added cells property, that defines the line segments connecting its vertices. By default, all vertices are connected sequentially following the order of the input vertices.

from geoh5py.objects import Curve

# Create the Curve object
curve = Curve.create(
    workspace,       # The target Workspace

Alternatively, the cells property can be modified, either directly or by assigning parts identification to each vertices:

# Split the curve into two parts
part_id = np.ones(n, dtype="int32")
part_id[:75] = 2

# Assign the part
curve.parts = part_id



Drillhole objects are different from other objects as their 3D geometry is defined by the collar and surveys attributes. The vertices and cells properties are only instantiated when interval or point log data are added.

from geoh5py.objects import Drillhole

# Create a simple well
total_depth = 100
dist = np.linspace(0, total_depth, 10)
azm = np.ones_like(dist) * 45.
dip = np.linspace(-89, -75, dist.shape[0])
collar = np.r_[0., 10., 10]

well = Drillhole.create(
    workspace, collar=collar, surveys=np.c_[dist, dip, azm]



The Surface object is also described by vertices and cells that form a net of triangles. If omitted on creation, the cells property is calculated using a 2D scipy.spatial.Delaunay triangulation.

from geoh5py.objects import Surface
from scipy.spatial import Delaunay

# Create a triangulated surface from points
surf_2D = Delaunay(xyz[:, :2])

# Create the Surface object
surface = Surface.create(
    vertices=points.vertices, # Add vertices



The Grid2D object defines a regular grid of cells often used to display model sections or to compute data derivatives. A Grid2D can be oriented in 3D space using the origin, rotation and dip parameters.

from geoh5py.objects import Grid2D

# Create the Surface object
grid = Grid2D.create(
    origin = [25, -75, 50],
    u_cell_size = 2.5,
    v_cell_size = 2.5,
    u_count = 64,
    v_count = 16,
    rotation = 90.0,
    dip = 45.0,



The BlockModel object defines a rectilinear grid of cells, also known as a tensor mesh. The cells center position is determined by cell_delimiters (offsets) along perpendicular axes (u, v, z) and relative to the origin. BlockModel can be oriented horizontally by controlling the rotation parameter.

from geoh5py.objects import BlockModel

# Create the Surface object
blockmodel = BlockModel.create(
    origin = [25, -100, 50],
    u_cell_delimiters=np.cumsum(np.ones(16) * 5), # Offsets along u
    v_cell_delimiters=np.cumsum(np.ones(32) * 5), # Offsets along v
    z_cell_delimiters=np.cumsum(np.ones(16) * -2.5),  # Offsets along z (down)
    rotation = 30.0



The Octree object is type of 3D grid that uses a tree structure to define cells. Each cell can be subdivided it into eight octants allowing for a more efficient local refinement of the mesh. The Octree object can also be oriented horizontally by controlling the rotation parameter.

from geoh5py.objects import Octree

octree = Octree.create(
        origin=[25, -100, 50],
        u_count=16,      # Number of cells in power 2
        u_cell_size=5.0, # Base cell size (highest octree level)
        w_cell_size=2.5, # Offsets along z (down)

By default, the octree mesh will be refined at the lowest level possible along each axes.



The geoh5 format allows storing data (values) on different parts of an Object. The data_association can be one of:

  • OBJECT: Single element characterizing the parent object

  • VERTEX: Array of values associated with the parent object vertices

  • CELL: Array of values associated with the parent object cells

Note: The length and order of the array provided must be consistent with the corresponding element of association.

The data types supported by geoh5py are:

  • Arrays

  • Integer

  • Text

  • Color_map


Add data

Data can be added to an Object entity using the add_data method.

# Create a straight Curve object
curve = Curve.create(
    workspace,       # The target Workspace
    vertices=np.c_[np.linspace(0, 100, 100), np.zeros(100), np.zeros(100)]

# Add a single string comment
    "my_comment": {
        "values": "hello_world"

# Add a vector of floats
    "my_cell_values": {
        "values": np.random.randn(curve.n_cells)

# Add multiple data vectors on a single call
data = {}
for ii in range(8):
    data[f"Period:{ii}"] = {
        "values": (ii+1) * np.cos(ii*curve.vertices[:, 0]*np.pi/curve.vertices[:, 0].max()/4.)

data_list = curve.add_data(data)
print([obj.name for obj in data_list])
['Period:0', 'Period:1', 'Period:2', 'Period:3', 'Period:4', 'Period:5', 'Period:6', 'Period:7']

If the association argument is omited, geoh5py will attempt to assign the data to the correct part based on the shape of the data values, either object.n_values or object.n_cells

The newly created data is directly added to the project’s geoh5 file and available for visualization:


Get data

Just like any Entity, data can be retrieved from the Workspace using the get_entity method. For convenience, Objects also have a get_data_list and get_data method that focusses only on their respective children Data.

my_list = curve.get_data_list()
print(my_list, curve.get_data(my_list[0]))
['Period:0', 'Period:1', 'Period:2', 'Period:3', 'Period:4', 'Period:5', 'Period:6', 'Period:7', 'my_cell_values', 'my_comment'] [<geoh5py.data.float_data.FloatData object at 0x7f2ca0f68bd0>]

Well Data

In the case of Drillhole objects, data are added as either interval log or point log values.

Point Log Data

Log data are used to represent measurements recorded at discrete depths along the well path. A depth attribute is required on creation. If the Drillhole object already holds point log data, geoh5py will attempt to match collocated depths within tolerance. By default, depth markers within 1 centimeter are merged (collocation_distance=1e-2).

depths_A = np.arange(0, 50.) # First list of depth

# Second list slightly offsetted on the first few depths
depths_B = np.arange(47.1, 100)

# Add both set of log data with 0.5 m tolerance
    "my_log_values": {
            "depth": depths_A,
            "values": np.random.randn(depths_A.shape[0]),
    "log_wt_tolerance": {
            "depth": depths_B,
            "values": np.random.randn(depths_B.shape[0]),
            "collocation_distance": 0.5
[<geoh5py.data.float_data.FloatData at 0x7f2cace41150>,
 <geoh5py.data.float_data.FloatData at 0x7f2cace41dd0>]


Interval Log Data

Interval log data are defined by constant values bounded by a start an end depth. A from-to attribute is expected on creation. Users can also control matching intervals by supplying a tolerance argument in meters (default tolerance: 1e-3 meter).

# Add some geology as interval data
    "interval_values": {
        "values": [1, 2, 3],
        "from-to": np.vstack([
            [0.25, 25.5],
            [30.1, 55.5],
            [56.5, 80.2]
        "value_map": {
            1: "Unit_A",
            2: "Unit_B",
            3: "Unit_C"
        "type": "referenced",
<geoh5py.data.referenced_data.ReferencedData at 0x7f2cdba4cbd0>


Property Groups

Data entities sharing the same parent Object and association can be linked within a property_groups and made available through profiling. This can be used to group data that would normally be stored as 2D array.

# Add another VERTEX data and create a group with previous
curve.add_data_to_group([obj.name for obj in data_list], "my_trig_group")
<geoh5py.groups.property_group.PropertyGroup at 0x7f2ca0ef2290>


# Update the geoh5 and re-write the Root