This Python package provides a python implementation of the method proposed by Moretti and Orlandini (2023) for extracting physically meaningful thalweg and ridge networks directly from high-resolution, unaltered digital elevation models (DEMs). The extracted networks form the basis for a new approach to terrain partitioning that preserves essential topographic features without relying on depression filling or grid coarsening. This work is a direct translation of the original Fortran implementation described in the published article :

Moretti, G., & Orlandini, S. (2023). Thalweg and ridge network extraction from unaltered topographic data as a basis for terrain partitioning. Journal of Geophysical Research: Earth Surface, 128, e2022JF006943. 

https://doi.org/10.1029/2022JF006943

Full documentation is available at:

https://IGNF.github.io/landmark/

landmark processes high-resolution DEMs to identify and extract interconnected thalweg and ridge networks without modifying the original elevation data. These networks are derived from a slopeline-based analysis and provide a physically consistent representation of convergent (thalweg) and divergent (ridge) terrain structures.

Key functionalities include:

• Detection of ridge points as local high-divide locations uncrossed by slopelines.
• Identification of endorheic and exorheic basins, and their connections through spilling saddles.
• Construction of a fully connected thalweg network, including within nested depressions.
• Derivation of a ridge network structured around dispersal areas.
• Support for terrain partitioning using extracted networks as breaklines, enabling unstructured mesh generation for hydrological or geomorphological modeling.

This tool is particularly useful in mountainous or lowland terrains where preserving depressions is essential for a meaningful representation of surface processes.

To install the package, clone the repository and install it in editable mode.

git clone https://github.com/IGNF/landmark.git
cd landmark
pip install -e .

git clone git@github.com:IGNF/landmark.git
cd landmark
pip install -e .

You can use LANDMARK directly in Python by calling the main function:

from landmark.main_landmark_geotiff import landmark_processing

landmark_processing(
    mnt_path="path/to/input_dem.tif",
    output_dir="path/to/output",
    a_spread=100000,
    a_out=100000,
    hso_th=5,
    curvature_slope=True,
    n_pts_calc_slope=5,
    no_data_values=[-9999, 0]
)

This will create two GeoPackage files containing slopelines and ridgelines respectively.

You can also run LANDMARK as a command-line interface (CLI) after installing the package.

landmark path/to/dem.tif \
    --output_dir path/to/output \
    --a_spread 100000 \
    --a_out 100000 \
    --hso_th 5 \
    --curvature_slope \
    --n_pts_calc_slope 5 \
    --no_data_values -9999 0

Two GeoPackage files are created in the output directory:

• slopelines_se_HSO_<input_name>.gpkg – extracted thalweg network
• ridgelines_se_HSO_<input_name>.gpkg – extracted ridge network

These can be directly opened in QGIS or any GIS software for further analysis.

The input Digital Elevation Model must:

• Be a GeoTIFF file.
• Use a projected coordinate system with metric units (e.g., UTM, Lambert 93).
  Do not use a geographic CRS with degrees (such as WGS84 / EPSG:4326), as the algorithm relies on metric distances for all processing.
  Using degrees will result in incorrect and unusable outputs.

This project is licensed under the MIT License. See the LICENSE.TXT file for details.