Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
170 lines (105 sloc) 8.5 KB


Pandarus is a GIS software toolkit for regionalized life cycle assessment (LCA). It is designed to work with brightway LCA framework, brightway2-regional, and Constructive Geometries. A separate library, pandarus-remote, provides a web API to run Pandarus on a server.

In the context of life cycle assessment, regionalization means the introduction of detailed spatial information for inventory activities and impact assessment characterization maps. As these will have different spatial scales, GIS functionality is required to overlay these two maps. Pandarus can do the following:

  • Overlay two vector datasets, calculating the areas of each combination of features using the Mollweide projection.
  • Calculate the area of the geometric difference (the areas present in one input file but not in the other) of one vector dataset with another vector dataset.
  • Calculate statistics such as min, mean, and max when overlaying a raster dataset with a vector dataset.
  • Normalize raster datasets, including use of compatible nodata values
  • Vectorization of raster datasets

The outputs from Pandarus can be used in LCA software which does not include a GIS framework, thus speeding the integration of regionalization into the broader LCA community. There is also a conference presentation video introducing Pandarus and its integration into LCA software.

Pandarus is open source under the BSD 3-clause license.

Usage example

In addition to this documentation, there is also a Jupyter notebook usage example.

Intersecting two vector datasets

The main capability of the Pandarus library is to efficiently and correctly intersect the set of spatial features from one vector dataset with the spatial features from another vector dataset. In regionalized life cycle assessment, the first dataset would be inventory locations (polygons, lines, or points), and the second dataset would be regions with site-dependent characterization factors.


.. autofunction:: pandarus.intersect

Calculating areas

Because Pandarus was designed for global data sets, the Mollweide projection is used as the default equal-area projection for calculating areas (in square meters). Although no projection is perfect, the Mollweide has been found to be a reasonable compromise [1].

[1]Usery, E.L., and Seong, J.C., (2000) A comparison of equal-area map projections for regional and global raster data

Projections through the calculation chain

The function intersect calls intersection_dispatcher, which in turns calls intersection_worker, which itself calls get_intersection.

In both intersect and intersection_dispatcher, spatial units from both the first and second datasets are unprojected. Inside the function intersection_worker, spatial units from the first dataset are projected to WGS 84. get_intersection calls Map.iter_latlong on the second dataset, which returns spatial units projected in WGS 84. Area and linear calculations are done on the intersection of spatial units from both the first and second spatial datasets, and are projected to the Mollweide CRS. This projection is done at the time of areal or length calculations.

Lines and points that intersect two vector features


When calculating the lengths of lines (or number of points) remaining outside a intersected areas, we have the problem that lines can lie along the edge of two vector features, and hence the section of the line would be counted twice. If therefore need to adjust our formula for calculating the lengths (or number of points) outside an intersected area. The key insight is that we don't want the actual remaining area, but the right relative remaining area - these value are all normalized to one anyway.

The formula for allocating lengths is therefore:

( total\_actual\_length - total\_intersected\_length ) \cdot \frac{\sum individual\_intersected\_lengths}{total\_intersected\_length}

In the example above, the formula would give the answer 1.5:

1.5 = ( 3 - 2 ) \cdot \frac{2 + 1}{2}

The same procedure is followed for multipoints (geometries that have more than one point), except that instead of calculating lengths you just count the number of points.

Calculating area outside of intersections


For many regionalized methodologies, it is important to know how much area/length from one spatial dataset lies outside a second spatial dataset entirely. The function calculate_remaining calculates these remaining areas.

.. autofunction:: pandarus.calculate_remaining

Using intersections spatial dataset as a new spatial scale

.. autofunction:: pandarus.intersections_from_intersection

Calculating raster statistics against a vector dataset

Pandarus can calculate mask a raster with each feature from a vector dataset, and calculate the min, max, and average values from the intersected raster cells. This functionality is provided by a patched version of rasterstats.

The vector and raster file should have the same coordinate reference system. No automatic projection is done by this function.


.. autofunction:: pandarus.raster_statistics

Manipulating raster files

Pandarus provides some utility functions to help manage and manipulate raster files. Raster files are often provided with incorrect or missing metadata, and the main pandarus capabilities only work on vector files. Unfortunately, however, many raster files require manual fixes outside of these functions.

.. autofunction:: pandarus.clean_raster

.. autofunction:: pandarus.round_raster

.. autofunction:: pandarus.convert_to_vector


Why the name Pandarus?

The software overlays (matches) two different maps, and Pandarus was a bit of a matchmaker himself. Plus, ancient names are 172% more science-y.


Pandarus can be installed directly from PyPi using pip or easy_install, e.g.

pip install pandarus

However, it is easy to run into errors if geospatial libraries like fiona, rasterio, and shapely are compiled against different versions of GDAL. One way to get an installation that is almost guaranteed is to use Conda:

conda create -n <virtualenv name> python=3.6
activate <virtualenv name>  # Could also be source activate <virtualenv name>
conda install -c conda-forge -c cmutel pandarus

You should replace <virtualenv name> with a name you will remember.

Pandarus is only compatible with Python >= 3.5. Source code is on GitHub.


Pandarus uses the following libraries:


Contributions are welcome via pull requests and issues on Github. Please ensure that pull requests include tests and documentation. Pandarus follows the Contributor Covenant code of conduct.

Technical Reference

.. toctree::
   :maxdepth: 2