This release will create a DOI via Zenodo which will be the DOI referenced in the upcoming publication about this model.

Fixed a bug where some package data wasn't properly installing from PyPI. Now, pip install ADVECTOR works like a charm!

First release of the 3D ADVECTOR.

This release fixes a bug in which particles whose release_date falls between the penultimate and ultimate output timestamps in a chunk will fail to be properly released into the simulation. Namely, they will be released with a null initial position, rather than their initial position from the sourcefile. In practice, this means if your particle release dates were randomly distributed, some small percentage of particles would fail to release, and would return from the kernel with exit code "NULL_LOCATION." The smaller your chunks are, the larger the percentage of particles which would return with this failure.

Adds support for devices running opencl 2.1. Small fix to an unusual int32 overflow bug on chunking calculations.

This release modifies partial derivative calculations so that 2nd-order taylor advection scheme works as expected. Before this version, particles would drift outwards from circular streamlines in both the eulerian and taylor2 advection schemes. Now, this drift is almost entirely negated from the taylor2 advection scheme, i.e. it is working properly.

This release fixes a major bug in V0.0, namely that the spacing of the coordinate grids was calculated incorrectly in the kernel. This leads to particles incorrectly selecting the nearest grid cell to themselves, with the error increasing along coordinates up to 1 full grid cell error at latitude 90, longitude 180. This resulted in particles visibly behaving incorrectly at shorelines, and invisibly querying incorrect current/wind vectors, southwest of their true locations.

This version can be considered the first functional 2D version of ADVECTOR.

• It can work with HYCOM files.
• It was proven able to reproduce results that were previously output by the MATLAB Plastic Dispersal Model, and with more particles.
• It can run on a supercomputer such a SURFSARA's Cartesius.
• It is only able to run on one GPU at a time.
• Results should be properly formatted, and a log should be output telling you about warnings.

Note : It is not possible to run ADVECTOR v0.0 with several GPUs or in a distributed architecture.
This should be planned for v0.1