Collaborative moisture tracking with WAM2layers v3

Atmospheric moisture tracking is a valuable technique for understanding the physical processes that drive (extreme) precipitation and drought in our changing climate. By following where precipitated moisture originally evaporated (backtracking) or where evaporated moisture eventually precipitates (forward tracking) we can gain valuable insights into the connection of large-scale weather systems and hydrometeorological events, land-atmosphere interactions, or the impact of land-use changes on water availability.

The WAM2layers model is an Eulerian moisture tracking code that solves the water balance equation for tagged moisture in gridded model output data. With the increasing resolution of weather and climate models, however, data handling and performance have become serious constraints. Over the past year, we have worked on a new version of WAM2layers (github.com/WAM2layers/WAM2layers), in which we tackle these computational challenges and make a substantial upgrade to the user- and developer-friendliness of the model. The most important changes are summarized below.

Usability: the new version of WAM2layers separates configuration from code. This makes it possible to run many different model simulations without modifying the source code. The model can now be run with a single command, supplying a configuration file as an input argument. It is even possible to use the model without copying the code. Simply install the wam2layers Python package from PyPI.

Modularity: we have made a stricter separation between preprocessing steps, the actual tracking code, and utilities for analysing the results. This is important, for example, for working with multiple datasets. So far, we've worked with ERA5 data. Adding support for other datasets requires no modifications to the tracking code, only a separate preprocessing script.

Documentation: the new version of the model comes with documentation on ReadTheDocs. The documentation includes theory, installation instructions, a complete user guide, and contributing guidelines.

Collaborative development: previous versions of the model were already available on GitHub, but further development often happened offline and without coordination. From the start of this project, we have opened up the development process such that everyone can ask questions, raise issues, and open pull requests. The brand-new documentation includes instructions for anyone willing to contribute. We believe this shift represents a modern perspective on collaborative research practice.