Why Fastscapelib?

There are already many Landscape Evolution Models (LEMs) available as well as a few great open-source frameworks for landscape evolution modeling such as Landlab and Badlands, so why do we need yet another tool?

Fastscapelib’s philosophy can be summarized as “write algorithms once, reuse it everywhere”.

The last decades have seen a great number of LEM alternatives proposed (see, e.g., Barnhart et al. [BGST19]), yet most alternatives depend on a few, common concepts and algorithms, e.g., for routing flow on the topographic surface. It is quite usual to see the same – sometimes complex – algorithms re-implemented over and over, with only minor differences from one implementation to another.

Flexibility, Reusability and Efficiency

Fastscapelib’s goal is to provide a core, “bare-bones” library that is focused on the three following aspects: flexibility, reusability and efficiency.

Maximizing reusability is made via a flexible grid system that is decoupled from flow routing and through first-class APIs in both C++ and Python. Support could be easily extended to other languages like R or Julia thanks to the Xtensor library (more details here). In addition, Fastscapelib makes as few as possible assumptions about how a LEM should be organized (e.g., variable names, units, etc.).

Optimizing for both flexibility and efficiency is hard (there are always trade-offs) but we did our best trying to achieve it at the expense of a fairly complex internal C++ implementation although with extensive unit tests and carefully designed API and abstractions.

What is Fastscapelib for?

Fastscapelib provides some core structures (grids, meshes, flow graphs) and solvers (diffusion, Stream Power Law) for building custom LEMs and/or for integrating it with other existing codes, models (e.g., geodynamics, climate) or frameworks written in various languages.

Fastscapelib will eventually replace the current Fortran backend of the Fastscape high-level Python package.

What is Fastscapelib not for?

Fastscapelib is a bare-bones library. It doesn’t provide any ready-to-use LEM nor anything for setting up simulations, driving it and processing or visualizing the outputs. Instead, you can freely use your tools of choice for those tasks.

Fastscapelib will never provide a list of model components that is as exhaustive as in Landlab.

Roadmap

Fastscapelib is still at an early stage of development and only provides a limited number of features regarding flow routing, bedrock erosion, sediment transport, etc.

TODO.