Can we develop a modular, multi-process, efficient landscape evolution model? Can we ensure that all algorithms are implicit and O(n) complexity so that they can be used to perform ensemble simulations or be coupled to large-scale tectonic models? Can we develop new mathematical formulations (partial differential equations) to represent a wide range of surface processes and associated physical/chemical features?

The Earth Surface Process Modeling section at the GFZ specializes in the development of state-of-the-art landscape evolution models that predict the evolution of the Earth’s surface geometry on geological time scales under a wide range of forcings. These processes include fluvial erosion, transport and deposition, glacial erosion, hill slope processes (including landslides), marine transport and deposition and the formation of the regolith by weathering. We have developed or improved differential equations to represent these processes as well as highly efficient numerical methods to solve them.

In particular, we have developed an implicit, O(n) complexity algorithm for solving the Stream Power Law, taking into account the effect of sediment transport and erosion and the variability of flow/discharge. We have also developed an O(n) complexity algorithm to compute flow path on a surface that contains arbitrary local minima (sinks). We have developed models for glacial erosion that are also implicit and O(n) as well as a regolith formation model that include the effect of the formation of hardened duricrust. More recently, we developed, in collaboration from colleagues at AWI, an arctic delta model.

We have also developed models to predict the evolution of biodiversity on an evolving landscape and a model for predicting grain size in a sedimentary system.

We are currently developing a modular modeling framework that allows to solve relevant differential equations on a flow graph with high efficiency and flexibility. This new framework (FastScapeLib) can be used on any spatial discretization (i.e., on 1D, regular or irregular 2D meshes and on a sphere) making use of simple local kernels with multi-processor efficiency.

Much of the programming effort is performed by Benoît Bovy, the section software engineer. Contributions are from many section members. Coordination is by Jean Braun.

Project duration: Ongoing project, started in 2016

Funding agencies: Internal funding mostly

Section researchers involved: Dr. Benoit Bovy, many section members and Prof. Jean Braun

Cooperations:

• University Sofia-Antipolis, France
• ETH Zurich, Switzerland
• University of Lausanne, Switzerland
• AWI
• University of Rennes, France