Hoppa till huvudinnehåll

Kalendarium

14

October

Numerical Analysis Seminar - Gregory Tucker, Univ. of Colorado at Boulder

Tid: 2024-10-14 10:15 till 11:00 Seminarium

Gregory Tucker, Univ. of Colorado at Boulder Title: A community-based numerical modeling system for Earth-surface dynamics

Abstract. The earth's surface forms a complex dynamic interface between the solid earth and the overlying fluid flows that help shape terrain. Research in earth-surface dynamics relies heavily on numerical models of landscape and seascape evolution. One of the key challenges lies in cyberinfrastructure: modeling codes must be continually refined, adapted, and often coupled together to address emerging research questions. This seminar presents an overview of the Community Surface Dynamics Modeling System (CSDMS; https://csdms.colorado.edu/), a collection of open-source tools, protocols, and resources designed to facilitate numerical modeling of diverse earth-surface processes. An online Model Repository enables community members to share codes, and an API standard known as the Basic Model Interface (BMI) simplifies model coupling and re-use. CSDMS also coordinates the development of Landlab, a modular Python-language modeling library for writing 2D gridded numerical codes. Educational resources turn out to be essential for encouraging adoption and community contributions to the codebase. The value of a standardized, modular modeling toolkit is illustrated with an application in which an ensemble of 34 long-term erosion models, representing alternative working hypotheses, is used to reconstruct post-glacial landscape history at a site in the northeastern United States where buried hazardous materials are vulnerable to future erosion. In this example, parameter optimization and comparison with data are used to test the explanatory power of each model, revealing the importance of relatively subtle variations in lithology as well as super-linearity in the functional relationship between water flow and erosion. More broadly, the CSDMS experience illustrates how community-based libraries and protocols can facilitate model-based research.



Om händelsen
Tid: 2024-10-14 10:15 till 11:00

Plats
MH:309A

Kontakt
alexandros [dot] sopasakis [at] math [dot] lth [dot] se

Sidansvarig: webbansvarig@math.lu.se | 2017-05-23