Earth systems modelling

This course is structured into four hands-on modules. Three modules are dedicated to using and developing different types of models typically used by Earth Scientists. Many models rely on solving Partial Differential Equations (PDE). They either represent an efficient simplification thereof, e.g., in the form of box modelling (module 2, or directly write and solve (un)coupled sets of PDE’s (module 3). A new branch of modelling constitutes data-driven modelling, which exploits artificial intelligence to build models based on observations (module 2). These first three modules integrate the learning of more advanced programming with acquiring the basics of different methods used for modelling. The last module focuses on the process of modelling itself, such that you can translate Earth Science problems into effective modelling strategies. Together these modules complete the programming, modelling and data skill learning lines.

You will develop and write your own box model from scratch. Step-by-step you will develop a complete numerical model, which quantifies to what depth the water level would fall if the Mediterannean Sea was disconnected from the Atlantic Ocean. This provides one example of how a numerical model can help to answer Earth Sciences questions that can not be answered based upon field observations or theoretical calculations alone.

You will develop and use data-driven models. You will learn the basics of different machine learning models, including how neural networks, random forests, and long short-term memory algorithms work. You will train models using hydrological data and optimise their performance on different data sets. Finally, you build a machine learning model from scratch to simulate a new hydrological data set.

You will develop or program various codes from scratch. Starting from physical conservation laws you will write codes to simulate 1D and 2D flow; heat, chemical and pressure diffusion; and coupled momentum and continuity equations with variable viscosity. Your developments are based on a staggered grid, finite difference formulation, which facilitates understanding of numerical mathematics for Earth Science students. Through gradually increasing code complexity you will be introduced to spatial and temporal discretisation, initial and boundary conditions, and basic numerical solvers. Applications to different Earth Science directions demonstrate the universality of your learnings and broad applicability of the codes you will develop.

Finally, your learnings will be applied to perform a numerical modelling research project. The thermomechanical code used for this module is an extension of the building blocks you have programmed yourself in the PDE-module. You can choose to apply this code to simulate subduction over millions of years to understand earthquake sizes, flow of a mountain glacier, or tsunami propagation over a lake. You are tasked to formulate a research question in one of these Earth Science domains based on literature and develop and execute a modelling strategy to answer it. You will execute your research project step-by-step accumulating into a poster presentation.