Introduction to Practical Stochastic Modelling

Dates: Monday, 1 March 2027 to Wednesday, 3 March 2027
Time: Each day from 9:00 am to 5:00 pm

• Pfaffenwaldring 5b (5.01 und 5.02)
• Pfaffenwaldring 5a, second floor

This course is organised by the GS SimTech and SFB 1313

Via registration form below. Registration will soon be open.

The short course "Introduction to Practical Stochastic Modelling" will be organised as a hands-on team workshop. Each day will start with a one-hour introductory lecture, followed by structured discussion, implementation, and testing tasks in Python. Short team presentations and a joint debriefing will conclude each day.

The course will use a simple lumped rainfall–runoff model as a recurring example, and use SFB 1313-specific problems for team work. The topics will include sources and representations of uncertainty, Monte Carlo forward uncertainty quantification, sensitivity analysis, Bayesian parameter inference, conditional prediction, and Sequential Importance Resampling for data assimilation. Bayesian model comparison will be included if time permits.

Scientific models contain uncertainty in their parameters, inputs, observations, initial states, and sometimes in the equations used to describe the system. Practical stochastic modelling begins by asking where uncertainty enters a model, what it represents, and how available knowledge or beliefs can be expressed through probability distributions.

In this three-day hands-on course, participants will develop a complete stochastic-modelling workflow using SFB 1313-specific applications (from Area A, B, C), while the lectures will use a simple lumped rainfall-runoff model as a recurring example. They will assign probability distributions to uncertain quantities, propagate uncertainty by Monte Carlo simulation, analyse parameter sensitivities, and use Bayes' theorem for parameter inference and conditional prediction. Sequential Importance Resampling will then be introduced as a practical method for assimilating streamflow observations that become available over time. If time permits, the workflow will be extended to uncertainty-aware comparison of alternative rainfall-runoff model equations.

Each day starts with a one-hour introductory lecture. Participants then work in small teams to discuss modelling choices, implement the methods in Python, test their solutions, and interpret the results. Short team presentations and a joint debriefing conclude each day. Optional extensions address low-dimensional representations of temporally or spatially varying uncertainty.

Prerequisites  Basic familiarity with Python is helpful. No advanced background in probability or statistics is required.