Single-Course Engelsk 5 ECTS

Data analysis, prediction and data-driven modelling of environmental systems

Overall Course Objectives

The overall objective of this course is to enable you to work with spatial and sensor data and use them to construct models of environmental systems in an independent manner. This is an advanced modelling course that combines concepts from e.g. 12104, 12320, 02450 and/or 02456 with the intention of making large, open data sources useful for prediction and analysis of environmental systems.

We will consider rainfall-runoff and pollution problems in the course, but the methods and challenges encountered in data management and model development are widely applicable for modelling and forecasting in environmental management (e.g. groundwater modelling, predictions of flooding, predictions of pollutant loads in rivers, etc.)

You will learn to
• Obtain spatial (e.g. elevation) and time series data (e.g. river water levels) from public sources and analyze them using scripts and GIS,
• Recognize and manage the faults and uncertainties included in the data sources
• Conceptualize, implement and validate model descriptions for environmental processes, and
• Integrate classical process descriptions and machine learning techniques to generate robust predictions

The course is aimed at students that are interested in advanced environmental modelling concepts and are interested in pursueing a scientific or industrial career in environmental data science. The course will be centered around real-world cases. It is assumed that you have a basic understanding of programming in scripting languages, GIS, and hydrological processes.

See course description in Danish

Manage and analyze large environmental data sets by means of scripting and GIS tools
Understand and analyse limitations of environmental observation data
Construct, implement and improve rainfall runoff models using scripting languages
Implement numerical parameter estimation procedures on self-written models
Apply systematic debugging procedures to detect programming mistakes in data management and modelling code
Formulate machine learning architectures that are applicable for predicting environmental processes
Understand and validate the assumptions underlying model calibration procedures
Conceptualize and implement combined physics-based and data-driven model architectures for environmental processes
Argue for modelling choices based on process understanding
Assess model performance with a focus on the modelling purpose
Communicate modelling results and limitations in different formats

Obtaining open data from public sources. Mathematical description of environmental processes. (Physics-informed) machine learning for environmental processes.

6 – 20 (Tues 8-12)

12320/12104/02402/02450/02002/02003, Programming in scripting languages (e.g. Python), GIS , and hydrological processes

Lectures and hands-on group work on 2-3 assignments with weekly question sessions

Roland Löwe

Contact or read more about Roland

See course in the course database.

Registration

Language	Engelsk
Duration	13 weeks
Institute	Sustain
Place	DTU Lyngby Campus
Course code	12345
Course type	Candidate
Semester start	Week 6
Semester end	Week 20
Days	Tues 8-12