Bachelor level course in Physical Geography at Marburg University
Monitoring biodiversity and its change through space and time is a key challenge for ecogeographical research. Linking up-to-date remote sensing data with machine learning has been shown to be a powerful complementation of labor-intensive field work in order to derive area-wide biodiversity indices, which also capture larger spatiotemporal scales.
Using LiDAR and Sentinel data as example, this course will be a foundation for acquiring, managing, and modelling different types of geodata and will provide a solid baseline for workflow automation with remote sensing data in R.
Intended learning outcomes
At the end of this course you should be able to
- feel comfortable with the usage of R,
- know the basics of remote sensing theory and machine learning,
- handle LiDAR and Sentinel data,
- use random forest for spatial predictions,
- include and process different types of remote sensing data in a single workflow, and
- apply all of the above to a real-world example.
Course features
Given the actual COVID-19 situation, this course is intended as a blended learning module in our study program and will be conducted online-only upon further notice.
The instructors will set up a digital meeting room for the official duration of the course on a weekly basis (Wednesdays, 10:15 - 14:00, starting 22.4.2020). Course units will consist of one topic (see table below), different ways of their practical application in R and an exercise for applying your new skills. The topics will be presented as online readers, screencasts, videos or collections of blog posts and examples of other researchers. You can do the exercises at your own pace and we will discuss them on the following Wednesday.
Further there will we a forum and chat group to discuss organizational matters and questions. We expect to resume presence teaching sometime during the semester. Participants will be informed by email as soon as possible when the situation changes.
Organizational matters and further details will be discussed during our first digital kick-off session, for which all participants (those who joined the course via MARVIN) will be invited by email.
Course overview
Session overview
| Session | Topic | Content |
|---|---|---|
| R Basics | ||
| 01 | The very basics | Course introduction, expectations, organisational matters, data and information, R, R Studio |
| 02 | More basics | Working environment, data types, object types, operators, indexing, plotting, reading and writing |
| 03 | Working with spatial data | Raster data, vector data, coordinate reference systems, reading and writing spatial data, spatial operators, mapping |
| Iteration I: Start simple | ||
| 04 | Remote Sensing Theory | Sensor types, spectral properties, satellites, data acquisition, temporal aspects |
| 05 | Remote Sensing in R | Satellite data processing, simple lidar parameters, data management |
| 06 | Modelling Theory | Machine Learning Basics, Random Forest, Validation |
| 07 | Modelling in R | Preprocessing, classification trees, random forest, validation |
| Iteration II: Reality | ||
| 08 | Advanced Remote Sensing in R | Spectral Indices, spatial indices, spatial operations, resampling, merging |
| 09 | Built-in hold | |
| 10 | Advanced Modelling I | Sampling strategies, caret, cross validation |
| 11 | Advanced Modelling II | Predictions, outlook |
| 12 | Built-in hold | |
| 13 | Wrap up | Get feedback from your peers and instructors, tell us how you self-assess your skills and happy holidays |
Deliverables
The graded course certificate will be based on an individual assignment (“Hausarbeit”), which builds upon the processing workflow established during the course.
Preparation and prerequisites
Basic knowledge of R and of handling spatial data is beneficial, but not required.
Instructors
