Doctoral School

Applied Machine Learning for Environmental Sciences: A Hands-On Workshop Using iMESC

Target audience

PhD candidates and early-career researchers from the Faculty of Sciences and the Faculty of Bioscience Engineering. The course is tailored for researchers who are already actively involved in data analysis. Expected background knowledge includes experience in handling, formatting, and structuring datasets in Microsoft Excel or similar data applications. Participants should preferably—but not restrictively—possess their own empirical research datasets to deploy and analyze during the practical workflow sessions.

Organizing and scientific committee

Ellen Pape (UGent, Department Biology)

Abstract

This course democratizes machine learning (ML) for environmental scientists using iMESC, an interactive, code-free application. Tailored for researchers dealing with complex, high-dimensional ecological and climate data, it bridges the gap between traditional statistics and predictive modeling. Participants will rapidly progress from foundational theory to executing autonomous ML workflows directly on their own empirical research datasets.

Objectives

By the end of this intensive workshop, participants will be able to:

  • Differentiate between traditional statistical paradigms and machine learning approaches.
  • Independently install, configure, and navigate the iMESC application environment.
  • Deploy and interpret unsupervised clustering algorithms for pattern recognition in ecological data.
  • Implement supervised predictive algorithms to model environmental trends and classifications.
  • Validate and score model accuracy, accounting for data biases. 
  •  Design, execute, and troubleshoot end-to-end data science workflows using their own datasets.

Dates and venue

23 November - 25 November 2026

Meeting Room 0.1 (Campus Sterre, S8)

Programme

Day 1: Foundations, System Setup, and Data Ingestion (8 Hours)


Morning Session (09:00 – 13:00 | 4 hours)

  • Topic 1: Foundations of ML vs. Traditional Statistics: Conceptual introduction contrasting classical frequentist/Bayesian statistics with machine learning paradigms. Understanding when to shift from hypothesis testing to predictive modeling.
  • iMESC Software Installation: Guided technical setup, environment configuration, and initial troubleshooting on participants' personal devices.
  • Topic 2: Framing the Research Problem: Conceptual framework on how to correctly structure an environmental or marine biology research question into a machine learning task (classification, regression, or clustering).

Lunch Break (13:00 – 14:00)

Afternoon Session (14:00 – 18:00 | 4 hours)

  • Topic 3: Data Ingestion Architecture: Learning how iMESC structures, reads, and processes environmental variables. Best practices for data cleaning, handling missing values, and formatting matrices.
  • Exploring the Data in iMESC: Hands-on lab focused on loading data into the app, exploring user interface dashboards, and executing initial descriptive data summaries.


Day 2: Algorithmic Implementations: Unsupervised & Supervised Learning (8 Hours)


Morning Session (09:00 – 13:00 | 4 hours)

  • Topic 4: Deep Dive into Unsupervised Learning (Clustering): Theoretical foundations of pattern recognition and clustering algorithms without pre-labeled target variables.
  • Topic 5: Unsupervised Practices & Own Examples: Hands-on laboratory session where candidates use benchmark data and their own imported datasets to identify natural ecological communities or environmental zones.


Lunch Break (13:00 – 14:00)

Afternoon Session (14:00 – 18:00 | 4 hours)

  • Topic 6: Supervised Learning & Validation Metrics: Theoretical framework for predictive modeling (classification and regression). Detailed study of validation metrics, scoring, and cross-validation techniques.
  • Topic 7: Supervised Practices & Own Examples: Practical application lab where participants deploy predictive algorithms to map or predict environmental metrics using iMESC, testing models directly on their own data.

Day 3: Advanced Modeling Architecture & Research Dissemination (8 Hours)

Morning Session (09:00 – 13:00 | 4 hours)

  • Topic 8: Advanced Modeling Approaches: Dedicated methodological lecture and practice tackling complex environmental data constraints:
  • Spatial Cross-Validation: Accounting for spatial autocorrelation to prevent model overfitting.
  • Time-Series: Handling temporal dependencies and tracking shifts over time
  • Experimental Designs: Integrating structured experimental layouts into machine learning workflows.


Lunch Break (13:00 – 14:00)


Afternoon Session (14:00 – 18:00 | 4 hours)

  • Topic 9: Participant Presentations & Peer Review: Interactive capstone session where participants present the machine learning workflows they constructed over the 3-day course using their own datasets. Includes group troubleshooting, peer feedback, and expert review by the lecturers to finalize their research or monitoring designs.

Registration

Registration fee

Free of charge for Doctoral School members.

Number of participants

Maximum 20

Language

English

Training method

  • Lectures ex-cathedra: 4 hours (Conceptual foundations, algorithmic frameworks, and evaluation theory).
  • Practical exercises / Guided software labs: 8 hours (Step-by-step navigation of iMESC, testing supervised and unsupervised models with benchmark data).
  • Interactive data troubleshooting & workflow implementation: 8 hours (Active group coaching and peer discussion while applying workflows to personal datasets).
  • Presentations by course participants: 4 hours (Presentation of the research question and implemented ML workflow)

Evaluation method

Evaluation is based on 100% attendance and active participation in all practical sessions. To pass, candidates must successfully import an environmental dataset into the iMESC environment and demonstrate the initial execution of a verified machine learning workflow during Day 3 of the workshop.

After successful participation, the Doctoral School Office will add this course to your curriculum of the Doctoral Training Programme in Oasis. Please note that this can take up to one to two months after completion of the course.