Supervised Machine Learning Readiness

This activity is designed for use in an upper-level undergraduate or graduate course with a focus on data in any Earth system science discipline. While the specific topics center on precipitation and the measurement of atmospheric variables, the content is presented in an introductory manner to ensure accessibility for students from a broad range of Earth systems science backgrounds.

Experience interrogating tabular datasets, interpreting graphs, and writing summaries is recommended. Learners should also have demonstrated proficiency in the NSF Unidata eLearning core digital competencies.

The final two modules in the learning series require Python. If students will be independently installing Python on their personal computers, they will require the ability to follow the installation instructions provided.

The modules so far have been delivered as a series of lab activities. However, the first module (Foundations) could be delivered as independent pre-work for a flipped classroom setting. The final module (Analysis) is designed as a small-group (2-4 learners) activity.

This series introduces supervised machine learning as a data analysis tool in the Earth systems sciences. Learners are introduced to a generalized framework for how a machine learning model gets developed, from problem framing, to data handling, to development and testing.

An emphasis is placed on critical judgement skills, encouraging learners to reflect and assess each choice made in model development. Learners must synthesize information from the data and scientific context to make recommendations on the utility of machine learning for a given task.

Learners will gain experience with scientific writing and interacting with a Jupyter Notebook environment. In the final module (Analysis), learners also have the opportunity to practice skills in building consensus in small groups.

This three-part series introduces supervised machine learning as a data analysis tool in the Earth systems sciences without requiring any advanced math, statistics, or programming skills.

How to Access Supervised Machine Learning Readiness

The learning series is available on NSF Unidata eLearning with a free account.
https://elearning.unidata.ucar.edu/course/view.php?id=13

Module 01 - Foundations

No-code eLearning

Estimated Time

• 1 hour

Learning Objectives

• Define machine learning in terms of its goals or purpose
• Explain the steps of a general supervised machine learning analysis
• Distinguish scenarios that are and are not appropriate for supervised machine learning analysis

Preview eLearning (read-only): Module 01 - eLearning (HTML)

Module 02 - Applications

Jupyter Notebook

Estimated Time

• 1 hour

Learning Objectives

• Describe the data used in the scenario, including discussion of the context to the physical world
• Create and make strategic refinements to a machine learning model following a guided low-code workflow
• Describe a multifaceted classification model evaluation, including accuracy, precision, and recall scores
• Justify a decision on the utility of a machine learning model based on evaluation metrics, physical context, impacts on the Earth Systems Science scenario

Preview Jupyter Notebook (read-only): Module 02 Jupyter Notebook (HTML)

Module 03 - Analysis 

Jupyter Notebook

Estimated Time

• 1 hour

Learning Objectives

• Develop a clear and comprehensive problem statement for a given machine learning scenario, considering the goals, constraints, and context of the analysis
• Summarize the characteristics of a provided dataset
• Summarize and justify the decisions made during the processes of model development, including feature selection and intermediate evaluation metrics
• Evaluate the model's performance in solving the initial problem, and recommend potential refinements or improvements for future iterations

Preview Jupyter Notebook (read-only): Module 03 Jupyter Notebook (HTML)

Module 01 - Foundations:
This eLearning module can be completed in a lab setting or as independent pre-work for a flipped classroom setting.

Module 02 - Applications:
This module (as well as Module 03) is ideal for classrooms that are already using a Jupyter Hub. For setting up a Jupyter workspace individually, follow these instructions: Set up the Jupyter Workspace (Microsoft Word 2007 (.docx) 184kB May19 25)
Note that some of the rich HTML in markup cells may not render correctly in Google Colab. The Jupyter Notebook modules begin with an introductory video that introduces the scenario as well as how to interact with a Jupyter workspace.Learners will complete exercises in a separate handbook document, and a corresponding rubric is available.

Module 03 - Analysis:
This module is set up the same as Module 02, but is recommended to be completed in small (2-4 learners) groups to allow learners the opportunity to practice building consensus.

Module 01:
A 10-question multiple-choice quiz is included in the eLearning module. Each question is linked to one of the listed learning objectives.

Modules 02 + 03:
Students complete handbook activities with short answers. Handbooks are graded with the associated rubric. Each handbook activity is linked to one of the listed learning objectives.

Preview Module 02 Handbook: 02 Applications - Machine Learning Model Handbook.docx (Microsoft Word 2007 (.docx) 175kB May19 25)

Preview Module 02 Rubric: 02 Applications - Rubric - Machine Learning Model Handbook.docx (Microsoft Word 2007 (.docx) 178kB May19 25)

Preview Module 03 Handbook: 03 Analysis - Machine Learning Model Handbook.docx (Microsoft Word 2007 (.docx) 176kB May19 25)

Preview Module 03 Rubric: 03 Analysis - Rubric - Machine Learning Model Handbook.docx (Microsoft Word 2007 (.docx) 179kB May19 25)