Climate AI Homework Series

Created by Ryan Albelda
Carnegie Mellon University – Spring 2025
Special Topics in AI Engineering and Policy for Climate Mitigation and Resilience

This is the inaugural offering of the course, marking its first introduction. Prior to this homework series, no previous assignments had been developed, making these the pioneering exercises for the course.

---

About This Homework Series

This collection includes six assignments designed for graduate students exploring the intersection of AI and climate policy.

• Audience: Students with limited Python experience but general code literacy.
• Duration: Each assignment was designed to be completed in 7–10 days.
• Scope: Homework 5 and 6 are intentionally shorter in length and complexity.

---

Homework Summaries

1. Homework 1

• Topic: Explore Python for climate data
• Objectives:
  • Review foundational Python skills and libraries for climate data analysis
  • Apply basic Python operations and programming logic
  • Load, explore, and visualize climate datasets with scatter plots and histograms
  • Apply and interpret linear regression and standardization techniques
• Climate Dataset: January temperatures data from various cities in Pennsylvania. Data range from 1949 to 2024. Data from NOAA

---

2. Homework 2

• Topic: Linear Regression and Simple Error Metrics
• Objectives:
  • Understand key statistical tools used in climate data analysis
  • Calculate and interpret the Pearson correlation coefficient
  • Explain the R-squared value for model fit
  • Use Mean Squared Error (MSE) as a model accuracy metric
  • Fit a linear model using scikit-learn’s LinearRegression
• Climate Dataset: Yearly global carbon emissions from Oak Ridge National Laboratory

---

3. Homework 3

• Topic: Predicting Crop Yield: A Machine Learning Approach
• Objectives:
  • Handle and interpolate missing data
  • Summarize key statistics and trends for each variable
  • Visualize a correlation matrix as a heatmap
  • Research and assess an AI use case in the transportation sector
• Climate Dataset: Publicly available FAO and World Bank crop yield data

---

4. Homework 4

• Topic: Water Data to Explore SVM and KNN
• Objectives:
  • Apply classification and regression techniques to real-world water quality data
  • Distinguish between binary and multi-class classification using water and Iris datasets
  • Use K-Nearest Neighbors (KNN) and Support Vector Machines (SVM) to classify potable water
  • Evaluate models using R², MSE, and RMSE
  • Compare performance across SVM, KNN, Decision Tree, Random Forest, and Linear Regression
  • Explore segmentation and clustering (e.g., K-Means) to enhance predictions
• Climate Dataset: Dataset for Assessing Water Quality for Drinking and Irrigation (Ajayi et al., 2022)

---

5. Homework 5

• Topic: Build a Neural Network with PyTorch
• Objectives:
  • Clean and preprocess data: convert columns, drop rows, and separate features/targets
  • Use train_test_split() and StandardScaler, convert to PyTorch tensors
  • Define a neural network with torch.nn.Module, using activation functions like F.relu() and torch.sigmoid()
  • Train with nn.MSELoss and the Adam optimizer; evaluate using accuracy, precision, recall
  • Experiment with model improvement strategies
• Climate Dataset: Residential electrical power consumption (4-year dataset)

---

6. Homework 6

• Topic: Computer Vision and CNN Practice
• Objectives:
  • Compare geospatial maps (e.g., flood inundation vs. flood susceptibility)
  • Understand SAR sensor data, DEMs, image patching, and visual encoding
  • Explain training processes, loss behavior, and hyperparameter roles
  • Propose model improvements, considering computational cost and complexity
• Climate Dataset: CCAI Tutorial on Flood Monitoring – Part I: People

---