Remote Sensing for Water Resources and Civil Engineering

Why this course

Satellites now image every field, canal, and reservoir on Earth every few days. Data is no longer the bottleneck. What's scarce is people who can turn pixels into defensible decisions, and this course trains them. Students learn the physics of how sensors see and the classical and deep-learning methods that extract information. New for 2027, they also learn how geospatial foundation models change the workflow: instead of training a model from scratch, they adapt one already pretrained on petabytes of imagery. Every method is tested on a real water, agriculture, or infrastructure problem.

What’s new for Spring 2027

• Open to graduate students from any program: engineering, geography, agriculture, sustainability, earth science, and data science.
• A new module on geospatial foundation models: Prithvi-EO-2.0, Clay, TerraMind, and AlphaEarth satellite embeddings.
• Cloud-first labs in Google Earth Engine and Colab, with nothing to install and no lab computer required.
• Hands-on work with the lab’s own hyperspectral (HySpex VS620) and drone imagery.

Who it’s for

Graduate students from any program who want to work with Earth observation data.

Prerequisites

• Some programming experience, ideally in Python. A week-0 “Python for geospatial” primer is provided.
• No prior GIS or remote sensing course is required.

What you’ll be able to do

1. Explain how active and passive sensors measure the Earth, and the trade-offs among spatial, spectral, temporal, and radiometric resolution.
2. Find, access, and preprocess multispectral, hyperspectral, SAR, and drone data using cloud-native tools.
3. Build and rigorously evaluate classification and regression models, from spectral indices to deep networks.
4. Adapt geospatial foundation models through embeddings, linear probing, or fine-tuning, and choose among these and training from scratch.
5. Fuse data across sensors and platforms, and explain what a model has learned.
6. Deliver an application to a water, agriculture, or infrastructure problem, and communicate its uncertainty.

Course plan

1. Weeks 1–3

   Foundations: how sensors see

   • The electromagnetic spectrum and surface reflectance
   • Active vs. passive sensing; satellite, aerial, and drone platforms
   • Spatial, spectral, temporal, and radiometric resolution
   • The cloud-native stack: STAC catalogs, Cloud-Optimized GeoTIFFs, Google Earth Engine, Python
   • Radiometric and atmospheric correction
2. Weeks 4–6

   Classical image analysis

   • Spectral indices and threshold-based classification
   • Supervised and unsupervised classification (random forest, k-means)
   • Accuracy assessment done right, including spatial cross-validation
   • Hyperspectral analysis: spectral unmixing and spectral angle mapping
3. Weeks 7–9

   Deep learning for Earth observation

   • Convolutional networks and U-Net semantic segmentation
   • Time series with recurrent networks and transformers
   • SAR–optical data fusion
   • Explainable AI for remote sensing
4. Weeks 10–12

   Geospatial foundation models New

   • Self-supervised pretraining: masked autoencoders and contrastive learning
   • Fine-tuning Prithvi-EO-2.0 with TerraTorch
   • Embeddings as products: Clay and AlphaEarth satellite embeddings (runs on a CPU)
   • Choosing among embeddings, linear probing, fine-tuning, and training from scratch
   • Benchmarks, domain shift, compute cost, and responsible use
5. Weeks 13–15

   Drones and water applications

   • Drone imaging, structure from motion, and point clouds
   • Evapotranspiration (OpenET) and irrigation water use
   • Water quality and land use change
   • Infrastructure monitoring, for example canal defects
   • Project presentations

The project

Students work in groups of up to three on a topic aligned with their research interests, chosen with the instructor, that must use remotely sensed data. The project runs all semester, with a literature review, progress updates, and a final technical article.

Readings

• Campbell, Wynne & Thomas, Introduction to Remote Sensing, 6th ed. (suggested)
• Amigo (ed.), Hyperspectral Imaging, Data Handling in Science and Technology vol. 32 (suggested)
• Current papers on geospatial foundation models, e.g., Prithvi-EO-2.0 (Szwarcman et al.)