20 · Satellite I: data and local processing

Tuesday, Nov 3, 2026

Materials for this session are not published yet. They appear here before class.

Objectives

By the end of this session you can:

• Apply the resolution checklist — spatial, temporal, spectral — to a research question.
• Compute a band index and say what it proxies.
• Choose between an off-the-shelf product and classifying your own.
• Run the MOSAIKS shortcut: satellite features as a table, then plain regression.
• Apply the three-problem rubric: small-unit validity, transferability, resolution mismatch.

What we cover

• What satellites have measured: GDP from nightlights, deforestation, urban growth, crop conditions.
• The resolution triad as a requirements checklist.
• Band algebra: NDVI in one build.
• Off-the-shelf products vs classify-your-own; NAIP’s free 1-meter imagery for the US.
• Big downloads without tears: scrape the file listing, loop download–extract–delete so the disk survives, cache the big intermediate with saveRDS(), and have the script beep when the hour-long job finishes.

Everything today runs on your own machine, on a scene you downloaded. That works until the question needs a decade of a continent, which is where the next session starts.

Verification habit. Ground-truth against sites you know.