Module 7: Data Sources & Acquisition

7.3 Remote Sensing Data Sources

Where to find free (and commercial) satellite imagery — Landsat, Sentinel, Planet, drones, and beyond.

Lesson 35 of 100·20 min read

Key takeaways

Landsat and Sentinel provide free, globally consistent Earth observation data updated every few days.

Commercial providers (Planet, Maxar) offer daily and sub-metre imagery at price.

Modern platforms (Earth Engine, Planetary Computer, AWS Open Data) provide petabyte-scale analysis-ready data.

Introduction

Satellite imagery used to be expensive and slow to obtain. Today, anyone with a laptop can tap into 50+ years of global Earth observation for free. This lesson covers the main sources you'll use, from venerable Landsat to the fastest commercial constellations.

Free public satellites

Landsat (USGS / NASA)

Missions: Landsat 1 (1972) — Landsat 9 (2021). Nine satellites over 50+ years.
Resolution: 30 m multispectral, 15 m panchromatic.
Revisit: ~16 days; combined L8 + L9 = ~8 days.
Bands: 11 (visible, NIR, SWIR, thermal).
Cost: Free from USGS EarthExplorer, AWS Open Data, Google Cloud, Microsoft Planetary Computer.
Strengths: Longest time series of any satellite; calibration quality; scientific pedigree.

Sentinel-2 (ESA Copernicus)

Resolution: 10 m (RGB, NIR), 20 m (red-edge, SWIR), 60 m (atmospheric).
Revisit: ~5 days with S2A + S2B.
Bands: 13.
Coverage: Global land + coastal zones.
Cost: Free.
Strengths: Higher resolution than Landsat; dense time series; great for vegetation and land-cover monitoring.

Sentinel-1 (SAR)

Synthetic Aperture Radar — sees through clouds, day and night.
Resolution: 5–40 m depending on mode.
Polarisations: VV, VH (or HH/HV).
Uses: Flood mapping, deforestation, ship detection, glacier motion.

Sentinel-3, Sentinel-5P

S3 — ocean colour, sea-surface temperature, land surface temperature.
S5P — atmospheric: NO₂, SO₂, CO, CH₄, O₃.

MODIS / VIIRS

250 m–1 km resolution but daily global coverage.
Useful for large-area, rapidly-changing phenomena (fire, chlorophyll, land surface temperature).

SRTM / ASTER GDEM / Copernicus DEM

Global digital elevation models.
SRTM (2000): 30 m; NASADEM (2020 reprocessing): 30 m.
Copernicus DEM (2021): 30 m / 90 m global, WorldDEM-based.

Commercial providers (high-res, daily)

Planet

PlanetScope: ~3 m resolution, daily global coverage since 2017.
SkySat: 50 cm resolution, tasked.
Subscription pricing.

Maxar (WorldView / GeoEye)

31 cm resolution at nadir — currently the highest commercial.
Task-based pricing; free low-res archive samples.

Airbus (SPOT, Pléiades, Pléiades Neo)

50 cm – 1.5 m resolution.
Strong European archive.

Newer entrants

Capella — SAR constellation.
ICEYE — SAR; daily revisit.
Umbra — sub-metre SAR.
BlackSky — optical, rapid tasking.

Analysis-ready data (ARD)

Raw satellite data requires atmospheric correction, cloud masking, and radiometric calibration. Analysis-Ready Data products have this processing applied, letting you skip straight to analysis.

Landsat Level-2 (L2SP) — surface reflectance, thermal temperature.
Sentinel-2 L2A — bottom-of-atmosphere reflectance with cloud masks.
HLS (Harmonised Landsat Sentinel) — co-registered, harmonised time series.

Modern platforms

Google Earth Engine

40+ years, 100+ petabytes analysis-ready.
JavaScript / Python API; server-side computation.
Free for research and non-commercial; commercial plans available.
Best for: "compute over a continent" workflows.

Microsoft Planetary Computer

STAC-indexed archives (Sentinel, Landsat, MODIS, Copernicus DEM, USGS NLCD).
Python Hub with pre-configured Dask clusters.
Free public access.

AWS Open Data

Landsat, Sentinel-1/2, SRTM available on S3.
Pay nothing to read; pay for EC2 / egress if used.

NASA Earthdata

Authoritative MODIS, VIIRS, ICESat-2, ASTER.
earthaccess Python library simplifies auth.

Drone / UAV data

For local high-resolution, drones are hard to beat:

Resolution: 1–10 cm.
Revisit: at will.
Limits: regulation, weather, battery, area coverage.
Typical outputs: orthophoto (GeoTIFF), DSM (raster), dense point cloud (LAS/LAZ).

Tools: OpenDroneMap, Pix4D, DroneDeploy, Agisoft Metashape.

Choosing imagery for a task

Goal	Suggested source
Monthly vegetation monitoring, regional	Sentinel-2 or Landsat
Daily fire monitoring globally	MODIS / VIIRS
Weekly deforestation, tropical	PlanetScope or Sentinel-2
Building damage assessment	Maxar sub-metre
Cloud-piercing flood mapping	Sentinel-1 SAR
Centimetre-level construction site	Drone
50-year land-use change	Landsat archive

Pre-processing pipeline

A typical remote sensing workflow:

Search — by bbox, time, cloud cover. Use STAC.
Download — L2 / ARD if available.
Mask clouds, shadows, nodata.
Mosaic / composite multiple acquisitions.
Compute indices (NDVI, NDWI — see Module 14).
Analyse / export.

Module 14 covers remote sensing in depth.

Self-check exercises

1. Landsat vs Sentinel-2 for a wildfire burn scar analysis — which would you pick?

Both work; Sentinel-2 usually wins. 10 m resolution captures fine burn patterns; 5-day revisit gives more cloud-free chances within a month; SWIR bands (B11, B12) are excellent for char detection. Landsat's thermal bands are useful for active-fire monitoring but post-fire burn scars are easier at 10 m.

2. You need cloud-free imagery of a tropical study area but every Sentinel-2 scene is cloudy. What's the solution?

Sentinel-1 SAR — it sees through clouds and works day or night. You get different information (backscatter, not reflectance), but for wet/dry, flooded/unflooded, or built/unbuilt discrimination SAR is extremely useful. Combine with sparse cloud-free optical scenes for richer analysis.

3. What's the difference between Level-1 and Level-2 Landsat products?

Level-1 is top-of-atmosphere reflectance / digital numbers — pre-atmospheric correction. Level-2 (L2SP surface reflectance) has atmospheric effects removed and is calibrated for surface properties. For most analysis, L2 is what you want. Level-1 is appropriate only when you do your own atmospheric correction.

Summary

Free: Landsat, Sentinel-1/-2/-3/-5P, MODIS, SRTM.
Commercial: Planet (daily 3 m), Maxar (sub-metre), SAR providers.
Platforms like GEE and Planetary Computer collapse discovery, access, and compute.
Always use Analysis-Ready Data when the mission offers it.