Every Earth-observation sensor is characterised by four resolutions. Spatial resolution is the ground size of a pixel; spectral resolution is the number and width of wavelength bands; temporal resolution is how often the sensor revisits a location; radiometric resolution is how finely it distinguishes brightness levels. No sensor maximises all four — engineering forces trade-offs — so choosing imagery means matching these resolutions to the question at hand.
Spatial: how small an object can you see
Rules of thumb: an object needs to span several pixels to be recognised. 10 m Sentinel-2 pixels map field-scale agriculture and land cover; 3 m PlanetScope separates smallholder plots; 30–50 cm commercial imagery resolves individual buildings, vehicles and road markings.
Finer pixels cost more per square kilometre and cover narrower swaths, which is why global monitoring runs at moderate resolution while site intelligence buys sub-metre.
Temporal: how often you look
Revisit governs what dynamics you can capture. A 16-day Landsat cycle suits seasonal land-cover work; Sentinel-2’s ≈5-day revisit supports within-season crop monitoring; daily smallsat constellations catch construction progress and short-lived events; geostationary weather satellites image continuously.
Clouds effectively degrade optical revisit — a 5-day nominal cycle may yield one clear view a month in the tropics, an argument for radar in wet climates.
Spectral and radiometric: how much you measure
More and narrower bands mean finer material discrimination: multispectral sensors (4–13 broad bands) handle most mapping; hyperspectral sensors (hundreds of narrow bands) identify specific minerals and species. Radiometric depth — 12-bit sensors distinguishing 4,096 brightness levels versus 8-bit’s 256 — decides how much subtlety survives in shadows and bright targets.
Applications should drive the specification: crop stress wants frequent multispectral looks; mineral exploration wants hyperspectral detail; emergency mapping wants whatever is soonest and cloud-free.
Frequently asked questions
What resolution do I need to see buildings?
Individual buildings become clearly mappable at around 1 m and comfortable at 30–50 cm. Building presence (built-up areas) can be detected far coarser — even at 10 m — but footprint extraction needs sub-metre imagery.
Why not always use the highest resolution?
Cost, coverage and volume. Sub-metre imagery over a province costs orders of magnitude more than 10 m data, produces terabytes to process, and its narrow swaths take longer to cover large areas. Fit-for-purpose beats maximum.
Related topics
Need this done professionally?
GLOBEIR delivers this work for government and enterprise clients.