Spatial Resolution vs Spectral Resolution

Short answer

Spatial resolution is how small a thing on the ground one pixel represents (a 10 m pixel averages a 10 × 10 m patch). Spectral resolution is how finely in wavelength the sensor samples light — how many bands it has and how narrow each one is. They answer different questions: spatial resolution decides whether you can see a feature; spectral resolution decides whether you can identify what it is made of. For geology the two pull in different directions, and the right sensor depends entirely on which question you are asking.

Defining the four resolutions precisely

Remote sensing has four resolutions; conflating any two is a common error.

• Spatial: ground sampling distance (GSD) per pixel — 10 m, 30 m, 0.5 m. Smaller = finer spatial detail.
• Spectral: number of bands and their bandwidth. A panchromatic sensor has one wide band; Sentinel-2 has 13; an imaging spectrometer (hyperspectral) has hundreds of bands ~5–10 nm wide. Narrower bands resolve diagnostic absorption features.
• Radiometric: bit depth — how many brightness levels per band (8-bit = 256 levels, 12-bit = 4096). Higher bit depth resolves subtle contrast.
• Temporal: revisit interval — how often the sensor re-images a location (Sentinel-2: ~5 days combined; Landsat: 16 days per satellite).

This article focuses on the first two because they embody the central design trade-off.

Why you cannot maximise both

Every detector element collects photons for a brief integration time over a fixed aperture. The total signal it gathers is shared three ways:

1. Pixel size — a smaller pixel images a smaller ground area, collecting fewer photons.
2. Band width — a narrower band admits a thinner slice of the spectrum, again fewer photons.
3. Swath / dwell — a wider swath or faster orbit gives each pixel less dwell time.

Fewer photons mean a lower signal-to-noise ratio (SNR). To keep SNR usable, designers trade these against each other. That is why hyperspectral instruments with hundreds of narrow bands tend to have coarser pixels or narrow swaths, while very-high-resolution commercial satellites (sub-metre) usually carry only a panchromatic band plus four broad multispectral bands. You buy spectral richness with spatial coarseness, and vice versa. Pan-sharpening partly cheats this by fusing a sharp panchromatic band with coarser colour bands, but it is a cosmetic enhancement, not new spectral information — never measure spectra off a pan-sharpened product.

The sensors geologists actually use

The pattern is clear: the instruments prized for mineral identification (ASTER, hyperspectral) accept 30 m or coarser pixels in exchange for the narrow SWIR bands where clays, carbonates and sulphates show diagnostic absorptions near 2.2 and 2.35 µm. Conversely, when you need to delineate a small outcrop or a road cut, you reach for high spatial resolution and accept broad bands.

Worked example: choosing for two geological tasks

Task A — map argillic alteration over a 50 km² prospect. Identification is the goal, so spectral resolution wins. Sentinel-2's band 11 (~1.61 µm) and band 12 (~2.19 µm) support a clay ratio (B11/B12), and ASTER's six SWIR bands or a PRISMA hyperspectral scene resolve the specific clay species. The 20–30 m pixels are perfectly adequate because alteration zones are large relative to a pixel. Workflow: obtain Level-2A surface-reflectance imagery (already atmospherically corrected), mask cloud and vegetation, compute the ratio, then validate against field spectra or known occurrences.

Task B — map a fault scarp and offset terrace risers. This is a delineation problem; spatial resolution wins. A 0.5 m commercial scene or lidar-derived hillshade shows the geomorphology that a 30 m pixel smears into nothing. Spectral content barely matters here.

The lesson: ask "identify the material" or "locate the feature" first, then pick the resolution axis that answers it.

Reprojection and alignment caveat

When you stack bands of different native GSD (Sentinel-2's 10/20/60 m, or ASTER VNIR vs SWIR), resample to a common grid before band math. Use gdalwarp -tr 20 20 -r cubic for continuous reflectance, and ensure all bands share one CRS (Sentinel-2 tiles are delivered in UTM zones; a project crossing a zone boundary needs reprojection with gdalwarp -t_srs). Misaligned bands produce edge artefacts in every ratio and index.

Common pitfalls and why they happen

• Confusing the two resolutions. "High resolution" said without qualification usually means spatial; a stakeholder may assume you can identify minerals because the picture looks sharp. Sharpness ≠ spectral discrimination.
• Measuring spectra off pan-sharpened imagery. Pan-sharpening invents plausible colour at fine spatial scale; the band values are no longer radiometrically faithful.
• Ignoring the SWIR. Most diagnostic mineral absorptions are in the SWIR (1.5–2.5 µm). A sensor with rich visible bands but no SWIR cannot separate most alteration minerals.
• Comparing scenes across dates without correction. Sun angle, season, atmosphere and soil moisture shift reflectance; use surface-reflectance (Level-2) products and matched conditions before differencing.
• Mixing native GSDs in band math. Always resample to a common grid first.

QA and validation checklist

• State both spatial GSD and band set for every product used.
• Use atmospherically corrected surface-reflectance (Level-2) inputs for any spectral analysis.
• Cloud/shadow/vegetation masks applied before ratios or classification.
• All bands resampled to one grid and CRS before band math.
• Spectral interpretations validated against field spectra, samples or known occurrences, and labelled as candidate evidence, not confirmed lithology.

Bathyl perspective

We pick the sensor from the question, not the other way round: spectral resolution to tell rocks apart, spatial resolution to draw their boundaries, and often both — a coarse spectral classification anchored by high-resolution structure. Every interpreted layer ships with its sensor, GSD, band set and processing level so the reader knows exactly what the image could and could not have seen.

Related reading

• Landsat vs Sentinel-2 for Geology
• Sentinel-2 Band Combinations for Terrain
• How GIS Is Used in Geology
• Remote sensing and Earth data

Sources

• USGS Landsat missions
• Copernicus Sentinel-2 mission
• NASA Earthdata: identifying minerals using ASTER
• GDAL gdalwarp
• QGIS coordinate reference systems