GISCI 241 Chap.5 The Four Dimensions of Resolution
The Four Dimensions of Resolution
Resolution describes several different sampling dimensions. Spatial resolution concerns ground detail and pixel support, spectral resolution concerns the width and number of wavelength bands, radiometric resolution concerns sensitivity to intensity differences, and temporal resolution concerns revisit frequency. This chapter prevents the common mistake of ranking imagery with one word such as high resolution.
It uses pixel-area and bit-depth calculations to show nonlinear relationships and compares sensors against the dimension that the research question actually needs.
What this chapter covers
- 01
Spatial resolution, pixel size and mixed boundaries
- 02
Spectral resolution and narrow wavelength intervals
- 03
Radiometric sensitivity and digital bit depth
- 04
Temporal revisit and usable observation frequency
- 05
Tradeoffs among coverage, detail and data volume
- 06
Resampling and the false appearance of new detail
Worked example · free
Compare pixels over one wetland
- +1Convert 1.44 square kilometres to 1,440,000 square metres.
- +1A 30 metre pixel covers 900 square metres, giving 1,600 ideal cells.
- +1A 10 metre pixel covers 100 square metres, giving 14,400 ideal cells.
- +1The threefold change in side length produces a ninefold change in cell count over fixed area.
- +1Real boundaries, projection and classification prevent exact ideal counts and limit accuracy.
Key terms
- Ground sample distance
- The ground length represented by one image pixel under the stated acquisition geometry.
- Spatial support
- The area or volume over which a recorded value represents the phenomenon.
- Spectral bandwidth
- The wavelength interval included in a sensor band.
- Bit depth
- The number of binary digits used to encode intensity levels in a band.
- Revisit interval
- The time between opportunities for a sensor to observe the same location.
- Resampling
- Estimating raster values on a new grid without creating new original observations.
- Saturation
- A condition where signals exceed the sensor's recordable range and collapse to a maximum value.
The Four Dimensions of Resolution FAQ
What does high resolution mean?
The phrase is incomplete unless it names a dimension. Small pixels mean finer spatial sampling, narrow bands mean finer spectral sampling, more digital levels mean finer radiometric encoding, and short revisit means finer temporal sampling. A dataset can be strong in one dimension and weak in another.
Does resampling improve spatial resolution?
Resampling can put data on a smaller output grid and make edges appear smoother, but it cannot create ground detail that the original sensor never observed. It interpolates from existing pixel support. Preserve the original resolution in metadata and avoid describing the new cell size as new observational detail.
Does greater bit depth mean greater accuracy?
Greater bit depth allows more possible digital levels, so smaller intensity differences can be encoded. It does not prove those levels represent meaningful physical differences. Sensor noise, calibration, atmospheric correction and saturation affect usable information. Evaluate signal quality as well as the theoretical level count.
Does resampling to smaller cells improve spatial resolution?
It can produce a finer output grid and smoother display, but it cannot create ground detail absent from the original observation. Interpolation redistributes existing values and changes pixel support. Describe the source resolution, resampling method and intended operation, then validate whether the target was actually resolved rather than inferring quality from the new cell size.
Exam move
Create a four-row resolution table and fill it for every sensor example. For spatial resolution, calculate cell area and ask whether the target is separately resolved or mixed. For spectral resolution, sketch band positions and widths. For radiometric resolution, calculate two to the power bit depth and distinguish encoding from accuracy.
For temporal resolution, compare nominal revisit with usable cloud-free observations and event duration. Practise sensor-choice questions in which one dimension dominates, then add a second sensor when local detail and rapid detection conflict. Finish by identifying which resampling or compositing step changes support and how that affects comparison.
Work pixel-area calculations until the square relationship is automatic: halving the cell side produces four times as many ideal cells over a fixed area, while reducing it by a factor of three produces nine times as many. Add irregular boundaries and mixed pixels to explain why ideal counts differ from mapped counts.
For radiometry, compare 8-bit, 10-bit, 12-bit and 16-bit capacities, then write a separate sentence about noise and calibration so encoding is never presented as accuracy. Draw broad and narrow spectral bands over the same response feature and decide which can distinguish it. For temporal resolution, contrast scheduled revisit with usable revisit after cloud and quality screening.
Build two sensor-choice cases where different dimensions conflict, such as rapid landslide detection versus detailed mapping or narrow-band vegetation discrimination versus regional coverage. Rank the requirements before selecting data. Finish each case with a two-sensor strategy only when the second source answers a genuinely different scale or timing need.
Review resampling last: a finer output grid changes display and support representation, but it does not create observation that the original sensor never recorded.