Remote Sensing I: Physical Principles, Spectral Signatures and the Four Resolutions

This is a section of Part A, the Common Section, of the Geomatics Engineering (GE) paper. Remote sensing is the first discipline heading of Part A, and the syllabus opens it with “Basic concept, Physical Principles of Remote Sensing, Spectral signature, Resolutions: Spectral, Spatial, Temporal, Radiometric”. That is the physics of the signal before any sensor is chosen: the electromagnetic spectrum and the regions sensors use, the radiation laws that fix where the Sun and the Earth emit, what the atmosphere scatters and absorbs and where its windows lie, how energy splits into reflection, absorption and transmission at the ground, and the characteristic reflectance curves of vegetation, soil and water. It ends with the four resolutions, including the ground sampling distance and the number of grey levels a sensor records. Platforms, sensors and data products are the next chapter. The numericals are Wien's law, Rayleigh's λ⁻⁴, the ground sampling distance and 2ⁿ grey levels.

1. The basic concept and the electromagnetic spectrum

Remote sensing acquires information about an object without contact, by recording electromagnetic energy reflected or emitted from it. The chain is: energy source → propagation through the atmosphere → interaction with the target → return through the atmosphere → recording by a sensor → processing and interpretation → application. Energy travels at c = λν (c ≈ 3 × 10⁸ m/s), and each photon carries E = hν = hc/λ, so short wavelengths carry more energy per photon — one reason thermal and microwave sensors need larger detectors or longer dwell times to collect enough signal.

The regions remote sensing uses (boundaries are conventional)
RegionWavelengthWhat is sensed
Visible: blue, green, red0.4–0.5, 0.5–0.6, 0.6–0.7 µmreflected sunlight
Near infrared (NIR)0.7–1.3 µmreflected; vegetation vigour
Shortwave infrared (SWIR)1.3–3 µmreflected; moisture, minerals
Mid-wave infrared3–5 µmmixed; hot targets such as fires
Thermal infrared (TIR)8–14 µmemitted by the Earth
Microwaveabout 1 mm – 1 memitted (passive) or backscattered (radar)

2. Physical principles: radiation laws and the atmosphere

Every body above absolute zero emits. A blackbody absorbs and emits perfectly; Planck's law gives its spectral exitance, and two consequences are examined. Stefan–Boltzmann: total exitance M = σT⁴ with σ = 5.67 × 10⁻⁸ W m⁻² K⁻⁴ (so doubling T multiplies M by 16). Wien's displacement law: λ_max = 2898 µm·K/T. The Sun (about 6000 K) peaks near 0.48 µm, in the visible; the Earth (about 300 K) near 9.7 µm, which is why the 8–14 µm window is the thermal band. Real surfaces have emissivity ε < 1 (M = εσT⁴), and by Kirchhoff's law ε equals absorptance at each wavelength; a radiometer therefore reads a radiant temperature T_rad = ε^(1/4) T_kin, lower than the true kinetic temperature.

Atmospheric scattering
TypeParticle size vs λDependenceEffect
Rayleighmuch smaller (gas molecules)∝ λ⁻⁴blue sky; haze in the blue band
Miecomparable (dust, smoke, aerosols)weaker, roughly λ⁻¹ to λ⁰hazy, polluted days
Non-selectivemuch larger (cloud droplets)independent of λclouds look white

Blue light at 0.4 µm is scattered (0.7/0.4)⁴ = 9.4 times as strongly as red at 0.7 µm. Absorption by H₂O, CO₂ and O₃ closes much of the spectrum; the atmospheric windows used by sensors are roughly 0.3–1.3 µm, around 1.6 and 2.2 µm, 3–5 µm, 8–14 µm, and microwaves longer than about 1 cm, which pass through cloud — the reason radar is the all-weather sensor.

⚠️ Rayleigh is λ⁻⁴, not λ⁻²
Squaring the ratio instead of raising it to the fourth power gives 3.1 instead of 9.4. And do not confuse the scattered-light haze that lifts the dark end of the blue histogram with absorption: haze adds radiance, absorption removes it.

3. Interaction at the surface and spectral signatures

Energy incident on a surface is conserved: E_I(λ) = E_R(λ) + E_A(λ) + E_T(λ). Spectral reflectance ρ(λ) = E_R/E_I, plotted against wavelength, is the spectral signature (reflectance curve) of a material. A smooth surface reflects specularly, a rough one diffusely; a surface that scatters equally in all directions is Lambertian, the assumption most corrections make. “Rough” is relative to wavelength: a surface rough to visible light can be smooth to radar.

The three curves every question uses
MaterialVisibleNIRSWIR
Healthy green vegetationlow; chlorophyll absorbs blue (~0.45 µm) and red (~0.67 µm); small green peak (~0.55 µm)high (40–50 % or more) from leaf mesophyll structure; the sharp rise near 0.7 µm is the red edgefalls, with leaf-water absorption dips near 1.4, 1.9 and 2.7 µm
Dry bare soilmoderate, rising gently with λmoderatemoderate; lowered by moisture and organic matter
Clear waterlow, highest in blue-green; turbidity raises italmost zero — water absorbs NIRalmost zero
🧠 One band pair separates most of the scene
Red is low and NIR high for vegetation; both are moderate for soil; NIR is lower than red for water. That single contrast is why a standard false colour composite shows vegetation red, why NDVI works, and why water bodies are mapped from an NIR band alone.

4. The four resolutions and the ground sampling distance

What each resolution measures
ResolutionMeaningSet by
Spatialthe smallest ground area resolved: pixel size, GSD, IFOV projected on the groundaltitude, focal length, detector size
Spectralthe number, width and position of the bandsfilters, dispersing optics
Radiometricthe smallest difference in radiance recorded: n bits give 2ⁿ levels, 0 to 2ⁿ − 1quantisation and signal-to-noise ratio
Temporalthe revisit interval over the same placeorbit, swath, off-nadir pointing, number of satellites

The instantaneous field of view (IFOV) β is the angle one detector sees; at nadir from altitude H the ground cell is GSD = H·β (β in radians). For a detector of pitch p behind a lens of focal length f, β = p/f, so GSD = p·H/f. A 7 µm detector, f = 0.7 m, H = 500 km: GSD = 7 × 10⁻⁶ × 500 000/0.7 = 5 m. A drone camera with 4 µm pixels, f = 8 mm, flown at 120 m: GSD = 4 × 10⁻⁶ × 120/0.008 = 0.06 m. Data volume grows with every resolution: a 6000 × 6000 scene in 4 bands at 12 bits is 36 × 10⁶ × 4 × 12/8 = 216 × 10⁶ bytes.

⚠️ Higher resolution means a smaller number
A 5 m sensor has higher (finer) spatial resolution than a 30 m one. And the resolutions trade against each other: narrower bands collect less energy, so a hyperspectral sensor usually has coarser pixels, and a wide swath that improves revisit usually coarsens the pixel.

Key takeaways

  • c = λν and E = hc/λ; the reflective region is 0.4–3 µm, the thermal window 8–14 µm, and microwaves pass through cloud.
  • Wien: λ_max = 2898/T µm (Sun ≈ 0.48 µm, Earth ≈ 9.7 µm); Stefan–Boltzmann: M = εσT⁴; radiant temperature = ε^(1/4) × kinetic temperature.
  • Rayleigh scattering ∝ λ⁻⁴ (molecules), Mie for aerosols of size ≈ λ, non-selective for cloud droplets.
  • Vegetation: low red, high NIR, red edge near 0.7 µm; water: near zero in NIR; soil: moderate and rising.
  • Spatial, spectral, radiometric (2ⁿ levels) and temporal resolution; GSD = H·IFOV = pH/f.

Practice questions (14)

Attempt each one before opening the answer. Every explanation names the tempting wrong option as well as the right one, because that is where marks are lost.

  1. Using Wien's constant 2898 µm·K, the wavelength of peak emission of the Earth's surface at 300 K, in µm (to two decimal places), is ____.

    Numerical answer — type the value.

    Show answer

    Answer: 9.66

    λ_max = 2898/T = 2898/300 = 9.66 µm, inside the 8–14 µm thermal window. For the Sun at 6000 K the same law gives 0.483 µm, in the visible.
  2. Assuming pure Rayleigh scattering, the ratio of the scattering of blue light at 0.4 µm to that of red light at 0.7 µm (to one decimal place) is ____.

    Numerical answer — type the value.

    Show answer

    Answer: 9.4

    Rayleigh scattering ∝ λ⁻⁴, so the ratio is (0.7/0.4)⁴ = 1.75⁴ = 3.0625² = 9.38, i.e. 9.4. Squaring gives 3.1, the λ⁻² slip; the ratio 1.75 alone ignores the power entirely.
  3. A surface at a kinetic temperature of 300 K has an emissivity of 0.90 in the thermal band. The radiant temperature a radiometer records, in kelvin (to one decimal place), is ____.

    Numerical answer — type the value.

    Show answer

    Answer: 292.2

    Equating exitances, σT_rad⁴ = εσT_kin⁴, so T_rad = ε^(1/4) T_kin = 0.90^0.25 × 300 = 0.97400 × 300 = 292.2 K. Multiplying by ε itself (270 K) forgets the fourth root.
  4. Clouds appear white in visible imagery because cloud droplets cause

    1. non-selective scattering
    2. Rayleigh scattering
    3. selective absorption of blue light
    4. specular reflection of red light only
    Show answer

    Answer: A — non-selective scattering

    Droplets are much larger than visible wavelengths, so they scatter all visible wavelengths about equally — non-selective scattering — and equal amounts of all colours look white. Rayleigh scattering by molecules favours blue and gives the blue sky instead.
  5. The high reflectance of healthy green vegetation in the near-infrared is due mainly to

    1. the internal cell structure (spongy mesophyll) of the leaf
    2. absorption by chlorophyll
    3. the water content of the leaf
    4. Rayleigh scattering in the canopy air
    Show answer

    Answer: A — the internal cell structure (spongy mesophyll) of the leaf

    NIR is scattered at the many cell-wall and air-space interfaces of the mesophyll, so a healthy leaf reflects strongly; stress that collapses those cells lowers NIR first. Chlorophyll controls the visible (it absorbs blue and red), and leaf water controls the SWIR absorption dips.
  6. In a near-infrared band, which of the following would normally appear darkest?

    1. A clear, deep lake
    2. A healthy wheat field
    3. Dry sandy soil
    4. A deciduous forest
    Show answer

    Answer: A — A clear, deep lake

    Water absorbs almost all incident NIR within a short depth, so clear deep water is nearly black in an NIR band. Vegetation is bright in NIR and dry sand is moderately bright, which is why NIR is the band used to delineate water bodies.
  7. A pushbroom sensor has detectors of 7 µm pitch behind optics of focal length 0.7 m and orbits at 500 km. Its ground sampling distance at nadir, in metres, is ____.

    Numerical answer — type the value.

    Show answer

    Answer: 5

    IFOV = p/f = 7 × 10⁻⁶/0.7 = 10⁻⁵ rad. GSD = H × IFOV = 500 000 × 10⁻⁵ = 5 m. Equivalently GSD = pH/f = 7 × 10⁻⁶ × 500 000/0.7 = 5 m. Forgetting to convert µm to m gives an absurd 5 × 10⁶ m.
  8. A drone camera with a pixel pitch of 4 µm and a focal length of 8 mm is flown at 120 m above flat ground. The ground sampling distance, in centimetres, is ____.

    Numerical answer — type the value.

    Show answer

    Answer: 6

    GSD = pH/f = (4 × 10⁻⁶ m × 120 m)/(8 × 10⁻³ m) = 4.8 × 10⁻⁴/8 × 10⁻³ = 0.06 m = 6 cm. The photo scale is f/H = 1:15 000, and 4 µm × 15 000 = 6 cm is the same result.
  9. A sensor quantises its signal to 11 bits. The number of distinct grey levels it can record is ____.

    Numerical answer — type the value.

    Show answer

    Answer: 2048

    n bits give 2ⁿ levels: 2¹¹ = 2048, numbered 0 to 2047. The maximum DN, 2047, is often given by mistake for the number of levels.
  10. A scene of 6000 × 6000 pixels is recorded in 4 spectral bands at 12 bits per pixel per band, with no compression. Its size, in megabytes (1 MB = 10⁶ bytes), is ____.

    Numerical answer — type the value.

    Show answer

    Answer: 216

    Bits = 6000 × 6000 × 4 × 12 = 1.728 × 10⁹; bytes = 1.728 × 10⁹/8 = 2.16 × 10⁸ = 216 MB. Storing each 12-bit value in a 16-bit word, as many formats do, would make it 288 MB, but the question asks for the packed size.
  11. A satellite revisits the same ground point every 16 days. This describes its

    1. temporal resolution
    2. radiometric resolution
    3. spectral resolution
    4. spatial resolution
    Show answer

    Answer: A — temporal resolution

    Temporal resolution is the repeat interval, set by the orbit, the swath, pointing ability and the number of satellites. Radiometric resolution is about grey levels, spectral about bands, and spatial about the ground size of a pixel.
  12. Which of the following regions are atmospheric windows commonly used for Earth observation?

    1. 0.4–0.7 µm
    2. 8–14 µm
    3. Microwaves of a few centimetres
    4. Around 6.3 µm
    Show answer

    Answer: A — 0.4–0.7 µm; B — 8–14 µm; C — Microwaves of a few centimetres

    The visible, the 8–14 µm thermal window and centimetre microwaves are all transmitted well; microwaves even pass cloud. Around 6.3 µm water vapour absorbs strongly — that region is used by meteorological water-vapour channels precisely because the surface cannot be seen through it.
  13. Which of the following statements about spectral signatures are correct?

    1. Increasing soil moisture lowers soil reflectance across the reflective spectrum
    2. Vegetation shows absorption dips near 1.4 and 1.9 µm due to leaf water
    3. Turbid water reflects more visible light than clear water
    4. Stressed vegetation shows higher NIR reflectance than healthy vegetation
    Show answer

    Answer: A — Increasing soil moisture lowers soil reflectance across the reflective spectrum; B — Vegetation shows absorption dips near 1.4 and 1.9 µm due to leaf water; C — Turbid water reflects more visible light than clear water

    Water in the pores absorbs, so wet soil is darker at all reflective wavelengths; the 1.4 and 1.9 µm dips are liquid-water absorption in leaves; and suspended sediment back-scatters visible light, brightening turbid water. Stress damages the mesophyll, so NIR reflectance falls, not rises.
  14. Which of the following changes, taken alone, make the ground sampling distance of a nadir-looking camera smaller (finer)?

    1. Using a longer focal length
    2. Flying lower
    3. Using smaller detector elements
    4. Increasing the number of bits per pixel
    Show answer

    Answer: A — Using a longer focal length; B — Flying lower; C — Using smaller detector elements

    GSD = pH/f, so it falls as f increases, as H decreases and as p decreases. The number of bits sets radiometric resolution — how finely brightness is divided — and has no effect on the ground size of a pixel.