Visible and Infrared Radiometers
CSIR NET Earth Sciences Part C satellite meteorology: imager channels, brightness temperature, scanning, calibration, resolution and reading VIS, IR, WV images.
- 19 min
On this page · 15 sections
- Prerequisites
- The Hook: The Cloud That Vanished at Night
- What a Radiometer Measures
- Inside the Radiometer
- Calibration: Counts to Radiance
- The Four Resolutions
- Instruments You Should Know
- Reading the Three Basic Images
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Planck, Wien, Stefan–Boltzmann and Kirchhoff, absorption bands and the atmospheric window, from Radiation, Scattering, Radiative Transfer and the Greenhouse Effect.
- The two orbits and their resolution limits, from Polar-Orbiting and Geostationary Satellites.
- Spectral signatures and image types used in geology, from Remote Sensing, GIS, Photogrammetry and GPS.
The Hook: The Cloud That Vanished at Night
What a Radiometer Measures
A radiometer measures radiance: power per unit area per unit solid angle, usually per unit wavelength (W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\)) or per unit wavenumber. It looks at the Earth through a narrow spectral band (a channel), and each channel is chosen for one physical reason:
- Window channels look through the clear atmosphere to the surface or cloud top.
- Absorption channels look into a gas band and see only the layer where that gas emits (water vapour, CO\(_2\), ozone).
In the reflected solar part of the spectrum (below about 3 μm) the signal is sunlight scattered back to space; the useful quantity is reflectance (or albedo), and the channel works only by day. In the thermal infrared (beyond about 4 μm) the signal is emitted by the surface, clouds and gases themselves; the useful quantity is brightness temperature, and the channel works day and night. The 3.5–4.0 μm band sits in between: by day it contains both reflected sunlight and emitted radiation, which makes it a powerful but tricky channel.
Standard channels of a meteorological imager
| Channel (approximate centre) | Region | Senses | Main uses |
|---|---|---|---|
| 0.6–0.65 μm | visible (VIS) | reflected sunlight | cloud and fog by day, cloud thickness, texture, snow, dust, smoke |
| 0.8–0.9 μm | near-infrared (NIR) | reflected sunlight; vegetation reflects strongly | land–water boundary, vegetation index, flood mapping |
| 1.6 μm | short-wave infrared (SWIR) | reflected sunlight; ice and snow absorb | separating ice cloud and snow (dark) from water cloud (bright); cloud phase |
| 3.7–3.9 μm | mid-infrared (MIR) | emitted plus reflected (day); emitted only (night) | night fog and low cloud, fire hot spots, droplet size, sea surface temperature at night |
| 6.2–7.3 μm | water vapour (WV) | emission by upper- and mid-tropospheric water vapour | upper-level moisture, jets, troughs, dry intrusions, water vapour winds |
| 8.7 μm | infrared window (weak absorption) | surface and cloud emission | cloud phase, dust, volcanic ash (with 10.8 and 12) |
| 9.6 μm | ozone band | ozone emission | total ozone, stratospheric intrusions |
| 10.3–11.3 μm | thermal infrared window (TIR1) | surface and cloud-top emission | cloud-top temperature, surface temperature, day and night imagery |
| 11.5–12.5 μm | split window (TIR2) | as TIR1, with slightly more water vapour absorption | sea surface temperature correction, dust, ash, thin cirrus |
| 13.3 μm and the 15 μm band | CO\(_2\) band | CO\(_2\) emission from different levels | cloud-top height (CO\(_2\) slicing), temperature sounding |
Brightness temperature
The brightness temperature \( T_B \) of a measured radiance \( L_\lambda \) is the temperature a black body would need to emit that radiance at that wavelength. Inverting Planck’s law,
$$B_\lambda(T) = \frac{c_1}{\lambda^5\left(e^{c_2/\lambda T} - 1\right)} \quad\Longrightarrow\quad \boxed{T_B = \frac{c_2}{\lambda\,\ln\!\left(1 + \dfrac{c_1}{\lambda^5 L_\lambda}\right)}}$$with \( c_1 = 2hc^2 = 1.191\times 10^{-16} \) W m\(^2\) sr\(^{-1}\) and \( c_2 = hc/k = 1.4388\times 10^{-2} \) m K. Three points follow:
- \( T_B \) equals the physical temperature only for a black body with no intervening absorption. Clear ocean at 10.8 μm comes close (emissivity about 0.99); thin cirrus does not.
- Because Planck’s function is much steeper at short wavelengths, a warm fraction of a pixel raises \( T_B(3.9) \) far more than \( T_B(10.8) \). This non-linearity is the basis of fire detection and of the night-time fog test.
- In the microwave the Rayleigh–Jeans limit applies, \( B \propto T \), so \( T_B = \varepsilon T \) for a surface of emissivity \( \varepsilon \): the reason the ocean looks cold to a microwave radiometer (next pages).
Worked example
A 10.8 μm channel measures 8.0 W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\). Then \( \lambda^5 = 1.47\times 10^{-25} \) m\(^5\), converting the radiance to SI units (8.0 W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\) = \( 8.0\times 10^{6} \) W m\(^{-3}\) sr\(^{-1}\)), \( c_1/(\lambda^5 L) = 1.191\times 10^{-16}/(1.47\times 10^{-25}\times 8.0\times 10^{6}) = 101.3 \), and
$$T_B = \frac{1.4388\times 10^{-2}}{10.8\times 10^{-6}\times\ln(102.3)} = \frac{1.4388\times 10^{-2}}{10.8\times 10^{-6}\times 4.628} \approx 288\ \text{K}$$a typical clear-sky land or sea surface. A radiance of about 2 W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\) corresponds to about 221 K, a deep convective cloud top.
Inside the Radiometer
graph LR
A["Scene radiance"] --> B["Scan mirror
(points the view)"]
B --> C["Telescope
(sets IFOV)"]
C --> D["Dichroic beam splitters
and band-pass filters"]
D --> E["Detectors
Si, InSb, HgCdTe"]
E --> F["Amplifier and
analogue-to-digital converter"]
F --> G["Counts to ground
calibrated to radiance"]
H["Space view and
onboard blackbody"] --> B| Element | Role | Note |
|---|---|---|
| Scan mirror | sweeps the view across the scene and periodically to calibration targets | the main moving part |
| Telescope | collects radiation and defines the instantaneous field of view (IFOV) | larger aperture = finer resolution and better signal |
| Beam splitters and filters | separate the beam into spectral channels | define the spectral response function of each channel |
| Detectors | turn photons into electrical signal | silicon for visible and near-IR; indium antimonide (InSb) for about 3.9 μm; mercury cadmium telluride (HgCdTe) for water vapour and thermal infrared |
| Cooler | infrared detectors are cooled (usually by a passive radiant cooler facing deep space) to reduce their own thermal noise | the cooler must never see the Sun |
| Electronics | amplify and digitise | number of bits sets the radiometric resolution |
Scanning methods
| Method | How the scene is built | Example |
|---|---|---|
| Cross-track (whiskbroom) | a rotating mirror sweeps across the ground track; the satellite’s motion supplies the next line | AVHRR on polar orbiters |
| Spin-scan | the whole spinning satellite sweeps one line per rotation; a mirror steps north–south between rotations | early geostationary satellites (first-generation Meteosat and GOES) |
| Two-axis mirror on a three-axis-stabilised satellite | the satellite body stays pointed at the Earth; a mirror scans east–west and steps north–south | modern geostationary imagers including INSAT-3D |
| Pushbroom | a long line of detectors images a whole cross-track line at once | high-resolution land imagers |
A spinning satellite looks at the Earth for only a small fraction of each rotation, so it collects less signal; three-axis stabilisation allows longer dwell time, flexible scanning of small sectors and rapid scans.
Calibration: Counts to Radiance
The detector output is a digital count \( C \). For the thermal infrared, calibration is two-point and onboard, assuming a linear response (a small non-linear correction is applied in practice):
$$\boxed{L = G\,(C - C_{\text{space}})}, \qquad G = \frac{L_{bb}}{C_{bb} - C_{\text{space}}}$$- Cold point: a view of deep space, whose radiance at thermal wavelengths is effectively zero.
- Warm point: an onboard blackbody of accurately measured temperature \( T_{bb} \), giving \( L_{bb} = B_\lambda(T_{bb}) \).
Then convert \( L \) to \( T_B \) with the inverse Planck function.
The visible channels cannot use an onboard blackbody. They rely on pre-launch calibration, on solar diffusers on some instruments, and on vicarious calibration against stable Earth targets (bright desert sites, deep convective cloud tops, the Moon) and against better-calibrated sensors. The visible channels of AVHRR, for example, have no onboard calibrator, so their gains drift and are tracked by vicarious methods: one reason long visible-channel climate records need careful reprocessing.
The Four Resolutions
| Resolution | Meaning | Set by | Typical values |
|---|---|---|---|
| Spatial | ground size of one pixel | IFOV × distance; worse off nadir | polar imagers about 0.4–1 km; geostationary about 0.5–1 km (VIS), 2–4 km (IR), coarser for WV on older imagers |
| Spectral | number and width of channels | filters, interferometers | imagers 3 to about 36 channels; hyperspectral sounders thousands |
| Radiometric | smallest detectable change in signal | bits, detector noise; expressed as noise-equivalent temperature difference (NEΔT) for IR | 10-bit data give 1,024 levels; NEΔT of order 0.1 K for good IR channels |
| Temporal | time between views of the same place | orbit and scan strategy | geostationary 10–30 min (minutes in rapid scan); one polar orbiter about 12 h |
These trade against one another. A smaller IFOV collects less energy, so to keep the same NEΔT the detector must dwell longer or the band must be wider; and dwelling longer slows the scan. That is why geostationary infrared pixels are coarser than visible ones, and why the water vapour channel, which sees a weak, cold signal, was often given the coarsest pixel.
Instruments You Should Know
| Instrument | Platform | Channels | Notes |
|---|---|---|---|
| AVHRR (Advanced Very High Resolution Radiometer) | NOAA polar orbiters, MetOp | 6 (0.63, 0.86, 1.6, 3.7, 10.8, 12.0 μm), of which 5 are transmitted at a time (1.6 μm and 3.7 μm share a slot) | about 1.1 km at nadir; the long-running workhorse for SST, cloud and vegetation |
| VHRR on Kalpana-1 | Indian geostationary | 3 (visible, water vapour, thermal infrared) | the basis of Indian cloud imagery before INSAT-3D; VHRRs on the earlier INSATs had fewer channels (visible and thermal infrared on the first ones) |
| INSAT-3D Imager | Indian geostationary | 6: VIS (about 0.55–0.75 μm), SWIR (about 1.6 μm), MIR (about 3.9 μm), WV (about 6.5–7.1 μm), TIR1 (about 10.3–11.3 μm), TIR2 (about 11.5–12.5 μm) | 1 km VIS and SWIR, 4 km MIR and TIR, 8 km WV; added the split window and the MIR channel for fog and SST |
| INSAT-3D Sounder | Indian geostationary | 19 (18 infrared plus 1 visible) | the first sounder on an Indian satellite; temperature and humidity profiles over the Indian region |
| MODIS | NASA Terra and Aqua (polar) | 36 | 250 m to 1 km; a research standard for cloud, aerosol and land products |
| VIIRS | Suomi NPP and JPSS | 22 | successor to AVHRR and MODIS-type imaging |
| ABI and AHI | GOES-R series and Himawari-8/9 | 16 | current-generation geostationary imagers; full disc every 10 min |
Reading the Three Basic Images
Visible image
- Brightness depends on reflectance: thick clouds (cumulonimbus, nimbostratus, thick stratus and fog) are brightest; thin cirrus is faint; the sea is dark (except in sun glint); snow and desert sand are bright.
- Texture and shadows tell height: lumpy, shadow-casting tops are convective; smooth, uniform tops with sharp edges are stratus or fog.
- Useless at night. Illumination changes with Sun angle, so morning and evening images look dim.
Infrared (10.8 μm) image
- Grey scale is conventionally inverted: cold is white, warm is dark, so that clouds look white as in the visible.
- Brightness depends on temperature of the emitting surface: high, cold cloud tops are brightest; low cloud and fog are almost as dark as the ground, because their tops are nearly as warm.
- Works day and night with the same appearance; the basis of time loops, cyclone analysis and rainfall estimation.
- Enhancement curves colour particular temperature ranges (for example, the coldest tops of deep convection) to make structure visible; the Dvorak technique uses a standard one.
Water vapour (6.7 μm) image
- The channel lies in a strong water vapour band, so it sees only the upper and middle troposphere (roughly 300–600 hPa, lower in a dry atmosphere and higher in a moist one). It almost never sees the surface.
- Bright (cold) means a moist upper troposphere or high cloud; dark (warm) means dry air aloft, usually sinking.
- Shows the upper-level flow even in cloud-free air: jet streams (sharp moisture boundaries), troughs and ridges, dry intrusions behind cyclones, and the moisture plumes feeding monsoon systems. Successive WV images are tracked to give water vapour winds.
Combining channels
The classic bispectral (VIS plus IR) classification:
| Warm in IR | Cold in IR | |
|---|---|---|
| Bright in VIS | low thick cloud: stratus, stratocumulus, fog; snow or bright desert | thick high-topped cloud: cumulonimbus, nimbostratus, deep frontal cloud |
| Dark in VIS | clear land or sea | thin cirrus (cold but transparent to sunlight) |
Two-channel differences do more:
| Combination | What it reveals |
|---|---|
| \( T_B(10.8) - T_B(12.0) \) large and positive | thin cirrus, or a moist lower atmosphere over the sea |
| \( T_B(10.8) - T_B(12.0) \) negative | dust or volcanic ash (“reverse absorption”) |
| \( T_B(10.8) - T_B(3.9) \) positive at night | fog and low water cloud (lower emissivity at 3.9 μm) |
| \( T_B(3.9) \gg T_B(10.8) \) at a hot spot | fire (Planck non-linearity) |
| VIS bright, 1.6 μm dark | snow, sea ice or ice cloud (ice absorbs at 1.6 μm) |
| NDVI \( = (\text{NIR} - \text{Red})/(\text{NIR} + \text{Red}) \) | vegetation vigour |
These tests are used on the synoptic identification page.
Memory Tricks
Don’t Confuse
| Pair | Distinction |
|---|---|
| Brightness temperature vs physical temperature | the black-body equivalent of a radiance vs the actual temperature; equal only for \( \varepsilon = 1 \) and no atmosphere in the path |
| Reflectance vs emittance | fraction of incident sunlight returned (VIS, NIR) vs thermal emission (IR); 3.9 μm by day has both |
| Window channel vs absorption channel | sees the surface or cloud top vs sees only one layer of the atmosphere |
| WV image vs IR image | upper-tropospheric humidity, no surface, works in clear air vs surface and cloud tops |
| Imager vs sounder | few broad channels, fine pixels, pictures vs many narrow channels in absorption bands, profiles |
| IFOV vs swath | angular size of one pixel vs width of the whole scanned strip |
| Spectral vs radiometric resolution | number and width of channels vs smallest detectable signal change (bits, NEΔT) |
| Spin-scan vs three-axis stabilised | satellite rotation does the east–west scan vs a mirror does it, the body stays pointed |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. Use \( c_1 = 1.191\times 10^{-16} \) W m\(^2\) sr\(^{-1}\), \( c_2 = 1.4388\times 10^{-2} \) m K.
Level 1: Foundation
- Which channel of an imager works only by day: 0.65 μm, 6.7 μm or 10.8 μm?
- What two targets are used to calibrate a thermal infrared channel in orbit?
- In an inverted-grey-scale infrared image, why does a cumulonimbus top look white?
Answers
- 0.65 μm: it measures reflected sunlight.
- A view of deep space (near-zero radiance) and an onboard blackbody at a known temperature.
- Its top is very cold (often colder than 220 K), and the inverted scale shows cold as white.
Level 2: Part B Style
Practice Q1. A pixel is bright in the visible image and has a brightness temperature close to that of the surrounding clear ground in the 10.8 μm image. The most likely scene is:
(a) thin cirrus (b) cumulonimbus (c) fog or low stratus (d) clear sea
Answer: (c) fog or low stratus
Bright in VIS means a thick reflecting layer; warm in IR means its top is low. Cirrus is dark in VIS and cold in IR; cumulonimbus is bright and cold; clear sea is dark and warm.
Practice Q2. A 10.8 μm channel records a radiance of 8.0 W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\). The brightness temperature is closest to:
(a) 255 K (b) 271 K (c) 288 K (d) 302 K
Answer: (c) 288 K
In SI units \( L = 8.0\times 10^{6} \) W m\(^{-3}\) sr\(^{-1}\). \( c_1/(\lambda^5L) = 1.191\times 10^{-16}/((10.8\times 10^{-6})^5\times 8.0\times 10^{6}) = 101.3 \). \( T_B = c_2/(\lambda\ln 102.3) = 1.4388\times 10^{-2}/(10.8\times 10^{-6}\times 4.628) \approx 288 \) K. Forgetting the μm-to-m conversion of the radiance is the usual slip.
Practice Q3. In a 6.7 μm water vapour image, a long dark band lies along the poleward edge of a cirrus shield over north India in winter. It most likely marks:
(a) a moist low-level jet (b) dry descending air on the poleward side of the subtropical jet (c) a dense fog bank (d) a warm sea surface
Answer: (b)
Dark in WV means warm \( T_B \), i.e. the channel sees down to warmer, lower levels because the upper troposphere is dry. A sharp dark–bright boundary is typical of the jet stream, with dry sinking air on its poleward side. The WV channel cannot see fog or the sea surface.
Practice Q4. An imager produces 10-bit infrared data spanning a linear temperature range of 180 K to 330 K. The temperature step per count is about:
(a) 0.015 K (b) 0.15 K (c) 0.6 K (d) 1.5 K
Answer: (b) 0.15 K
\( 2^{10} = 1024 \) levels, 1023 steps: \( 150/1023 = 0.147 \) K. (c) would be 8-bit data (150/255 ≈ 0.59 K).
Level 3: Part C Style
Practice Q5. A 10.8 μm channel is calibrated in orbit. The space view gives 40 counts; the onboard blackbody at 290 K gives 840 counts. Assume a linear response with zero radiance from space. A cloud pixel reads 300 counts. Which statements are correct?
(A) The calibration gain is about 0.0104 W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\) per count. (B) The pixel radiance is about 2.7 W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\). (C) The pixel brightness temperature is about 233 K. (D) Because the response is linear in radiance, the pixel brightness temperature is \( 290\times(300-40)/(840-40) \approx 94 \) K.
(a) A only (b) A and B (c) A, B and C (d) A, B, C and D
Answer: (c) A, B and C
\( L_{bb} = B_{10.8}(290\ \text{K}) = 8.28 \) W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\). Gain \( = 8.28/(840-40) = 0.01035 \). Pixel radiance \( = 0.01035\times 260 = 2.69 \). Inverting Planck with \( L = 2.69\times 10^{6} \) W m\(^{-3}\) sr\(^{-1}\): \( c_1/(\lambda^5 L) = 301.3 \), \( T_B = 1.4388\times 10^{-2}/(10.8\times 10^{-6}\times\ln 302.3) = 1.4388\times 10^{-2}/(10.8\times 10^{-6}\times 5.711) \approx 233 \) K.
(D) is the classic error: the response is linear in radiance, not in temperature. Scaling counts to temperature gives a physically absurd 94 K.
Practice Q6. A 1 km pixel of land at 300 K contains a fire covering 0.1% of its area at 800 K. Assume black bodies. The brightness temperatures are about 331 K at 3.9 μm and 301 K at 10.8 μm. Which conclusions follow?
(A) The 30 K difference arises because Planck radiance at 3.9 μm rises much more steeply with temperature than at 10.8 μm. (B) The 10.8 μm channel alone would detect the fire reliably. (C) A large \( T_B(3.9) - T_B(10.8) \) with a high \( T_B(3.9) \) is a fire test. (D) The same test applied by day must allow for reflected sunlight at 3.9 μm.
(a) A and C (b) A, C and D (c) B and D (d) B, C and D
Answer: (b) A, C and D
At 3.9 μm, \( B(800)/B(300) \approx 2{,}200 \), so even 0.1% of the pixel adds about 30 K to \( T_B \). At 10.8 μm the ratio is only about 20, so the pixel warms by about 1 K, within natural variation of the background: (B) is wrong. By day, sunlight reflected at 3.9 μm also raises \( T_B \) (bright clouds, sun glint), so daytime fire tests need extra checks.
Practice Q7. Four pixels are observed at noon:
| Pixel | VIS reflectance | \( T_B \) (10.8 μm), K | \( T_B(10.8) - T_B(12.0) \), K |
|---|---|---|---|
| P | 0.80 | 205 | 0.5 |
| Q | 0.15 | 235 | 4.0 |
| R | 0.65 | 288 | 0.3 |
| S | 0.35 | 300 | −1.5 |
The best match is:
(a) P cumulonimbus; Q thin cirrus; R fog or low stratus; S dust storm (b) P thin cirrus; Q cumulonimbus; R dust storm; S fog (c) P cumulonimbus; Q dust storm; R clear sea; S thin cirrus (d) P fog; Q clear land; R cumulonimbus; S thin cirrus
Answer: (a)
P: very bright and very cold, an opaque deep convective top (a small split-window difference, as expected for an opaque cloud). Q: dark but cold, with a large positive split-window difference: semi-transparent ice cloud. R: bright but at near-surface temperature: low water cloud or fog. S: moderately bright, warm, and a negative split-window difference: dust (reverse absorption).
Quick Revision Table
| Idea | Fact or formula |
|---|---|
| Inverse Planck | \( T_B = c_2/[\lambda\ln(1 + c_1/\lambda^5L)] \); \( c_1 = 1.191\times 10^{-16} \), \( c_2 = 1.4388\times 10^{-2} \) (SI) |
| Microwave | Rayleigh–Jeans: \( T_B = \varepsilon T \) |
| Channels | VIS 0.65, NIR 0.86, SWIR 1.6, MIR 3.9, WV 6.7, O\(_3\) 9.6, TIR 10.8 and 12.0, CO\(_2\) 13.3 and 15 μm |
| WV channel | sees about 300–600 hPa; dark = dry aloft; no surface |
| IR calibration | \( L = G(C - C_{\text{space}}) \), \( G = L_{bb}/(C_{bb} - C_{\text{space}}) \); linear in radiance |
| VIS calibration | pre-launch, diffuser, vicarious (deserts, deep convection, Moon) |
| Detectors | Si (VIS, NIR), InSb (3.9 μm), HgCdTe (WV, TIR); IR cooled |
| Scanning | whiskbroom (AVHRR), spin-scan (early GEO), two-axis mirror (INSAT-3D, modern GEO), pushbroom |
| Bispectral | bright-cold Cb; bright-warm fog/St; dark-cold Ci; dark-warm clear |
| Differences | 11−12 > 0 cirrus; < 0 dust/ash; 11−3.9 > 0 at night fog; 3.9 ≫ 11 fire |
| INSAT-3D Imager | 6 channels: 1 km VIS/SWIR, 4 km MIR/TIR, 8 km WV |
| INSAT-3D Sounder | 19 channels (18 IR + 1 VIS) |
| AVHRR | 6 channels, 5 transmitted; about 1.1 km |
Key Takeaways
What’s Next
- Identifying Synoptic Systems, Cyclones and Fog: using these images to recognise weather systems.
- Retrieval of SST, Cloud-Top Temperature, Winds, Rainfall and Soundings: turning radiances into numbers.
- Revise optical phenomena caused by scattering in Atmospheric Optics.
- Back to the chapter overview.