Retrieval of SST, Cloud-Top Temperature, Winds, Rainfall and Soundings
CSIR NET Earth Sciences Part C satellite meteorology: split-window SST, cloud-top height, motion vectors, scatterometer winds, rainfall algorithms and sounding.
- 23 min
On this page · 16 sections
- Prerequisites
- The Hook: Turning a Radiance into a Weather Number
- The General Retrieval Problem
- Sea Surface Temperature
- Cloud-Top Temperature, Height and Pressure
- Winds
- Rainfall
- Soundings: Temperature and Humidity Profiles
- Summary: Which Channel Retrieves What
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Brightness temperature, channels and calibration, from Visible and Infrared Radiometers.
- Schwarzschild’s equation, optical depth and weighting functions, from Radiation, Scattering, Radiative Transfer and the Greenhouse Effect.
- Lapse rates, the hydrostatic equation and precipitable water, from Composition, Thermal Structure and Hydrostatics and Thermodynamics of Dry and Moist Air.
- Sea surface temperature and ocean–atmosphere exchange, from Ocean Circulation, Waves and Tides.
The Hook: Turning a Radiance into a Weather Number
The General Retrieval Problem
graph LR
A["State of atmosphere
and surface x"] -->|"Forward model
(radiative transfer)"| B["Radiances y"]
B -->|"Inversion
(regression, physical,
variational)"| C["Retrieved x"]
D["Prior information:
first guess, climatology,
model forecast"] --> C
E["Ground truth:
buoys, radiosondes,
rain gauges, radar"] -.->|"validation and
tuning"| C| Approach | Idea | Example |
|---|---|---|
| Statistical (regression) | fit coefficients between satellite radiances and matched ground truth | split-window SST, many rainfall algorithms |
| Physical | iterate a radiative-transfer model until computed radiances match the observed ones | physical sounding retrievals |
| Variational (1D-Var) and direct assimilation | combine radiances and a model forecast, weighted by their errors | modern NWP assimilates sounder radiances directly instead of retrieved profiles |
Sea Surface Temperature
What the satellite sees
An infrared radiometer sees the skin of the ocean, the top few micrometres to tens of micrometres. The skin is usually a few tenths of a kelvin cooler than the water a few metres below (the bulk SST measured by buoys and ships), because the ocean loses heat at its surface; by day, calm seas can develop a warm layer that reverses the sign. Satellite SST is therefore tuned or corrected against buoy data.
Two problems must be solved:
- Clouds. IR cannot see through them. Every IR SST product starts with a cloud mask (brightness thresholds, spatial uniformity, channel-difference tests). Undetected cloud gives SST that is too cold.
- Water vapour. Even in the 10–12 μm window, water vapour absorbs and re-emits at colder temperatures, so the measured \( T_B \) is lower than the true SST, by up to several kelvin in the moist tropics.
The split-window method
The 10.8 and 12.0 μm channels have slightly different water vapour absorption: more at 12.0 μm. For a weakly absorbing atmosphere, each channel’s deficit is roughly proportional to its absorption coefficient \( k_i \) and to the water vapour amount \( W \):
$$T_s - T_i \approx k_i W, \qquad i = 11, 12$$Subtracting to eliminate the unknown \( W \):
$$T_{11} - T_{12} = (k_{12} - k_{11})W \ \Rightarrow\ \boxed{T_s = T_{11} + \gamma\,(T_{11} - T_{12}),\qquad \gamma = \frac{k_{11}}{k_{12} - k_{11}}}$$The difference between two channels measures the correction that each needs. Operational algorithms use the form
$$\text{SST} = a\,T_{11} + b\,(T_{11} - T_{12}) + c \quad\text{(MCSST)}$$with \( a, b, c \) fitted to matched buoy data, and refinements:
| Algorithm | Idea |
|---|---|
| MCSST (multichannel SST) | linear split-window regression, as above |
| NLSST (non-linear SST) | the coefficient of \( T_{11} - T_{12} \) scales with a first-guess SST, and a term in \( (\sec\theta - 1) \) corrects for the longer slant path at large zenith angles |
| Triple-window (night) | adds the 3.7–3.9 μm channel, a more transparent window (less water vapour absorption) that is very sensitive to surface temperature but contaminated by sunlight by day |
| Dual-view | views the same sea through two path lengths (nadir and forward), so the atmosphere is corrected from the difference in path, as on the ATSR series |
Typical accuracy against buoys is a few tenths of a kelvin in cloud-free conditions.
Microwave SST
Low-frequency microwave channels (around 6.9 and 10.7 GHz) sense SST through non-precipitating clouds, which makes them valuable in persistently cloudy regions such as the monsoon Bay of Bengal. The price: coarse resolution (tens of kilometres), no data near coasts (land contamination of the large footprint) or in heavy rain, and sensitivity to wind-roughened sea emissivity, which must be corrected.
| Infrared SST | Microwave SST | |
|---|---|---|
| Layer seen | skin (micrometres) | a slightly deeper sub-skin layer |
| Resolution | about 1 km (polar), 4 km (GEO) | tens of km |
| Clouds | blocked | sees through non-raining cloud |
| Near coasts | yes | no |
| Main errors | cloud contamination, water vapour, aerosol (dust makes SST too cold) | rain, wind roughness, radio interference |
Cloud-Top Temperature, Height and Pressure
Opaque clouds
For a thick cloud that behaves as a black body at 10.8 μm, the cloud-top temperature (CTT) is simply the window brightness temperature. Height comes from matching CTT to a temperature profile (radiosonde or model). With a single lapse rate \( \Gamma \):
$$\boxed{z_{\text{top}} \approx \frac{T_s - T_{\text{top}}}{\Gamma}}$$For a surface temperature of 30 °C and \( \Gamma = 6.5 \) K km\(^{-1}\), a top at −50 °C is at about 12.3 km. Cold tops mark deep convection; operational thresholds such as 235 K (−38 °C) and colder for very deep convection are used in rainfall estimation and in cyclone and MCS analysis.
Where the simple method fails
| Problem | Effect | Remedy |
|---|---|---|
| Semi-transparent cirrus | warm radiation from below leaks through, so \( T_B > \) CTT; height assigned too low | CO\(_2\) slicing; the H\(_2\)O-intercept method; split-window emissivity tests |
| Low cloud under an inversion | the CTT may match two heights in the profile | use the inversion base; model profiles |
| Sub-pixel cloud | partial cloud mixes with warm surface | finer pixels, cloud-fraction retrievals |
| Tropopause and overshooting tops | temperature no longer falls with height above the tropopause | special treatment; overshooting tops can be colder than the tropopause |
For a cloud of emissivity \( \varepsilon \) at temperature \( T_c \) over a surface at \( T_s \), the radiance is
$$L = \varepsilon\,B(T_c) + (1 - \varepsilon)\,B(T_s)$$CO\(_2\) slicing uses pairs of channels in the 13–15 μm CO\(_2\) band, which see to different depths. Taking the ratio of the cloud’s radiance signals in two neighbouring channels cancels the emissivity (it is nearly the same for both), and the ratio depends only on the cloud-top pressure, which is found by matching with radiative-transfer calculations. It works best for high, thin cloud; the H\(_2\)O-intercept method does a similar job using the WV and window channels.
Products from cloud-top retrievals
- Cloud-top pressure and height, used for aviation and for assigning heights to motion vectors.
- Cloud amount and type.
- Outgoing longwave radiation (OLR), estimated from window radiances; low OLR (under about 200 W m\(^{-2}\) on daily or longer averages) is a standard proxy for deep tropical convection, used for the monsoon and the MJO.
Winds
Atmospheric motion vectors (cloud and water vapour winds)
Geostationary imagery gives the wind by tracking features in successive images:
- Select targets: cloud elements or water vapour gradients with good contrast.
- Track: find each target in the next images, usually by maximising the cross-correlation within a search box, over two or three images 15–30 minutes apart (5–15 min on newer imagers).
- Compute speed and direction from the displacement and the time interval.
- Assign a height, from the target’s cloud-top temperature or pressure (or the WV radiance for clear-air targets).
- Quality control: consistency between successive image pairs and with the model background, summarised as a quality indicator.
| Channel | Level of the winds |
|---|---|
| VIS | low-level cumulus, mainly over the sea by day |
| 3.9 μm | low-level cloud at night |
| IR window | low cloud and high cloud (cirrus, anvil edges) |
| WV | upper troposphere, from cloud tops and clear-air moisture features |
The largest source of error is height assignment: a correct velocity at the wrong height corrupts the analysis. Semi-transparent cirrus (assigned too low) and low cloud tracked as if it were the mid-level flow are the classic cases. Tracked features may also move with the system (for example, a growing anvil or a wave cloud) rather than with the wind.
Polar orbiters provide motion vectors only at high latitudes, where consecutive orbits overlap.
Scatterometer winds over the sea
A scatterometer is an active microwave radar that measures the normalised radar backscatter \( \sigma^0 \) of the sea surface. Backscatter at moderate incidence angles is dominated by Bragg scattering from centimetre-scale capillary and short gravity waves, whose amplitude adjusts within minutes to the local wind stress. \( \sigma^0 \) increases with wind speed and depends on the angle between the wind and the radar look direction, being largest looking upwind or downwind and smallest crosswind.
- A geophysical model function relates \( \sigma^0 \) to wind speed, wind direction relative to the look, incidence angle and polarisation.
- Viewing the same cell from several azimuths gives several \( \sigma^0 \) values; solving gives up to four candidate wind vectors (ambiguities), and the correct one is chosen using the model background and spatial consistency.
- Output: the 10 m neutral-stability equivalent wind vector over the ocean.
| Instrument | Band | Notes |
|---|---|---|
| ASCAT (MetOp) | C-band | less affected by rain |
| SeaWinds (QuikSCAT) | Ku-band | wide swath; more rain contamination |
| OSCAT (Oceansat-2), SCATSAT-1 | Ku-band | Indian scatterometers |
Limits: no winds over land or sea ice; heavy rain contaminates Ku-band retrievals; very high winds in intense cyclones saturate the signal; resolution of about 25 km (finer in coastal products).
Other satellite wind measurements
| Method | Measures |
|---|---|
| Passive microwave radiometers | surface wind speed from sea-surface emissivity, which rises with roughness and foam; direction only from polarimetric radiometers |
| Altimeter | wind speed along the nadir track, from the strength of the return |
| Doppler wind lidar (ESA’s Aeolus, 2018–2023) | line-of-sight wind profiles from the Doppler shift of laser light scattered by molecules and aerosols |
Rainfall
Infrared techniques
Rain itself is not seen in the infrared; cold, high cloud tops are used as a proxy for deep convection that produces rain. The methods are statistical.
GOES Precipitation Index (GPI) (Arkin and Meisner): every pixel colder than 235 K in the 10.8 μm channel is assigned a rain rate of 3 mm h\(^{-1}\). Over a large box (2.5° × 2.5°) and a long period,
$$\boxed{\text{GPI (mm)} = 3\ \text{mm h}^{-1}\times F \times t}$$where \( F \) is the mean fractional area colder than 235 K and \( t \) the period in hours.
| IR technique | Idea |
|---|---|
| GPI | fixed rate under a fixed threshold; good for monthly, large-area totals in the tropics |
| Cold cloud duration | rain proportional to the time a pixel stays below a threshold; used for African rainfall monitoring |
| Griffith–Woodley | rain depends on the area and life-cycle stage of the cloud: growing clouds rain more than decaying ones |
| Hydro-Estimator | rate depends on the pixel’s temperature relative to its surroundings (convective cores) and is adjusted for moisture and orography from model data |
| INSAT-based products | IMD and ISRO produce rainfall estimates from INSAT-3D/3DR IR data, including a Hydro-Estimator type product and a multispectral algorithm calibrated with microwave rain |
Weaknesses of IR rainfall:
- Cirrus false alarms: cold, thick-looking anvils and cirrus that are not raining.
- Warm rain missed: shallow orographic or maritime clouds with warm tops, such as the Western Ghats in the monsoon, give heavy rain without tops colder than 235 K.
- It works on large areas and long periods; at single pixels and single hours it is poor.
Passive microwave techniques
Microwaves interact with the rain itself.
| Mechanism | Frequency | How it works | Works over |
|---|---|---|---|
| Emission | about 10–37 GHz | the ocean has low emissivity (about 0.5), so it looks radiometrically cold, \( T_B = \varepsilon T \); liquid rain drops emit and raise \( T_B \) | ocean only (land emissivity about 0.9 is already high, leaving no contrast) |
| Scattering | about 85–91 GHz (and 37 GHz) | large ice particles in the upper part of convective clouds scatter upwelling radiation away, lowering \( T_B \) | land and ocean |
Microwave estimates are more direct than IR but come from polar or low-inclination orbiters only a few times a day.
Spaceborne radar
| Mission | Radar | Notes |
|---|---|---|
| TRMM (1997–2015, USA–Japan) | Precipitation Radar, Ku-band (13.8 GHz) | first spaceborne rain radar; 35° inclination |
| GPM Core Observatory (from 2014) | Dual-frequency Precipitation Radar, Ku (13.6 GHz) and Ka (35.5 GHz) | light rain and snow, drop size information; 65° inclination; calibrates a constellation of microwave radiometers |
Radar reflectivity \( Z \) is converted to rain rate \( R \) by a power law, for example the Marshall–Palmer relation \( Z = 200R^{1.6} \) (\( Z \) in mm\(^6\) m\(^{-3}\), \( R \) in mm h\(^{-1}\)); 40 dBZ corresponds to about 11.5 mm h\(^{-1}\).
Merged products
Modern rainfall products combine the accuracy of microwave and radar with the frequent sampling of geostationary IR, and adjust to rain gauges: for example IMERG (GPM era), its predecessor TMPA (TRMM era), CMORPH and GSMaP, and merged satellite–gauge products prepared for India.
graph TD
A["Spaceborne radar
(accurate, narrow, rare)"] --> B["Calibrates passive
microwave estimates"]
B --> C["Microwave rain moved
between overpasses using
geostationary IR motion"]
D["Geostationary IR
(frequent, indirect)"] --> C
C --> E["Merged half-hourly
rainfall"]
F["Rain gauges"] --> ESoundings: Temperature and Humidity Profiles
The physics
Looking down through a non-scattering, clear atmosphere, the radiance in a channel of wavelength \( \lambda \) is (from Schwarzschild’s equation)
$$\boxed{L_\lambda = B_\lambda(T_s)\,\tau_\lambda(p_s) + \int_{p_s}^{0} B_\lambda[T(p)]\,\frac{\partial\tau_\lambda(p)}{\partial \ln p}\,d\ln p}$$where \( \tau_\lambda(p) \) is the transmittance from pressure level \( p \) to the top of the atmosphere. The factor \( W(p) = \partial\tau_\lambda/\partial\ln p \) is the weighting function: it tells which layer contributes most to the radiance. It peaks where the optical depth from the top is about 1.
- A channel at the centre of a strong absorption band becomes opaque high up: its weighting function peaks in the upper troposphere or stratosphere.
- A channel in the wing of the band sees deeper: its weighting function peaks lower.
- A set of channels across a band therefore samples the atmosphere at different heights.
Temperature sounding uses a gas whose concentration is known and uniform, so that the radiance depends only on temperature: CO\(_2\) (the 15 μm and 4.3 μm bands) in the infrared, O\(_2\) (50–60 GHz) in the microwave. Humidity sounding then uses water vapour bands (6.3 μm in the infrared, 183 GHz in the microwave) with the temperature known.
Retrieval
The weighting functions are broad and overlap, so a handful of channels cannot fix a detailed profile: the problem is ill-posed. Retrievals add a first guess (climatology, a regression estimate or a model forecast) and adjust it physically to fit the radiances. Modern NWP skips the retrieval step and assimilates radiances directly.
| Infrared sounders | Microwave sounders | |
|---|---|---|
| Bands | CO\(_2\) 15 and 4.3 μm; H\(_2\)O 6.3 μm; window | O\(_2\) 50–60 GHz; H\(_2\)O 183 GHz |
| Vertical resolution | better (especially hyperspectral) | coarser |
| Clouds | blocked; need clear pixels or cloud clearing | see through most non-precipitating cloud |
| Horizontal resolution | finer | coarser |
| Instrument | Type | Channels |
|---|---|---|
| HIRS | infrared filter sounder (TOVS and ATOVS) | 20 (19 infrared plus 1 visible) |
| AMSU-A | microwave temperature sounder | 15 |
| MHS (successor of AMSU-B) | microwave humidity sounder | 5 |
| AIRS (Aqua) | hyperspectral infrared grating spectrometer | 2,378 |
| IASI (MetOp) | hyperspectral infrared interferometer | 8,461 |
| ATMS (Suomi NPP, JPSS) | microwave sounder | 22 |
| INSAT-3D Sounder | geostationary infrared sounder | 19 (18 infrared plus 1 visible) |
Products include temperature and humidity profiles, total precipitable water (TPW), and stability indices such as the lifted index, useful for nowcasting thunderstorms over India in the pre-monsoon season. Geostationary sounders give hourly soundings over one region; polar sounders give global soundings twice a day.
Radio occultation
A GNSS receiver on a low-orbit satellite tracks a navigation satellite as it sets or rises behind the Earth’s limb. The signal is bent by the atmosphere’s refractive index gradient; the bending angle profile gives the refractivity profile,
$$N = 77.6\,\frac{p}{T} + 3.73\times 10^{5}\,\frac{e}{T^2}$$(\( p \) and \( e \) in hPa, \( T \) in K), from which temperature (in dry upper levels) and humidity (lower down, with a first guess) are derived. It has high vertical resolution, is insensitive to cloud, and needs no calibration against other instruments, which makes it valuable for climate monitoring; its horizontal resolution along the ray is coarse (a few hundred kilometres).
Summary: Which Channel Retrieves What
| Product | Main channels or technique | Main limitation |
|---|---|---|
| SST | IR split window (10.8, 12.0), 3.9 at night; microwave 6.9–10.7 GHz | clouds (IR); coarse, no coasts (microwave) |
| Cloud-top temperature and height | IR window; CO\(_2\) slicing; H\(_2\)O intercept | thin cirrus, inversions |
| Upper and lower winds | tracking in VIS, IR, WV; 3.9 at night | height assignment |
| Ocean surface wind | scatterometer (σ⁰ from several azimuths); microwave radiometer speed | land, ice, rain, saturation |
| Rainfall | IR cold-cloud methods; microwave emission (ocean) and scattering (ice); radar | IR indirect; microwave infrequent |
| Temperature profile | CO\(_2\) 15 and 4.3 μm; O\(_2\) 50–60 GHz; radio occultation | ill-posed, broad weighting functions |
| Humidity profile, TPW | H\(_2\)O 6.3 μm; 183 GHz; 22 GHz for TPW over sea | clouds (IR) |
Memory Tricks
Don’t Confuse
| Pair | Distinction |
|---|---|
| Skin vs bulk SST | micrometre surface layer seen by IR vs water at about 1 m or more measured by buoys |
| MCSST vs NLSST | linear split-window regression vs coefficient scaled by first-guess SST with a slant-path term |
| CTT vs cloud-top height | brightness temperature of an opaque top vs height found by matching CTT to a profile |
| AMV vs scatterometer wind | tracked cloud or vapour motion at an assigned height vs surface wind from sea-surface roughness |
| Emission vs scattering rain signal | liquid raises \( T_B \) over the radiometrically cold ocean vs ice lowers \( T_B \) at high frequencies, over land too |
| Retrieval vs radiance assimilation | profile derived first vs radiances fed directly to the model |
| IR vs microwave sounder | finer, better vertical detail, blocked by cloud vs coarser, sees through non-raining cloud |
| Radar reflectivity \( Z \) vs rain rate \( R \) | sixth moment of drop size vs volume flux of water; linked by an empirical power law |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. The SST coefficients used are illustrative, not those of any operational algorithm.
Level 1: Foundation
- Why does \( T_{11} \) usually exceed \( T_{12} \) over a clear tropical ocean?
- In the GPI, what rain rate is assigned to a pixel colder than 235 K?
- Which gases are used for satellite temperature sounding, and why?
Answers
- Water vapour absorbs more strongly at 12.0 μm, so that channel receives more radiation from colder, higher water vapour and less from the warm surface.
- 3 mm h\(^{-1}\).
- CO\(_2\) (15 and 4.3 μm) and O\(_2\) (50–60 GHz): their concentrations are known and nearly uniform, so the radiance depends on temperature alone.
Level 2: Part B Style
Practice Q1. In a linearised split window, \( T_s - T_i = k_iW \). If \( k_{12} = 1.5k_{11} \), \( T_{11} = 296.0 \) K and \( T_{12} = 294.5 \) K, the SST is:
(a) 294.5 K (b) 297.5 K (c) 299.0 K (d) 300.5 K
Answer: (c) 299.0 K
\( \gamma = k_{11}/(k_{12} - k_{11}) = 1/0.5 = 2 \). \( T_s = 296.0 + 2\times 1.5 = 299.0 \) K. Check: the atmospheric deficit is 3.0 K at 11 μm and 4.5 K at 12 μm, a ratio of 1.5. (b) uses \( \gamma = 1 \).
Practice Q2. A geostationary IR image shows a cloud element displaced 54 km east-northeast between two images 30 minutes apart. The wind at the assigned height is about:
(a) 18 m/s (b) 30 m/s (c) 54 m/s (d) 108 m/s
Answer: (b) 30 m/s
\( 54{,}000\ \text{m}/1800\ \text{s} = 30 \) m/s, blowing toward the east-northeast (a west-southwesterly wind).
Practice Q3. Over a 2.5° × 2.5° box for one day, the mean fraction of pixels colder than 235 K in three-hourly IR images is 0.40. The GPI rainfall is about:
(a) 1.2 mm (b) 9.6 mm (c) 28.8 mm (d) 72 mm
Answer: (c) 28.8 mm
\( 3\ \text{mm h}^{-1}\times 0.40\times 24\ \text{h} = 28.8 \) mm. (b) uses 8 hours (the number of images) instead of 24.
Practice Q4. A spaceborne radar measures 40 dBZ. Using \( Z = 200R^{1.6} \), the rain rate is about:
(a) 2.5 mm h\(^{-1}\) (b) 11.5 mm h\(^{-1}\) (c) 48.6 mm h\(^{-1}\) (d) 200 mm h\(^{-1}\)
Answer: (b) 11.5 mm h⁻¹
\( Z = 10^{40/10} = 10^4 \) mm\(^6\) m\(^{-3}\). \( R = (10^4/200)^{1/1.6} = 50^{0.625} \approx 11.5 \) mm h\(^{-1}\). (c) is the value for 50 dBZ.
Level 3: Part C Style
Practice Q5. A cirrus layer at 220 K with emissivity 0.5 at 10.8 μm lies over a surface at 300 K. Take the lapse rate as 6.5 K km\(^{-1}\) from the surface. Which statements are correct?
(A) The pixel brightness temperature is about 269 K. (B) Assigning height from \( T_B \) gives about 4.7 km, against a true height of about 12.3 km. (C) A motion vector from this cloud, assigned by its \( T_B \), would be placed far too low (near 4.7 km rather than about 12 km) although it moves with upper-tropospheric winds. (D) The brightness temperature is the emissivity-weighted mean of the two temperatures, 260 K.
(a) A and B (b) A, B and C (c) B and D (d) A, B, C and D
Answer: (b) A, B and C
\( B_{10.8}(220) = 1.91 \), \( B_{10.8}(300) = 9.67 \) W m\(^{-2}\) sr\(^{-1}\) μm\(^{-1}\). \( L = 0.5\times 1.91 + 0.5\times 9.67 = 5.79 \), which inverts to \( T_B \approx 269.2 \) K. Apparent height \( (300 - 269.2)/6.5 \approx 4.7 \) km; true \( (300 - 220)/6.5 \approx 12.3 \) km. The AMV would be put far too low, the classic height-assignment error. (D) averages temperatures instead of radiances; because Planck’s function is non-linear, the radiance-weighted value (269 K) is warmer than 260 K.
Practice Q6. A passive microwave radiometer views a calm tropical sea at 300 K. Take the ocean’s emissivity as 0.5 at 19 GHz and land’s as 0.9. Which statements are correct?
(A) The clear ocean appears at about 150 K at 19 GHz. (B) A layer of liquid rain over the ocean will raise the 19 GHz brightness temperature. (C) Over land at 300 K the background is about 270 K, so the same rain layer gives much less contrast. (D) At 89 GHz, ice in convective towers lowers the brightness temperature over both land and ocean.
(a) A and B (b) A, B and C (c) B, C and D (d) A, B, C and D
Answer: (d) all four
Rayleigh–Jeans: \( T_B = \varepsilon T \): \( 0.5\times 300 = 150 \) K over the sea, \( 0.9\times 300 = 270 \) K over land. Rain drops absorb and emit at nearly their own temperature, raising \( T_B \) above the cold ocean background but adding little over warm-looking land. Ice scattering at high frequency removes upwelling radiation and lowers \( T_B \) everywhere, which is why land rainfall algorithms use it.
Practice Q7. A sounder has three channels in the 15 μm CO\(_2\) band: one at the band centre, one partway into the wing and one far in the wing near the window. Which statements are correct?
(A) The band-centre channel’s weighting function peaks highest. (B) The far-wing channel is the most sensitive to surface temperature and low cloud. (C) Three channels can retrieve the temperature at three exact levels without any first guess. (D) The channels cannot retrieve temperature through thick cloud; a 50–60 GHz microwave sounder can, except in heavy precipitation.
(a) A and B (b) A, B and D (c) B, C and D (d) A, B, C and D
Answer: (b) A, B and D
Absorption is strongest at the band centre, so optical depth 1 is reached high up; wing channels see deeper, and the far wing nearly sees the surface. Weighting functions are broad and overlap, so the inversion is ill-posed and needs prior information: (C) is wrong. Infrared is blocked by thick cloud; oxygen-band microwave channels pass through non-precipitating cloud.
Quick Revision Table
| Idea | Fact or formula |
|---|---|
| Retrieval | forward model + inversion + prior; regression, physical, variational |
| Split window | \( T_s = T_{11} + \gamma(T_{11} - T_{12}) \), \( \gamma = k_{11}/(k_{12} - k_{11}) \); MCSST \( aT_{11} + b(T_{11} - T_{12}) + c \) |
| SST refinements | NLSST (first-guess, \( \sec\theta - 1 \)); triple window (night, 3.9 μm); dual view |
| Skin vs bulk | skin usually a few tenths K cooler |
| Microwave SST | 6.9–10.7 GHz; through cloud; coarse; no coasts |
| Cloud top | CTT = \( T_B \) for opaque; \( z = (T_s - T_{\text{top}})/\Gamma \); cirrus too low; CO\(_2\) slicing, H\(_2\)O intercept |
| Semi-transparent | \( L = \varepsilon B(T_c) + (1-\varepsilon)B(T_s) \) |
| AMVs | track VIS, IR, WV, 3.9; 15–30 min; height assignment the main error |
| Scatterometer | Bragg scattering; \( \sigma^0 \) vs speed and relative direction; several looks; ambiguities; ASCAT C-band, QuikSCAT and OSCAT Ku-band |
| GPI | 3 mm h\(^{-1}\) × fraction below 235 K × hours |
| Microwave rain | emission 10–37 GHz (ocean, raises \( T_B \)); scattering 85–91 GHz (ice, lowers \( T_B \), land and ocean) |
| Radar | TRMM PR Ku (1997); GPM DPR Ku + Ka (2014); \( Z = 200R^{1.6} \) |
| Merged rain | IMERG, TMPA, CMORPH, GSMaP |
| Sounding | \( W = \partial\tau/\partial\ln p \); peak near \( \tau \approx 1 \); CO\(_2\), O\(_2\) for T; H\(_2\)O for q |
| Sounders | HIRS 20; AMSU-A 15; MHS 5; AIRS 2,378; IASI 8,461; ATMS 22; INSAT-3D 19 |
| Radio occultation | \( N = 77.6p/T + 3.73\times 10^5 e/T^2 \); cloud-insensitive, fine vertical resolution |
Key Takeaways
What’s Next
- Satellite Meteorology: Quick Revision: the whole chapter on one page.
- Revisit radiative transfer and weighting functions in Radiation, Scattering, Radiative Transfer and the Greenhouse Effect.
- Use retrieved rainfall in water-balance problems from Humidity, Precipitation and Water Balance.
- Back to the chapter overview.