Elements of Remote Sensing
CSIR NET Earth Sciences: EM spectrum, atmospheric windows, spectral signatures, NDVI, orbits, resolutions, radar, InSAR, image classification and GIS.
- 19 min
On this page · 17 sections
- Prerequisites
- The Hook: Seeing Through the Monsoon
- What Remote Sensing Is
- Electromagnetic Radiation
- Spectral Signatures
- Platforms and Orbits
- Sensors and Resolution
- Microwave Remote Sensing
- From Images to Information
- GIS and GNSS
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Blackbody radiation (Planck, Wien and Stefan–Boltzmann laws) and wave optics, revised in Physical Sciences.
- Kepler’s third law for satellite orbits from Milky Way, Solar System and Earth’s Orbit.
- The applications this page serves: hydrology, ecology, mineral exploration and natural hazards.
The Hook: Seeing Through the Monsoon
What Remote Sensing Is
Remote sensing is acquiring information about an object without physical contact, by measuring electromagnetic radiation reflected or emitted from it. A complete system has: an energy source (Sun or the sensor itself), propagation through the atmosphere, interaction with the target, the sensor on a platform, transmission and processing, interpretation, and application.
graph LR
A[Energy source: Sun or active sensor] --> B[Atmosphere: scattering and absorption]
B --> C[Target: reflection, absorption, emission]
C --> D[Atmosphere again]
D --> E[Sensor on platform]
E --> F[Ground station: processing]
F --> G[Interpretation and analysis]
G --> H[Application: maps, monitoring, decisions]- Passive sensors record natural radiation: reflected sunlight (visible, near- and shortwave infrared) or emitted thermal infrared and microwave.
- Active sensors send their own energy and record the return: radar (microwave), LiDAR (laser), radar altimeters. Active sensors work day and night; microwave ones also through cloud.
Electromagnetic Radiation
\( c = \lambda\nu \) and \( E = h\nu \): longer wavelength, lower photon energy (which is why thermal and microwave sensors need larger detectors or longer dwell times to collect enough signal).
The spectrum used in remote sensing
| Region | Wavelength | Main use |
|---|---|---|
| Ultraviolet | 0.01–0.4 μm | limited (strong atmospheric scattering and ozone absorption) |
| Visible (blue, green, red) | 0.4–0.7 μm | colour images, water and land cover |
| Near infrared (NIR) | 0.7–1.3 μm | vegetation vigour, water boundaries |
| Shortwave infrared (SWIR) | 1.3–3 μm | moisture, minerals (clays, carbonates), snow versus cloud |
| Mid-wave infrared (MWIR) | 3–8 μm | hot targets: fires, lava |
| Thermal infrared (TIR) | 8–14 μm | surface temperature, emissivity |
| Microwave | 1 mm–1 m | radar, passive microwave (soil moisture, sea ice) |
Blackbody laws
$$\boxed{\lambda_{\max} = \frac{2898\ \mu\text{m K}}{T}} \quad \text{(Wien)} \qquad \boxed{M = \sigma T^4} \quad \text{(Stefan–Boltzmann)}$$- Sun (about 5,800 K): \( \lambda_{\max} \approx 0.50 \) μm, in the visible.
- Earth’s surface (about 300 K): \( \lambda_{\max} \approx 9.7 \) μm, in the thermal window.
- A forest fire (about 1,000 K): \( \lambda_{\max} \approx 2.9 \) μm, so fires are detected in the 3–5 μm band.
Real surfaces emit less than a blackbody by their emissivity \( \varepsilon \) (between 0 and 1): \( M = \varepsilon\sigma T^4 \). Water and dense vegetation have \( \varepsilon \) close to 1; bare metal is low. A thermal sensor measures brightness temperature, which equals the real (kinetic) temperature only if \( \varepsilon = 1 \).
The atmosphere
Scattering:
| Type | Particle size relative to \( \lambda \) | Wavelength dependence | Result |
|---|---|---|---|
| Rayleigh | much smaller (gas molecules) | \( \propto \lambda^{-4} \) | blue sky; haze in blue bands |
| Mie | comparable (dust, smoke, aerosols) | weaker, favours longer \( \lambda \) than Rayleigh | hazy, whitish sky in polluted air |
| Non-selective | much larger (cloud droplets) | none | white clouds |
Absorption by H\(_2\)O, CO\(_2\) and O\(_3\) blocks parts of the spectrum. The transparent ranges are atmospheric windows: the visible and near infrared (about 0.4–1.3 μm), parts of the SWIR (around 1.6 and 2.2 μm), 3–5 μm, 8–14 μm, and the microwave region beyond about 1 cm. Sensors are designed to look through these windows; the 6–7 μm water-vapour band is used deliberately by weather satellites to map humidity.
Energy balance at the target
$$E_I(\lambda) = E_R(\lambda) + E_A(\lambda) + E_T(\lambda)$$Incident energy is reflected, absorbed or transmitted, in proportions that depend on the material and wavelength. Spectral reflectance \( \rho_\lambda = E_R/E_I \) plotted against wavelength is the spectral signature or reflectance curve.
Smooth surfaces reflect specularly (mirror-like: calm water); rough surfaces reflect diffusely (Lambertian). Roughness is judged relative to wavelength: a surface smooth to radar may be rough to visible light.
Spectral Signatures
| Surface | Signature |
|---|---|
| Healthy vegetation | low in blue and red (chlorophyll absorption near 0.45 and 0.67 μm), a small green peak near 0.55 μm, a steep red edge near 0.7 μm, and high NIR reflectance (scattering by leaf mesophyll); dips at 1.4, 1.9 and 2.7 μm from leaf water |
| Stressed vegetation | red reflectance rises and NIR falls; the red edge shifts to shorter wavelengths |
| Clear water | low reflectance, falling to almost zero in the NIR (water absorbs NIR strongly); turbid or algae-rich water reflects more in the visible |
| Dry bare soil | reflectance rises gradually with wavelength; moisture, organic matter and roughness lower it; iron oxides cause absorption in the visible and near infrared |
| Snow | very high in the visible, low in SWIR near 1.6 μm (this separates snow from cloud, which stays bright) |
| Clay and carbonate minerals | diagnostic SWIR absorption features near 2.2 μm (Al–OH) and 2.3 μm (carbonate, Mg–OH) |
Vegetation indices
$$\boxed{\text{NDVI} = \frac{\rho_{\text{NIR}} - \rho_{\text{Red}}}{\rho_{\text{NIR}} + \rho_{\text{Red}}}}$$NDVI ranges from \( -1 \) to \( +1 \): dense green vegetation about 0.6–0.9, sparse vegetation or crops early in the season about 0.2–0.5, bare soil near 0.1, water and snow negative or near zero. With NIR reflectance 0.45 and red 0.05: \( \text{NDVI} = 0.40/0.50 = 0.80 \). NDVI time series feed crop forecasting, drought monitoring and land-degradation mapping.
Related indices: NDWI (water, using green or SWIR with NIR), NDSI (snow, green and SWIR) and the soil-adjusted vegetation index (SAVI) for sparse cover.
Platforms and Orbits
| Orbit | Altitude | Period | Use |
|---|---|---|---|
| Geostationary | about 35,786 km above the equator (orbital radius about 42,164 km) | one sidereal day; stays over one point | continuous weather watching (INSAT and Kalpana series, Meteosat, GOES); coarse spatial resolution |
| Sun-synchronous, near-polar | about 600–900 km | about 97–102 min | Earth observation (Landsat, IRS, Sentinel-2): passes over each latitude at the same local solar time (usually mid-morning) |
| Low inclined orbits | a few hundred km | about 90 min | specific missions |
Sun-synchronous orbits are slightly retrograde (inclination about 98°), so that the oblateness of the Earth makes the orbital plane precess eastward by about 1° a day, keeping pace with the Earth’s motion around the Sun. Constant illumination makes images from different dates comparable.
Orbital period from Kepler’s third law:
$$T = 2\pi\sqrt{\frac{a^3}{GM}}$$For an altitude of 800 km (\( a = 7{,}171 \) km, \( GM = 3.986\times 10^{14} \) m\(^3\) s\(^{-2}\)): \( T \approx 100.7 \) min, about 14.3 orbits a day.
Other platforms: aircraft (aerial photography, airborne hyperspectral and LiDAR), drones (UAVs), and ground-based spectroradiometers for calibration.
Sensors and Resolution
Sensor types
| Type | Note | Examples |
|---|---|---|
| Photographic camera | film or digital frame; photogrammetry | aerial survey cameras |
| Whiskbroom (across-track) scanner | rotating mirror sweeps one detector per band across the track | Landsat MSS and TM |
| Pushbroom (along-track) scanner | a linear array of detectors images a whole line at once; longer dwell time, better signal; no moving mirror | IRS LISS sensors, SPOT, Sentinel-2 |
| Multispectral | a few broad bands | LISS-III, Landsat OLI |
| Hyperspectral (imaging spectrometer) | tens to hundreds of narrow, contiguous bands; a near-continuous spectrum per pixel | mineral mapping |
| Thermal scanner | TIR bands | surface temperature |
| Radar (SAR) | active microwave | Sentinel-1, RISAT, NISAR |
| LiDAR | active laser ranging | DEMs, canopy height, ice-sheet elevation |
The four resolutions
| Resolution | Meaning | Example |
|---|---|---|
| Spatial | smallest ground area resolved; ground sampling distance ≈ IFOV × altitude | IFOV 0.1 mrad at 800 km → 80 m |
| Spectral | number and width of bands | panchromatic (1 broad band) vs hyperspectral (hundreds of narrow) |
| Radiometric | number of brightness levels, \( 2^n \) for \( n \) bits | 8-bit → 256 levels; 11-bit → 2,048; 12-bit → 4,096 |
| Temporal | revisit interval | Landsat 16 days; geostationary minutes |
There are trade-offs: finer spatial resolution usually means a narrower swath and longer revisit; narrower bands collect less energy, so pixels must be larger.
Indian sensors. The Indian Remote Sensing programme began with IRS-1A (1988). Resourcesat carries LISS-III (23.5 m), LISS-IV (5.8 m) and AWiFS (56 m); Cartosat provides high-resolution panchromatic stereo for mapping; Oceansat observes ocean colour and winds; RISAT and EOS-04 carry C-band SAR; and the NISAR mission (NASA–ISRO, launched 2025) carries L- and S-band SAR. Data are distributed by the National Remote Sensing Centre (NRSC, Hyderabad) through the Bhuvan portal. Landsat (since 1972) and Europe’s Sentinel satellites provide free global data.
Microwave Remote Sensing
| Band | Approximate wavelength | Note |
|---|---|---|
| X | about 2.4–3.8 cm | high resolution; little vegetation penetration |
| C | about 3.8–7.5 cm (Sentinel-1 about 5.6 cm) | general land and sea-ice mapping |
| S | about 7.5–15 cm | — |
| L | about 15–30 cm | penetrates vegetation canopy and dry sand; biomass, deformation |
| P | about 30–100 cm | deepest penetration; forest biomass |
Longer wavelength = deeper penetration into vegetation, dry soil and sand.
Radar backscatter depends on surface roughness (rough = bright), dielectric constant (wet soil and metal are bright), geometry (slopes facing the radar are bright), and structure (double-bounce between buildings and ground, or flooded trees and water, is very bright). Calm water and smooth tarmac reflect specularly away from the sensor and appear dark: the basis of flood and oil-spill mapping.
Geometric distortions of side-looking radar: foreshortening (slopes facing the sensor appear compressed), layover (the top of a steep mountain appears before its base) and radar shadow (slopes facing away receive no signal).
A synthetic aperture radar (SAR) uses the platform’s motion to synthesise a long antenna and achieve fine azimuth resolution independent of range.
Interferometric SAR (InSAR) compares the phase of two SAR images. In repeat-pass InSAR, each fringe (a \( 2\pi \) phase cycle) represents a change in range of half a wavelength along the line of sight:
$$\boxed{\Delta r = \frac{\lambda}{2}\ \text{per fringe}}$$For C-band (\( \lambda \approx 5.6 \) cm) each fringe is about 2.8 cm of line-of-sight motion. InSAR maps earthquake displacement, volcanic inflation, landslides and subsidence from groundwater pumping.
From Images to Information
Visual interpretation
The elements of image interpretation: tone/colour, size, shape, texture, pattern, shadow, site (topographic position) and association (things found together).
Standard false colour composite (FCC): NIR displayed as red, red as green, green as blue. Healthy vegetation appears red (high NIR), clear water dark blue to black, urban areas cyan or blue-grey, bare soil light tones.
Aerial photographs and photogrammetry
Vertical photographs are taken along flight lines with about 60% forward overlap (for stereo) and 20–30% side overlap.
$$\boxed{\text{Scale} = \frac{f}{H - h}} \qquad \boxed{d = \frac{r\,h}{H}}$$where \( f \) is the focal length, \( H \) the flying height above datum, \( h \) the terrain (or object) height, \( d \) the relief displacement of the top of an object of height \( h \) whose top is at radial distance \( r \) from the principal point. Relief displacement is radial outward from the nadir and zero at the nadir; it is what makes stereoscopic height measurement possible. A 152 mm camera at 3,040 m above the terrain gives a scale of 1:20,000.
Object heights from stereo pairs use the parallax difference: \( h = H\,\Delta p/(P + \Delta p) \), where \( P \) is the absolute stereoscopic parallax at the base of the object.
Digital image processing
- Pre-processing: radiometric correction (sensor calibration, atmospheric correction, conversion to reflectance) and geometric correction (registration to map coordinates with ground control points, then resampling by nearest neighbour, bilinear or cubic convolution).
- Enhancement: contrast stretching, spatial filtering (smoothing, edge detection), band ratios (cancel illumination effects), principal component analysis.
- Classification:
- Supervised: the analyst chooses training areas of known classes; the computer assigns pixels by minimum distance, parallelepiped or maximum likelihood (uses class means and covariances).
- Unsupervised: the computer clusters pixels by spectral similarity (k-means, ISODATA); the analyst labels clusters afterwards.
- Accuracy assessment with an error (confusion) matrix against reference (ground-truth) data.
- Change detection between dates.
Accuracy assessment
With rows = classified and columns = reference:
- Overall accuracy = sum of the diagonal / total.
- Producer’s accuracy of a class = correct / column total (reference); its complement is the omission error.
- User’s accuracy = correct / row total (classified); its complement is the commission error.
- Kappa coefficient corrects overall accuracy for chance agreement:
GIS and GNSS
A geographic information system (GIS) stores, analyses and displays spatial data. Raster data are grids of cells (images, DEMs); vector data are points, lines and polygons (wells, roads, districts). Typical analyses: overlay (for example combining slope, lithology, rainfall and land use into a landslide susceptibility map), buffering, network analysis and terrain analysis from a digital elevation model (DEM).
Global navigation satellite systems (GNSS) (GPS, GLONASS, Galileo, BeiDou, and India’s regional NavIC) give position by ranging to at least four satellites (three coordinates plus receiver clock error). Remote sensing, GIS and GNSS together provide the mapping backbone for every application in this chapter.
| Application | Remote sensing input |
|---|---|
| Groundwater prospecting | lineaments, geomorphology, land cover; GRACE gravity for storage change |
| Mineral exploration | alteration minerals in SWIR and hyperspectral data; structural mapping |
| Land use and land cover | multispectral classification and change detection |
| Agriculture | crop acreage and yield forecasting, drought (NDVI) |
| Hazards | SAR flood maps, InSAR deformation, cyclone tracking, landslide inventories, fire detection |
| Oceans | sea-surface temperature, ocean colour (chlorophyll), altimetry, winds |
| Cryosphere | snow cover, glacier retreat and glacial lakes (GLOF risk) |
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Passive: uses Sun or emitted radiation | Active: provides its own energy (radar, LiDAR) | |
| Geostationary: 35,786 km, fixed over equator | Sun-synchronous: about 800 km, near-polar, constant local time | |
| Whiskbroom: scanning mirror, across-track | Pushbroom: linear array, along-track, no mirror | |
| Spatial resolution: pixel size | Radiometric resolution: number of grey levels | |
| Rayleigh: \( \lambda^{-4} \), tiny particles | Non-selective: large droplets, white clouds | |
| Brightness temperature | Kinetic temperature (equal only if \( \varepsilon = 1 \)) | |
| Producer’s accuracy: omission | User’s accuracy: commission | |
| Supervised: training areas first | Unsupervised: clustering first, labelling after |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper.
Level 1: Foundation
- Which atmospheric window is used for measuring land-surface temperature?
- In a standard FCC, which colour indicates healthy vegetation?
- How many brightness levels can a 10-bit sensor record?
Answers
- 8–14 μm (thermal infrared), where Earth’s emission peaks.
- Red.
- \( 2^{10} = 1{,}024 \).
Level 2: Part B Style
Practice Q1. The peak wavelength of radiation emitted by a surface at 290 K is closest to:
(a) 0.5 μm (b) 2.9 μm (c) 10 μm (d) 29 μm
Answer: (c) 10 μm
\( \lambda_{\max} = 2898/290 = 9.99 \) μm. (b) is the peak for 1,000 K.
Practice Q2. A pixel has reflectance 0.30 in the NIR and 0.10 in the red band. Its NDVI is:
(a) 0.20 (b) 0.33 (c) 0.50 (d) 3.0
Answer: (c) 0.50
\( (0.30 - 0.10)/(0.30 + 0.10) = 0.20/0.40 = 0.50 \). (d) is the simple ratio NIR/red, a different index; (a) is only the numerator.
Practice Q3. For mapping the ground surface beneath a forest canopy using radar, the most suitable band is:
(a) Ka (b) X (c) C (d) L
Answer: (d) L
Penetration increases with wavelength; L-band (about 24 cm) passes through much of the canopy, while X- and Ka-band scatter from the leaves at the top.
Practice Q4. A vertical aerial photograph is taken with a 152 mm lens from 3,040 m above a flat plain. A tower 40 m tall stands on the plain, and its top appears 76 mm from the principal point. Which pair is correct (photo scale on the plain; relief displacement of the tower top)?
(a) 1:20,000; 1.0 mm (b) 1:2,000; 1.0 mm (c) 1:20,000; 0.1 mm (d) 1:200,000; 10 mm
Answer: (a)
Scale \( = f/H = 0.152/3040 = 1/20{,}000 \).
Relief displacement \( d = rh/H = 76\times 40/3040 = 1.0 \) mm (outward from the nadir). (b), (c) and (d) slip a factor of 10 in the scale or in \( d \).
Level 3: Part C Style
Practice Q5. An error matrix for a land-cover map (rows = classified, columns = reference) is:
| Classified \ Reference | Water | Forest | Urban | Row total |
|---|---|---|---|---|
| Water | 46 | 4 | 0 | 50 |
| Forest | 2 | 78 | 12 | 92 |
| Urban | 2 | 8 | 48 | 58 |
| Column total | 50 | 90 | 60 | 200 |
Which statement is correct?
(a) Overall accuracy 86%, kappa about 0.78, user’s accuracy for forest about 85% (b) Overall accuracy 86%, kappa about 0.86, producer’s accuracy for urban about 83% (c) Overall accuracy 78%, kappa about 0.86, user’s accuracy for forest about 87% (d) Overall accuracy 86%, kappa about 0.36, producer’s accuracy for forest about 85%
Answer: (a)
Overall: \( (46 + 78 + 48)/200 = 172/200 = 0.86 \).
Chance agreement: \( p_e = (50\times 50 + 92\times 90 + 58\times 60)/200^2 = 14{,}260/40{,}000 = 0.3565 \).
\( \kappa = (0.86 - 0.3565)/(1 - 0.3565) = 0.5035/0.6435 = 0.78 \).
User’s accuracy (forest) = 78/92 = 0.848; producer’s accuracy (forest) = 78/90 = 0.867; producer’s (urban) = 48/60 = 0.80; user’s (urban) = 48/58 = 0.83.
(b) confuses user’s and producer’s accuracy for urban and gives kappa = overall accuracy; (d) gives \( p_e \) as kappa and swaps forest’s two accuracies.
Practice Q6. A repeat-pass C-band interferogram (\( \lambda = 5.6 \) cm) over a subsiding city shows 3 complete fringes between the edge of the city and its centre. Which statements are correct?
- The line-of-sight displacement difference is about 8.4 cm.
- The same displacement would produce about 0.7 fringe in an L-band (\( \lambda = 23.6 \) cm) interferogram.
- Each fringe corresponds to a full wavelength of displacement.
- InSAR can measure this because the pulse travels to the ground and back, doubling the path change.
(a) 1, 2 and 4 only (b) 1 and 3 only (c) 2 and 3 only (d) 1, 2, 3 and 4
Answer: (a) 1, 2 and 4 only
- 1 true: \( 3\times\lambda/2 = 3\times 2.8 = 8.4 \) cm.
- 2 true: L-band fringe = 11.8 cm; \( 8.4/11.8 = 0.71 \).
- 3 false: a fringe is half a wavelength.
- 4 true: the two-way path is why the factor is \( \lambda/2 \).
Practice Q7. An Earth-observation satellite is in a circular orbit 800 km above the surface (\( R_E = 6{,}371 \) km, \( GM = 3.986\times 10^{14} \) m\(^3\) s\(^{-2}\)). Its detectors have an instantaneous field of view of 0.1 mrad. Which pair is correct (orbital period; ground sampling distance at nadir)?
(a) about 101 min; 80 m (b) about 101 min; 8 m (c) about 84 min; 80 m (d) about 24 h; 80 m
Answer: (a)
\( a = 7.171\times 10^{6} \) m; \( T = 2\pi\sqrt{a^3/GM} = 2\pi\sqrt{3.688\times 10^{20}/3.986\times 10^{14}} = 2\pi\times 961.9 \) s \( = 6{,}044 \) s \( \approx 100.7 \) min.
GSD \( = 0.1\times 10^{-3}\times 800{,}000 = 80 \) m.
(c) is the period at the Earth’s surface (no altitude); (d) is geostationary; (b) slips a factor of 10 on mrad.
Quick Revision Table
| Idea | Fact or formula |
|---|---|
| Spectrum | VIS 0.4–0.7, NIR 0.7–1.3, SWIR 1.3–3, MWIR 3–8, TIR 8–14 μm; microwave 1 mm–1 m |
| Windows | VIS–NIR, SWIR (1.6, 2.2 μm), 3–5, 8–14 μm, microwave |
| Wien | \( \lambda_{\max} = 2898/T \) μm; Sun 0.5, Earth about 10 μm |
| Scattering | Rayleigh \( \lambda^{-4} \) (blue sky), Mie (aerosols), non-selective (clouds) |
| Emissivity | \( M = \varepsilon\sigma T^4 \); brightness T ≤ kinetic T |
| Vegetation | low blue and red, red edge 0.7 μm, high NIR; water bands 1.4, 1.9 μm |
| NDVI | \( (\text{NIR} - \text{R})/(\text{NIR} + \text{R}) \), −1 to +1 |
| Orbits | geostationary 35,786 km; sun-synchronous about 800 km, about 98°, about 100 min |
| Resolutions | spatial (IFOV × H), spectral, radiometric (\( 2^n \)), temporal |
| Scanners | whiskbroom (mirror, across-track) vs pushbroom (array, along-track) |
| Indian sensors | IRS-1A 1988; LISS-III 23.5 m, LISS-IV 5.8 m, AWiFS 56 m; RISAT, NISAR (L, S); Bhuvan |
| Radar | X, C, S, L, P; long = deeper; smooth water dark; foreshortening, layover, shadow |
| InSAR | one fringe = \( \lambda/2 \) line-of-sight |
| FCC | NIR→red, red→green, green→blue; vegetation red |
| Photo | scale \( f/(H - h) \); relief displacement \( d = rh/H \) |
| Accuracy | OA = diagonal/N; producer’s (column), user’s (row); \( \kappa = (p_o - p_e)/(1 - p_e) \) |
| GNSS | at least 4 satellites; NavIC is India’s regional system |
Key Takeaways
What’s Next
- Environmental Earth Sciences: Quick Revision: the whole chapter on one page.
- Natural Hazards: revisit the hazards that remote sensing monitors.
- Graphs of Functions and Scientific Plots: reading spectral curves and data plots quickly for Part A.
- Back to the chapter overview.