Interpretation of Remote Sensing Data
Dr. Gopi Kandaswamy
TCS Research and Innovation
Copyright © 2011 Tata Consultancy Services
Limited
1
Spectral Imaging
Measuring electromagnetic energy at varying wavelengths as it
interacts with materials
The nature of the material causes electromagnetic energy to be
reflected, refracted or absorbed in a way that is unique to each
material
These interactions when plotted form a unique shape that is
called the spectral signature of the material
Vegetation interacts with solar radiation in a different way than
other natural materials
2
The Electromagnetic Spectrum
3
* Images downloaded from multiple websites
The Infra-Red Spectrum
Division Wavelength Frequency Characteristics
Near infra-red 0.75 to 1.4 um 214 to 400 THz Atmospheric water
absorption
Short wave infra- 1.4 to 3 um 100 to 214 THz Atmospheric water
red absorption
increases
significantly at 1.45
um. 1.53 to 1.56
um is used in long
distance
communication
Medium wave infra- 3 to 8 um 37 to 100 THz Heat seeking
red missiles are
designed to work in
this spectrum
Long wave infra-red 8 to 15 um 20 to 37 THz Thermal imaging
region
4
The Vegetation Spectrum 1/2
5
* Images downloaded from multiple websites
The Vegetation Spectrum 2/2
Strong absorption in red and blue wavelength
Reflects in green wavelength
Strong reflectance in near Infra-red wavelength
Strong absorption in wavelengths where atmospheric water is
present
Different plant materials, water content, pigments, carbon and
nitrogen content cause further variation across the spectrum
6
Sensor Considerations 1/2
Solar reflected optical spectrum spans 0.4 um to 2.5 um
This is readily measured by most satellites and airborne
sensors
Partitioned into 4 distinct wavelength ranges
– Visible: 400 to 700 nm
– Near infra-red: 700 to 1300 nm
– Shortwave infra-red 1: 1300 to 1900 nm
– Shortwave infra-red 2: 1900 to 2500 nm
Transition from NIR to SWIR1 is marked by 1400 nm
atmospheric water absorption
Transition from SWIR1 to SWIR2 is marked by 1900 nm
atmospheric water absorption
7
Sensor Considerations 2/2
Near infra-red (0.7 to 1.0 um): Silicon
Short wave infra-red (1 to 3 um): InGaAs covers up to 1.8 um
Medium wave infra-red (3 to 5 um): InSb, HgCdTe, PbSe
Long wave infra-red (7 to 15 um): HgCdTe and
microbolometers
8
Vegetation Indices
Broadband greenness
– Measure of the vigor or health of green vegetation
– Sensitive to chlorophyll concentration, leaf area, foliage clumping,
canopy architecture
– Exploit the strong reflective property in NIR and strong absorption
in red wavelengths
– Examples are NDVI, SRI, EVI, ARVI and SGI
9
NDVI (Normalized Difference Vegetation Index)
NDVI = (NIR + VIS) / (NIR –
VIS)
Fraction of reflected light
needs to be calculated
Need to know the amount of
incident light
Values are between -1 and 1
Healthy vegetation falls
between 0.2 and 0.8.
Absolute and relative NDVI
Can use Red instead of VIS
10
* Images downloaded from multiple websites
NDVI as an Indicator of Drought
NDVI Anomaly; diff between
current NDVI and average
NDVI over long time period
Exceptionally lush
vegetation in Aug 1993 over
Northern Great Plains due
to flooding of Missouri river
Very low rainfall in Eastern
US resulted in a strong
negative anomaly
11
* Images downloaded from multiple websites
Enhanced Vegetation Index (EVI)
NDVI can get saturated
when there is too much
chlorophyll as in Evergreen
forests
NDVI is also influenced by
reflectance from soil and
particles in air
EVI corrects for the above
errors; EVI = 2.5[(NIR –
RED) / (NIR + 6RED –
7.5BLUE +1)]
12
* Images downloaded from multiple websites
Other Broadband VIs
ARVI (Atmospherically Resistant Vegetation Index)
– Corrects for aerosols in the atmosphere
– Uses reflectance measurement in blue to correct for scattering
effects
– ARVI = (NIR – (2RED – BLUE)) / (NIR + (2RED – BLUE))
SGI (Sum Green Index)
– Detects changes in vegetation greenness
– Useful for detecting forest disturbances as it is highly sensitive to
canopy opening
– Mean reflectance across 500 nm to 600 nm
– Value ranges from 0 to 50 %
– Common range for green vegetation in 10 to 25 % reflectance
13
Narrowband Greenness
Similar to broadband greenness; measure of overall amount
and quality of photosynthetic material in vegetation
Uses reflectance in red-edge region; between 690 and 740 nm
Using red-edge makes these VIs more sensitive to smaller
changes in vegetation health than broadband greenness
Intended to be used with high resolution imagery like those
from advanced multi-spectral or hyperspectral cameras
Red Edge NDVI, Modified Red Edge Simple Ratio, Modified
Red Edge NDVI, Vogelmann Red Edge Index 1,2,3, Red Edge
Position Index
14
Common Narrowband Greenness VIs
Red Edge NDVI: (750nm – 705nm) / (750nm + 705nm)
– Makes use of the sensitivity of Red Edge to canopy foliage
content, gap fraction and senescence
– Values range from -1 to 1
– Common range for green vegetation in 0.2 to 0.9
Modified Red Edge Simple Ratio Index:
– = (750nm – 445nm) / (705nm – 445nm)
– Uses Red Edge and incorporates correction for leaf specular
reflection
– Values range from 0 to 30
– Common range for green vegetation is 2 to 8
15
Common Narrowband Light Use Efficiency VIs
Light Use Efficiency: Quantifies vegetation’s ability to use
incident light for photosynthesis
Photochemical Reflectance Index:
– Exploits role of Xanthophylls in photoprotection
– (530nm – 570nm) / (530nm + 570nm)
– Values range from -1 to 1
– Healthy vegetation falls between -0.2 and 0.2
– Indicates light use efficiency and stress; more specifically drought
or nitrogen stress
– Highly affected by LAI; cannot be used when LAI is > 4
– Low chlorophyll content results in high PRI
– High chlorophyll content results in low PRI when LAI is high
16
Normalized Difference Nitrogen Index
High reflectance at 1510nm indicates high nitrogen
concentration and high biomass
Reflectance at 1680nm is due to biomass without Nitrogen
influence
This property is exploited in NDNI
– log (1/1510nm) – log (1/1680) / log (1/1510nm) + log (1/1680nm)
– Values range from 0 to 1
– Common range for green vegetation is 0.02 to 0.1
17
Senescence related VIs
Levels of lignin and cellulose increase just before senescence
Lignin and cellulose are flammable
Normalized Difference Lignin Index
– log (1/1754nm) – log (1/1680nm) / log (1/1754nm) + log
(1/1680nm)
– Values range from 0 to 1
– Common range for healthy green vegetation in 0.005 to 0.05
Cellulose Absorption Index
– Quantifies exposed surfaces that contain dried plant material
– Strong absorption in 2000nm and 2200nm indicate high
concentration of cellulose
– Used in crop residue monitoring
– = 0.5 (2000nm + 2200nm) – 2100nm
18
Plant Senescence Reflectance Index
– Exploits ratio of carotenoids to chlorophyll that indicates canopy
stress, canopy senescence and fruit ripening
– = (680nm – 500nm) / 750nm
– Values range from -1 to 1
– Common range for green vegetation is from -0.1 to 0.2
19
Canopy Water Content
Provides measure of water content in foliage
Higher water content often indicates healthier vegetation and
fire resistance
Water Band Index
– 900nm / 970nm
– Common range of green vegetation is 0.8 to 1.2
Normalized Difference Water Index
– = (857nm – 1241nm) / (857nm + 1241nm)
– Values range from -1 to 1
– Common range for green vegetation is -0.1 to 0.4
20
Using QGIS for NDVI
Add multi-spectral images using “Add Raster Layer”
21
Add Red and NIR images
22
Specify coordinate system
23
Both layers loaded and selected
24
Open Raster Calculator
Calculate NDVI using formula NDVI = (NIR – R) / (NIR + R)
25
Output is in grayscale
26
Change color scheme; right click on “ndvi” layer and select
properties
27
Select Singleband Pseudocolor and edit the color scheme
28
False color NDVI final image
29
RGB NDVI
30
NDVI EVI
31
