Environmental Remote Sensing

Types of remotes sensing

EM Spectrum
Gravity (GRACE)
Magnetic field
Radioactivity (and more)

Users of RS rata

Weather
Land management
Marine shipping
Military
Earth and environmental science

Earth System Components

Atmosphere
Hydrosphere
Geosphere (lithosphere)
Biosphere
Cryosphere
Anthroposphere

Advantages of RS

Better Coverage in time and space
Better access to data
Better quality
Some measurements are only possible from space

Positive contributions of RS to Environmental Science

Environmental monitoring
Natural disaster monitoring
Agriculture and food security
Resources exploration and management
Geographic Information Systems

What is frequency measured in

Hertz (Hz, kHZ, MHz)

What is remote sensing?

The identification, observation and measurement of an object without coming into direct contact with it

Examples of remote sensing

MRI, CT scans, MODIS, Landsat

Wavelength is measured in

Meters (centimetres, nanometres)

Frequency definition

Number of cycles of a wave passing a fixed point per unit of time

𝑐=𝑓𝜆

speed of light= frequency*wavelength (Wave theory)

𝑓∝1/𝜆

Frequency (f) and wavelength (𝜆) for any given wave are related inversely (Wave theory)

wave theory of light

the idea that light actually consists of waves, not particles

quantum theory of light

the theory stating that light has both a wave nature and a particle nature.

EM radiation is composed of

1905 Einstein suggested that EM radiation is composed of discrete units or quanta (i.e. photons)

𝑬=𝒉𝒇

Planck-Einstein relation
Energy(J)=planks constant*frequency

Combination of wave and particle theory

𝑬∝𝒉/𝝀

𝑬∝𝒉/𝝀

Energy is proportional to planks constant/wavelength.

Given 𝑬∝𝒉/𝝀 what can we say about energy and wavelength

The energy of quantum is inversely proportional to its wavelength. i.e shorter wave length more energy.

blackbody radiation

The electromagnetic radiation emitted from a heated solid

Planck spectrum (blackbody spectrum)

All normal (baryonic) matter emits electromagnetic radiation when it has a temperature above absolute zero

peak wavelength

Single wavelength where the source's radiometric emission spectrum reaches its maximum

Signal to noise ratio (SNR)

Measurement of the level of desired signal to background noise

which wavelengths are harder to detect and why

Naturally emitted long wave radiation are more difficult to sense than shorter wavelengths. Because they have less energy per photon so you need to detect more to over come the SNR

As naturally emitted radio waves have less energy what happens to the resolution? (Planck relation)

You need a larger spatial resolution to overcome the SNR my gathering more signal and keeping the noise constant

What are the three types of atmospheric interaction?

Absorption
Refraction
Scattering
This measn atmospheric corrections maybe required

Absorption

Gas molecules absorb certain wavelengths and can re-radiate at different wavelengths. Amount increases with path length. The photon has a two-way journey there and back.

Refraction

Bends light at contact between two layers that transmit light at different speeds (think water in a glass)

Scattering

Redirection of EM waves by particles or molecules

Rayleigh Scattering

Scattering of EM radiation due to particles being small relative to radiation's wavelength. Selective scattering as intensity depends on frequency or wavelength

Intensity of scattering is proportional to EITHER 1/the wavelength^4 or frequency^4

Rayleigh scattering equation

Non-selective scattering

scattering of light caused by atmospheric particles larger than the wavelength being scattered(e.g. smoke, water droplets). Scattering is not wavelength dependent

What happens at 1400 and 1900 nm Wavelength

All radiation is absorbed by water, so the only use of satellites at this wavelength would be to measure water

Main absorbers of incoming shortwave radiation

Water Vapour
Carbon dioxide
oxygen

Atmospheric path length

Varies with location and time of day

Three types of interactions that can occur when energy strikes or is incident upon a surface

Absorption
Transmission
Reflection

Types of reflection

Specular- Mirror like
Diffuse- all different direction

Types of resolution

Spatial
Temporal
Radiometric
Spectral

Radiometric resolution

Detection of subtle changes in radiation
Expressed in number of bits. (e.g. 8 bit is 256 level- typical of computer files.

Spectral resolution

Ability to precisely measure wavelength of light (colour for visible spectrum).
Different landcover different reflective signatures.

Temporal Resolution

Pinpoint changes in time
determined by: altitude, orbital inclination, swath width

Key orbital parameters

Altitude- Determines orbital period, height above surface
Eccentricity- Deviation from circular orbit
Orbital inclination- Determines the latitudinal component

Physics of orbit

Balance between force and centripetal force. There is a force pulling satellites and earth together and one pulling them apart, they need to be balanced.

Common RS orbits

Low earth orbit
Geostationary orbit
sun-synchronous orbit

Low earth orbit

LEO- 400-1000 KM low FOV high spatial resolution.
Typically in polar orbit as it is best to map whole earth.
Global coverage in strips
revisit time usually 1-32 days

Geostationary Orbit

GEO- satellite's orbital period matches Earth's rotation period
Large FOV poor spatial resolution.
High temporal resolution
Used for weather and communications, multiple needed for global coverage, poles never visible

Sun-synchronous orbit

Desirable to observe Earth's surface at same local time for consistency Landsat passes equator at 10 am
Orbital inclination of ~98 degrees
Makes use of gravitational anomalies to cause orbit to precess slightly

Airborne RS platforms

Aircraft
Drones
UAVs

Aircraft pros and cons

Higher spatial resolution- 0.5m but smaller areas can schedule flight whenever but fuel is expensive so uneconomical for small areas

UAVs

Higher spatial resolution ~1cm
Limited coverage (batteries) 10km^2
Rapid deploy and survey 30 mins 1 flight
Fast and cost effective

Underwater remote sensing

Platforms- Ships submarines AUV
Acoustic sensors- ADCP bathymetry-single or multibeam
No GPS underwater

Sensor types

Passive active
Active

Passive sensors

Uses natural radiation from the sun or other emitted signal

Active sensor

Illuminates the subject from an artificial energy source
e.g. RADAR and LIDAR
using radio waves and laser pulses respectively

Imaging satellites

Crrates a picture by scanning across linear array of detectors while array moves through space 'Swath'. Typically produces raster data

Non-imaging satellites

Measures a set point locations
Requires interpolation and processing

Differences between Landsat 7 and 8/9

8/9 add a coastal aerosol band and cirrus cloud detection band
8/9 30m for thermal bands and 15 m for panchromatic
12 bit radiometric resolution compared to 8bit

When was Landsat launched

1972

Landsat 7 orbit and day cycle

Sun synchronous orbit
16 day cycle

MSS
ETM+
OLI
TIRS

Multispectral scanner
Enhanced thematic mapper +
Operational Land Imager
Thermal Infra-Red Sensor

OLI and TIRS bit and landsat 8/9 date launched

16 bit (opposed to 8 bit of L7)
L8 11 Feb 2013
L9 27 September 2021
Scene size 170km N-S
183km E-W
30m res
15 panchromatic

Pansharpening

a process of merging high-resolution panchromatic and lower resolution multispectral imagery to create a single high-resolution colour image

Level 1 or level 2 data

Level 2 has been corrected more than level 1.
Level 2 is Earth's surface and Level 1 is atmospheric processes (e.g. wildfires)

Scan Line Corrector

31st May 2003 SLC failed it compensates for the forward motion of the satellite only 78% of pixels remaining

Sentinel 2 MSI

Mulitspectral imager
12 bit data
5 day repeat cycle
2A launched 23rd June 2015
2B 7th March 2017
2C september 2024
Wide swath 290km

MODIS

Moderate Resolution Imaging Spectroradiometer
12 bit radiometric sensitivity 36 spectral bands
Two bands at 250m
Five bands at 500m
Remaining 29 bands at 1km
Orbit of 705KM achieves a swath of 2,330 km global coverage every 1-2 days

False colours

Humans can only see in the visible light spectrum and computers use RGB. So to see invisible radiation the bands are plugged into the RBG channels

specular reflection

a reflection produced by a smooth surface in which parallel light rays are reflected in parallel

diffuse reflection

Reflection that occurs when parallel rays of light hit a rough surface and all reflect at different angles

anisotropic reflection (Non-Lambertian)

Anisotropic- Different values measured depending on direction measured
Reflectance is usually anisotropic
Varies with angle of illumination
Varies with angle of view

Bidirectional Reflectance Distribution Function

distribution of reflected radiation according to illumination and observation angles

why is infrared reflected by plants

Photons at longer wavelengths do not carry enough energy for photosynthesis

wavelengths absorbed the most by plants

Red and blue are absorbed the most
400-700 nm range useful for photosynthesis
<400 nm = damage cells and tissues, but are also filtered out by stratospheric ozone

Band Ratios

Mathematical combination of spectral radiation in two or more observed spectral bands

NDVI

Normalized Difference Vegetation Index
Vegetation health and density
(NIR-red)/(NIR+red)

Health plants and NDVI

Healthy plant has a higher NDVI as it absorbs visible light and reflects most NIR

NDSI

Normalised Difference Snow Index
Snow cover
Needs Ground Truth
(SWIR-Red)/(SWIR+red)

NDWI

Normalized Difference Water Index
Maps water bodies
NDWI=(green-NIR)/(green+NIR)

Uses of machine learning in RS

Generates models
statistical research
Symbolic regression etc

What are spectral signatures

Different substances absorb, transmit and reflect radiation with different wavelengths. Hence they can be distinguished by spectral signatures.
Can be used to map land use can be do for all 32 bands of MODIS

What are Spectral classification of images vital for and their classification

Natural resource mapping at large scales
Monitoring environmental change
Classification based on single bands are rarely useful because of class overlap
multispectral classifications are better

unsupervised classification

Spectral info to group pixels into clusters
a pixel is more likely to have a similar value to a closer pixel than one further away

Advantages of unsupervised classification

No prior knowledge required
Low opportunity for human error
Unique classes are recognised as distinct units

Disadvantages of unsupervised classification

Spectrally homogenous classes may not correspond to informational categories of interest
Limited control over menu of classes
Spectral properties of specific informational classes will change over time making it problematic to compare classes between images.

Step of supervised classification

Identify landcover types
Create training sites
derive spectral signatures for each class from multispectral data
Select rules for group membership
Assign all pixels in the image to one of the groups
Maximum likelihood often used

Maximum likelihood

Location, shape and size of clusters is determined from statistical properties of training data
All image pixels assigned based on probability of belonging to a class

Land use and Land Cover LULC schema

Common global schemes
International Geosphere-Biosphere Programme (IGBP)
UN FAO Landcover Cover Classification System (LCCS)

Deep Learning in RS

A type of Machine Learning algorithm mimicking the biological structure if the brain

Strengths of active remote sensing

Make observations during the night
Many can see through clouds

LiDAR

Light detection and ranging
uses lasers
coherent light means it has a fine spatial resolution allows tightly focussed beam to be emitted
by using phase information precise ranging can be achieved

LiDAR types

Profiling Lasers
Scanning Lasers

Profiling lasers

downward (nadir) pointing
Closely spaced measurements along track but larger spacings between track
Measures range to surface precisely
If sensor elevation and orientation are known surface elevation can be calculated
Spatial resolution depends on altitude

Spatial resolution for ICESat 1 and 2

ICESat1 70 m
ICESat2 17m

Scanning lasers

More common with airborne systems
Uses a swath to collect more data
Integrated scanning mechanis
mintensive processing required

LiDAR point cloud

A matrix based structure
each point has an XYZ
and attributes like time, intensity, colour etc
typically has elevation data but can have reflectance values

extracting information from LiDAR

by using full waveform multiple vertical layers can be measured
Useful for vegetation and ‘seeing’ through shallow water

LiDAR return waveforms

First Return
Intermediate Return
Ground Return (Final Return)

Biomass and LiDAR

Peaks in intensity from different parts of land cover and vegetation
Delay is measured in nanoseconds

Detailed analysis of point cloud can be used to

Determine vegetation type
Estimate above-ground biomass

NASA GEDI

Global Ecosystem Dynamics Investigation Orbiting on ISS (2018-present)
+/-51 degrees latitude