Principles of Remote Sensing 10: RADAR 3 Applications of imaging RADAR Dr. Mathias (Mat) Disney UCL Geography Office: 113, Pearson Building Tel:
AGENDA Single channel data Radar penetration Multi-temporal data Vegetation, and modelling Agriculture & water cloud model Forest structure and coherent models Multi-parameter
Observations of forests... C-band (cm-tens of cm) – low penetration depth, leaves / needles / twigs L-band – leaves / branches P-band – can propagate through canopy to branches, trunk and ground C-band quickly saturates (even at relatively low biomass, it only sees canopy); P-band maintains sensitivity to higher biomass as it “sees” trunks, branches, etc Low biomass behaviour dictated by ground properties
Surfaces - scattering depends on moisture and roughness Note - we could get penetration into soils at longer wavelengths or with dry soils (sand) Surfaces are typically –bright if wet and rough –dark if dry and smooth What happens if a dry rough surface becomes wet ? Note similar arguments apply to snow or ice surfaces. Note also, always need to remember that when vegetation is present, it can act as the dominant scatterer OR as an attenuator (of the ground scattering)
Eastern Sahara desert SIR-A Penetration 1 – 4 m Landsat
Safsaf oasis, Egypt SIR-C L-band 16 April 1994Landsat Penetration up to 2 m
Single channel data Many applications are based on the operationally-available spaceborne SARs, all of which are single channel (ERS, Radarsat, JERS) As these are spaceborne datasets, we often encounter multi- temporal applications (which is fortunate as these are only single-channel instruments !) When thinking about applications, think carefully about “where” the information is:- –scattering physics –spatial information (texture, …) –temporal changes
Multi-temporal data Temporal changes in the physical properties of regions in the image offer another degree of freedom for distinguishing them but only if these changes can actually be seen by the radar for example - ERS-1 and ERS-2:- –wetlands, floods, snow cover, crops –implications for mission design ?
Wetlands in Vietnam - ERS Oct 97 Jan Mar May 99 Sept 99 Dec 99 Jan 00 Feb 00
Wetlands...
SIR-C (mission 1 left, mission 2 centre, difference in blue on right)
Floods... Maastricht A two date composite of ERS SAR images 30/1/95 (red/green) 21/9/95 (blue)
Snow cover... Glen Tilt - Blair Atholl ERS-2 composite red = 25/11/96 cyan=19/5/97 Scott Polar Research Institute
Agriculture Gt. Driffield Composite of 3 ERS SAR images from different dates
OSR - Oil seed rape WW - Winter wheat
ERS SAR East Anglia
Radar modelling Surface roughness Volume roughness Dielectric constant ~ moisture Models of the vegetation volume, e.g. water cloud model of Attema and Ulaby, RT2 model of Saich Multitemporal SHAC radar image Barton Bendish
Water cloud model A – vegetation canopy backscatter at full cover B – canopy attenuation coefficient C – dry soil backscatter D – sensitivity to soil moisture σ 0 = scattering coefficient m s = soil moisture θ = incidence angle L = leaf area index Vegetation
Values of A, B, C, D ParameterValueUnits / description A dB B1.945Fractional canopy moisture C dB D0.262Fractional soil moisture
Response to moisture Source: Graham 2001
Detection? SAR image In situ irrigation Source: Graham 2001
Simulated backscatter r 2 = 0.81
Canopy moisture r 2 = 0.96
Applications Irrigation fraud detection Irrigation scheduling Crop status mapping, e.g. disease, water stress
Multi-parameter radar More sophisticated instruments have multi-frequency, multi-polarisation radars, with steerable beams (different incidence angle) Also, different modes –combinations of resolutions and swath widths SIR-C / X-SAR ENVISAT ASAR, ALOS PALSAR,...
Flevoland April 1994 (SIR-C/X-SAR) (L/C/X composite) L-total power (red) C-total power (green) X-VV (blue)
Thetford, UK AIRSAR (1991) C-HH
Thetford, UK AIRSAR (1991) multi-freq composite
Thetford, UK SHAC (SAR and Hyperspectral Airborne Campaign) playpage&Itemid= 66&op=page&Sub Menu=66 Disney et al. (2006) – combine detailed structural models with optical AND RADAR models to simulate signal in both domains Drat optical model + CASM (Coherent Additive Scattering Model) of Saich et al. (2001) Coherent RADAR modelling
Thetford, UK SHAC (SAR and Hyperspectral Airborne Campaign) playpage&Itemid= 66&op=page&Sub Menu=66 Disney et al. (2006) – combine detailed structural models with optical AND RADAR models to simulate signal in both domains Drat optical model + CASM (Coherent Additive Scattering Model) of Saich et al. (2001) Coherent RADAR modelling
Optical signal with age for different tree density (HyMAP optical data)
Coherent (polarised) modelled RADAR signal (CASM)
OPTICAL RADAR
An ambitious list of Applications... Flood mapping, Snow mapping, Oil Slicks Sea ice type, Crop classification, Forest biomass / timber estimation, tree height Soil moisture mapping, soil roughness mapping / monitoring Pipeline integrity Wave strength for oil platforms Crop yield, crop stress Flood prediction Landslide prediction
CONCLUSIONS Radar is very reliable because of cloud penetration and day/night availability Major advances in interferometric SAR Should radar be used separately or as an adjunct to optical Earth observation data? ALOS
Speckle filtering –Mean –Median –Lee –Lee-Sigma –Local Region –Frost –Gamma Maximum a Posteriori (MAP) –Simulated annealing: modelling what the radar backscatter would have been like without the speckle
Original SAR data Frost filter Gamma MAP filter Simulated annealing Retford, UK ERS-2 SAR data April – September 1998
Original SAR data Frost filter
Gamma MAP filter Simulated annealing Recommendation : use these two
Discussion question What sort of radars are preferred for the following applications to be successfully realised and what is the physical basis? –Forest mapping –Flood extent –Soil moisture in vegetated areas –Snow mapping