...

L. Pulvirenti, P. Castracane, G.d*Auria, N. Pierdicca Dip. Electronic

by user

on
Category:

forestry

81

views

Report

Comments

Transcript

L. Pulvirenti, P. Castracane, G.d*Auria, N. Pierdicca Dip. Electronic
Modelling the GNSS Reflectometry
Signal over Land: sensitivity to soil
moisture and biomass
N. Pierdicca1, L. Guerriero2 ,
E. Santi3, A. Egido4
1
2
3
4
DIET - Sapienza Univ. of Rome, Rome, Italy
DISP - University of Tor Vergata, Rome, Italy
CNR/IFAC, Sesto Fiorentino. Italy
Starlab, Barcelona, Spain
Introduction: GNSS-R
 GNSS Reflectometry performs bistatic
measurements, with most of the signal coming from
around the specular direction
 Specular reflection and
diffuse scattering from the
Earth surface are combined
 This is similar to the radar
altimeter, which however
works in monostatic
configuration
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
2
Introduction
ESA has funded two projects aiming at
 evaluating the potential of GNSS signals for remote sensing
of land bio-geophysical parameters (soil moisture and
vegetation biomass), through ground based (LEIMON, 2009)
and airborne (GRASS, 2011; in progress) experimental
campaigns
 developing a simulator to theoretically explain experimental
data and predict the capability of airborne and spaceborne
GNSS-R systems for moisture and vegetation monitoring
This presentation resumes some issues addressed
and some experience gained on land GNSS-R signal
modeling
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
3
Content





The projects
The problem
Simulator description
Validation results and sensitivity
Conclusions
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
4
LeiMon project
The crane
The GNSS
receiver
antenna
The monitored fields
(West and East side)
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
5
The flight track and
monitored fields
GRASS project
The airplane
The antenna
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
6
L
e
i
m
o
n
v
O
v
e
r
v
i
e
w
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
7
Coherent vs. incoherent power
Incoherent scattering
diffused in any
direction
Eincoh=E-<E>
Wrq (incoh) 
Wt 
p 2
4  
3
D tp ( i , i ) Dqr ( s ,  s )
rt2 rr2
TX
TX
 opqdA
pq
Coherent reflection
along specular direction
Ecoh=<E>
RX
°(i,s)
Wrq ( coh)  pq
2 Dtp DqrWt p
4 2 rt  rr 2
 Different dependence on ranges (rt and rr) and resolution (dA)
makes the relative magnitude of incoherent and coherent
components varies with receiver height, besides dependence
on surface roughness, vegetation.
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
Signal correlation time
 Incoherent component fluctuates (speckle) with correlation time
Tc dependent on system configuration
 In LeiMon ground based steady receiver
→
TC15 s
 In GRASS airborne receiver
→
TC7-8 ms
 Long coherent integration
(TI>>TC) can reduce incoherent,
zero mean, component when
possible
 Alternatively, long incoherent
integration is required to
mitigate fading, still preserving
the incoherent signal power
direct
≈TC
reflected
Confirmed by the slow fluctuation
of LeiMon signal (red line)
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
Incoherent component model
Absent or homogeneous vegetation
cover.
RTE solution considerind attenuation
and multiple scattering by a discrete
medium (Tor Vergata model)
Indefinite mean surface plane
with roughness at wavelength
scale.
Bistatic scattering of locally
incident plane waves by AIEM
(Fung, Chen)
 Contributions from single independent surface elements,
whose dimension can be assimilated to the roughness
correlation length (order of the wavelength) , add
incoherently.
 Then, the incident wave can be assumed locally plane
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
Coherent component model
Scattering of spherical
wave by Kirchoff approxi.
(Eom & Fung, 1988)
•
•
•
•
This a canonical problem in electromagnetism (antenna above a
dissipative plane, Sommerfeld equality, Exact Image Theory, and so on)
Signal is determined by a large portion of the mean surface, at least
the first Fresnel zone, from few meters (ground) to tens of meters
(airborne), to kms (spaceborne)
Kirchoff approximation can be useful. The incident wave must be
assumed to be spherical (as in Fung & Eom, 1988)
We have removed some constraints of Fung & Eom, i.e.,
consideration of identical transmitting and receiving antennas at
the same distance from the surface, and restriction to
backscattering and specular scattering cases.
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
The Bistatic Radar Equation
The mean power of received signal vs. delay t and frequency f is
modeled by integral Bistatic Radar Equation which includes time
delay domain response 2t’-t and Doppler domain response
S2 (f’-f ) of the system (Zavorotny and Voronovich, 2000).
Y (tˆ, fˆ )
|Y |2
PT
GT , GR
RR, RT
Ti

2
S2
dA
2
Ti 2 PT 2 GT GR 2 (t '-t ) S 2 ( f '- f ) 0

 dA
3 
2 2
(4 )
RR RT
Processed signal power at the receiver vs. delay t and frequency f.
The transmitted power of the GPS satellite.
The antenna gains of the transmitting and the receiving instrument.
The distance from target on the surface to receiving and transmitting
antennas.
The coherent integration time used in signal processing.
Bistatic scattering coefficient provided by the electromagnetic model
The GPS correlation (triangle) function
The attenuation sinc function due to Doppler misalignment. The longer
the time Ti the narrower the filter in Doppler space
Differential area within scattering surface area A (the glistening zone).
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
12
Model validation: LeiMon angular trend
Reflected (down antenna) over direct (up antenna)
April 8th SMC=30%
August 26th SMC=10%
•
•
•
•
East field Z=3cm
West field Z=1.75cm
East field Z=0.6 cm
West field Z=1cm
• Wetter and smoother fields exhibits higher down/up
• Simulator reproduces quite well LR signal versus  at
incidence angles  ≤45°.
• Higher angle may suffer from poor antenna characterization
(pattern, polarization mismatch)
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
13
LeiMon coherent & incoherent: soil
April 8th SMC=30%
August 26th SMC=10%
•
•
East Z=3cm
East Z=0.6 cm
total
coherent
 Comparison of LR theoretical simulations and data
shows that incoherent component strongly
contributes to total signal when soil is rough.
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
14
Model vs LeiMon data: bare soil LR
 Overall LR LeiMon model vs data
comparison over bare soil
10%<SMC<30%
0.6<z<3cm
 Slight model overestimation of
the DOWN/UP ratio but good
correlation
LR
-8
-12
-14
-16
LR
-8
15 West
-18
15 East
25East
25West
-20
-20
-18
-16
-14
-12
Leimon data DN/UP [dB]
-1035East -8
35West
 Without considering incoherent
component lower values (rough
surfaces) would be strongly
underestimated
Simulated DN/UP (only coherent) [dB]
SImulated DN/UP [dB]
-10
-10
-12
-14
15 West
-16
15 East
25East
25West
-18
35East
35West
-20
-16
-14
-12
-10
Serie1
Leimon DN/UP data [dB]
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
15
-8
Model vs GRASS data: overall LR
RMSE2dB
Bias 1dB
 Overall LR GRASS model vs data (DOWN/UP) comparison over bare
soil and forest
 Good performance over forest and slight model underestimation
over bare soil (but still good correlation)
 Note that soil underestimation can be easily reduce by changing
correlation length
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
16
Model vs data: RR component
LeiMon experiment
GRAA experiment
 The Simulator underestimates the signal at RR polarization
especially when it is expected to be low (e.g. rough soil)
 Besides the possible model errors (underestimation of crosspolarized incoherent scattering) e limit due to instrumental
noise seems to saturate observations below -18 dB (LeiMon)
and -22 dB (GRASS)
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
Sensitivity to SMC
 Reflected power at 35° incidence in Left Right (LR) polarization exhibits a
good sensitivity÷correlation (0.3 dB/%÷0.76) with SMC of the West field
(the one covered by sunflower), whilst the correlation with SMC is poor
on the East field (which was always bare, but with change in roughness).
 By using the ratio RR/LR, we observed a significant negative
sensitivity÷correlation (0.2 dB/%÷0.84) to the SMC on both fields
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
18
LeiMon LR data bivariate fitting
Best fitting of LR versus z and mv, for the East field data,
by equation:
LR  a0  a1  a2 mv  a3mv
• Data (blue diamonds) and
fitting surface (gray)
• Slopes of the surface
measures sensitivity
Sensitivity to z:
-4 to -2 dB/cm
Sensitivity to mv : 1 to 3 dB/10%
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
LeiMon: LR sensitivity to crop biomass
At 35°, the coherent
component is
attenuated by about 1
dB each 1kg/m2, which
would predict a large
sensitivity to PWC
(Ulaby et al., 1983;
Jackson et al., 1982;
O’Neill,1983)
t  e -20.1PWCsec
 The Simulator predicts a quite large incoherent component
according to the short coherent integration time (1 msec),
which explains the saturation effect with PWC in the data.
 The model reproduces the measured (low) sensitivity (0.3
dB/kg·m-2 , about 2 dB for the whole PWC range) thanks to
the consideration of both component
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
20
GRASS: LR sensitivity to forest biomass
• Coherent component is
dominant.
• Incoherent component
is reduced by the
longest coherent
integration time (20
msec) which filter a
narrower Doppler band
• Data and simulations
agree in showing a
fairly good sensitivity
to biomass (1dB every
100 m3/ha)
• The observed Highest Tree Volume corresponds to a Dry
Biomass of about 110 t/ha (beyond the SAR saturation limit )
• Sensitivity to even highest biomass to be verified
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
Conclusions
 A simulator has been developed which provides DDM’s,
waveforms and peak power (reflected/direct) of a GNSS-R
system looking at bare or vegetated soils (LHCP and RHCP real
antenna polarization)
 It singles out coherent and incoherent signal components.
 Validated using data over controlled experimental sites (bare
soil, sunflower, forest) (LeiMon and GRASS).
 Simulator results and experimental data show a fair agreement at
LR polarization and angles <45° (the antenna beamwidth)
 The incoherent component may be high in the ground based
LeiMon configuration, whereas was reduced by coherent
integration in airborne GRASS
 Sensitivity to SMC is significant and well reproduced by the
simulator
 Sensitivity to vegetation is reproduced and it is quite low when
the incoherent contribution cannot be reduced.
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
22
AIT-CeTeM – Telerilevamento a Microonde, Bari 4,5 dicembre 2012
23
Fly UP