View
219
Download
2
Tags:
Embed Size (px)
Citation preview
Aquarius/SAC-D Soil Moisture Retrieval and Applications
T. J. Jackson1 and H. Karszenbaum2
1 USDA ARS Hydrology and Remote Sensing Lab, Beltsville, MD USA2 Instituto de Astronomia y Fisica del espacio, Buenos Aires, Argentina
With contributions fromM. Cosh, R. Bindlish, P. O’Neill
6th Aquarius Science Team MeetingJuly 19-21, 2010
Outline
• Passive, active, and combined microwave soil moisture algorithms
• Basis of passive microwave soil moisture algorithms
• Strategy for developing an Aquarius baseline soil moisture algorithm
• Validation• Recent field campaigns (CanEx)• Updates of related Aquarius/SAC-D projects
Passive and Active Soil Moisture Algorithms• Passive microwave
– Several passive algorithms have been developed, applied and validated using AMSR-E and WindSat
• Translating these to L-band is possible and should result in improved estimates
– Recent implementation of SMOS algorithms• Still early but looks good (See Yann’s talk!)
– All algorithms are based on the same radiative transfer equation ( model). They vary in their approaches to parameter estimation and related assumptions.
• Single channel, Iterative optimization, Polarization indices…
• The optimal algorithm for a satellite mission will vary with the sensor system configuration (i.e. Aquarius is quite different from SMOS or AMSR-E)
Passive and Active Soil Moisture Algorithms• Active microwave
– No robust retrieval technique has been developed and validated for use with high resolution SAR data
• Two decades of C-band (ERS, Radarsat, ...) and one decade of L-band (JERS, PALSAR)
• Statistical and semi-empirical methods for specific sites/conditions, i.e. the Dubois model.
• SMAP and SAOCOM are focusing on this problem.
– Operational products have been developed using coarse resolution (50 km) C-band scatterometers
• Temporal change technique.
• Extended period of observations required to develop frequency specific calibration of each footprint!
• Considered an index as opposed to actual soil moisture.
Passive and Active Soil Moisture Algorithms• Passive and active microwave synergy
– There have been a few very limited studies.
– SMAP has this capability but the primary focus at the moment is on resolution enhancement.
• Utilizing the radar data directly in the retrieval algorithm is under consideration.
– Mixing and matching passive and active data from existing satellites as a means of simulating Aquarius may not be useful.
• AMSR-E/SMOS and ASCAT/QUIKSCAT: vegetation and roughness effects are very important and vary with frequencies and polarizations. Need concurrent observations.
• SMOS and ALOS: issues with disparate scales need to be resolved. Need concurrent observations.
(or worth the effort at this point).
Aquarius Soil Moisture Algorithms
• Our approach will be to build from proven techniques and add in the capabilities of Aquarius/SAC-D– Start with a description of the radiative transfer model used
in passive microwave methods for soil-vegetation conditions.
skyfpfpB TeTeTfp
]1[ ,,,,,,
T
Te fpB
fp
,,
,,
2
2
2
,,sincos
sincos1
r
rsoilfHe
,,,,,,,,,,,,,,,,, ]]1[1][1][1[ vfpsurf
fpvfpsurf
fpvfpvfpfp eee
]cosexp[]1[1 2,,,,,, gfp
surffp
soilfp hee
Brightness temperature (TB) is a function of emissivity (e) and
physical temperature (T).
The second term is small resulting in a simple relationship for e
The observed e is the result of contributions from the soil surface
(esurf) modified by the scattering () and attenuation () of the
vegetation (v)
esurf is the soil emission
(esoil) modified by the surface roughness (h)
The esoil is a function of the soil dielectric properties (r)
Observations are made at a specific polarization (p), frequency
(f), and angle ()
A dielectric mixing model relates r to soil moisture based on sand and clay
fractions
From Brightness Temperature to Soil Moisture
The contributing depth of the soil is on the order of
0.25*wavelength. For 1.4 GHz or L-band this is 5 cm
Unknowns
• Problem: There are more unknowns than measurements: under determined system of equations
• Solutions attempt to reduce the dimensionality of the problem. Some approaches are:
1. Multiple angle observations.2. Assuming polarization independence (or defining the
dependence) of parameters.3. Assuming frequency independence of parameters.4. Estimating parameters using ancillary information.5. Localized calibration.
• Initial selection for this study:– Single Channel Algorithm (SCA) – Alternatives: three algorithms being evaluated by SMAP
Soil Moisture Retrieval Algorithms
X
X
Not compatible with Aquarius/SAC-D
skyfpfpB TeTeTfp
]1[ ,,,,,,
T
Te fpB
fp
,,
,,
2
2
2
,,sincos
sincos1
r
rsoilfHe
,,,,,,,,,,,,,,,,, ]]1[1][1][1[ vfpsurf
fpvfpsurf
fpvfpvfpfp eee
]cosexp[]1[1 2,,,,,, gfp
surffp
soilfp hee
Brightness temperature (TB) is a function of emissivity (e) and
physical temperature (T).
The second term is small resulting in a simple relationship for e
The observed e is the result of contributions from the soil surface
(esurf) modified by the scattering () and attenuation () of the
vegetation (v)
esurf is the soil emission
(esoil) modified by the surface roughness (h)
The esoil is a function of the soil dielectric properties (r)
Observations are made at a specific polarization (p), frequency
(f), and angle ()
A dielectric mixing model relates r to soil moisture based on sand and clay
fractions
The contributing depth of the soil is on the order of
0.25*wavelength. For 1.4 GHz or L-band this is 5 cm
Aquarius Radiometer
Aquarius Radar
SAC-D 36.5 GHz or NIRST
Aquarius Radar
Potential of Aquarius/SAC-D Measurements in Soil Moisture Retrieval
Land Surface Temperature (LST)
• Required for all algorithms (TB to emissivity)
– MWR 36.5 GHz V algorithm• Heritage from SSM/I, TMI, AMSR-E, and WindSat
• Potential mission product
• Data integration issues?
– Numerical Weather Forecast Model products• SMOS and SMAP approach
• Several options
– NIRST LST product (research)
Currently conducting a comparison of three NWF temperature products (MERRA, ECMWF, and NCEP) and impacts on soil moisture retrieval.
SCAL Band passive H pol.
Aquarius/SAC-D Soil Moisture Retrieval
L Band passive V pol.
L Band radar
Forecast model LST
MWR 36.5 V
NIRST TIR
NDVI-Climatological
NDVI-MODIS
Alternatives
Ver. X.1
Ver. X.2
Inputs Algorithm Product
.
.
.
.
.
• Approaches using various inputs and algorithms will be systematically developed and evaluated.
SCAL Band passive H pol.
Aquarius/SAC-D Soil Moisture RetrievalVer. 1
L Band passive V pol.
L Band radar
Forecast model LST
MWR 36.5 V
NIRST TIR
NDVI-Climatological
NDVI-MODIS
Alternatives
Ver. 1.1(BASELINE)
Ver. 1.2
Inputs Algorithm Product
• Ver. 1 represents the adaptation of heritage passive microwave X and C-band algorithms to L-band.
• Model LST will be used until MWR 36.5 V data are validated and available in the integrated data set.
Update of our current AVHRR to MODIS almost complete
TBDL Band passive H pol.
Aquarius/SAC-D Soil Moisture RetrievalVer. ?
L Band passive V pol.
L Band radar
Forecast model LST
MWR 36.5 V
NIRST TIR
NDVI-Climatological
NDVI-MODIS
Ver. ?
Inputs Algorithm Product
• Ver. ? will attempt to utilize both passive and active L-band data in a modified (or new) retrieval algorithm.
• This is the long-term objective of the project.
Aquarius Soil Moisture Validation• Exploit existing resources (satellite/model/in
situ)– AMSR-E, SMOS, SMAP– Argentina collaboration
• In Situ– Sparse networks
• Usually geographically distributed, public domain, real time
• Scaling is a major challenge
– Dense networks• Not many• Compatibility and standardization
Sparse and Dense In Situ Soil Moisture Networks (Examples)
DENSE: ARS Soil Moisture Validation Watersheds
Land Cover Conditions in the WatershedsAZ-WG OK-LW GA-LR ID-RC
SPARSE:. The USDA SCAN
AMSR-E Soil Moisture Validation
• Some AMSR-E experiences– Wide range of results with different algorithms– Validation period of record is a factor
AMSR-E Validation: 4 Soil Moisture Algorithms - 7 Years of Data
Little Washita (Dsc)
0
0.1
0.2
0.3
0.4
0.5
0 0.1 0.2 0.3 0.4 0.5
Observed Soil Moisture (m3/m3)
Es
tim
ate
d S
oil
Mo
istu
re (
m3 /m
3 )
JAXA
NASA
SCA
LPRM
AMSR-E Validation: Season and Anomalies Can Impact Performance
SCA Bias
-0.100-0.080-0.060-0.040-0.0200.0000.0200.0400.0600.0800.100
9/1
/20
02
3/1
/20
03
9/1
/20
03
3/1
/20
04
9/1
/20
04
3/1
/20
05
9/1
/20
05
3/1
/20
06
9/1
/20
06
3/1
/20
07
9/1
/20
07
Time
VS
M (
m3
/m3
)
LW_d
LR_d
WG_d
LW_a
LR_a
WG_a
0.000
0.020
0.040
0.060
0.080
0.100
9/1
/20
02
3/1
/20
03
9/1
/20
03
3/1
/20
04
9/1
/20
04
3/1
/20
05
9/1
/20
05
3/1
/20
06
9/1
/20
06
3/1
/20
07
9/1
/20
07
VS
M (m
3/m
3)
Time
SCA SEE
LW_d
LR_d
WG_d
LW_a
LR_a
WG_a
Recent SMAP Related Field Campaigns
(Active and Passive Aircraft/Satellite Observations)
• CanEx: June 2-16, 2010
• SJV10: June 29, October 25, 2010, TBD
• SMAPEx: July 6-10, Dec. 4-8, 2010, and two additional in 2011.
CanEx SM 2010: CSA and NASA• Primary mission objectives: (1) validation of SMOS
brightness temperature and soil moisture products and (2) concurrent time series of active and passive microwave observations for SMAP passive, active, and combined soil moisture algorithms.
• Flight dates: – June 2-15 Kenaston (KEN) (7 dates) (agricultural)– June 16 BERMS (1 date) (forest)
• Other mission considerations– Both domains include two independent SMOS grid
footprints.– Coordination with SMOS over passes.– Calibration and scaling of permanent in situ networks
CanEx Permanent In Situ Networks
Environment Canada
Kenaston
University of Guelph
Environment CanadaBERMS
CanEx SMOS Pixel Centers and Aircraft Coverage
SMOS products are at ~40 km resolution and gridded at ~16 km.
Each area includes ~ 2 independent pixels (~33 by 70 km).
Aircraft logistics limit the size of the coverage domain.
CanEx Campaign Soil Moisture Sites
Ground sampling sites included most permanent sites, the BERMS temporary network, and additional sites selected to be representative of domain, provide spatial coverage, and support multiple scaling objectives.
59
50
BERMS Temporary Network• Will provide a longer record for SMOS/BERMS and to establish scaling of the limited permanent network.
NASA G-III
Environment Canada Twin Otter
CanEx Campaign Aircraft
UAVSAR: L-band fully polarimetric radar (Swath 20 km, resolution 6 m).Multiple lines were required to provide coverage between 35 and 45o.
L-band dual polarization radiometer, 40o (single beam, resolution 3 km). Multiple lines were required to provide coverage .
6.9, 19, 37 and 89 GHz radiometers, 53o (single beam, res. are 1.3 km for 6.9 GHz and 0.8 km for the others)
CanEx SummaryDate Saskatoon
Daily Rainfall (mm)
Prince Albert
Daily Rainfall (mm)
Aircraft Flights Site
Satellite Coverage Note
May 22-30 67.3 68.4
June 1
2 0.7 KEN SMOS
3 9.8 14.8 SMOS
4 2.2
5 6.8 KEN SMOS
6 0.2 KEN
7 5.5 2.8
8 16.9 7.2 SMOS Partial Twin Otter
9 KEN
10 5.5 4.0 SMOS
11 13.2
12
13 0.4 KEN SMOS
14 0.1 KEN
15 0.3 KEN SMOS Only UAVSAR
16 0.2 BERMS SMOS
Significant Rain
Re-wetting
25o
65o
CanEx Kenaston Area
Composite radar image (HH-red, VV-blue, HV-green)
June 5, 2010
~70 Data Sets!
Australia Campaign (SMAPEx)• Objective: concurrent multiple season
active and passive microwave observations for passive, active, and combined soil moisture algorithms. (Supported by ARC) (Leads: J. Walker/R. Panciera Univ. Monash/Melbourne)
• Details– 4, 1-week campaigns– Semi-arid, irrigated & dryland crops, grazing– Airborne: L-band radiometer & radar,
VIS/IR/NIR/SWIR and Thermal IR– Ground: Soil Moisture, Vegetation
Biomass/VWC/LAI/VIS&NIR, Roughness, Soil Temperature
• Monitoring Network: 29 semi-permanent soil moisture stations (on SMAP 36km/9km/3km nested grids)
YA
YB
40km
July 6-10, 2010Dec. 4-8, 2010TBD, 2011
MERIT Airborne Facility
PLMR:Polarimetric L-band Multibeam Radiometer
Frequency/bandwidth: 1.413GHz/24MHz
Polarisations: V and H
Resolution: 1km at 3km flying height,
Incidence angles: +/- 7°, +/-21.5°, +/- 38.5° across track
Antenna type: 8x8 patch array
PLIS:Polarimetric L-band Imaging Scatterometer:
Frequency/bandwidth:1.26GHz/30MHz
Polarisations: VV, VH, HV and HH
Resolution: 10m
Incidence angles 15º -45º on both sides of aircraft
Antenna type: 2x2 patch array
6 x TIR
L-band Radar (focused SAR)
L-band Radiometer (6 beams)
6 x Vis/NIR
6 x SWIR
SMAP Major Field Campaignsver. 06/10
Year/Quarter
1 2 3 4
2008SMAPVEX08
2009 SMOS
2010SMAPEx
CanEx-SMSMAPEx
2011SMAPEx
Aquarius GCOM-WCanEx-FT
2012 SAOCOMSMAPVEX12
2013
2014 SMAP
2015 SMAPVEX15 SMAPVEX15 SMAPVEX15
• SMAPVEX08– High priority design/algorithm issues
• SMAPEx (Australia)– 4 one-week campaigns to span four
seasons– Aircraft Radar/Radiometer
• CanEx-SM (Canada)– Two-week soil moisture campaign– Aircraft Radar/Radiometer
• CanEx-FT (Canada)– Two-week freeze/thaw campaign– Aircraft Radar/Radiometer
• SMAPVEX12– Major hydrology campaign – Long duration– Aircraft Radar/Radiometer
• SMAPVEX15– Extended campaign for both SM and
FT – Long duration– Aircraft Radar/RadiometerSatellite Launch in Red
L-Band Satellite Synergy• Continuity for L-band measurements of SMOS,
Aquarius, SAOCOM, and SMAP.
• Significant contributions to SMAP– Aquarius experience with RFI– Aquarius TB and o for SMAP L2 soil moisture algorithms
0
0.2
0.4
0.6
0.8
1
1.2
2006 2008 2010 2012 2014 2016 2018 2020
SMAP
Aquarius
SMOS
SAOCOM
Estimated Mission Timeline
Summary• Aquarius/SAC-D provides opportunities to explore new
approaches to soil moisture retrieval.– First space borne data that can be used to assess the synergy of L-band passive
and active observations for improving remote sensing of soils and vegetation.
• Our approach will build from heritage satellite-based low frequency passive microwave algorithms
• These utilize a unique element of Aquarius/SAC-D, 1.4 and 36.5 GHz radiometers. – Information from the MWR 36.5 V GHz is used to derive land surface
temperature (LST), which is then employed in computing emissivity. – The retrieved LST is another potential Aquarius/SAC-D product.
• These results will serve as a baseline for research on a combined passive-active soil moisture algorithm.– Improvements in corrections for roughness, vegetation, and transient water
effects.