Content uploaded by Deborah K Smith
Author content
All content in this area was uploaded by Deborah K Smith on Nov 24, 2014
Content may be subject to copyright.
19
V. Barale, J.F.R. Gower and L. Alberotanza (eds.), Oceanography from Space, revisited.
© Springer Science+Business Media B.V. 2010
Passive Microwave Remote Sensing of the
Ocean: an Overview
Chelle L. Gentemann, Frank J. Wentz, Marty Brewer, Kyle Hilburn, and
Deborah Smith
Remote Sensing Systems, Santa Rosa, CA, USA
Abstract. Passive microwave observations from satellites provide measu-
rements of sea surface temperature (SST), wind speed, water vapor, cloud
liquid water, rain rate, and sea ice that have lead to significant advances in
meteorological and oceanographic research as well as improvements in
monitoring and forecasting both weather and climate. Future instruments
are planned to measure sea surface salinity. The calibration of passive mi-
crowave radiometers has continued to improve, along with the retrieval al-
gorithms. The production of accurate geophysical retrievals depends on
the close development of both calibrated brightness temperatures and re-
trieval algorithm design in concert. Data must be carefully screened for
near-land emissions, radio frequency interference, rain scattering (for SST,
wind, and vapor retrievals), and high wind events (SST retrievals only).
1. Introduction
Global geophysical measurements from passive microwave radiometers
provide key variables for scientists and forecasters. The daily measure-
ments of Sea Surface Temperature (SST), wind speed, water vapor, cloud
liquid water, rain rate, and, in the future, Sea Surface Salinity (SSS) over
the oceans has provided data sets used to significantly improve our under-
standing of the Earth system. The data are used extensively in numerical
weather prediction, hurricane forecasting, climate monitoring, ecosystem
forecasting and fisheries; as well as for climate, weather, oceanographic,
metorological and ecosystem research. The measurement accuracy is tied
to the evolution of both the calibration methods and retrieval algorithms.
2. Background
Designed to measure rainfall, the first Passive MicroWave (PMW) radi-
ometer was launched in December 1972 on the Nimbus-5 satellite. After a
20 C.L. Gentemann et al.
short gap, PMW radiometers have been continuously observing the oceans
since the launch of Nimbus-7 in 1978. This instrument was followed by
the Special Sensing Microwave Imager (SSM/I) series. More recently,
several other PMW radiometers have been launched on National Aeronau-
tics and Space Administration (NASA), Japan Aerospace eXploration
Agency (JAXA), and European Space Agency (ESA) satellites (Table 1).
Table 1. PMW radiometer mission characteristics
Satellite Sensor Launch
Failure
Frequency (GHz) Coverage
Nimbus-5 ESMR 12/1972
5/1977 19.4 Global
Nimbus-7 SMMR 10/1978
8/1987 6.6, 10.7, 18.0, 21.0, 37.0 Global
SEASAT SMMR 6/1978 10/1978
6.6, 10.7, 18.0, 21.0, 37.0 Global
DMSP F08
SSM/I 7/1987 12/1991
19.4, 22.2, 37.0, 85.5 Global
DMSP F10
SSM/I 12/1990
11/1997
19.4, 22.2, 37.0, 85.5 Global
DMSP F11
SSM/I 12/1991
5/2000 19.4, 22.2, 37.0, 85.5 Global
DMSP F13
SSM/I 5/1995 Present
19.4, 22.2, 37.0, 85.5 Global
DMSP F14
SSM/I 5/1997 8/2008 19.4, 22.2, 37.0, 85.5 Global
DMSP F15
SSM/I 12/1999
Present
19.4, 22.2, 37.0, 85.5 Global
TRMM TMI 12/1997
Present
10.7, 19.4, 21.3, 37.0, 85.5 40S-40N
ADEOS-II
AMSR 12/2002
10/2003
6.9, 10.7, 18.7, 23.8, 36.5, 89.0 Global
AQUA AMSR-E
5/2002 Present
6.9, 10.7, 18.7, 23.8, 36.5, 89.0 Global
Coriolis WindSat
6/2003 Present
6.8, 10.7, 18.7, 23.8, 37.0 Global
DMSP F16
SSMI/S 10/2003
Present
- Global
DMSP F17
SSMI/S 11/2006
Present
- Global
SMOS MIRAS 11/2009
- 1.4 Global
GPM GMI (7/2013)
- 10.7, 18.7, 23.8, 36.5, 89.0 65S-65N
SAC-D Aquarius
(5/2010)
- 1.4 Global
GCOM-W
AMSR2
(2/2012)
- 6.9, 7.3, 10.7, 18.7, 23.8, 36.5, 89.0
Global
C2 MIS (5/2016)
- 6.8, 10.7, 18.7, 23.8, 37.0, 89.0 Global
The Electrically Scanning Microwave Radiometer (ESMR) on Nimbus-
5 had only one channel at 19.35 GHz and was capable of measuring both
rainfall and sea ice detection.
From October 1978 through July 1987, the Nimbus-7 Scanning Multi-
channel Microwave Radiometer (SMMR) measured at 6.6, 10.7, 18.0,
21.0, and 37 GHz in both the horizontal and vertical polarizations
(Gloersen et al., 1984). SMMR geophysical retrievals were compromised
by non-negligible switch leakages (Han and Kim, 1988), rendering the
Passive Microwave Remote Sensing of the Ocean: an Overview 21
SMMR measurements useful for detection of sea ice but not accurate
enough for geophysical retrievals.
The Defense Meteorological Satellite Program (DMSP) satellite series
launched the first SSM/I on F-08 in June 1987. This was followed by
SSM/Is on F-09 through F-15. The DMSP satellites orbit the earth in 102
minutes, at approximately 833 km with an inclination of 98.8 ° (Hollinger
et al., 1990). The F-series alternate between early and late morning Local
Equator Crossing Times (LECTs). The SSM/I instrument measures at
19.4, 22.2, 37.0, and 85.5 GHz. Both vertical and horizontal polarizations
are measured for all channels except the 22.2 GHz which only measures
the vertical. SSM/I was the first satellite PMW radiometer to have exter-
nal calibration accomplished by viewing a mirror that reflects cold space
and a hot reference absorber once each scan, every 1.9 seconds. The cold
space is a known 2.7 K while the hot absorber temperature is monitored
with thermistors. The frequent calibration minimizes receiver gain fluc-
tuation contributions to the signal but does not correct radiometer nonlin-
earity (if it exists). This well-calibrated instrument’s measurements are
used to determine wind speed, water vapor, cloud liquid water, rain rates,
and sea ice concentration over global oceans.
In December 1997, NASA launched the Tropical Rainfall Measuring
Mission (TRMM) carrying the TRMM Microwave Imager (TMI), a PMW
radiometer measuring at 10.7, 19.4, 21.3, 37.0, and 85.5 GHz. Similar to
SSM/I, all channels measure both vertical and horizontal polarizations, ex-
cept the 21.3 GHz which only measures in the vertical (Kummerow et al.,
1998). Designed to measure the tropics and sample the diurnal cycle, the
satellite was launched with an orbital inclination of 35° at an altitude of
350 km (later changed to 400 km to extend satellite life). This equatorial
orbit yields coverage from 39N to 39S. The satellite is sun-asynchronous,
processing through the diurnal cycle every 23 days. Again, similar to
SSM/I, the feed horns and main reflector rotate, with a period of 1.9 sec-
onds, about an axis parallel to the local spacecraft nadir. The stationary
hot reference absorber and cold calibration reflector are positioned so that
they pass between the feed horns and main reflector once per scan. The
temperature of the warm load is monitored by three thermistors while the
cold reflector views the cosmic microwave (MW) background at 2.7 K. At
fairly regular intervals the platform yaws from forward (aft) viewing direc-
tion to aft (forward). Each scan consists of 104 discrete samples spaced by
8 km. In addition to the geophysical variables measured by SSM/I, TMI is
able to measure SST. TMI suffered calibration problems due to an emis-
sive reflector, for which corrections were developed and implemented.
NASA’s AQUA satellite carries the JAXA’s Advanced Microwave
Scanning Radiometer - Earth Observing System (AMSR-E). The AQUA
22 C.L. Gentemann et al.
satellite was launched in May 2002 into a polar, sun-synchronous orbit at
an altitude of 705 km, with a LECT of 1:30 AM/PM. AMSR-E has 12
channels corresponding to 6 frequencies: 6.9, 10.7, 18.7, 23.8, 36.5, and
89.0 GHz, all except 23.8 measure both vertical and horizontal polariza-
tions (Parkinson, 2003). The calibration is completed similar to SSM/I and
TMI using a cold reflector and hot absorber with 8 thermistors. The
AMSR-E hot absorber has large thermal gradients not well measured by
the thermistors. A correction for this error in the calibration reference
point has been developed and implemented. In addition to the geophysical
variables measured by SSM/I, AMSR-E is able to measure SSTs. Almost
global coverage is attainable in 2 days (Figure 1).
The Naval Research Laboratory (NRL) launched the Coriolis satellite in
January 2003. The sun-synchronous orbit is at an altitude of 840 km with
a LECT at 6:00 AM/PM (Gaiser et al., 2004). Coriolis carries the Wind-
Sat instrument, a fully polarimetric PMW radiometer intended to retrieve
wind direction in addition to wind speed. The fully polarimetric channels
are at 10.7, 18.7, and 37.0 GHz, but the instrument also has channels at 6.8
and 23.8 that only measure the vertical and horizontal polarizations. Cali-
bration is similar to SSM/I with a cold reflector and hot absorber measured
by 6 thermistors.
DMSP satellites F16 and forward carry the Special Sensor Microwave
Imager/Sounder (SSMIS). F16 was launched in October 2003 into a sun-
synchronous orbit at an altitude of 830 km and a LECT of 8 AM/PM.
SSMIS has 24 channels, several of which are similar to the SSM/I set
(19.35, 22.2, and 37.0 GHz). The additional channels are intended for at-
mospheric sounding. The calibration is completed similar to SSM/I using
a cold reflector and hot absorber. SSMIS has two main problems, an emis-
sive antenna and non-uniform hot absorber. Corrections for these issues
have been developed and implemented.
Future PMW radiometers include JAXA’s Global Change Observation
Mission – Water (GCOM-W) AMSR2, the National Polar Orbiting Earth
observing System of Systems (NPOESS) C2 satellite will carry the Mi-
crowave Imager Sounder (MIS), and NASA’s Global Precipitation Mis-
sion (GPM) will carry the GPM Microwave Imager (GMI). For all these
instruments, the planned calibration is similar to SSM/I using a cold reflec-
tor and hot absorber.
GCOM-W is to be launched in February 2012 into NASA’s A-Train
satellite formation in a sun-synchronous orbit with an altitude of 700 km
and a LECT of 1:30 AM/PM. AMSR2 is similar to AMSR-E but has an
improved hot absorber and an additional channel at 7.3 GHz to minimize
Radio Frequency Interference (RFI). With a launch date set for February
Passive Microwave Remote Sensing of the Ocean: an Overview 23
2012, it is hoped that the AQUA AMSR-E remains healthy until then to al-
low for satellite inter-calibration.
Fig. 1. AMSR-E geophysical retrievals 1-2 October 2009. Small amounts of
missing data due to rain events are visible in the SST and wind retrievals.
Two other future instruments, the European Space Agency’s Soil Mois-
ture and Ocean Salinity (SMOS) Microwave Imaging Radiometer using
Aperture Synthesis (MIRAS) and the Satélite de Aplicaciones Científicas-
24 C.L. Gentemann et al.
D (SAC-D) Aquarius are intended to measure ocean salinity and only have
a single channel at 1.4 GHz. SMOS launched in November 2009 into a
sun-synchronous orbit at 800 km with an LECT of 6:00 AM/PM. Aquar-
ius is scheduled to be launched in May 2010 into a sun-synchronous orbit
at 650 km with a LECT of 6:00 AM/PM. Both of these instruments are
designed to provide measurements of ocean salinity.
3. Calibration
To create a climate quality, inter-calibrated dataset of PMW geophysical
retrievals, it is necessary to start the process using radiometer counts and
work towards calibrated geophysical retrievals. Table 2 describes the steps
to produce a calibrated brightness temperature (TB). First, it is necessary
to reverse engineer the antenna temperatures (TAs) or TBs back to radi-
ometer counts. Often there are small provider added corrections or ad-
justments put into the TA or TBs which are sometimes undocumented.
For example, SSMI/S had five TB version changes in the first two years of
data. Therefore, the first step is to reverse these steps and remove any cor-
rections. Starting from radiometer counts, the first two steps in the calibra-
tion procedure are crucial to accurately determining other errors.
Table 2. Calibration steps for PMW radiometers
Geolocation
analysis
Attitude
adjustment
Along-
scan
correction
Absolute
calibration
Hot load
correction
Antenna
emissivity
SSM/I NRL/RSS No Yes APC No1 0
TMI Goddard Dynamic Yes APC No 3.5%
AMSRE
RSS Fixed Yes APC Yes 0
AMSRA
RSS Dynamic Yes APC Yes 0
WindSat
NRL/RSS Fixed Yes APC Yes 0
SSMIS RSS No Yes APC Yes 0.5-3.5%
To ensure that any subsequent collocations or comparisons that are per-
formed are correct, it is necessary to do a geolocation analysis. The cor-
rection to the geolocation is different than a correction for erroneous satel-
lite pointing information (roll/pitch/yaw). This is a correction for the
mounting of the instrument on the satellite. Pointing is usually off by
1 Errors due to hot load are removed when doing the zonal TB inter-calibration
Passive Microwave Remote Sensing of the Ocean: an Overview 25
about 0.1 ° from the satellite specified roll/pitch/yaw. The geolocation
correction uses ascending minus descending TA to ensure that islands do
not ‘move’. The geolocation analysis has been performed by a number of
groups, NRL and Remote Sensing Systems (RSS) both contributed to
SSM/I, TMI was completed by Goddard, and other instruments as speci-
fied in Table 2.
Corrections from this point onward are determined by comparisons be-
tween the satellite TA measurements and TAs simulated using a radiative
transfer model (RTM). Using collocated environmental information, RTM
simulated TBs are determined. These TBs are then transformed into TAs
using the instrument, channel specific antenna patterns.
After the pointing is corrected, the spacecraft reported roll/pitch/yaw are
then examined for errors using comparisons of the observed minus RTM
TAs. Spacecraft pointing is determined by a number of different methods,
the preferred being a star tracker. Another method is horizon balancing
sensor. For SSM/I no pointing information was given, so it was assumed
to be correct. TMI has a dynamic pointing correction that changes within
an orbit because the horizon sensor used prior to the orbit boost is not as
accurate as a star tracker. After boost, the horizon sensor was disabled and
pointing was determined from two on-board gyroscopes, also not as accu-
rate as a star tracker. AMSR-E had no pointing problems, as the AQUA
had a star tracker. The AMSR on ADEOS-II needed a dynamic correction,
while WindSAT needed a simple fixed correction to the roll/pitch/yaw.
Once instrument mounting errors and satellite attitude errors have been
corrected for, an along-scan correction is completed. It is very important
to complete the first two corrections first because TA is dependent on inci-
dence angle. Not correcting for pointing errors would result in an errone-
ous cross-scan biasing. As the mirror rotates, at the edge of the earth scene
the view will begin to contain obstructions such as the satellite itself or
part of the cold mirror. Additionally, during the scan, the antenna side-
lobe pattern may result in contributions from different parts of the space-
craft. Therefore the difference between the TA and RTM simulated TAs
are again used to examine the data for along-scan biases. This correction
is needed for every instrument.
The antenna pattern correction (APC) is then completed. Pre-launch, an
APC is determined, consisting of the spill over and cross-polarization val-
ues. After launch the spill over and cross-polarization values are adjusted
so that the measured TAs matches the simulated TAs. This correction is
needed for all instruments. Next, a correction for the hot load thermal gra-
dients and antenna emissivity are developed. These are only needed for
specific instruments. The determination of TB from counts for PMW ra-
diometers is completed using two known temperatures to infer the scene
26 C.L. Gentemann et al.
temperature. For each scan, the feedhorns view a mirror that reflects cold
space, a known 2.7 K, a hot absorber, measured by several thermistors, and
Earth scenes. Assuming a linear response, the Earth scene temperatures
are then determined by fitting a slope to the two known measurements as
shown in Figure 2. This 2-point calibration system continuously compen-
sates for variations in the radiometer gain and noise temperatures. This
seemingly simple calibration methodology is fraught with subtle difficul-
ties. The cold mirror is relatively trouble-free, as long as lunar contamina-
tion is flagged. Occasionally, the cold mirror will not reflect deep space,
but the moon instead. These data must be removed.
Fig. 2. Calculation of Earth scene brightness temperatures using the radiometer
counts and calibration points (cold mirror and hot absorber) known temperatures.
The hot absorber has been more problematic as the thermistors often do
not adequately measure thermal gradients across the hot absorber. For ex-
ample, a hot load correction is required for AMSR-E because of a design
flaw in the AMSR-E hot load. The hot reference load acts as a blackbody
emitter and its temperature is measured by precision thermistors. Unfortu-
nately, during the course of an orbit, large thermal gradients develop
within the hot load due to solar heating making it difficult to determine the
average effective temperature from the thermistor readings. The thermis-
tors themselves measure these gradients and may vary by up to 15 K be-
tween themselves at any time for AMSR-E. Several other instruments
have had similar, but smaller, issues. RTM simulations are used to deter-
Passive Microwave Remote Sensing of the Ocean: an Overview 27
mine an effective hot load temperature which is a regression of the meas-
ured hot load thermistor temperatures. The follow-on instrument, AMSR2
on GCOM-W, has an improved hot absorber design that should mitigate
these issues.
Finally, the main reflector is assumed to be a perfect reflector with an
emissivity of 0.0, but this is not always the case. For example, a bias rec-
ognized in the TMI measurements was attributed to the degradation of the
primary antenna. Atomic oxygen present at TMI’s low altitude (350 km)
led to rapid oxidization of the thin, vapor-deposited aluminum coating on
the graphite primary antenna, resulting in a much higher antenna emissiv-
ity than expected. The measured radiation is comprised of the reflected
earth scene and antenna emissions. Emissivity of the antenna was deduced
during the calibration procedure to be 3.5%. The antenna emissivity cor-
rection utilizes additional information from instrument thermistors to esti-
mate the antenna temperature, thereby reducing the effect of the temporal
variance. This emissivity is constant for all the TMI channels. SSMI/S
has an emissive antenna where the emissivity appears to increase as a
function of frequency, changing from 0.5 – 3.5 %.
4. Retrieval algorithm
Geophysical retrievals from PMW radiometers are commonly determined
using a radiative transfer model to derive a regression algorithm (Wentz,
1998). A schematic of the derivation of the regression coefficients is
shown in Figure 3. A large ensemble of ocean-atmosphere scenes is first
assembled. The specification of the atmospheres comes from quality-
controlled radiosonde flights launched from small islands (Wentz, 1997).
One half of these radiosonde flights are used for deriving the regression
coefficients, and the other half is withheld for testing the algorithm. A
cloud layer of various columnar water densities ranging from 0 to 0.3 mm
is superimposed on the radiosonde profiles. Underneath these simulated
atmospheres, we place a rough ocean surface. SST is randomly varied
from 0 to 30 °C, the wind speed is randomly varied from 0 to 20 ms-1, and
the wind direction is randomly varied from 0 to 360°.
Atmospheric brightness temperatures and transmittance are computed
from these scenes and noise, commensurate with measurement error which
depends on spatial resolution, is added. The noise-added simulated bright-
ness temperatures along with the known environmental scene are used to
generate multiple linear regression coefficients. Algorithm testing is un-
dertaken by repeating the process using the withheld scenes.
28 C.L. Gentemann et al.
Fig. 3. Derivation of regression coefficients
5. Geophysical retrievals
Wind speed
Ocean surface winds are crucial to transferring heat, gases, energy and
momentum between the atmosphere and ocean. Winds also determine the
large scale ocean circulation and transport, power global weather patterns,
and play a key role in marine ecosystems. Hurricanes, typhoons, and mid-
latitude winter storms all contain high wind speeds that threaten interna-
tional shipping and the lives and property of people along the coasts.
Passive Microwave Remote Sensing of the Ocean: an Overview 29
Ocean surface winds change rapidly in both time and space and satellite
sampling and accuracy make these observations the most useful wind data
available for research and forecasting over the global oceans.
Surface wind speeds (at 10 m height, without directions) are routinely
estimated from passive microwave radiometers (SSM/I, AMSR-E, TMI,
SSMIS) on a spatial scale of roughly 25 km. Wind speeds in the range of
0 to 30 ms-1 are simultaneously retrieved along with SST, water vapor,
cloud liquid water and rain rates using an algorithm that exploits the po-
larization signature of wind induced sea surface emissivity (Wentz, 1997).
Radiometer winds are quite accurate under typical ocean conditions when
no rain is present, however when even a little rain exists, the wind speeds
are unusable. Validations of radiometer winds in rain-free conditions have
been performed. Comparisons with ocean buoy and weather model winds
show root-mean-square differences near (model winds) or less than 1 ms-1
(buoy winds) in rain-free conditions (Mears et al., 2001; Meissner et al.,
2001). Since 1996, there have been three or more radiometers in polar or-
bits simultaneously, resulting in good spatial and temporal sampling, yield-
ing over 95% Earth ocean surface coverage in a given day.
WindSat is a passive fully-polarimetric microwave radiometer designed
to measure ocean surface vector winds. It has been found to have wind ac-
curacies close to that of scatterometers for winds between 6 and 20 ms-1,
with significant wind direction uncertainty below 6 ms-1 (Bettenhausen et
al., 2006). WindSat vector winds have been poor in rainy conditions until
recently when a new WindSat algorithm has been developed that improves
WindSat winds even in rain (Meissner and Wentz, 2009). The quality of
these new winds is similar to QuikScat in all but very heavy rain and very
low winds. Excellent agreement (to within 0.5 ms-1) is found between pas-
sive radiometer wind speeds, polarimetric radiometer wind vectors and
scatterometer vector winds despite the different measuring methods of
each instrument (Wentz and Meissner, 2007). Only a few small regions of
difference exist that seem to be related to the 37 GHz observations of the
ocean surface and atmosphere.
Combined surface wind data sets have recently become more available
and are very useful in atmospheric and oceanographic research due to the
lack of data gaps. One example, the Cross Correlated Multi-Platform
(CCMP) winds (Atlas et al., 2009), use carefully inter-calibrated PMW
wind speeds from radiometers and wind vectors from scatterometers.
Simple interpolation schemes are unable to adequately represent fast-
moving storms in mid-latitude regions when making a merged wind prod-
uct with no gaps. An advanced 4-dimensional variational analysis method
is used in the CCMP to merge the satellite winds with the European Center
for Medium-range Weather Forecasting (ECMWF) Re-Analysis (ERA)-40
30 C.L. Gentemann et al.
model wind vectors, providing a gridded wind product consisting of an
analyzed wind field every six hours for 20 years.
The satellite winds used in the CCMP include over 20 years of SSMI
winds. A recent study showed that these carefully inter-calibrated SSM/I
winds have no spurious trends. Wentz et al. (2007) found agreement be-
tween ocean buoy trends and the SSM/I trends for many buoy types and
different ocean regions. The overall difference in wind trend (SSM/I mi-
nus buoy) is -0.02 ms-1/decade. This gives one confidence in using the
passive microwave winds in climate studies.
Water Vapor
Over 99% of the atmospheric moisture is in the form of water vapor, and
this vapor is the principal source of the atmospheric energy that drives the
development of weather systems on short time scales and influences the
climate on longer time scales. Tropospheric water vapor measurements
are an important component to the hydrological cycle and global warming
(Held and Soden, 2006; Trenberth et al., 2005). The microwave measure-
ment of water vapor can also be used as a proxy to detect global warming
of the lower troposphere with a signal-to-noise ratio that is five times bet-
ter than the AMSU method of measuring the temperature change (Wentz
and Schabel, 2000).
Satellite microwave measurements near the 22.2 GHz vapor absorption
line provide the most accurate means to determine the total amount of va-
por in the atmosphere. Quality controlled radiosondes from stations on
small islands or ships are used for validation of the columnar water vapor
retrievals. Simulations show that retrievals are accurate to 0.1 mm total
columnar water vapor. Comparisons of AMSR-E water vapor retrievals
with ship based radiosondes show an error of 2.2 - 0.5 mm (Szczodrak et
al., 2006) which includes errors due to differences between a radiosonde
point measurement and the larger AMSR-E footprint.
Cloud Liquid Water
Cloud water links the hydrological and radiative components of the cli-
mate system. Cloud water can be retrieved from passive microwave
measurements because of its strong spectral signature and polarization sig-
nature (Wentz, 1997). Passive microwave observations provide a direct
estimate of the total absorption along the sensor viewing path. At 18 and
37 GHz, clouds are semi-transparent allowing for measurement of the total
columnar absorption. The absorption is related to the total amount of liq-
Passive Microwave Remote Sensing of the Ocean: an Overview 31
uid water in the viewing path, after accounting for oxygen and water vapor
absorption.
Validation of columnar cloud liquid water is a difficult undertaking.
The spatial variability of clouds makes comparisons between upward look-
ing ground based radiometers and the large footprint size of the downward
looking satellite retrievals problematic. The upward looking ground-based
radiometers also have very limited geographic distribution, making mean-
ingful validation over global conditions impossible. Generally, validation
is completed using a statistical histogram method (Wentz, 1997).
Rain Rate
Rainfall is the key hydrological parameter, so much so that changes in the
spatial distribution of rainfall have led to the collapse of civilizations
(Haug et al., 2003; O'Conner and Kiker, 2004). Rain is one of the most
difficult parameters to accurately retrieve using remote sensing because of
its extreme variability in space and time over a variety of scales. The most
accurate and physically-based rain retrieval techniques take advantage of
the interactions between microwave radiation and water, and both passive
and active microwave remote sensing techniques can be used to derive rain
rates over both ocean and land.
PMW observations respond to the presence of rain in the instrument
field-of-view with two primary signals: an emission signal and a scattering
signal (Petty, 1994). The ocean surface is roughly 50% emissive, so it
serves as a cold background around 150 K against which to observe rain.
Since the ocean is an expansive flat surface, the emission is strongly polar-
ized. For typical incidence angles and clear skies, vertical polarization
brightness temperatures are larger than horizontal polarization brightness
temperatures by as much as 100 K. The emission depends on the sea sur-
face temperature, salinity, and surface roughness.
Emission from small round rain and cloud drops is unpolarized, and the
liquid emission strongly decreases the polarization seen by the sensor.
Heavy rain can bring the difference between vertical and horizontal polari-
zation brightness temperatures down to zero. The emission signal has a
strong spectral signature that increases with frequency – that is, higher mi-
crowave frequencies are more affected by rain. The strength of the emis-
sion signal depends on the total amount of liquid water below the freezing
level, and this is related to the surface rain rate. The primary factors gov-
erning this relationship are: the height of the freezing level, the relative
portioning of cloud and rain water, and the effect horizontal inhomogene-
ity – the beamfilling effect (Hilburn and Wentz, 2008; Wentz and Spencer,
1998). The scattering signal measures a decrease in brightness tempera-
32 C.L. Gentemann et al.
tures due to the presence of ice above the freezing level (Spencer et al.,
1989). Usually the scattering signal is used over a warm background, and
is especially useful over land. The relationship of the scattering signal to
surface rain rate is less direct than it is for the emission signal.
The relationship between the emission signal and the rain rate is
strongly nonlinear. Since rain is horizontally inhomogeneous over satellite
footprints (which may range in diameter from 6 - 56 km), the measurement
represents an average over the satellite footprint. Averaging a highly vari-
able observable quantity, when the observable quantity is nonlinearly re-
lated to the desired quantity, results in a bias in the desired quantity. This
is the beamfilling effect, and it causes rain rates to be underestimated by
PMW radiances.
Different sensors have systematically different spatial resolutions and
the probability distribution function of liquid water in the footprint
changes systematically with the size of the footprint. For example, an in-
finitely small satellite footprint would model the variability of liquid in the
footprint with a delta function, whereas a satellite footprint the size of the
Earth models that variability with the global rain probability distribution
function – typically taken to be a mixed log-normal distribution. Fortu-
nately, real satellite footprints do not vary that much. The spatial resolu-
tion of SSM/I rain retrievals is nominally 32 km, and the spatial resolution
of AMSR rain retrievals is nominally 12 km. This means that SSM/I rain
retrievals require a larger beamfilling correction than AMSR rain retriev-
als, because SSM/I retrievals have more spatial averaging.
(Hilburn and Wentz, 2008) developed a new beamfilling correction by
simulating lower resolution SSM/I data with higher resolution AMSR data.
Rain retrievals were computed from the simulated SSM/I data at several
resolutions and compared to the AMSR rain retrievals at the highest possi-
ble resolution to deduce how the variability of liquid water changes sys-
tematically with footprint size. When the new correction was applied to
satellite data, rain rates agreed to within 3% (after removing sampling bi-
ases due to the different local times-of-day for each satellite). New inter-
calibrated rain rate retrievals have been successfully used to close the wa-
ter cycle (Wentz et al., 2007), show excellent agreement with rain gauges
on ocean buoys (Bowman et al., 2009), and correlate well with the TRMM
Precipitation Radar (Cecil and Wingo, 2009).
Sea Ice
PMW retrievals of sea ice form one of the most important climate data re-
cords in existence. The time series of sea ice, from 1979 – present, has
provided measurements of ice concentration and classification of sea ice
Passive Microwave Remote Sensing of the Ocean: an Overview 33
types (multiyear or first-year ice) on a daily basis. The PMW sea ice re-
trievals are vital because of their ability to see through clouds. Large ice
shelf breakup events, such as the Larsen Ice shelf breakup, have been wit-
nessed and monitored using PMW retrievals. Sea ice is important to the
global climate as it acts to regulate heat, moisture, and salinity in the polar
ocean. The recent increase in summer Arctic sea ice acts as a positive
feedback for global warming by changing the albedo.
There are two common retrieval algorithms for sea ice, the NASA team
algorithm and the bootstrap algorithm. Both algorithms use the polariza-
tion and gradient ratios to determine ice concentration and type. At 19
GHz the difference between the vertical and horizontal polarizations is
small for sea ice (both first-year and multiyear) and large for ocean. The
two polarizations are different for first-year versus multi-year ice at 37
GHz (Cavalieri et al., 1984). The primary error in the NASA team algo-
rithm is due to the effects of surface glazing and layering on these channel
ratios (Comiso et al., 1997). Newer team algorithms use the 89 GHz gra-
dient ratio to minimize these errors (Markus and Cavalieri, 2000). The
bootstrap algorithm uses the polarization and gradient ratios, combining
different channels, such as the 19 and 37 vertical polarization ratio
(Comiso et al., 1997). Both algorithms use different methodologies to fil-
ter weather effects.
Validation of the sea ice retrievals has been completed through inter-
comparison between different algorithms and comparison to visible and in-
frared satellite measurements. The NASA team algorithm and bootstrap
algorithm generally agree with each other but differ by 10 – 35 % in areas
within the ice pack (Comiso and Steffen, 2001).
Sea Surface Temperature
Sea surface temperature is a key climate and weather measurement rou-
tinely made each day by satellite infrared (IR) and PMW radiometers, in
situ moored and drifting buoys, and ships of opportunity. These measure-
ments are used to create daily spatially-complete global maps of SST that
are then used for weather prediction, ocean forecasts, and in coastal appli-
cations such as fisheries forecasts, pollution monitoring, and tourism.
They are also widely used by oceanography, meteorology, and climate sci-
entists for research. Prior to 1998, SSTs were only available globally from
IR satellite retrievals, but with the launch of TMI, PMW retrievals became
possible. While IR SSTs have a higher resolution than PMW SSTs (1 – 4
km as compared to 25 km), their retrieval is prevented by clouds giving
PMW SSTs improved coverage since they are retrieved through clouds.
34 C.L. Gentemann et al.
Between 4 and 11 GHz the vertically polarized TB of the sea-surface
has an appreciable sensitivity to SST. In addition to SST, TB depends on
the sea-surface roughness and on the atmospheric temperature and mois-
ture profile. Fortunately, the spectral and polarimetric signatures of the
surface-roughness and the atmosphere are quite distinct from the SST sig-
nature, and the influence of these effects can be removed given simultane-
ous measurements at multiple frequencies and polarizations. Both TMI
and AMSR-E measure multiple frequencies that are more than sufficient to
remove the surface-roughness and atmospheric effects. Sea-surface
roughness, which is tightly correlated with the local wind, is usually pa-
rameterized in terms of the near-surface wind speed and direction. The
additional 7 GHz channel present on AMSR-E and not TMI, provides im-
proved estimates of sea-surface roughness and improved accuracy for
SSTs less than 12°C (Gentemann et al., in press). All channels are used to
simultaneously retrieve SST, wind speed, columnar water vapor, cloud
liquid water, and rain rate (Wentz and Meissner, 2000). SST retrieval is
prevented only in regions with sun-glitter, rain, and near land. Since only
a small number of retrievals are unsuccessful, almost complete global cov-
erage is achieved daily. Any errors in retrieved wind speed, water vapor,
cloud liquid water can result in errors in retrieved SST.
Buoy measurements from the Tropical Atmosphere Ocean / Triangle
Trans-Ocean Buoy Network (TAO/TRITON) and the Pilot Research
Moored Array in the Tropical Atlantic (PIRATA) are used to validate the
PMW SSTs. Table 3 shows the mean difference, mean satellite minus
buoy SST difference and standard deviation (STD) for each of the buoy ar-
rays. Comparisons with TMI data from 1 January 1998 through 9 June
2005 show that the TAO and PIRATA arrays have very small mean biases,
-0.09 C and –0.09 C, and STD of 0.67 C and 0.60 C respectively. Com-
parisons with AMSR-E data (1 May 2002 through 9 June 2005) show the
TAO and PIRATA arrays have very small biases (-0.03 C and -0.01 C) and
STD (0.41 C and 0.35 C, respectively).
Table 3. Nighttime satellite – buoy SST errors, bias and standard deviation (STD).
TOGA TAO/TRITON PIRATA
Satellite Collocations
Bias STD Collocations
Bias
STD
TMI 84072 -0.09
0.67 11669 -0.09
0.60
AMSR-E 21461 -0.03
0.41 2837 -0.00
0.35
Passive Microwave Remote Sensing of the Ocean: an Overview 35
Sea Surface Salinity
The first measurements of SSS from space will be from the SMOS and
Aquarius. SSS is important to ocean circulation, the global hydrological
cycle, and climate. Monitoring SSS will provide information on geophysi-
cal processes that affect SSS and the global hydrological cycle, such as the
sea ice freeze/thaw cycle, evaporation and precipitation over the ocean,
and land runoff. The Aquarius mission will attempt to measure SSS with a
150 km spatial resolution and a measurement error of < 0.2 PSS-78 (Prac-
tical Salinity Scale of 1978) (Lagerloef et al., 2008).
At 1.4 GHz, retrievals are sufficiently sensitive to SSS to allow for ac-
curate retrieval of SSS. The retrievals depend on the dielectric constant of
sea water, the wind-induced sea-surface emissivity and scattering charac-
teristics, atmospheric absorption, particularly that due to rain, and Faraday
rotation. Additional contributions from near-land emissions, galactic
background radiation reflection, and reflected solar radiation present in-
creased difficulties.
6. Erroneous retrievals
Rain Contamination
The retrievals for SST, wind speed, and vapor must be flagged as bad data
in the presence of rain. This is usually done by looking at the simultane-
ous retrieval of rain rate. Occasionally, sub-pixel rain cells contaminate
these retrievals but are not flagged as rain. These can be seen as anoma-
lously warm or cold SSTs or anomalously high wind values, usually only
affecting 1-2 pixels in a region where other data nearby has been flagged
as rain contaminated. In working with PMW data, area-rain flagging is
necessary to remove these anomalously affected cells near rain. Only then
can climatological results be trusted.
Near land emission
Near land, the lobes to the side of the main beam can result in side-lobe
contamination. This contamination results in geographic dependent re-
trieval errors unless the data are flagged as erroneous. This contamination
impacts all the geophysical retrievals from PMW radiometers to differing
extents depending on the land emission signal at the frequencies included
in the various retrieval algorithms. For example, because the 10.7 GHz
channels is affected more by land emissions, the land contamination at
10.7 GHz is larger than at 6.9 GHz, resulting in a warm bias and small in-
36 C.L. Gentemann et al.
crease in standard deviation for both TMI and AMSR-E measurements
near land, but the effect is larger in the TMI retrievals.
To estimate the side-lobe contamination in the TMI PMW SST retriev-
als we have compared contemporaneous Visible Infrared Radiometer
Scanner (VIRS) IR SST retrievals in coastal regions, using data from
January 1998 through December 1998. VIRS is an infrared radiometer
carried on the TRMM satellite alongside TMI. VIRS SSTs were deter-
mined to have a standard deviation of 0.7 °C when compared to Reynolds
Optimal Interpolated SSTs (Ricciardulli and Wentz, 2004).
Fig. 4. Estimate of bias due to side-lobe contamination near land for 10.7 GHz
SST retrievals.
To investigate how the effect of land contamination on the TMI SSTs
diminishes away from land, the distance from land for each data point was
calculated. The effect of land contamination can be seen in the mean dif-
ference, TMI minus VIRS SST (Figure 4). The mean difference away
from land is roughly 0.12 C, which is approximated by the difference ex-
pected between a skin (VIRS) and subskin (TMI) measurement of SST. As
the distance to land decreases, the mean difference increases, with a
maximum magnitude of 0.72 K, indicating that the bias due to land con-
tamination is on the order of 0.6 K. From Figure 5, it is clear that biasing
exists mostly for retrievals less than 100 - 150 km from land. These results
are specific to the 10.7 GHz SST retrieval from TMI. Although AMSR-E
has land contamination also, the impact is less at 6.9 GHz, the primary
channel used for AMSR-E SSTs.
Passive Microwave Remote Sensing of the Ocean: an Overview 37
Fig. 5. Land contamination bias derived from TMI VIRS comparisons. This
global average shows that by removing data within 100 - 150 km of land, side-
lobe contamination will be removed.
Radio Frequency Interference
RFI is arguably the fastest growing source of errors in microwave SSTs
and wind speeds. The RFI impact on water vapor, cloud liquid water, and
rain rate is less, but growing. RFI errors are largely caused by media
broadcasting activities (including television and radio) from commercial
satellites in geostationary orbits. Geostationary RFI results when signals
broadcast from these communication satellites reflect off the Earth’s ocean
surface into a PMW instrument’s field of view. Ground-based instrumen-
tation in the microwave range is also producing RFI, some sources of
which have been identified and characterized. Both these types of anthro-
pogenic RFI are increasing in magnitude and extent. While it is relatively
straightforward to identify and flag data affected by large RFI contamina-
tion, less-obvious RFI contamination can be difficult to identify. The spa-
tial and temporal nature of the RFI removal must be carefully monitored to
avoid spurious trends in climate data records. The addition of new com-
munication satellites, more power, more ground coverage, and the use of
more frequencies near PMW instrument measurement bandwidths signify
that sources of RFI will continue to change and increase in the future.
The RFI errors resulting from geostationary broadcast sources are pri-
marily dependent on communication broadcast frequency, power and di-
rection, PMW instrument bandwidth, signal glint angle, and ocean surface
roughness. The observation bandwidths of PMW instruments are typically
38 C.L. Gentemann et al.
wider than the protected bands allocated for PMW remote sensing. Thus,
PMW instruments can receive RFI from legal activity using nearby fre-
quency bands allocated for communication and other commercial uses.
AMSR-E and WindSat are the two PMW instruments most affected by
RFI, while SSM/I and TMI both appear to be relatively unaffected. This is
likely because the lower frequency channels of AMSR-E and WindSat,
particularly the 10.7 and 18.7 GHz measurement channels, are sensitive to
frequencies used extensively for media broadcasting. WindSat has more
significant RFI than AMSR-E due to wider observation bandwidth. Ob-
serving more bandwidth tends to yield less noise, but also leads to more in-
terference from frequencies further from the channel’s center observation
frequency. For example, at 18.7 GHz, WindSat receives interference from
DirecTV nationwide broadcast beams. AMSR-E, with narrower band-
width at 18.7 GHz, does not appear to be significantly affected by nation-
wide broadcast frequencies, but does receive RFI from DirecTV spot
beams, which broadcast at frequencies closer to the center observation fre-
quency of the 18.7 GHz channel.
Power and direction are also important factors affecting RFI. Satellite
media broadcasts appear to direct most signal power very carefully to spe-
cific markets. Powerful signals can result in large RFI errors within cer-
tain regions. To serve smaller but growing geographically dispersed mar-
kets, media satellites also broadcast wide, low power beams to cover much
larger, less populated regions. These lower power beams induce more sub-
tle RFI effects that can be difficult to detect and remove. Assuming the
Earth observation point is within the footprint of a geostationary broadcast,
the magnitude of RFI is highly dependent on the glint angle, or how close
the observation reflection vector comes to pointing at the RFI source.
RFI induced errors in AMSR-E ocean products were investigated over
the entire 7 year mission data set. The effects of the different sources of
RFI are listed in Table 4, including which PMW passes are affected and
the time period of interference. Because most geostationary broadcast
power is directed toward the northern hemisphere, many broadcast beams
only reflect into2 the descending pass AMSR-E field-of-view.
From the start of the AMSR-E mission in 2002, HotBird, which is posi-
tioned over 13.0° East longitude, and Astra, located at 19.2° East, have
steadily increased RFI in European waters over time. DirecTV-10 at
102.8° West and DirecTV-11 at 99.2° West have produced RFI in Ameri-
can waters since 2007, and Atlantic Bird 4A at 7.2° West has been con-
tributing to Mediterranean Sea RFI since 2009. Also from beginning of
mission in 2002, SkyBrazil has directed power toward the southern hemi-
sphere, therefore reflecting into ascending passes of ASMR-E and produc-
ing RFI off the coasts of southern Brazil and Argentina.
Passive Microwave Remote Sensing of the Ocean: an Overview 39
Table 4. Sources of RFI
Source
Region affected Frequency
(GHz)
Effect on data
(
decreases)
Affected
overpass
Period of
interference
HotBird Europe 10.7 SSTs
Winds Descending
Pre 2002 – present
Astra Europe 10.7 SSTs
Winds Descending
Pre 2002 – present
Atl.Bird 4A Mediterranean 10.7 SSTs
Winds Descending
Apr 2009 – present
DirecTV-10
USA
18.7
SSTs
Wind
vapor
cloud
rain
Descending
Sep 2007 - present
DirecTV-11
USA
18.7
SSTs
Wind
vapor
cloud
rain
Descending
July 2008 - present
SkyBrazil SE American Coast
10.7 SSTs
Winds Ascending Pre 2002 – present
ground-based
Ascension Island 6.9 SSTs
no wind effect Both Pre 2002 – present
ground-based
Gulf of Aden 10.7 SSTs
Winds Both Mar 2009 - present
ground-based
Coastal Netherlands
Coastal Norway 6.9 SSTs
no wind effect Both 2004 - present
ground-based
Mumbai 6.9 SSTs
no wind effect Both 2003 - present
Ground-based RFI sources are also growing stronger and more numer-
ous over time. Unlike the Geostationary RFI, the ground-based RFI af-
fects both ascending and descending swaths, though to different extents.
This is likely due to differing levels of RFI activity at the AMSR-E local
observation times of 1:30AM or 1:30PM. Although errors caused by these
ground-based sources cover fairly small regions, the size and intensity of
these RFI effects have been increasing over the years. Ground-based RFI
sources can operate intermittently, sometimes even sporadically. The most
prominent regions include coastal Netherlands and Norway, coastal Mum-
bai, the Gulf of Aden through the waters south of Oman, and waters
around Ascension Island.
40 C.L. Gentemann et al.
Fig. 6. RFI induced wind (left) and SST (right) errors shown in descending pass
difference plots for years 1, 3, 5 and 7 of the ASMR-E mission (starting July,
2002, 2004, 2006, 2008) over North America and Europe where the RFI has in-
creased most in coverage and intensity over the years. The striping is caused by
the shifting orbital pattern of the most intense geostationary glint angles.
Regions of RFI are located by differencing AMSR-E SSTs derived us-
ing all SST channels (6.9 GHz – 36.5 GHz) from those derived without 6.9
GHz (10.7 GHz – 36.5 GHz), as well as by differencing winds derived us-
ing all wind channels (10.7 GHz – 36.5 GHz) from those derived without
10.7 GHz (18.7 GHz – 36.5 GHz). An example is shown in Figure 6.
Passive Microwave Remote Sensing of the Ocean: an Overview 41
Since most geostationary sources affect the AMSR-E descending
passes, this plot shows the wind (North America) and SST (Europe) de-
scending orbit difference maps. The wind RFI around North America
caused by DirecTV outlines U.S coastal waters and the Great Lakes (both
pictured), with some subtle effects detected as far as Hawaii and possibly
the Canary Islands off the coast of Africa (neither shown). The SST RFI
around Europe shows consistently increasing extent and intensity over
the years.
A ground RFI source off the Netherland coast has concurrently in-
creased power to become more prominent as seen by the small distinctive
dot forming over the years. The ground source produces SST errors of op-
posite sign compared to the geostationary RFI in the region. In this small
region, two prominent sources of RFI error tend to cancel each other, po-
tentially complicating detection and removal. The striping visible in Fig-
ure 6 is not due to any cross-swath problem with the SSTs or wind speeds,
but is due to the glint angle geometry which results in a heavily stripped
glint angle pattern caused by AMSR-E’s ground track repeat pattern every
233 orbits.
Glint angles and broadcast footprints are together highly predictive of
potential RFI bias. Therefore, to remove RFI errors from the AMSR-E
SST and wind products we calculate the signal glint angles using the longi-
tude of the geostationary orbits. These glint angles, together with analysis
of broadcast footprints, are used to remove retrievals with high probability
of RFI error.
7. Conclusions
PMW retrievals of wind speed, water vapor, cloud liquid water, rain rate,
sea ice, and SST have provided key information for research, climate, and
operational applications. For research and operational applications, the
daily global coverage provided by PMW retrievals are a significant ad-
vance over the pre-satellite era which relied on ship and buoy observa-
tions. For climate monitoring, the careful inter-calibration of the PMW ra-
diometers and consistent (single algorithm) processing of the entire data
set has provided an accurate 22 year time series of PMW retrievals.
42 C.L. Gentemann et al.
Acknowledgements
The AMSR-E SSTs are from Remote Sensing Systems, processed using
the version 5 algorithm, and available at www.remss.com. This work was
funded by the NASA grants NNG04HZ29C, NNG07HW15C,
NNH08CC60C, and NNH09CF43C.
References
Atlas RM, Hoffman RN, Ardizzone J, Leidner SM, Jusem JC (2009) Development
of a new cross-calibrated, multi-platform (CCMP) ocean surface wind
product. Paper presented at AMS 13th Conference on Integrated Observing
and Assimilation Systems for Atmosphere, Oceans and Land Surface (IOAS-
AOLS), Phoenix, AZ, USA
Bettenhausen MH, Smith CK, Bevilacqua RM, Wang NY, Gaiser PW, Cox SK
(2006) A Nonlinear Optimization Algorithm for WindSat Wind Vector
Retrievals. IEEE Trans Geosci Rem Sen 44(3):597-608
Bowman KP., Homeyer CR, Stone DG (2009) A Comparison of Oceanic
Precipitation Estimates in the Tropics and Subtropics. Journal of Applied
Meteorology and Climatology, 48:1335-1344
Cavalieri DJ, Gloersen P, Campbell JW (1984) Determination of Sea Ice
Parameters With the Nimbus 7 SMMR. J Geophys Res 89:5355-5369
Cecil DJ, Wingo M (2009) Comparison of TRMM Rain-Rate Retrievals in
Tropical Cyclones. J Meteorol Society Japan 87A:369-380,
doi:10.2151/jmsj.87A.369
Comiso JC, Cavalieri DJ, Parkinson CL, Gloersen P (1997) Passive Microwave
Algorithms for Sea Ice Concentration: a Comparison of Two Techniques.
Rem Sens Environ, 60(3):357-384
Comiso JC, Steffen K (2001) Studies of Antarctic Sea Ice Concentrations From
Satellite Data and Their Applications. J Geophys Res 106(C12):31361-31385
Gaiser PW, St.Germain KM, Twarog E, Poe G, Purdy W, Richardson D,
Grossman W, Jones WL, Spencer D, Golba G, Cleveland J, Choy L,
Bevilacqua RM, Chang P (2004) The WindSat Spaceborne Polarimetric
Microwave Radiometer: Sensor Description and Early Orbit Performance.
IEEE Trans Geosci Rem Sen 42(11):2347-2361
Gentemann CL, Meissner T, Wentz FJ (2009) Accuracy of satellite sea surface
temperatures at 7 and 11 GHz. IEEE Trans Geosci Rem Sen, in press
Gloersen P, Cavalieri DJ, Chang ATC, Wilheit TT, Campbell WJ, Johannessen
OM, Katsaros KB, Kunzi KF, Ross DB, Staelin D, Windsor EPL, Barath FT,
Gudmandsen P, Langham E, Ramseier RO (1984) Summary of Results From
the First NIMBUS 7 SMMR Observations. J Geophys Res 89:5335-5344
Passive Microwave Remote Sensing of the Ocean: an Overview 43
Han D, Kim ST (1988) Effects of switch leakages upon Nimbus-7 SMMR
calibration. 100705, pp. 17, Goddard Space Flight Center, Greenblet, MD,
USA
Haug GH, Gunther D, Peterson LC, Sigman DM, Hughen KA, Aeschlimann B
(2003) Climate and the Collapse of Maya Civilization. Science 299:1731-
1735
Held IM, Soden BJ (2006) Robust Responses of the Hydrological Cycle to Global
Warming. J Climate 19:5686-5699
Hilburn KA, Wentz FJ (2008) Intercalibrated Passive Microwave Rain Products
From the Unified Microwave Ocean Retrieval Algorithm (UMORA). Journal
of Applied Meteorology and Climatology, 47):778-794
Hollinger J, Peirce JL, Poe G (1990) SSM/I Instrument Evaluation. IEEE Trans
Geosci Rem Sens 28(5):781-790
Kummerow CD, Barnes W, Kozu T, Shieu J, Simpson JJ (1998) The Tropical
Rainfall Measuring Mission (TRMM) Sensor Package. J Atmos Oceanic
Technol 15(3):809-817
Lagerloef GSE, Colomb FR, Le Vine D, Wentz FJ, Yueh SH, Ruf CS, Lilly J,
Gunn Y, Chao A, deCharon A, Feldman G, Swift C (2008) The
Aquarius/SAC-D Mission: Designed to Meet the Salinity Remote-Sensing
Challenge. Oceanography 21:68-81
Markus T, Cavalieri DJ (2000) An Enhancement of the NASA Team Sea Ice
Algorithm. IEEE Trans GeoscI Rem Sens 38:1387-1398
Mears CA, Smith DK, Wentz FJ (2001) Comparison of Special Sensor Microwave
Imager and Buoy-Measured Wind Speeds From 1987 – 1997. J Geophys Res
106(C6):11719-11729
Meissner T, Smith DK, Wentz FJ (2001) A 10-Year Intercomparison Between
Collocated Special Sensor Microwave Imager Oceanic Surface Wind Speed
Retrievals and Global Analyses. J Geophys Res 106(C6):11731-11742
Meissner T, Wentz FJ (2009) Wind Vector Retrievals Under Rain With Passive
Satellite Microwave Radiometers. IEEE Trans Geosci Rem Sens 47(9):3065-
3083
O'Conner T, Kiker GA (2004) Collapse of the Mapungubwe Society:
Vulnerability of Pastoralism to Increasing Aridity. Climatic Change 66:49-66
Parkinson CL (2003) Aqua: An Earth-Observing Satellite Mission to Examine
Water and Other Climate Variables. IEEE Trans Geosci Rem Sens 41(2):173-
183
Petty GW (1994) Physical Retrievals of Over-Ocean Rain Rate From
Multichannel Microwave Imagery. Part I: Theoretical Characteristics of
Normalized Polarization and Scattering Indices. Meteorol Atmos Physics
54:79-99
Ricciardulli L, Wentz FJ (2004) Uncertainties in Sea Surface Retrievals From
Space: Comparison of Microwave and Infrared Observations From TRMM. J
Geophys Res 109 C12013 doi:10.1029/2003JC002247
Szczodrak M, Minnett PJ, Gentemann CL (2006) Comprison of AMSR-E
retrievals of total water vapor over ocean. paper presented at 2006 Ocean
Sciences Meeting, Honolulu, HI, USA
44 C.L. Gentemann et al.
Trenberth KE, Fasullo J, Smith L (2005) Trends and Variability in Column-
Integrated Atmospheric Water Vapor. Climate Dynamics 24:741-758
Wentz FJ (1997) A Well Calibrated Ocean Algorithm for Special Sensor
Microwave / Imager. Journal of Geophysical Research 102(C4):8703-8718
Wentz FJ (1998) AMSR Ocean Algorithm. 110398, pp. 66, Remote Sensing
Systems, Santa Rosa, CA
Wentz FJ, Meissner T (2007) AMSR_E Ocean Algorithms. 051707, pp. 6, Remote
Sensing Systems, Santa Rosa, CA, USA
Wentz FJ, Ricciardulli L, Hilburn KA, Mears CA (2007) How Much More Rain
Will Global Warming Bring? Science 317:233-235
Wentz FJ, Schabel MC (2000) Precise Climate Monitoring Using Complementary
Satellite Data Sets. Nature 403(6768):414-416
Wentz FJ, Spencer RW (1998) SSM/I Rain Retrievals Within a Unified All-
Weather Ocean Algorithm. J Atmos Sci 55(9):1613-1627
