Content uploaded by Kenneth S Casey
Author content
All content in this area was uploaded by Kenneth S Casey
Content may be subject to copyright.
323
V. Barale, J.F.R. Gower and L. Alberotanza (eds.), Oceanography from Space, revisited.
© Springer Science+Business Media B.V. 2010
The Past, Present, and Future of the AVHRR
Pathfinder SST Program
Kenneth S. Casey
1
, Tess B. Brandon
1
, Peter Cornillon
2
, and Robert Evans
3
1
NOAA National Oceanographic Data Center, Silver Spring, MD, USA
2
University of Rhode Island Graduate School of Oceanography, Narragan-
sett, RI, USA
3
University of Miami Rosenstiel School of Marine and Atmospheric Sci-
ence, Miami, FL, USA
Abstract. With origins dating back to 1990, the Advanced Very High
Resolution Radiometer (AVHRR) Pathfinder Sea Surface Temperature
(SST) Program has experienced a 20-year history of reprocessing space-
based observations to create accurate, consistent, climate data records.
Both scientific and programmatic aspects of this history are reviewed and
summarized in this paper, along with a review of the currently available
Pathfinder SST data. In addition, a look forward to the next generation of
Pathfinder currently under development is presented.
1. Introduction
With today’s modern satellite sensors, many oceanic parameters are capa-
ble of being observed from space. Of these, only sea surface temperature
(SST) from the Advanced Very High Resolution Radiometer (AVHRR)
series on board the NOAA polar-orbiting satellites enjoys an unbroken,
nearly 30-year long history of observation from the same class of instru-
ment. These AVHRR observations date back to 1981, when the first five-
channel AVHRR flew on board the NOAA-7 platform. The AVHRR se-
ries continues to fly today, and will continue perhaps another ten years into
the future on board the remaining NOAA polar orbiting satellites and the
European METOP platforms.
Early in the life of this series of instruments, NASA Headquarters con-
ceived the Pathfinder program to support the needs of the U.S. Global
Change Research Program (King and Greenstone, 1999). These Path-
finder efforts were to provide unprecedented access to large volumes of
consistently processed satellite datasets, in advance of the launch of the
The original publication is available at www.springerlink.com.
(http://www.springerlink.com/content/t2l4275xk08t5m51/)
Casey,K.S., T.B. Brandon, P. Cornillon, and R. Evans (2010).
"The Past, Present and Future of the AVHRR Pathfinder SST
Program", in Oceanography from Space: Revisited, eds. V.
Barale, J.F.R. Gower, and L. Alberotanza, Springer. DOI:
10.1007/978-90-481-8681-5_16.
324 K.. S. Casey et al.
NASA Earth Observing System (EOS) satellites. Climate-quality datasets
would be generated and made available, providing valuable experience in
reprocessing, archiving, and distributing large satellite data sets. Thus,
these Pathfinder efforts would “find the path” forward in preparation for
the ambitious EOS series of instruments and missions that were to follow.
In October of 1990, NOAA and NASA signed a Cooperative Agree-
ment to commence work on three joint NOAA/NASA Pathfinder projects.
These joint projects would use existing NOAA datasets, focusing on
AVHRR Global Area Coverage (GAC) data, the Television and Infrared
Observation Satellite (TIROS) Operational Vertical Sounder (TOVS) data,
and the Geostationary Operational Environmental Satellite (GOES) data.
Several months later, in 1991, a project involving Special Sensor Micro-
wave/Imager (SSM/I) data, which was archived by NOAA under an
agreement with the U.S. Navy, became the fourth NOAA/NASA Path-
finder. In 1992 a NASA/EPA/USGS Landsat Pathfinder project was
formed, as was the first NASA-only Pathfinder, using Scanning Mul-
tichannel Microwave Radiometer (SMMR) data. Funding for these ac-
tivities came mainly from NASA.
Each of the Pathfinder projects would form a Science Working Group
(SWG) that would determine the specific products to be created based on
scientific need, identify community-consensus algorithms to generate
those products, and make recommendations on product validation, storage,
maintenance, and required data services. A NASA Distributed Active Ar-
chive Center (DAAC) would then make the data available, as would the
EOS Data and Information System (EOSDIS).
Because the AVHRR data stream was perceived to be of value to sev-
eral different areas in the Earth sciences, three product science working
groups were formed under the AVHRR SWG umbrella: an atmosphere
SWG, a land SWG and an oceans SWG (OSWG in the following). The of-
ficial goals of the AVHRR Pathfinder projects, as presented on 08 January
1992 by Mary James of NASA Goddard Space Flight Center to a group
that included the project chairpersons, were:
• To produce long term global datasets for research, modeling, and
trend analysis. Datasets were to be consistently processed with the best
available community consensus algorithms.
• To produce multiple geophysical products from a common input
stream using integrated processing concepts.
The Past, Present, and Future of the AVHRR Pathfinder SST Program 325
• To assemble a consistent, low maintenance, readily accessible ar-
chive for AVHRR data with browse and on-line access capabilities.
In December of 1994 a Pathfinder NASA Research Announcement was
released, marking the transition from the first phase to a second phase of
the Pathfinder datasets. The NRA emphasized building long-term datasets
and addressing consistency issues that arise when linking together multiple
sensors over time. This second phase ended in 2001, which marked the
formal end of NASA support for the NOAA/NASA Pathfinder efforts.
However, because of the successes obtained and momentum built over the
previous decade, some of these efforts continued operating under different
mechanisms. For example, the AVHRR Pathfinder Atmosphere
(PATMOS) project continued as a NOAA National Environmental Satel-
lite Data and Information Service (NESDIS) effort (Stowe et al., 2002).
The OSWG met for the first time in February of 1991. As with the
PATMOS project, the Pathfinder SST activities also continued with the
goal of building on the successes of the previous decade. In the case of
Pathfinder SST, the program continued through the efforts of the
NOAA/NESDIS National Oceanographic Data Center (NODC), beginning
in 2002. Details on the history, current status, and future directions of
Pathfinder SSTs constitute the remainder of this paper.
2. History of the Pathfinder SST Program
Reprocessing of AVHRR data to create long, accurate, and consistent SST
records was one component of the overall NOAA/NASA Pathfinder Pro-
gram. To remain consistent with terminology, the activities that took place
between 1990 and 2001 are referred to as the “NOAA/NASA AVHRR
Oceans Pathfinder SST Project.” During that period, NASA was the pri-
mary sponsor of the effort. Beginning in 2002 NODC became involved
and a variety of NOAA funding sources began supporting the efforts. Be-
ginning in 2009, new NASA funding supported some specific aspects of
the program. The overall set of activities, including both the early 1990-
2000 and the 2002 and beyond periods is termed simply the “Pathfinder
SST Program” (see Table 1).
The AVHRR Ocean Pathfinder SST Project was an outgrowth of the
work of the NASA funded SST Archive Science Working Group
(SASWG) formed in early 1987. The objective of this group was “to de-
termine the needs of the scientific community for SST fields and the pos-
sibility of meeting these needs with existing sources of data.” The SASWG
326 K.. S. Casey et al.
met several times over the next two years and on 1 June 1989 delivered its
final report detailing two useful classes of SST products and the steps nec-
essary to produce these fields. One class of products, with fine spatial
resolution and high relative accuracy, would address the needs of feature-
related studies and the other, with high absolute accuracy, would address
the needs of heat flux related studies. The SASWG final report identified
the AVHRR GAC data stream as appropriate for the generation of both
SST products outlined in the report.
Table 1. Summary of Overall Pathfinder SST Program Periods.
The Pathfinder SST Program
NOAA/NASA AVHRR
Oceans Pathfinder SST
Project (Versions 1-4)
NOAA-supported Pathfinder
SST Program (Version 5)
Version 6
(in preparation)
1990-2001 2002-2008 2009-2010
Following the SASWG report, Stan Wilson (then of NASA Headquar-
ters but currently with NOAA/NESDIS) assembled a subgroup of the
SASWG plus several others within NASA with expertise in the processing
of GAC data to explore the actual production of the desired fields. This
group included a representative from the terrestrial science community in
addition to those from the oceanographic community. Two observations
emerged from this meeting: first, much of the required GAC data existed
only at NOAA so a collaboration with NOAA would be required as part of
any reprocessing effort and second, other Earth science communities could
benefit from ready access to a complete global GAC data set. Based on
these observations and the desire to learn how to handle large, multidisci-
plinary, open-ended data sets for the upcoming EOSDIS, Stan Wilson ini-
tiated a data transcription effort and a Cooperative Agreement with
NOAA. The Cooperative Agreement, entitled “Early EOSDIS Pathfinder
Data Set Activity”, was signed on 15 October 1990. This agreement was
the formal start of the Pathfinder Program, as described in Section 1.
However, despite the Cooperative Agreement, the available funding,
and the desire of the NOAA staff (Bud Booth and Levin Lauriston) to sup-
port the nascent Pathfinder group with the data transcription, little progress
was made over the next 13 months. To determine the nature of the prob-
The Past, Present, and Future of the AVHRR Pathfinder SST Program 327
lem, in November 1990 Peter Cornillon and Otis Brown met with John
Knauss, who was the NOAA Administrator at that time. In that meeting it
became clear that the delays resulted form a misunderstanding within
NOAA of a comment made by Knauss. When shown what NASA was do-
ing with satellite data, Knauss made a statement to the effect, “Why can’t
NOAA do these things?” This statement was misinterpreted by NOAA
staff to mean that the data were not to be provided to others and that
NOAA was to undertake the processing steps. Knauss had not intended his
statement to be interpreted this way and clarified the situation with his
staff after the meeting with Cornillon and Brown. The data transcription
began in earnest and progress continued steadily on the Pathfinder Pro-
gram from that point forward.
The Pathfinder OSWG first met in February of 1991 at Goddard Space
Flight Center in Greenbelt, Maryland, USA. Peter Cornillon of the Uni-
versity of Rhode Island chaired the OSWG, which included members from
NOAA, NASA, and universities (see Appendix). The OSWG met ap-
proximately annually over the next four years continually evolving and
improving the algorithms and reprocessing techniques. During this same
period, the chairmen of the Pathfinder groups met approximately annually
to coordinate the work of the different Pathfinder projects and subgroups
met when needed to deal with data transcription and calibration issues. To
give a sense of the technical challenges faced by the OSWG to simply ma-
nipulate the large volumes of data, Figure 1 shows the first part of an e-
mail sent to Cornillon on 22 February 1991 containing a draft report from
the first OSWG meeting. Note the sender’s concern that the text might be
garbled because he had sent it at 9600 baud, which was considered to be
very fast at the time. Much of the information emerging from the various
groups was distributed by fax or over the Internet at extremely slow data
rates by today’s standards.
Sea surface temperature retrieval algorithms proposed by the OSWG
were tested and implemented under the direction of Robert Evans at the
University of Miami where the data were processed from Level 0
(L0)/Level 1b (L1b) through to Level 3 (L3) equal-area datasets on an in-
tegerized sinusoidal grid. These data were then sent to the NASA/Caltech
JPL Physical Oceanography Distributed Active Archive Center
(PO.DAAC) where they were converted to equal-angle, regularly gridded
datasets in HDF4-Raster format and provided to users. The first four ver-
sions of Pathfinder SST were created at a reduced resolution of 9.28 km,
due to computational resource limitations of the 1990s.
NASA funding for the initial effort ended in 2001-2002 and in 2002
NODC reinvigorated the Pathfinder SST activities with the goal of produc-
328 K.. S. Casey et al.
ing a Version 5 dataset at the full GAC resolution. Modest levels of
NOAA funding from various sources including NODC base resources and
the NOAA Coral Reef Conservation Program kept the Pathfinder SST
Program alive from 2002 through 2006. During this period, the Version 5
data were produced at the University of Miami and then sent to NODC for
final quality assurance, archiving, and provision to users. NODC provided
the Version 5 data to the PO.DAAC as well, to maintain continuity for us-
ers of earlier versions of the Pathfinder SST data who had grown accus-
tomed to accessing it from the PO.DAAC systems. From 2007 through
2009, the NOAA Scientific Data Stewardship Program supported efforts to
transition the reprocessing from the University of Miami to NODC, where
it could be supported over the long term.
Fig. 1: Excerpt from a 22 February 1991 email to Peter Cornillon, demonstrating
the slow data transmission rates available at the time. Despite the sender’s con-
cerns, the transmission was not garbled.
The Past, Present, and Future of the AVHRR Pathfinder SST Program 329
Beginning in 2009, new NASA funding augmented the continued
NOAA support, with the goal of adding two new components to the Path-
finder SST Program. The first addition was the ability to generate error es-
timates, a critical component necessary to make the future Pathfinder Ver-
sion 6 data compliant with the standards of the Group for High Resolution
SST (GHRSST, Donlon et al., 2007). The second addition was to enable
Pathfinder reprocessing of not just AVHRR GAC data, but also a large
collection of 1 km resolution Local Area Coverage (LAC) and High Reso-
lution Picture Transmission (HRPT) data from the AVHRR instruments.
By the end of 2010, the GAC reprocessing system is expected to be func-
tional at NODC and the LAC/HRPT at the University of Rhode Island.
Long-term stewardship, archiving, and provision of the Version 6 data to
users will take place at NODC.
3. Earlier Pathfinder SST Versions
Over its twenty-year history, the Pathfinder SST Program has produced
five distinct versions and is actively developing the sixth. The Pathfinder
SST algorithm is based on the Non-Linear SST (NLSST) algorithm
(Walton et al., 1998) and was originally chosen by the SWG because of
the algorithm’s adequate performance, operational nature, and its accep-
tance by the scientific user community (Kilpatrick et al., 2001). The basic
form of the algorithm is
SST
sat
= a + bT
4
+ c(T
4
-T
5
)SST
guess
+ d(T
4
-T
5
)[sec(
θ
) - 1],
(1)
where
SST
sat
= the satellite-derived SST estimate,
T
4
and T
5
= brightness temperatures in the 10.8 and 11.4
µ
m
AVHRR bands (channels 4 and 5, respectively),
SST
guess
= a first-guess SST value,
θ
= the satellite zenith angle, and
a, b, c, and d = coefficients estimated from regression between
collocated and coincident in situ and satellite measurements.
Numerous modifications were made over time to improve the perform-
ance of the algorithm. A review of the algorithm evolution from Version 1
330 K.. S. Casey et al.
through 4 is provided below, based on information provided in an early
version of the Pathfinder Users Manual (Vazquez et al., 1995) and Casey
and Cornillon (1999). For more details, with an emphasis on Pathfinder
Version 4, see Kilpatrick et al. (2001). Version 5 and Version 6 are dis-
cussed in more detail in the following sections.
Version 1
The first version of Pathfinder implemented several key enhancements to
the NLSST algorithm. The form of the NLSST equation was slightly
modified to include a time dependent term. Extra care was taken with the
first steps in which the digital counts are converted to radiance and bright-
ness temperatures. The first of these steps involves the linear transforma-
tion of counts to radiance based on the space-view and sensor base plate
onboard calibration information. Next, a non-linear correction factor was
applied based on pre-launch calibration data. Finally, lookup tables based
on the sensor’s operating temperature were used to convert the channel 4
and 5 radiances to brightness temperatures.
Pathfinder Version 1 also included improvements to the navigation of
the satellite observations over previous AVHRR SST data products. These
improvements focused on the clock drift and spacecraft and sensor atti-
tude. Clock drifts caused uncertainties in the along-track position and
were corrected using a database of satellite clock time and Earth time off-
sets determined by comparing precise time measurements recorded at the
University of Miami Domestic Communication Satellite (DOMSAT, the
communication satellite NOAA used to transfer AVHRR data) station with
the time information in HRPT passes received at Wallops Island, VA, and
Fairbanks, AK. Averaging over multiple geographic locations and times
along the orbital track mitigated attitude errors.
Further improvements introduced in Pathfinder Version 1 were imple-
mented during the next phase of reprocessing, in which SST values are
calculated. Three sets of algorithm coefficients (a, b, c, and d in Eq. 1)
were generated using the satellite-in situ matchup database: one for high
atmospheric water vapor regimes where the T
4
– T
5
brightness temperature
difference is greater than or equal to 1.8 ºC, one for low water vapor re-
gimes where the T
4
– T
5
brightness temperature difference is less than 0.7
ºC, and one for moderate levels of water vapor where the T
4
– T
5
bright-
ness temperature difference is greater than or equal to 0.7 ºC and less than
1.8 ºC. In Version 1, the coefficients were calculated on an annual basis,
using 12 months of matchup data.
Finally, an extensive set of quality tests were implemented in Version 1,
including checks for gross cloud contamination, uniformity tests, satellite
The Past, Present, and Future of the AVHRR Pathfinder SST Program 331
zenith angle tests, stray sunlight tests, and a check against a reference SST
field based on the Reynolds Optimally Interpolated SST (OISSTv1; Rey-
nolds and Smith, 1994). Pixels on the edges of a scan line or in the first or
last scan line of an orbital piece were excluded, and those subject to sun
glint were also identified. Based on the combined results of these many
tests, Version 1 Pathfinder SSTs were then assigned a quality level of be-
tween 0 (worst) and 4 (best). Pathfinder Version 1 data spanning 1987 to
1993 were released to the public beginning in 1995.
Version 2
A second version of Pathfinder was developed to better correct for tempo-
ral changes observed in the match-up statistics. The time-dependent term
was removed and coefficients were generated on a monthly instead of an-
nual basis. The coefficients were calculated using a temporally weighted
five-month running window and were found to better correct the temporal
changes than the time-dependent term used in the NLSST equation in Ver-
sion 1. The central month is given a 100% weight, the adjacent months an
80% weight, and the months at the ends of the window a 50% weight. This
version of the Pathfinder data set was never released to the public.
Version 3
In the third version of Pathfinder SST, two sets of algorithm coefficients
(a, b, c, and d in Eq. 1) were generated using the satellite-in situ matchup
database instead of the three sets used in Version 1. The two sets were
found to better reduce the overall bias between satellite and buoy SST.
One set was for high atmospheric water vapor regimes where the T
4
– T
5
brightness temperature difference is greater than or equal to 0.7 ºC and the
other was for low water vapor regimes where the T
4
– T
5
brightness tem-
perature difference is less than 0.7 ºC. To avoid discontinuities, in cases
where the T
4
– T
5
brightness temperature difference was between 0.5 ºC
and 0.9 ºC the SST was calculated using a weighted combination of SST
values determined using both sets of coefficients. Pathfinder Version 3
data were first released to the public in 1997 and spanned only 1991-1994.
They were quickly superseded by Pathfinder Version 4, which was re-
leased only a short time later.
Version 4
In Version 4, a more robust algorithm coefficient scheme was introduced,
and decision trees were implemented to better eliminate cloud-
332 K.. S. Casey et al.
contaminated pixels. Quality levels were expanded to 0 to 7 to allow for
greater user flexibility in determining trade-offs between data coverage
and data quality. Version 4 data became available in 1997 and eventually
covered 1985 to 2003. The Version 4 data marked the culmination of the
NOAA/NASA AVHRR Oceans Pathfinder period.
4. Current State of Pathfinder SST: Version 5
Beginning in 2002, the Pathfinder SST Program entered a new phase tar-
geted at producing an even more accurate, consistent, and finer resolution
SST data set. By this time, the term “Climate Data Record” had emerged
and the Pathfinder SST Program renewed its efforts toward delivering the
SST climate data record. Improvements in spatial resolution, the land
mask, and the way sea ice information was used in the quality level deter-
minations were implemented. The 1-degree resolution OISSTv1 first-guess
and reference SST field used in earlier Pathfinder versions was replaced
with a newer version of the product, the OISSTv2 (Reynolds et al., 2002)
For the earliest part of the AVHRR record from late 1981 through 1984, a
new 25 km Daily Optimally Interpolated SST (DOISST, Reynolds et al.,
2007) was used. A more quantitative analysis of the improvements
achieved along with a description and evaluation of new SST climatolo-
gies derived from Pathfinder Version 5 is available in Casey et al. (2009).
By 2009, production, distribution, and long-term stewardship of the
Pathfinder Version 5 SST data had become a joint University of Miami
and NODC effort. The University of Miami processed the data from
L0/L1b to L3, including a transformation from the integerized sinusoidal
grid to the regularly gridded data in HDF4-Scientific Data Set format.
These data were then transferred to NODC, where they were subjected to
additional quality assurance involving browse image generation, visual in-
spection of every image, and comparison against the HadSST2 (Rayner et
al., 2006) in situ dataset. Any problems identified were then analyzed, and
corrections implemented at the University of Miami where the data were
regenerated before re-inspection at NODC. Once cleared through the
quality assurance steps, the data were then archived at NODC and pro-
vided to users through its http, ftp, and OPeNDAP services. Access to the
entire Pathfinder Version 5 collection is available on line
1
. An example
image, generated from a climatological mean of Pathfinder Version 5 SST,
is shown in Figure 2.
1
at
http://pathfinder.nodc.noaa.gov
The Past, Present, and Future of the AVHRR Pathfinder SST Program 333
Fig. 2. Climatological mean SST for Week 50, based on Pathfinder Version 5
data.
The following steps provide the details on the current state of the Path-
finder processing system. Boldface items highlight the data and informa-
tion provided to the processing system.
Step 1:
Ingestion, calibration, and navigation of GAC data
1.1. Calibrate and convert AVHRR digital counts for channels 1 through 5
to radiances
1.1.1. Obtain AVHRR channels 1 through 5 radiometer count
data from NOAA and University of Miami collections of GAC
data.
334 K.. S. Casey et al.
1.1.2. For channels 1 and 2, use pre-launch calibration coefficients
to perform a linear counts-to-radiance conversion, followed by
a correction for temporal changes using sensor decay rate
data and then a correction for inter-satellite differences using
inter-satellite standardization data to a NOAA-9 reference,
both of which use Libyan desert target area data.
1.1.3. For channels 3, 4, and 5 use both the above pre-launch calibra-
tion data and onboard blackbody (space view and sensor
base plate) data to perform a non-linear counts-to-radiance
conversion.
1.2. Navigation, Clock, and Attitude Corrections
1.2.1. Apply satellite clock corrections using Earth time offset data
based on University of Miami DOMSAT recorded HRPT files.
1.2.2. Apply attitude corrections made using coastline comparison
data.
1.2.3. At this point, navigated, calibrated albedos/brightness tempera-
tures are available for all five channels. Note that channels 1
and 2 are used in the Pathfinder Matchup Database decision
trees [see Kilpatrick et al. (2001), Figure 7] and channel 3 is
used only in assignment of a quality indicator (see step 2.4.1).
Step 2:
SST Calculation
2.1. Convert channel 4 and 5 brightness temperatures to SST in ºC using
the Pathfinder algorithm (Equation 1), which requires a set of
monthly coefficients.
2.2. These coefficients are derived using the Pathfinder Buoy Matchup
Database, a set of in situ SST observations and collocated AVHRR
data. The in situ data consist mainly of drifting buoys, but during
1981-1984 bias-corrected ship-based observations are also used.
2.3. The calculation of SST in step 2.1 also requires a first-guess SST
field. This first-guess field is the OISSTv2 product for 1985-2009
and the DOISST for 1981-1984. Note that Versions 1-4 of Path-
finder used OISSTv1.
2.4. Quality Flag Assignment
2.4.1. A Channel 3, 4, and 5 brightness temperature test is per-
formed. The brightness temperatures were calculated in step
1.1.3.
2.4.2. The viewing angle is then evaluated using a satellite zenith
angle check.
The Past, Present, and Future of the AVHRR Pathfinder SST Program 335
2.4.3. Next, a reference field comparison check is made against the
OISSTv2/DOISST used in step 2.3.
2.4.4. A stray sunlight test is then performed which requires infor-
mation on whether the data in question are to left or right of
nadir.
2.4.5. An edge test is performed next, which checks the location of
the pixel within a scan line and the location of the scan line
within the processing piece (a “piece” is a subset of an entire
orbit file).
2.4.6. Then, a glint test is performed which requires a glint index
calculated according to the Cox and Munk (1954) formulation.
2.4.7. Finally, these steps are all combined into an overall quality
flag assignment for each pixel.
Step 3:
Spatial Binning
3.1. The GAC pixels, converted to SST in Step 2, are then binned into an
equal-area integerized sinusoidal grid.
3.2. To reduce discontinuities along the date line, a data-day based on a
spatial data-day definition (Podestá, 1995) is used.
3.3. A land mask is applied to the data, identifying pixels that fall on
land. In Pathfinder Version 5, an improved mask based on a 1 km
resolution Moderate Resolution Imaging Spectroradiometer
(MODIS) dataset (MOD12Q1) derived by the USGS Land Processes
Distributed Active Archive Center is used
2
. In earlier versions of
Pathfinder, the land mask was based on a Central Intelligence
Agency database of navigation hazards, but no citation or further in-
formation is available.
Step 4:
Temporal Binning
4.1. Step 4 begins by temporally binning the spatially binned pieces from
Step 3 into a single daytime and single nighttime file for each day.
In the case of overlapping satellite passes, only the highest quality
pixels available for each are used. If there is more than one contrib-
uting pixel at the highest quality, these values are averaged. In Path-
finder Version 5, temporally binned files are also created for 5-day,
7-day, 8-day, monthly, and yearly periods.
2
see
http://www-modis.bu.edu/landcover/userguidelc/lc.html for more info
336 K.. S. Casey et al.
4.2. An internal Pathfinder reference check comparison is then made to
an internal 3-week Pathfinder comparison field. A sea ice mask
based on weekly SSM/I and sea ice information in the OISSTv2
is used to exclude pixels from the computation of the internal refer-
ence field. Note that sea ice information is not used in this way in
the earlier versions of Pathfinder.
4.3. The SST, quality, and related fields are reformatted from equal-area
to equal-angle for distribution and archiving in HDF4-Scientific
Data Set format. Older Pathfinder versions were distributed in
HDF4 Raster format.
Step 5: Quality Assurance, Archiving, and Distribution
5.1. Utilizing checksums to ensure file integrity, the HDF4-SDS files are
then acquired by NODC from the University of Miami and browse
graphics are generated.
5.2. Each browse graphic is then visually inspected and problematic data
are sent back to the University of Miami for correction and reproc-
essing.
5.3. After passing the visual inspection, an in situ comparison against the
HadSST2 data set (Rayner et al., 2006) is conducted. Again, any er-
rors detected are investigated and, if the data are subjectively con-
firmed to be bad, they are returned to the University of Miami for
correction and reprocessing.
5.4. After passing the visual inspection and in situ comparisons, the Path-
finder data are formally archived by NODC.
5.5. The result of these steps is the Pathfinder Version 5 SST product.
The data are distributed by NODC
3
and made available to the
PO.DAAC for redistribution.
Pathfinder Version 5 data were first released by NODC to the public in
April of 2003. By December of 2003, data for 1985-2001 were available.
Over the next several years, additional data were added incrementally. In
April of 2009, Pathfinder data for 1981-1984 were made available for first
time in the Pathfinder SST Program history. In prior versions, the lack of
adequate in situ matchups and suitable first-guess SST field made produc-
tion of those early data from the NOAA-7 platform impossible. However,
these limitations were overcome through the use of not just drifting buoys
in the matchup database, but also by including bias-corrected ship observa-
tions. In addition, working iteratively, the DOISST, which itself uses
3
http://pathfinder.nodc.noaa.gov
The Past, Present, and Future of the AVHRR Pathfinder SST Program 337
Pathfinder SST data, was extended to include 1981-1984. The DOISST
then served as the first guess and reference SST field for this period. By
2010, Pathfinder Version 5 data were available for 1981-2009.
5. Future Directions in Pathfinder SST: Version 6
Despite the improvements implemented over the years, problems still re-
main in Pathfinder Version 5. These errors are being corrected, and im-
plemented in a new, Version 6 of Pathfinder. Version 6 will improve upon
Version 5 in several notable ways. First, known errors in the land mask
will be corrected, and Version 6 will utilize the DOISST throughout the
record. Use of the finer resolution DOISST is resulting in improved abil-
ity of the quality procedures to properly identify cloudy pixels and to re-
tain good pixels in the vicinity of coasts and strong gradient regions. Sea
ice information will also be used to a greater extent in Version 6, and will
be provided as part of the data set. Many spatial and seasonally varying
biases evident in the Pathfinder Version 5 comparisons with HadSST2 will
also be minimized through the use of new coefficients generated using a
latitudinal band scheme.
Another problem in Version 5 manifests itself as a jump in SST every
18 lines in the regularly gridded data. While not clearly visible in a single
image, edge detection algorithms that examine sequences of data identified
this problem. The problem is related to the mapping procedures by which
the equal-area integerized sinusoidal bins in which the data are generated
are mapped into the equal-angle, uniform grid space in which the data are
distributed. This problem has been resolved through the use of a refined
mapping procedure that will be used in Version 6.
Significantly, Version 6 will also conform to the newest GHRSST data
format, data content, and metadata requirements currently being devel-
oped, known as the GHRSST Data Specification Version 2 (GDS v2). The
GDS requires each SST pixel value to have an associated bias and uncer-
tainty error. These error estimates are being developed for Pathfinder Ver-
sion 6 using an error hypercube approach which partitions the matchup da-
tabase into a multi-dimensional lookup table. Pathfinder Version 6 will
also be in netCDF and will contain Climate and Forecast metadata attrib-
utes. For the first time, Pathfinder SST data will be made available in not
just collated L3 files, but also at the L2 swath and uncollated L3 process-
ing levels.
Another addition to Version 6 will be to include a large collection of
HRPT and LAC data collected at stations around the world. These full-
338 K.. S. Casey et al.
resolution, 1 km data sets will provide key inputs to ultra-high resolution
analysis systems and other fine scale applications. HRPT passes covering
the east and west coasts of the United States, the western Pacific off of Ja-
pan, and the waters around Australia, have already been identified for in-
clusion into the Pathfinder Version 6 processing system and other HRPT
station data are being actively sought.
6. Conclusion
For 20 years, the ongoing efforts of the AVHRR Pathfinder SST Program
have contributed, and continue to contribute to a wide range of marine ap-
plications. To date, over 65,000 unique users have accessed AVHRR
Pathfinder Version 5 SST data and applied them to a diverse set of appli-
cations and research topics. While it is difficult to accurately quantify the
use of Pathfinder SST data in the scientific literature due to varying ways
in which the data have been acknowledged or cited, searches of ScienceDi-
rect and the Web of Science indicate that number to be in the hundreds and
perhaps thousands. These studies include climate research areas such as
SST trend analysis (e.g., Casey and Cornillon, 2001), numerical modelling
applications (e.g., Shu et al., 2009), and the use of Pathfinder data as the
basis for higher-level SST products (e.g., Reynolds et al., 2007; Marullo et
al., 2007). A wide range of ecosystem-related studies have also relied on
Pathfinder SST data (e.g., Somoza et al., 2008; Halpern et al., 2008).
Many other scientific studies, too numerous to summarize, have utilized
Pathfinder SST across a wide range of disciplines and topics. Direct appli-
cations of the Pathfinder data to various societal benefits are also numer-
ous. For example, Pathfinder data have been used to characterize hump-
back whale distributions to reduce ship collisions, improve marine
protected area design and placement, plan optimal routing for yacht races,
evaluate performance of ocean and numerical weather prediction models,
and establish baselines for integrated ecosystem assessments and marine
spatial planning projects.
The improvements coming in Version 6 will enable an even greater
range of science and applications, enhance the compatibility of Pathfinder
data with numerous other GHRSST-compliant data streams, and will result
in a more accurate, consistent, and useful climate data record for SST.
The Past, Present, and Future of the AVHRR Pathfinder SST Program 339
Acknowledgements
The authors wish to thank their respective institutions for support during
the preparation of this manuscript and to Jorge Vazquez and Ed Armstrong
of the NASA Physical Oceanography DAAC for their long-term efforts to
support the Pathfinder SST community of users. Thanks also go to the
many NASA and NOAA program and project managers who kept the
AVHRR Pathfinder SST Program alive over the course of the last two
decades, including Eric Lindstrom and John Bates. Finally, special thanks
go to the last three Directors of the NOAA National Oceanographic Data
Center, most recently Margarita Conkright Gregg and before her, Zdenka
Willis and Lee Dantzler. All three deemed the Pathfinder effort suffi-
ciently worthy to contribute significant NODC resources over the years.
Without these contributions, the history of Pathfinder SST would have
come to an end circa 2001 and would have fallen well short of making the
most of the available AVHRR record.
References
Casey KS, Cornillon P (1999) A comparison of satellite and in situ based sea sur-
face temperature climatologies. J Climate 12:1848-1863
Casey KS, Cornillon P (2001) Global and regional sea surface temperature trends.
J Geophys Res 14(18):3801-3818
Casey KS, Evans R, Kilpatrick K, Kearns EJ, Brandon TB, Reynolds RW,
Vazquez-Cuervo J (2009) Toward the sea surface temperature climate data
record from space: An improved Advanced Very High Resolution Radiometer
Pathfinder dataset. In preparation for J Climate
Cox C, Munk W (1954) Measurement of the roughness of the Sea surface from
photographs of the sun’s glitter. J Optical Soc Amer 44(11):838-850
Donlon C, Robinson I, Casey KS, Vazquez-Cuervo J, Armstrong E, Arino O,
Gentemann C, May D, LeBorgne P, Piollé JF, Barton I, Beggs H, Poulter
DJS, Merchant CJ, Bingham A, Heinz S, Harris A, Wick G, Emery B, Min-
nett P, Evans R, Llewellyn-Jones D, Mutlow C, Reynolds RW, Kawamura H,
Rayner N (2007) The Global Ocean Data Assimilation Experiment High-
resolution Sea Surface Temperature Pilot Project. Bull Amer Meteorol Soc
88(8):1197-1213 doi:10.1175/BAMS-88-8-1197
Halpern BS, Walbridge S, Selkoe KA, Kappel CV, Micheli F, D'Agrosa C, Bruno
JF, Casey KS, Ebert C, Fox HE, Fujita R, Heinemann D, Lenihan HS, Madin
EMP, Perry MT, Selig ER, Spalding M, Steneck R, Watson R (2008) A
global map of human impact on marine ecosystems. Science 319(5865):948-
952 doi: 10.1126/science.1149345
340 K.. S. Casey et al.
Kilpatrick KA, Podestá GP, Evans R (2001) Overview of the NOAA/NASA Ad-
vanced Very High Resolution Radiometer Pathfinder algorithm for sea sur-
face temperature and associated matchup database. J Geophys Res
106(C5):9179-9197
King MD, Greenstone R, eds (1999) Earth Observing System (EOS) Reference
Handbook. National Aeronautics and Space Administration, EOS Project Sci-
ence Office, 355 pp
Marullo S, Nardelli BB, Guarracino M, Santoleri R (2007). “Observing the Medi-
terranean Sea from space: 21 years of Pathfinder-AVHRR sea surface tem-
peratures (1985 to 2005): re-analysis and validation. Ocean Science 2(2):299-
310
Podestá GP (1995) Case studies for SeaWiFS calibration and validation, part 3,
SeaWifs global fields: What’s in a day? NASA Technical Memorandum, no.
10456, 27 pp
Rayner NA, Brohan P, Parker DE, Folland CF, Kennedy JJ, Vanicek M, Ansell T,
Tett SFB (2006) Improved analyses of changes and uncertainties in sea sur-
face temperature measured in situ since the mid-nineteenth century: the
HadSST2 data set. J ClImate 19:446-469
Reynolds RW, Smith TM (1994) Improved global sea surface temperature analy-
sis using optimum interpolation. J Climate 7:929–948
Reynolds RW, Rayner NA, Smith TM, Stokes DC, Wang W (2002) An improved
in situ and satellite SST analysis for climate. J Climate 15:1609-1625
Reynolds, R.W., T.M. Smith, C. Liu, D.B. Chelton, K.S. Casey, and M.G. Schlax
(2007) Daily high-resolution-blended analyses for sea surface temperature.” J
Climate 20(22):5473-5496
Somoza, RD, Kampel M, Rudorff FD, Sousa RB, Cobas S (2008) Lobster (Panu-
lirus argus) captures and their relation with environmental variables obtained
by orbital sensors for Cuban waters (1997-2005). Brazilian J Oceanography
56(3):225-237
Stowe LL, Jacobowitz H, Ohring G, Knapp KR, Nalli NR (2002) The Advanced
Very High Resolution Radiometer (AVHRR) Pathfinder Atmosphere
(PATMOS) Climate Dataset: Initial Analyses and Evaluations.” J Climate
15(11):1243-1260
Shu YQ, Zhu J, Wang DX, Yan CX, Xiao XJ (2009) Performance of four sea sur-
face temperature assimilation schemes in the South China Sea. Continental
Shelf Res 29(11-12):1489-1501.
Vazquez JV, Tran AV, Sumagaysay R, Smith E, Hamilton M (1995) NOAA/
NASA AVHRR oceans Pathfinder sea surface temperature data set user’s
guide version 1.2. See
http://pathfinder.nodc.noaa.gov/AVHRR_Pathfinder_Oceans_UsersGuidev1.2_27Sept1995
.pdf
Walton CC, Pichel WG, Sapper JF (1998) The development and operational ap-
plication of non-linear algorithms for the measurement of sea surface tem-
peratures with the NOAA polar-orbiting environmental satellites. J Geophys
Res 103:27,999-28,012
The Past, Present, and Future of the AVHRR Pathfinder SST Program 341
Appendix
Original Members of the AVHRR Pathfinder Oceans SWG
Peter Cornillon, University of Rhode Island (Chair)
Robert Evans, University of Miami
Gene Feldman, NASA
Richard Legeckis, NOAA
Charles McClain, NASA
Richard W. Reynolds, NOAA
Charles Walton, NOAA
