Marine Geodesy

ISSN: 0149-0419 (Print) 1521-060X (Online) Journal homepage: https://www.tandfonline.com/loi/umgd20

Comparison of Global Mean Sea Level Time Series

from TOPEX/Poseidon, Jason-1, and Jason-2

D. Masters , R. S. Nerem , C. Choe , E. Leuliette , B. Beckley , N. White & M.

Ablain

To cite this article: D. Masters , R. S. Nerem , C. Choe , E. Leuliette , B. Beckley , N. White & M.
Ablain (2012) Comparison of Global Mean Sea Level Time Series from TOPEX/Poseidon, Jason-1,
and Jason-2, Marine Geodesy, 35:sup1, 20-41, DOI: 10.1080/01490419.2012.717862

To link to this article: https://doi.org/10.1080/01490419.2012.717862

Published online: 18 Dec 2012.

Submit your article to this journal

Article views: 650

View related articles

Citing articles: 45 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=umgd20
Marine Geodesy, 35(S1):20–41, 2012
Copyright © Taylor & Francis Group, LLC
ISSN: 0149-0419 print / 1521-060X online
DOI: 10.1080/01490419.2012.717862

Comparison of Global Mean Sea Level Time Series

from TOPEX/Poseidon, Jason-1, and Jason-2

D. MASTERS,1 R. S. NEREM,1 C. CHOE,1 E. LEULIETTE,2

B. BECKLEY,3 N. WHITE,4 AND M. ABLAIN5
1
Colorado Center for Astrodynamics Research, University of Colorado, Boulder,
Colorado, USA
2
Laboratory for Satellite Altimetry, National Oceanic and Atmospheric
Administration (NOAA), Silver Spring, Maryland, USA
3
SGT Inc., NASA GSFC, Greenbelt, Maryland, USA
4
Centre for Australian Weather and Climate Research and Wealth from Oceans
Flagship, CSIRO Marine and Atmospheric Research, Hobart, Tasmania,
Australia
5
CLS, Ramonville Saint-Agne, France

In the interest of improving the sea level climate data record, we compare the 19-year
global mean sea level (GMSL) time series derived from TOPEX-Poseidon, Jason-1, and
Jason-2 and produced by different research institutions. The GMSL time series are each
produced using varying techniques and different applied corrections, and subsequently,
the time series exhibit unique characteristics and sensitivity to seasonal and interannual
signals. We find that the different applied corrections affect the time series variability
much less than the technique for computing the GMSL from along-track versus gridded
sea surface height anomalies and the choice of a minimum depth criteria.

Keywords Global mean sea level, climate data record, altimetry, interannual variability

Introduction
Since the launch of TOPEX/Poseidon (T/P) in 1992, monitoring global mean sea level
(GMSL) using satellite altimetry has become an important barometer of climate change
(Bindoff et al. 2007). The subsequent launches of Jason-1 and Jason-2 into the same T/P
10-day repeating reference orbit and their tandem-phase, inter-satellite calibrations have
created a unique, 19-year climate record of GMSL. A few research groups, including
the University of Colorado (CU; Nerem et al. 2010), the Goddard Space Flight Center
(GSFC; Beckley et al. 2011a), the National Oceanic and Atmospheric Administration
(NOAA; Leuliette and Scharroo 2010), LEGOS/CNES/CLS (AVISO; Ablain et al. 2009),
and CSIRO (Church and White 2011), regularly produce estimates of the GMSL variations,
and the average of these can be considered a sea level climate data record. After align-
ing these time series, applying a 60-day boxcar smoother, and removing the annual and

Received 24 January 2012; accepted 23 April 2012.

Address correspondence to Dallas Masters, Colorado Center for Astrodynamics Research, Uni-
versity of Colorado, 431UCB, Boulder, CO 80309. E-mail: dallas.masters@colorado.edu

20
 Comparison of Global Mean Sea Level Time Series 21

Figure 1. The mean of five independently computed GMSL time series and the 95 percent confidence
interval about the mean. Reduction of the variability among the time series would improve confidence
in the GMSL as a climate data record.

semi-annual components, the mean of the five GMSL time series estimates is plotted in
Figure 1. The shaded 95 percent confidence interval about the mean represents the spread
of differences among the five GMSL time series. Although the five series all yield the
same 3.2 mm/yr trend (within the 0.4 mm/yr uncertainty), understanding and reducing the
differences among the series is desirable for improving the confidence of the climate record
for monitoring climate change and interpreting the remaining interannual signals that cause
most of the variation in the filtered and detrended mean series. Because all of the series are
based on the same underlying altimeter measurements and contain common errors from
using some of the same corrections, the homogenization of the various series does not
remove all error sources. Since it is now possible to begin closing the sea level budget
using combinations of altimetry, Gravity Recovery and Climate Experiment (GRACE),
Argo and terrestrial hydrology models, reducing the uncertainty in the GMSL time series
is increasingly important (Cazenave and Llovel 2010; Church et al. 2011; Leuliette and
Willis 2011; Llovel et al. 2011).
In this study, we make an initial comparison of the five GMSL time series produced from
the reference orbit phases of the T/P and Jason-1/2 missions to understand the differences
among them and the causes of these differences. Although it is possible to use the T/P and
Jason-1 interleaved orbit mission phases and the ERS-2, GFO, and Envisat altimeter data in
the computation of the time series, we focus here only on the reference orbit missions. The
nonreference orbit missions could be used as a nearly-independent check on the reference
22 D. Masters et al.

GMSL, as Leuliette and Scharroo (2010) suggest that Envisat is comparable to Jason-1 for
measuring GMSL.
We first review the geophysical corrections required to estimate sea surface height
anomalies (SSHA) from the altimeter measurements and the differences in the sets of
corrections used by each institution. Next, we review the different algorithms used by each
institution to compute the GMSL time series from the SSHA. We compute the differences
among the unfiltered and filtered GMSL time series and simultaneously estimate the annual,
semi-annual, trend, and mean components of the filtered series. We then compare the time
series with seasonal signals removed and detrended. Finally, we recompute the along-track
CU series with corrections and algorithms similar to those of the time series produced from
gridded SSHA to attempt to reduce the differences among the CU and grid-derived GMSL
time series.

Methods

Computation of the Sea Surface Height Anomalies

Global mean sea level, as we refer to it here, is an approximation to the true, physical global
mean sea level since the altimeter-based GMSL is computed from the per-cycle, area-
weighted average of the altimeter-estimated sea surface height anomalies between ±66◦
latitude. Additionally, the altimeter per-cycle sampling is not consistent due to seasonal sea
ice, rain flagging, and data drop outs. Although the T/P reference orbit inclination angle of
66.04 degrees limits the sampling of the oceans to ± 66◦ latitude, the measurements still
sample approximately 94 percent of the ocean surfaces (Tai and Wagner 2011).
The satellite altimeter measures the round-trip time of a microwave radar pulse reflect-
ing off of the ocean surface. To estimate the sea surface height, a number of geophysical
corrections must be applied to the altimeter range measurement, along with estimation of
the satellite orbit to centimeter-level precision. These corrections include instrument delays,
ionospheric and both wet and dry tropospheric delays, the response of the sea surface to
dynamic atmospheric pressure, sea state bias, and both ocean and solid earth tides. Standard
corrections are made available on the altimeter geophysical data records (GDR), but many
of these, such as orbits and wet troposphere, are replaced by more recent or improved
ancillary corrections that subsequently improve the GMSL time series. Here, we describe
some of the corrections important in GMSL processing.
Instrument delay corrections are standardized by the science and engineering teams
and undergo periodic updates as anomalies are resolved. Ionospheric delay is estimated
from the dual-frequency radar, and this parameter is usually smoothed along-track to
reduce noise (AVISO 1996). Wet tropospheric delay is often estimated from the on-board
microwave radiometer, but calibration and drift issues have led to experimentation using
tropospheric delays derived from forecast models, such as ECMWF or NCEP reanalyses
(Brown 2012). Along with orbits, the wet tropospheric delay is one of the larger sources
of error. Dry tropospheric delays are derived from these same forecast models and should
be corrected for improper sampling of the S2 atmospheric tide due to the 6-hour analysis
frequency (the S2 air tide aliasing correction is mentioned because it is not applied in
some GMSL time series processing) (Ponte and Ray 2002; Ray and Ponte 2003). The
dynamic atmospheric pressure effect on the sea surface is a sum of the mean pressure and
the modeled high-frequency ocean response to pressure and wind forcings (Pascual et al.
2008). The sea state bias (SSB) is a combination of the electromagnetic bias, resulting
 Comparison of Global Mean Sea Level Time Series 23

from the preferential radar reflection off of wave troughs than crests, and the wave-height
induced tracker bias. The SSB correction is one of the larger errors in altimetry and a source
of variability among GMSL time series, and subsequently it is also the focus of continual
investigation and refinement (Chambers et al. 2003; Tran et al. 2010a, 2010b). Ocean tides
are removed using various tide models, such as GOT4.7 or FES-2004, which are based upon
direct analysis of the altimeter data and hydrodynamic/assimilation modeling, respectively
(Lyard et al. 2006; Ray 1999). For each of these corrections, only those data within a defined
valid range are used, and this effectively masks altimeter measurements that fall outside
these ranges. Recommended validity ranges are published in the altimeter handbooks, but
each institution uses different validity ranges for the corrections (AVISO 1996; AVISO and
JPL 2008, 2009).
The sea surface height anomaly is estimated with respect to a mean surface. A mean sea
surface (MSS) is representative of the input datasets and the time period over which the data
are averaged, and this leads to differences among the computed mean sea surfaces as more
recent data are assimilated (Andersen and Knudsen 2009; Hernandez and Schaeffer 2000).
Since the different institutions employ different mean sea surfaces in the GMSL processing,
this selection is another source of variability among the time series. If the altimeters had
a fixed coverage area from cycle to cycle, then different choices of MSS would produce a
constant bias in time. However, the coverage area is not fixed; it varies seasonally with sea
ice coverage and due to data dropouts. The sea ice extent is particularly significant because
MSS models differ the most in data sparse regions, and therefore the choice of MSS can
produce nontrivial differences in global mean sea level, particularly in seasonal signals.
Correction and SSHA outlier removal criteria also vary among the GMSL time series.
Each of the applied geophysical corrections is usually constrained to a valid range, and
those data points not within this range remove the SSHA estimate from consideration. The
sea surface height anomalies are further constrained within a valid range, thus removing
outliers. Additionally, some GMSL series constrain the altimeter measurements to those
in the “open ocean” (e.g., greater than 120 meters depth) to avoid inaccurate tidal model
removal. Sampling of the oceans at ± 66◦ latitude takes slightly less than 10 days due
to the repeat cycle of the altimeter reference orbit. Therefore, the global mean of the sea
surface height anomalies is usually computed for each 10-day orbit repeat cycle (or longer
periods).
Algorithms to compute the global mean of the sea surface height anomalies vary as
well. The simplest method computes the area-weighted mean of the along-track sea surface
height anomalies measured per cycle (e.g., Nerem, 1995) as

φ ht (φ, λ)ω(φ)
gmslt = λ  , (1)
λ φ ω(φ)

where t is the cycle mid-time, λ/ϕ are longitude/latitude, and ht (ϕ,λ) are the valid along-
track SSHA estimates in each cycle. A weighted mean is required since the altimeter orbit
characteristics result in denser sampling at higher latitudes. Wang and Rapp (1994) sug-
gested a “inclination weighting” scheme that is a function of the latitude of the measurement
and the orbit inclination angle of the satellite, and Tai and Wagner (2011) simplified this
using a spherical Earth approximation to

ω(φ) = sin(I )2 − sin(φ)2 , (2)

where I is the inclination angle of the satellite orbit. This inclination weighting is usu-
ally applied to along-track SSHA values when computing the GMSL (Nerem 1995).
24 D. Masters et al.

However, Scharroo (2006) showed that this method under-weights along-track measure-
ments at latitudes higher than ± 50 degrees. Scharroo also showed that weighting along-
track measurements by the cosine of the latitude results in over-weighting the southern
ocean band from approximately -30 to -60 degrees. Therefore, selection of a weighting
scheme for along-track measurements has not been standardized and may play a role in
variability of the different GMSL time series that employ this technique.
The second method of computing the global mean of the sea surface height anomalies
spatially grids the anomalies prior to calculation of the area-weighted mean (weighted
by the cos(ϕ)). This effectively spatially smooths the anomalies prior to computation of
the mean and reduces measurement noise. Different techniques to grid the measurements
include simple box-averaging (all measurements within a grid cell are averaged together)
or more sophisticated methods, such as optimal interpolation (e.g., Hendricks et al. 1996;
Le Traon et al. 1998). Since no groups actively use optimal interpolation techniques to
produce GMSL time series, we focus here only on the simple box-averaging technique.
Choice of cell size obviously affects the resulting mean SSHA and is therefore a source of
variation among the time series using this technique.
Seasonal sea ice extents also affect the computed GMSL since higher latitude measure-
ments are unavailable when sea ice is within the footprint of the altimeter. GMSL computed
from along-track means normally use all available, ice-free measurements, while some grid-
ding techniques may remove grid cells that do not contain measurements in each season.
This effect may be evident in the seasonal amplitudes and phases of the GMSL time series.
To create a stable time series across the three satellite instruments during the nominal
reference orbit phases (the Jason satellites were launched into and sampled the same exact
orbit as T/P), the inter-satellite sea surface heights are calibrated during the tandem phases
when the pair of satellites sampled the same ground track within one minute of each other.
For T/P and Jason-1, the satellites were sampling the same ground track during 21 cycles
(T/P cycles 343–364 and Jason-1 cycles 1–21). A similar tandem phase of collocated
sampling from Jason-1 cycles 240–259 and Jason-2 cycles 1–20 is used to calibrate sea
surface heights measured by the two instruments. To link the GMSL time series of each
satellite, these estimated inter-satellite biases are respectively applied to the Jason-1 time
series and Jason-2 time series to create a 19-year GMSL climate record. An additional bias
is sometimes estimated for the change from the T/P side A altimeter to the redundant side
B altimeter in 1999, but the methods to link these two phases of T/P are not standardized
and are, to some extent, dependent on the corrections applied to T/P measurements (e.g.,
SSB models differ across the T/P Side A altimeter and Side B altimeter (Chambers et al.
2003; Tran et al. 2010a)).

Computation and Characteristics of the GMSL Time Series

In this study, we compare five GMSL time series from independent institutions: CU, GSFC,
NOAA, AVISO, and CSIRO. Each group computes and periodically updates an ongoing
GMSL time series. Each applies a different set of corrections and uses a different algorithm
(mean of along-track or gridded SSHA) to compute the GMSL time series, and these
differences are summarized and highlighted in Table 1. In general, the time series can be
categorized into two groups: those that compute the GMSL from along-track SSHA (CU
and GSFC) and those that compute GMSL after spatially gridding the SSHA (NOAA,
AVISO, CSIRO). CU and GSFC both mask shallow water measurements, while NOAA,
AVISO, and CSIRO do not, and this removes the shallow Indonesian seas and other areas
that have strong El Niño-Southern Oscillation (ENSO) variations. Additionally, the series
 Table 1
Differences among the GMSL corrections and algorithms used to compute the five independent GMSL time series. Significant differences are
highlighted in bold

Parameter TOPEX/Poseidon Jason-1 Jason-2

Cycles AVISO, CSIRO: TOPEX & Poseidon CU, NOAA: cycles 1-239; GSFC: AVISO: cycles: 11-present; all
cycles; CU, GSFC, NOAA: TOPEX; 14-239; AVISO: 11- 249; others: 1-present
CU, NOAA: 8-343; GSFC: 8-356; CSIRO: 1-257
AVISO: 8-353; CSIRO: 11-364
Orbit CU, NOAA: GSFC STD0905; GSFC: CU: GSFC STD0905; GSFC: GSFC STD1007; all others: GDR
GSFC STD1007; AVISO: GSFC
STD0809; CSIRO: MGDRB
Range & Corrections
GSFC: ECMWF Interim Reanarysis; all GDR GDR
Dry Troposphere others MGDRB; NOAA, AVISO, GSFC
correct S2 air tides
CU, GSFC, NOAA: TMR replacement; GSFC, NOAA: Enhanced JMR; all CU, AVISO: GDR; GSFC:
Wet Troposphere AVISO, CSIRO: drift correction others: GDR Recalibrated Enhanced AMR;
(Scharoo et al., 2004); AVISO: NOAA, CSIRO: Enhanced AMR
internal correction for ∼60-day
yaw/fix signals
Ionosphere All: dual-frequency from GDR; CU does not smooth along-track
Sea State Bias CU, GSFC, AVISO: CLS Collinear v. CU, GSFC, NOAA: CLS Collinear CU, NOAA: CLS Collinear v. 2009;
2009 (CU does not use updated v. 2009; AVISO, CSIRO: GDR GSFC, AVISO, CSIRO: GDR
SWH/wind) & BM4 for Poseidon;
NOAA: CLS Collinear v. 2006 with
updated SWH/wind; CSIRO:
MGDRB
Inverted Barometer Inverted barometer and high frequency fluctuations; CSIRO IB only Inverted barometer and high
frequency fluctuations
Range Bias All: Center of gravity correction All: GDR pseudo datation bias GSFC: Pseudo datation bias

25
(except CU) (0.29 ms)
(Continued on next page)
26
Table 1
Differences among the GMSL corrections and algorithms used to compute the five independent GMSL time series. Significant differences are
highlighted in bold. (Continued)

Parameter TOPEX/Poseidon Jason-1 Jason-2

Mean Sea Surface &
Corrections
Ocean Tide & Loading GSFC: GOT4.9; CU, NOAA, AVISO: GSFC: GOT4.8; CU, NOAA, AVISO:
Tide GOT4.7; CSIRO: GDR GOT4.7; CSIRO: GDR
Solid Earth Tide GDR
Mean Sea Surface CU, AVISO: CLS01; NOAA, GSFC: TOPEX collinear mean profile, CSIRO: GDR
Processing Corrections
Minimum Ocean Depth CU, GSFC: 120 m; NOAA, AVISO, CSIRO: 0 m
Inland Bodies Included Mediterranean Sea: all; Black Sea: CU, GSFC, NOAA, AVISO; Hudson Bay: GSFC, NOAA, AVISO; Caspian
Sea: AVISO
SSHA Outlier Removal CU: > 2 m; GSFC, NOAA: > 1 m
Processing Algorithm: CU, GSFC: along-track, per-cycle latitude/inclination weighted means; NOAA (3◦ xl◦ ), AVISO (2◦ x2◦ ):
box-gridded, per-cycle, area-weighted means; CSIRO: l◦ xl◦ box-gridded, monthly, area-weighted means
 Comparison of Global Mean Sea Level Time Series 27

do not all include the same set of inland water bodies: Mediterranean Sea, (all), Hudson Bay
(GSFC, NOAA, AVISO, CSIRO), Black Sea (CU, GSFC, NOAA, AVISO), and Caspian
Sea (AVISO). This may be another source of variability among the series.
Due to the overlap of measurements during the tandem calibration phases of the
altimeters, each group selects which of the altimeter measurements will be included in
the GMSL time series. CU and NOAA use the latter mission measurements in the tandem
phases. GSFC begins using Jason-1 at cycle 13, and AVISO splits the tandem phases
in half (Jason-1 starting at cycle 11 and Jason-2 starting at cycle 11). These different
choices affect the computation of the inter-mission biases that link measurements from the
altimeter missions to form the GMSL time series and could potentially lead to different
trends computed from the series.
The GDR-provided corrections are often periodically revised to fix discovered prob-
lems (e.g., radiometer drift) or improve certain corrections (e.g., orbits) with newer ver-
sions, but GMSL processing often incorporates these corrections prior to their release on
the GDRs. Since some corrections are continuously improving, not all GMSL time series
have the most up-to-date corrections implemented in processing. Although many have been
updated, CSIRO mostly uses corrections that are available on the altimeter GDRs, and this
may contribute to the larger differences between the CSIRO time series and the others,
especially during T/P. Here we highlight significant differences among the corrections and
processing listed in Table 1.
Orbits: CU and GSFC use the GSFC std0905 (Lemoine et al. 2010) and std1007
orbits (Lemoine et al. 2011), respectively, for all missions, while NOAA and AVISO use
the GSFC std0905 orbits for T/P and CNES GDR orbits for Jason-1 and Jason-2. The
std1007 GSFC replacement orbits used by GSFC are based on the current ITRF2008
terrestrial reference frame (Altamimi et al. 2011) while the std0905 orbits are based on
the earlier ITRF2005 realization (Altamimi et al. 2007). Although changing the reference
frame does not significantly affect the GMSL time series (e.g., replacing the Jason series
GDR orbits with newer releases of the GSFC std orbits), using a consistent reference frame
across all missions does have regional effects and also removes a source of inter-mission
biases (Beckley et al. 2010).
Wet troposphere: For the T/P mission, CU, GSFC, and NOAA use the T/P Microwave
Radiometer (TMR) Replacement Product (JPL 2006), while AVISO and CSIRO correct the
TMR drift following Scharroo et al. (2004). AVISO also corrects the TMR for approximate
60-day yaw/fix mode signals (Zlotnicki and Callahan 2002). GSFC, NOAA, and CSIRO
use the latest enhanced microwave radiometer corrections for Jason-1 and Jason-2, and
GSFC additionally corrects the -0.5 mm/yr Jason-2 radiometer drift with a preliminary
recalibration to be included in the GDR-D release (Beckley et al. 2011b; Brown 2012).
CU and AVISO use the standard GDR corrections for the Jason-1 and Jason-2 microwave
radiometers.
Ionospheric smoothing: Although it is a recommended technique to reduce noise on
the ionospheric delay estimate, CU does not smooth the ionospheric correction along-track.
Sea state bias: Since the SSB models are an active area of research and improvement,
there is considerable variability among the models used during T/P. The choice of SSB
model has a large impact on the resulting time series, especially during the T/P Side-A
altimeter anomaly (1996–1999), where there was a strong drift in the altimeter-derived
significant wave height (Chambers et al. 2003). CU employs the latest CLS collinear model
v.2009 (Tran 2010a) for all missions, while GSFC uses it for T/P and Jason-1, NOAA for
Jason-1 and Jason-2, and AVISO for T/P. The others use the standard SSB models available
on the GDRs.
28 D. Masters et al.

Ocean tides: Except for CSIRO, which uses the GDR-provided tide models, all of
the time series use the GOT tide model (version 4.7). GSFC uses updated versions of
the GOT model (GOT4.9 for T/P and GOT4.8 for Jason-1/2) that has the S2 constituent
reevaluated and subsequently reduces the amplitude of the 59-day signal in the Jason time
series (Beckley et al. 2011a).
Mean sea surface: CU and AVISO use the CLS01 mean sea surface (Hernandez and
Schaeffer 2000), NOAA uses the DNSC08 mean sea surface (Andersen and Knudsen 2009),
CSIRO uses the GDR-provided mean sea surfaces, and GSFC computes the cross-track
gradient using the DTU10 mean sea surface (an update to DNSC08) and the anomalies
with respect to the T/P collinear mean profile (Andersen and Knudsen 2010).
Minimum depth and SSHA limits criteria: Historically, the GMSL time series have
been produced over “open ocean” areas by applying a shallow water cutoff to avoid inaccu-
rate tidal modeling. CU and GSFC apply this constraint and discard SSHA measurements
over water less than 120 m deep, while the others do not constrain SSHA based upon water
depth. CU discards sea surface height anomalies that are greater than ± 2 m, while GSFC
and NOAA discard anomalies greater than ± 1 m.
Algorithm: The series can be differentiated by the algorithm used to compute the
mean, with CU and GSFC computing area-weighted means of along-track sea surface
height anomalies. NOAA, AVISO, and CSIRO first grid the anomalies before computing
the area-weighted mean of the gridded values. NOAA uses a 3◦ x1◦ (longitude x latitude)
grid, and AVISO uses a 2◦ x2◦ grid. CSIRO computes the mean of gridded anomalies on a
1◦ x1◦ grid and at monthly intervals rather than the 10-day cycle interval used by all of the
other groups. This is to accommodate comparison with other monthly climate data sets,
such as the PSMSL global tide gauge data set (Church and White 2011).

Time Alignment and GIA Correction of the GMSL Time Series

Four of the time series are produced at 10-day orbit repeat cycle intervals, but the cycle
time of each GMSL estimate varies slightly among the series. This is due to computation
of the time of each GMSL estimate as either the mid-time of the cycle (CU) or the average
time of the SSHA estimates used in that cycle (NOAA, GSFC, AVISO). Because each of
the latter applies different masks to the SSHA estimates based upon the constraints in
Table 1, the computed mean time varies slightly among the time series. Since the four series
are fundamentally based on per-cycle GMSL estimates, we chose to align all-time series to
the single mid-cycle time base used by CU. The CSIRO time series is computed at monthly
intervals, so it was linearly interpolated to the CU time base for comparison purposes.
Finally, some of the published time series include the glacial isostatic adjustment (GIA)
correction (CU, AVISO, CSIRO), while two series are published without the correction
applied (GSFC, NOAA). Before comparison, we applied a 0.3 mm/yr GIA correction to
the GSFC and NOAA series (Peltier 2009; Tamisiea 2011).

Results

Comparison of the Unfiltered GMSL Time Series

The aliasing of a semi-diurnal signal causes a known, approximately 59-day signal in the
GMSL time series. The cause has been linked to uncorrected movement of the T/P center of
mass due to flexure of the solar panels during heating/cooling cycles, although other sources
of the T/P 59-day signal remain (Ablain et al. 2010; Leuliette et al. 2010; Zelensky et al.
 Comparison of Global Mean Sea Level Time Series 29

2010). The Jason satellites do not experience this movement, but the signal was assimilated
into the GOT4.7 ocean tide model that is often used to correct the Jason measurements.
Therefore, GMSL time series are normally filtered to remove this signal, but the raw time
series can be compared and show important differences in the processing. Three of the
groups make available unfiltered GMSL time series at the native 10-day cycle resolution:
CU, GSFC, and NOAA. Since AVISO applies a low-pass filter prior to publishing their time
series and CSIRO computes a monthly time series, they were not included in the analysis
of unfiltered time series.
The unsmoothed CU, GSFC, and NOAA GMSL time series are plotted in Figure 2. The
mean of each series has been removed and an offset applied for presentation clarity. Each
series has the same trend and dominant annual signal, but the 59-day signal is strongest
throughout the entire CU time series and during the Jason-era of the NOAA time series. The
inter-mission calibration phases when both satellites were flying less than a minute apart
and sampling the same ground track are shown in the gray time slices to highlight potential
inter-mission biases (these slices are present in subsequent figures). The differences among
the time series are plotted in Figure 3. The 59-day signal is weaker in the GSFC time series
than either CU or NOAA, and this is due to the GSFC series employing an updated GOT

Figure 2. The three unsmoothed GMSL time series from CU, GSFC, and NOAA (offset for clarity).
The gray periods indicate the tandem calibration phases of the missions. The CU series exhibits much
stronger 59-day signals than either GSFC or NOAA due to CU not smoothing the ionospheric cor-
rection along-track. Additionally, both GSFC and NOAA correct the T/P dry tropospheric correction
for S2 atmospheric tidal aliasing. The GSFC series also uses an updated version of the GOT (v4.9 for
T/P and v4.8 for Jason 1/2) ocean tide model that reduces the 59-day signal in the Jason era.
30 D. Masters et al.

Figure 3. The differences of the unsmoothed GMSL series from Figure 2. CU-GSFC shows a
stronger 59-day signal in the T/P era due to GSFC using the updated GOT4.9 and correcting for the
S2 air tides, while CU does not. The CU-GSFC difference does not have a strong annual component
that is evident in the CU-NOAA and GSFC-NOAA differences, and this due to the NOAA series using
gridded rather than along-track SSHA and not editing shallow water measurements. The magnitude
of the 59-day signal difference decreases in the Jason era since the Jason-1/2 GDR dry tropospheric
correction used in all series is properly corrected for the S2 air tide.

tide model that partially corrects assimilation of the T/P 59-day signal into the Jason series
(Ablain et al. 2010; Zelensky et al. 2010). Additionally, the unsmoothed differences contain
a strong annual component that is most likely due to along-track versus gridded SSHA in
the computation of the GMSL time series.

Comparison of 59-day Smoothed Time Series

We smoothed the GMSL series using a 60-day boxcar filter to reduce the 59-day noise
signal. Figure 4 shows the resulting GMSL time series after smoothing. The AVISO series
lacks any remaining high-frequency noise since it is published in a filtered form and has been
filtered again with the 60-day boxcar filter for consistency. The CU and GSFC series exhibit
the same features, and likewise the NOAA and AVISO series are similar due to the similarity
in the algorithms and shallow water cut-off, as described previously. The differences among
all of the smoothed series are plotted in Figure 5. In general, these differences also indicate
that the CU and GSFC series are distinctly different from the NOAA and AVISO series.
The plot of CU-GSFC shows a semi-annual signal remains between the series during T/P,
while an annual signal remains during the Jason-era. It also indicates a potential difference
 Comparison of Global Mean Sea Level Time Series 31

Figure 4. The smoothed (60-day boxcar filtered) GMSL time series. The CU and GSFC series
exhibit the same features, and likewise the NOAA and AVISO series are similar due to the similarities
in the respective algorithms. The CSIRO series is smoother than the others since it is computed at
monthly intervals instead of at the 10-day cycle frequency. Differences in the interannual signals are
also evident.

in the T/P to Jason-1 biases applied in the two series due to the observed discontinuity
during the T/P to Jason-1 tandem phase. The CU-NOAA, CU-AVISO, GSFC-NOAA, and
GSFC-AVISO plots indicate a strong annual signal remains in the differences, along with
a slowly varying signal over the entire time period. The NOAA-AVISO plot shows that
these two series are very similar in the Jason-era, but the T/P time period is different due
to the corrections applied. The NOAA and AVISO indicate potential differences in the
inter-mission biases at both the T/P to Jason-1 transition and Jason-1 to Jason-2 transition.
The long-period signal in the CSIRO differences is most likely the result of CSIRO using
the MGDRB orbits for TOPEX while the others use the updated GSFC std orbits.

Estimation and Removal of the Annual, Semi-Annual, and Trend Components

We simultaneously estimated the annual, semi-annual, trend, and mean components of
each series using least-squares. The component parameters are summarized in Table 2. As
expected from Figure 4, the annual and semi-annual components agree closely for the CU
and GSFC series both in amplitude and phase, and likewise for the NOAA and AVISO
series. The CSIRO series annual signal agrees with CU and GSFC, but the semi-annual
signal phase (219 deg) is between the CU/GSFC and NOAA/AVISO series. The linear
32 D. Masters et al.

Figure 5. The differences among the smoothed GMSL series shown in Figure 4. The amplitude of
the CU-GSFC difference is small and indicates the similarities in the processing (mean of along-
track SSHA) in these two series. The remaining annual signals are most likely the result of different
spatial editing that would enhance the seasonal differences. Likewise, the NOAA-AVISO difference
shows that these two series are similar due to the gridding technique employed. The T/P era of the
NOAA-AVISO difference is much larger than the Jason-1/2 eras, and this is most likely due to the
different corrections applied during T/P. NOAA-AVISO also shows slight offsets at the T/P to Jason-1
and Jason-1 to Jason-2 tandem phases, and these may be indicative of different inter-mission bias
estimates.

Table 2
Least-squares estimated annual, semi-annual, and trend components of the GMSL time
series

Semi-Annual
Annual Component Component
Linear
Amplitude Phase Amplitude Phase Trend
Time Series (mm) (deg) (mm) (deg) (mm/yr)
CU 4.6 284 1.1 206 3.19
GSFC 4.2 285 1.3 193 3.12
NOAA 5.8 299 1.3 231 3.14
AVISO 6.2 299 1.2 237 3.21
CSIRO 4.9 288 1.1 219 3.15
 Comparison of Global Mean Sea Level Time Series 33

Figure 6. The GMSL series after removing the annual and semi-annual components (offset for
clarity). The interannual signals are evident in these time series, especially the strong fluctuations
during ENSO events in 1997–98 and 2010–11. The different sensitivities to interannual variations
among the series is most likely the result of the mean of along-track or gridded SSHA algorithm used.

trends range from 3.12 mm/yr for GSFC to 3.21 mm/yr for AVISO, with no correlation
among the seasonal signals or trends.
We removed the estimated seasonal signals from the smoothed series (leaving the linear
trend and interannual signals), and these are plotted (offset for clarity) in Figure 6. The
detrended signals (with seasonal signals removed) are plotted in Figure 7. These detrended
series show the interannual variations in the GMSL due to the ENSO (e.g., strong peaks in
1997–98 due to the strong El Niño and abrupt drop due to the strong 2010–11 La Niña). But
the relative magnitudes of these interannual variations are different among the series due to
the shallow water editing and along-track versus gridding algorithm. Additionally, the CU
and GSFC detrended series both show a steep drop in GMSL at the beginning of the T/P
mission from 1993 to 1996. The NOAA detrended series also has a sharp increase in GMSL
just before 1997 that is absent from the other series, and this could possibly be an outlier.
The differences in the detrended series (Figure 7) are plotted in Figure 8. The seasonal
signals present in Figure 5 have been reduced and only differences in interannual variations
remain. In addition to the tandem phases highlighted in gray in 2002 and 2008, the change
from T/P Side A to Side B is also highlighted in Figure 8 by a thin gray line at 1999.
In general, the T/P mission exhibits the largest variability among the series due to the
varied corrections applied to the older T/P measurements. The Jason-1 period shows the
best agreement among the series and is relatively stable (i.e., strong interannual variations
34 D. Masters et al.

Figure 7. The detrended series with seasonal signals removed from Figure 6 (offset for clarity). A
third shaded line in 1999 marks the transition from T/P Side A to Side B and a potential period of
bias offsets among the series. The remaining interannual signals vary among the time series due to
differences in the algorithm and applied corrections. The CU, GSFC, and NOAA series exhibit a
steep drop in residual GMSL during the start of T/P, as well as larger amplitude interannual signals
throughout the series.

may not be as prevalent during this phase). The Jason-2 period shows more variability than
Jason-1, although the Jason-2 record is yet only half as long.
The plot of CU-GSFC in Figure 8 shows an anomalous offset around the beginning
of 2000, which may be due to solar maximum on CU’s unfiltered ionospheric correction.
Small offsets at the transitions (T/P Side A to Side B, T/P to Jason-1, and Jason-1 to Jason-
2) may indicate different inter-mission bias estimation practices and may cause different
estimated trends among the time series.

Modifying CU Processing to Reduce Differences among the NOAA and AVISO Series
To test the sensitivity of the GMSL series to certain corrections and processing algorithm
parameters, we recomputed the CU series with a set of corrections and an algorithm similar
to the NOAA and AVISO series. Since the CSIRO series is computed much differently than
the other series and uses older corrections, we compare the recomputed CU series only
to the NOAA and AVISO series. For each recomputed CU GMSL series, we estimated
the seasonal and trend components (as in Table 2), and these newly estimated parameters
are summarized in Table 3. For clarity, the nominal CU GMSL series is listed again first,
followed by the recomputed GMSL series (denoted “CU a” through “CU e”). For each
 Comparison of Global Mean Sea Level Time Series 35

Figure 8. The differences among the series after removing the seasonal signals (Figure 7) and
detrending. CU-GSFC and NOAA-AVISO each have the smallest differences and indicate the sim-
ilarities between the series because they are derived from the mean of along-track versus gridded
SSHA. Residual annual signals in CU-GSFC are most likely due to different outlier removal and
included inland water bodies. The T/P era of NOAA-AVISO shows considerable variability compared
to the Jason-1/2 eras, and this is most likely the result of using different wet tropospheric and sea state
bias corrections and the handling of the switch from T/P Side A to Side B altimeter. The long-period
signal in the CU-NOAA and CU-AVISO series are due to computing the mean from along-track
versus gridded SSHA (as is evident in Figure 9). The largest differences among all of the series occur
at strong ENSO events in 1997–98 and 2010–11, indicating that the choice of algorithm and other
constraints plays a large role in determining the sensitivity of each series to interannual signals.

recomputed CU series, in Figure 9 we plot the difference with the nominal CU, NOAA,
and AVISO series.
CU a: Because the NOAA and AVISO series are computed with the GDR orbits for
Jason-1 and Jason-2, we first replaced the GSFC std0905 orbits with the GDR orbits. This
resulted in an insignificant change to the seasonal components and trend of the entire series,
but a small trend is evident in the Jason-era of the CU-CU a plot. This indicates a slight
trend in the difference of the GSFC std0905 orbits versus GDR orbits.
CU b: To test the effect noted in Scharroo (2006), we tested the choice of cosine of lati-
tude weighting (applied to ungridded, along-track SSHA) instead of the inclination/latitude
weighting used in the nominal CU series. This resulted in an increase in the annual
amplitude to 5.0 mm, a small decrease in phase of the semi-annual signal, and a drop
in the linear trend to 3.16 mm/yr. Only slight interannual differences are evident in the plot
of CU-CU b in Figure 9.
36 D. Masters et al.

Table 3
Least-squares estimated annual, semi-annual, and trend components of the recomputed CU
GMSL time series and the NOAA and AVISO time series

Semi-Annual
Annual Component Component
Linear
Amplitude Phase Amplitude Phase Trend
Time Series (mm) (deg) (mm) (deg) (mm/yr)
CU (nominal) 4.6 284 1.1 206 3.19
CU a (J1/2 GDR 4.6 283 1.1 204 3.18
orbit)
CU b (cosine 5.0 285 1.0 198 3.16
weighted)
CU c (no min. 4.9 292 1.2 215 3.21
depth)
CU d (3◦ x1◦ grid, 5.0 288 1.3 221 3.22
cosine weighted)
CU e (3◦ x1◦ grid, 5.5 296 1.3 230 3.25
cosine weighted,
J1/2 GDR orbit, no
min. depth)
NOAA 5.8 299 1.3 231 3.14
AVISO 6.2 299 1.2 237 3.21

CU c: We recomputed the CU series without restricting the minimum depth to 120 m.

This resulted in an increase of the seasonal signal amplitudes and phases. The linear
trend also increased to 3.21 mm/yr. From the plot of CU-CU c in Figure 9, we note that
allowing the shallow water measurements into the GMSL computation increases sensitivity
to ENSO in 1997-98 (El Niño) and 2010-11 (La Niña). We also note that the differences
among NOAA-CU c and AVISO-CU c are reduced.
CU d: To more closely simulate the NOAA and AVISO processing, we computed
the GMSL series after gridding the nominal CU SSHA estimates on a 3◦ x1◦ grid and
computing the area-weighted mean. This resulted in a reduction of the annual amplitude
to 5.0 mm and a decrease in the annual phase to 288 degrees. The semi-annual amplitude
increased to 1.3 mm, and the semi-annual phase increased to 221 degrees. The linear trend
also increased to 3.22 mm/yr. As expected, the plot of NOAA-CU d in Figure 9 shows a
reduction in the difference relative to NOAA-CU c.
CU e: Finally, all of the individual modifications were applied to the CU processing
to best simulate the NOAA processing (GDR orbits for the Jason-era, no minimum depth
criterion, area-weighted means calculated on a 3◦ x1◦ grid). As shown in Table 3, the CU e
seasonal signal amplitudes and phases now more closely agree with the NOAA and AVISO
seasonal signal parameters (appended to Table 3 from Table 2). The annual amplitude of
CU e was 5.5 mm, lower than both NOAA (5.8 mm) and AVISO (6.2 mm), but the other
estimated parameters agree well. The linear trend of CU e was 3.25 mm/yr, higher than
NOAA (3.14 mm/yr) and only slightly higher than AVISO (3.21 mm/yr). From Figure 9,
it is evident that NOAA-CU e and AVISO-CU e both minimize the series differences,
 Comparison of Global Mean Sea Level Time Series 37

Figure 9. Differences among the nominal CU, NOAA, and AVISO detrended series (from Figure 7)
and the recomputed CU series based upon Table 3. In CU-CU a, we note that using the GDR
orbit rather than the GSFC std0905 orbit in Jason-1/2 has a minor effect and slight negative trend.
Weighting the along-track SSHA by the cosine of the latitude (CU b) also has a minor effect, mostly
during the ENSO events. A larger effect is noted in CU-CU c when the minimum depth criterion
is relaxed, especially at the 1997-98 El Niño. The largest single effect is in CU-CU d and results
from computing the mean of gridded rather than along-track SSHA. A long period signal is evident
in CU-CU d that is also evident in the NOAA-CU [a-c] and AVISO-CU [a-c] differences. When
all of the modifications are applied in CU e, the CU-CU e difference is greatest and many of the
features that were evident in the CU-NOAA and CU AVISO differences in Figure 8 are now visible.
Additionally, the NOAA-CU e difference has been minimized, especially in the Jason-1/2 eras. An
unexplained long period signal is still evident in the AVISO-CU e difference.

especially during the Jason era, but there are still residual differences yet to be explained.
These interannual differences are most likely due to some the applied corrections and are
subjects for future work.

Conclusions and Future Work

We have compared the GMSL time series produced by five institutions and note the
significant differences in the applied corrections and algorithms used to compute the mean
sea surface height anomalies. The five series can be separated into two groups based upon
the algorithm to compute the mean SSHA. The largest differences among the GMSL time
series result from using gridded versus non-gridded sea surface height anomalies when
calculating the means. Both CU and GSFC do not grid the anomalies, while NOAA,
AVISO and CSIRO compute a mean from gridded anomalies. When the CU processing is
recomputed using gridded anomalies, the differences between the estimated seasonal and
trend parameters is reduced. Additionally, applying cosine weighting rather than inclination
weighting to the CU processing further reduces the differences with NOAA and AVISO.
38 D. Masters et al.

We have also shown that the minimum depth criterion used by CU and GSFC also
strongly changes the GMSL series relative to using SSHA measured in shallow water areas.
This may indicate that these shallow water areas have strong seasonal and interannual
changes in sea level relative to open oceans. Removing the depth criterion from CU
processing also brings the estimated seasonal and trend parameters into better agreement
with those of NOAA and AVISO and more accurately represents the true GMSL time series.
By simulating the NOAA algorithm and corrections and recomputing the CU series, we
demonstrate that the difference between the CU series and both the NOAA and AVISO series
can be substantially reduced and the estimated seasonal and trend parameters brought into
agreement. The residual differences after applying these changes to the CU series indicate
effects of the individual corrections applied to each series. Future work should investigate
these smaller effects and the best choice among the applied corrections when computing
the GMSL time series.
This initial study of the differences among the GMSL time series has illuminated the
fact that the variability among the different series is due to both the individual center’s
processing rationale of the altimeter measurements (i.e., how to interpret the sea surface
height data), and the application of topical suites of geophysical models, orbits, and range
corrections. Indeed, the significant agreement in the time series due to the international
cooperation between centers and the Ocean Surface Topography Science Team is a model
for other climate records. The accuracy of the GMSL estimates continues to improve
as the sea surface height climate data record evolves due in part to advances in space
geodetic techniques resulting in more accurate terrestrial reference frame realizations and
improved models of the static and time-varying gravity fields from GRACE observations. In
order to better isolate differences and explain the variability among the GMSL time series
as the climate data record (CDR) evolves, the community should come to a consensus
and propose standardized selection criteria of processing details, such as depth mask,
observation weighting, editing criteria, etc., for the construction of altimeter-derived mean
sea level estimates. This will increase confidence in the satellite altimetry sea level record
for use in monitoring climate change and provide a more accurate accounting of the sea
level budget.

Acknowledgements
This work was supported by two grants from NASA (Ocean Surface Topography Science
Team and the MEaSUREs Project). Additional support from GSFC, NOAA, CNES, CLS,
and CSIRO is also acknowledged.

References
Ablain, M., A. Cazenave, G. Valladeau, and S. Guinehut. 2009. A new assessment of the error budget
of global mean sea level rate estimated by satellite altimetry over 1993–2008. Ocean Science 5:
193–201. http://www.ocean-sci.net/5/193/2009/os-5-193-2009.html
Ablain, M., S. Phillips, L. Carrere, G. Valladeau, J.-F. Legeais, E. Bronner, and N. Picot. 2010. Anal-
ysis of 58.74-day signal observed on the MSL derived from Jason-1&2 and TOPEX data. Pre-
sentation to the Ocean Surface Topography Science Team. Lisbon, Portugal. http://www.aviso.
oceanobs.com/fileadmin/documents/OSTST/2010/oral/20 Wednesday/morning/Ablain.pdf
Altamimi, Z., X. Collilieux, J. Legrand, B. Garayt, and C. Boucher. 2007. ITRF2005: A new release
of the international terrestrial reference frame based on time series of station positions and earth
orientation parameters. Journal of Geophysical Research 112(B9). doi: 10.1029/2007JB004949
 Comparison of Global Mean Sea Level Time Series 39

Altamimi, Z., X. Collilieux, and L. Métivier. 2011. ITRF2008: an improved solution of the interna-
tional terrestrial reference frame. Journal of Geodesy 85(8):457–473. doi: 10.1007/s00190-011-
0444-4
Andersen, O., and P. Knudsen. 2009. DNSC08 mean sea surface and mean dynamic topography
models. Journal of Geophysical Research (Oceans) 114: 11001. doi: 10.1029/2008JC005179
Andersen, O., and P. Knudsen. 2010. DTU10 mean sea surface and mean dynamic topography– im-
provement in the Arctic and coastal zone. Presentation to the Ocean Surface Topography Science
Team, Lisbon, Portugal. http://www.aviso.oceanobs.com/fileadmin/documents/OSTST/2010/
oral/19 Tuesday/ANDERSEN.pdf
AVISO/Altimetry. 1996. AVISO User Handbook for Merged TOPEX/POSEIDON Prod-
ucts. AVI-NT-02-101, edition 3.0. http://www.aviso.oceanobs.com/fileadmin/documents/data/
tools/hdbk tp gdrm.pdf
AVISO and JPL. 2008. AVISO and PODAAC user handbook: IGDR and GDR Jason products. SMM-
MU-M5-OP-13184-CN, Edition 4.1. http://www.aviso.oceanobs.com/fileadmin/documents/
data/tools/hdbk j1 gdr.pdf
AVISO and JPL. 2009. OSTM/Jason-2 Products Handbook. SALP-MU-M-OP-15815-CN, Edition
1.4. http://www.aviso.oceanobs.com/fileadmin/documents/data/tools/hdbk j2.pdf
Beckley, B. D., R. D. Ray, S. A. Holmes, N. P. Zelensky, F. G. Lemoine, X. Yang, S. T. Brown,
S. Desai, G. T. Mitchum, and J. Hausman. 2011a. Integrated Multi-Mission Ocean Altimeter
Data for Climate Research TOPEX/Poseidon, Jason-1 and OSTM/Jason-2 User’s Handbook:
Ver. 1. California Institute of Technology. ftp://podaac-ftp.jpl.nasa.gov/allData/merged alt/
preview/L2/docs/multi alt handbook.pdf
Beckley, B. D., N. P. Zelensky, X. Yang, S. Holmes, F. G. Lemoine, R. Ray, S. Desai, S. Brown,
B. Haines, G. Mitchum. 2011b. Assessment of current global and regional mean sea level
estimates based on the TOPEX/Poseidon, Jason1 & 2 climate record. Presentation to the Ocean
Surface Topography Science Team, San Diego, CA. http://www.aviso.oceanobs.com/fileadmin/
documents/OSTST/2011/poster/beckley cv5.pdf
Beckley, B. D., N. P. Zelensky, S. Holmes, F. G. Lemoine, S. Luthcke, R. Ray, S. Desai, S. Brown, G.
Mitchum, and N. Tran. 2010. Assessment of global and regional mean sea level estimates based
on ITRF2008. Presentation to the Ocean Surface Topography Science Team, Lisbon, Portugal.
http://www.aviso.oceanobs.com/fileadmin/documents/OSTST/2010/beckley.pdf
Bindoff, N. L., J. Willebrand, V. Artale, A, Cazenave, J. Gregory, S. Gulev, K. Hanawa, C. Le Quéré, S.
Levitus, Y. Nojiri, C. K. Shum, L. D. Talley, and A. Unnikrishnan. 2007. Observations: Oceanic
climate change and sea level. In ed. S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis,
K. B. Averyt, M. Tignor, and H. L. Miller, Climate Change 2007: The Physical Sci-
ence Basis. Contribution of Working Group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press.
http://www.ipcc.ch/pdf/assessment-report/ar4/wg1/ar4-wg1-chapter5.pdf
Brown, S. 2012. Maintaining the long-term calibration of the Jason-2/OSTM advanced microwave
radiometer through inter-satellite calibration. IEEE Trans. Geosci. Remote Sens. Submitted for
publication.
Cazenave, A., and W. Llovel. 2010. Contemporary sea level rise. Annual Review of Marine Science
2: 145–173. doi: 10.1146/annurev-marine-120308-081105
Chambers, D., S. A. Hayes, J. C. Ries, and T. J. Urban. 2003. New TOPEX sea state bias models and
their effect on global mean sea level. Journal of Geophysical Research (Oceans) 108:3305. doi:
10.1029/2003JC001839
Church, J., and N. White. 2011. Sea-level rise from the late 19th to the early 21st century. Surveys in
Geophysics. doi: 10.1007/s10712-011-9119-1
Church, J., N. White, L. Konikow, C. Domingues, J. Cogley, E. Rignot, J. Gregory, M. van den Broeke,
A. Monaghan, and I. Velicogna. 2011. Revisiting the Earth’s sea-level and energy budgets from
1961 to 2008. Geophysical Research Letters 38(18). doi: 10.1029/2011GL048794
Hendricks, J. R., R. R. Leben, G. H. Born, and C. J. Koblinsky. 1996. Empirical orthogonal function
analysis of global TOPEX/POSEIDON altimeter data and implications for detection of global
sea level rise. Journal of Geophysical Research 101(C6):14131–14145. doi: 10.1029/96JC00922
40 D. Masters et al.

Hernandez, F., and P. Schaeffer. 2000. Altimetric mean sea surfaces and gravity anomaly
maps intercomparisons. AVISO Technical Report: AVI-NT-011-5242-CLS. ftp://ftp.space.dtu.
dk/pub/GOCINA/DOC/MSS comp.pdf
JPL. 2006. Topex Microwave Radiometer Replacement Product, Version 1.0. ftp://podaac-
ftp.jpl.nasa.gov/allData/topex/preview/L2/tmr replace/docs/TMR Replacement 1.0.pdf
Le Traon, P.-Y., F. Nadal, and N. Ducet. 1998. An improved mapping method of multisatellite altimeter
data. Journal of Atmospheric and Oceanic Technology 15(2):522–534.
Lemoine, F. G., N. P. Zelensky, D. S. Chinn, D. E. Pavlis, D. D. Rowlands, B. Beckley, S. B. Luthcke,
P. Willis, M. Ziebart, A. Sibthorpe, J. Boy, V. Luceri. 2010. Towards development of a consistent
orbit series for TOPEX, Jason-1, and Jason-2. Advances in Space Research 46: 1513–1540. doi:
10.1016/j.asr.2010.05.007
Lemoine, F. G., N. P. Zelensky, D. S. Chinn, J. W. Beall, and S. Melachroinos. 2011. Im-
provements in orbit determination for altimetric satellites. Abstract G54A-06 presented at
2011 Fall Meeting, AGU, San Francisco, CA, 5–9 December. http://earthdata.nasa.gov/sites/
default/files/webform/agu/Lemoine ALTIMPOD AGU2011 v04 0.pdf
Leuliette, E. W., and R. Scharroo. 2010. Integrating Jason-2 into a multiple-altimeter climate data
record. Marine Geodesy 33 (Supp. 1):504. doi: 10.1080/01490419.2010.487795
Leuliette, E. W., R. Scharroo, W. Smith, J. Lillibridge, and L. Miller. 2010. An investi-
gation into the source of the 59-day variations in Jason sea level. Presentation to the
Ocean Surface Topography Science Team, Lisbon, Portugal. http://www.aviso.oceanobs.com/
fileadmin/documents/OSTST/2010/oral/20 Wednesday/morning/Leuliette.pdf
Leuliette, E. W., and J. K. Willis. 2011. Balancing the sea level budget. Oceanography 24(2): 122–129.
doi: 10.5670/oceanog.2011.32
Llovel, W., M. Becker, A. Cazenave, S. Jevrejeva, R. Alkama, B. Decharme, H. Douville, M. Ablain,
and B. D. Beckley. 2011. Terrestrial waters and sea level variations on interannual time scale.
Global and Planetary Change 75(1–2):76–82. doi: 10.1016/j.gloplacha.2010.10.008
Lyard, F., F. Lefevre, T. Letellier, and O. Francis. 2006. Modeling the global ocean tides: modern
insights from FES2004. Ocean Dynamics 56(5–6):394–415. doi: 10.1007/s10236-006-0086-x
Nerem, R. S. 1995. Global mean sea level variations from TOPEX/POSEIDON altimeter data. Science
268: 708–710. doi: 10.1126/science.268.5211.708
Nerem, R. S., D. P. Chambers, C. Choe, and G. T. Mitchum. 2010. Estimating mean sea level
change from the TOPEX and Jason altimeter missions. Marine Geodesy 33(Supp. 1):435. doi:
10.1080/01490419.2010.491031
Pascual, A., M. Marcos, and D. Gomis. 2008. Comparing the sea level response to pressure and
wind forcing of two barotropic models: Validation with tide gauge and altimetry data. Journal
of Geophysical Research 113(C7). doi: 10.1029/2007JC004459
Peltier, W. R. 2009. Closure of the budget of global sea level rise over the GRACE era: The importance
and magnitudes of the required corrections for global glacial isostatic adjustment. Quaternary
Science Reviews 28: 1658–1674. doi: 10.1016/j.quascirev.2009.04.004
Ponte, R. M., and R. D. Ray. 2002. Atmospheric pressure corrections in geodesy and oceanog-
raphy: A strategy for handling air tides. Geophysical Research Letters 29: 2153. doi:
10.1029/2002GL016340
Ray, R. 1999. A Global Ocean Tide Model From TOPEX/Poseidon Altimetry: GOT99.2. NASA
Techincal Memo 1999-209478. NASA Goddard Space Flight Center. http://hdl.handle.net/2060/
19990089548
Ray, R., and R. Ponte. 2003. Barometric tides from ECMWF operational analyses. Annales Geophys-
icae 21(8):1897–1910. doi: 10.5194/angeo-21-1897-2003
Scharroo, R., J. Lillibridge, W. Smith, and E. J. Schrama. 2004. Cross-calibration and long-term
monitoring of the microwave radiometers of ERS, TOPEX, GFO, Jason, and Envisat. Marine
Geodesy 27(1):279. doi: 10.1080/01490410490465265
Scharroo, R. 2006. Presentation to the Ocean Surface Topography Science Team, Venice, Italy.
Tai, C.-K., and C. Wagner. 2011. Sampling errors of the global mean sea level derived from
TOPEX/Poseidon altimetry. Acta Oceanol. 30(6):12–18. doi: 10.1007/s13131-011-0156-x
 Comparison of Global Mean Sea Level Time Series 41

Tamisiea, M. E. 2011. Ongoing glacial isostatic contributions to observations of sea level change.
Geophysical Journal International 186(3):1036–1044. doi: 10.1111/j.1365-246X.2011.05116.x
Tran, N., S. Labroue, S. Philipps, E. Bronner, and N. Picot. 2010a. Overview and update of the sea
state bias corrections for the Jason-2, Jason-1 and TOPEX missions. Marine Geodesy 33(Supp
1):348. doi: 10.1080/01490419.2010.487788
Tran, N., D. Vandemark, S. Labroue, H. Feng, B. Chapron, H. L. Tolman, J. Lambin, and N. Picot.
2010b. Sea state bias in altimeter sea level estimates determined by combining wave model and
satellite data. Journal of Geophysical Research 115(C3):C03020. doi: 10.1029/2009JC005534
Wang, Y. M., and R. Rapp. 1994. Estimation of sea surface dynamic topography, ocean tides,
and secular changes from TOPEX altimeter data. Document No. 430. Department of Geode-
tic Science and Surveying, Ohio State University. http://sealevel.colorado.edu/files/pubs/wang
rapp report 430.pdf
Zelensky, N., B. Beckley, F. Lemoine, S. Klosko, R. Ray, S. Holmes, D. Rowlands, S. Luthcke,
D. Chinn, and O. Bordyugov. 2010. The puzzling 59-day altimeter data signal and possi-
ble causes. Presentation to the Ocean Surface Topography Science Team, Lisbon, Portugal. http://
www.aviso.oceanobs.com/fileadmin/documents/OSTST/2010/oral/20 Wednesday/morning/20
1150 zelensky aud3.pdf
Zlotnicki, V., and P. Callahan. 2002. TOPEX and JASON microwave radiometer assessment
against the DMSP SSM/I and TRMM TMI. Presentation to the Ocean Surface Topography
Science Team, Biarritz, France. http://www.aviso.oceanobs.com/fileadmin/documents/OSTST/
2002 b/zlotnicki2.pdf