ArticlePDF Available

The Berkeley Earth Land/Ocean Temperature Record

  • Berkeley Earth

Abstract and Figures

A global land–ocean temperature record has been created by combining the Berkeley Earth monthly land temperature field with spatially kriged version of the HadSST3 dataset. This combined product spans the period from 1850 to present and covers the majority of the Earth's surface: approximately 57 % in 1850, 75 % in 1880, 95 % in 1960, and 99.9 % by 2015. It includes average temperatures in 1∘×1∘ lat–long grid cells for each month when available. It provides a global mean temperature record quite similar to records from Hadley's HadCRUT4, NASA's GISTEMP, NOAA's GlobalTemp, and Cowtan and Way and provides a spatially complete and homogeneous temperature field. Two versions of the record are provided, treating areas with sea ice cover as either air temperature over sea ice or sea surface temperature under sea ice, the former being preferred for most applications. The choice of how to assess the temperature of areas with sea ice coverage has a notable impact on global anomalies over past decades due to rapid warming of air temperatures in the Arctic. Accounting for rapid warming of Arctic air suggests ∼ 0.1 ∘C additional global-average temperature rise since the 19th century than temperature series that do not capture the changes in the Arctic. Updated versions of this dataset will be presented each month at the Berkeley Earth website (, last access: November 2020), and a convenience copy of the version discussed in this paper has been archived and is freely available at (Rohde and Hausfather, 2020).
Content may be subject to copyright.
Earth Syst. Sci. Data, 12, 3469–3479, 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.
The Berkeley Earth Land/Ocean Temperature Record
Robert A. Rohde1and Zeke Hausfather1,2
1Berkeley Earth, Berkeley, CA 94705, USA
2Breakthrough Institute, Oakland, CA 94612, USA
Correspondence: Robert A. Rohde (
Received: 31 December 201 – Discussion started: 2 June 2020
Revised: 28 September 2020 – Accepted: 5 October 2020 – Published: 17 December 2020
Abstract. A global land–ocean temperature record has been created by combining the Berkeley Earth monthly
land temperature field with spatially kriged version of the HadSST3 dataset. This combined product spans the
period from 1850 to present and covers the majority of the Earth’s surface: approximately 57 % in 1850, 75 %
in 1880, 95 % in 1960, and 99.9 % by 2015. It includes average temperatures in 1×1lat–long grid cells for
each month when available. It provides a global mean temperature record quite similar to records from Hadley’s
HadCRUT4, NASA’s GISTEMP, NOAA’s GlobalTemp, and Cowtan and Way and provides a spatially complete
and homogeneous temperature field. Two versions of the record are provided, treating areas with sea ice cover
as either air temperature over sea ice or sea surface temperature under sea ice, the former being preferred for
most applications. The choice of how to assess the temperature of areas with sea ice coverage has a notable
impact on global anomalies over past decades due to rapid warming of air temperatures in the Arctic. Account-
ing for rapid warming of Arctic air suggests 0.1 C additional global-average temperature rise since the 19th
century than temperature series that do not capture the changes in the Arctic. Updated versions of this dataset
will be presented each month at the Berkeley Earth website (, last access: Novem-
ber 2020), and a convenience copy of the version discussed in this paper has been archived and is freely available
at (Rohde and Hausfather, 2020).
1 Introduction
Global land–ocean temperature indices combining 2 m sur-
face air temperature over land with sea surface temperatures
(SSTs) over oceans are commonly used to assess changes in
the Earth’s climate. While it is a less physically meaning-
ful metric than Earth system total heat content, it is well-
measured with reliable data extending back to ca. 1850 for
oceans (Kennedy et al., 2011b) and as far back as ca. 1750
for land (Rohde et al., 2013a), and it is the part of the Earth
system most relevant for impacts on human civilization. Sea
surface temperatures are used in lieu of marine air tempera-
tures due to scarcity and inhomogeneity of marine air temper-
ature data (Kent et al., 2013), though it is only an imperfect
proxy and may be subject to slightly slower warming rates
than marine air temperatures in recent decades (Cowtan et
al., 2015; Richardson et al., 2016; Jones, 2020).
A number of prior groups have developed global land–
ocean surface temperature indexes, including NASA’s GIS-
TEMP (Hansen et al., 2010; Lenssen et al., 2019),
Hadley/UEA’s HadCRUT4 (Morice et al., 2012), NOAA’s
GlobalTemp (Smith et al., 2008; Vose et al., 2012; Huang
et al., 2020), and the Japan Meteorological Agency (JMA)
(Ishihara, 2006). Additionally, Cowtan and Way (2014) pro-
vide a spatially interpolated variant of HadCRUT4 featuring
greater spatial coverage, hereafter denoted CW2014. These
series differ in a number of respects. They all largely utilize
the same set SST measurements drawn from the ICOADS
database (Freeman et al., 2017) and most of the same land
temperature records contained in the Global Historical Cli-
matological Network – Monthly database (GHCNm) (Law-
rimore et al., 2011), though HadCRUT4 (and by exten-
sion CW2014) includes a more modest number of land sta-
tions than GISTEMP and GlobalTemp, which recently transi-
Published by Copernicus Publications.
3470 R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record
tioned to using the much larger GHCNm v4 database (Menne
et al., 2018).
Both GISTEMP and GlobalTemp utilize NOAA’s pair-
wise homogenization algorithm to detect and correct inho-
mogeneities such as station moves or instrument changes in
land stations (Menne and Williams, 2009), though NASA
applies an additional satellite nightlight-based urbanity cor-
rection (Hansen et al., 2010). GISTEMP and GlobalTemp
both use NOAA’s Extended Reconstructed Sea Surface Tem-
perature (ERSST) version 5 (Huang et al., 2017) for SSTs,
HadCRUT4 and CW2014 use HadSST3 (Kennedy et al.,
2011a, b), and JMA uses COBE-SST (Ishii et al., 2005).
HadCRUT4 and JMA include no spatial interpolation out-
side of 5×5latitude–longitude grid cells, while Global-
Temp includes some interpolation over land but has nearly
complete ocean temperature fields with the primary excep-
tion that sea ice regions are masked as missing. GISTEMP
and CW2014 spatially interpolate temperatures out to re-
gions with no direct station coverage (GISTEMP using a sim-
ple linear interpolation technique, while CW2014 uses krig-
ing). The upcoming HadCRUT5 will transition to HadSST4
and include spatial interpolation (Morice et al., 2020).
Here we describe the global land–ocean surface tempera-
ture product from Berkeley Earth that combines the Berke-
ley Earth land temperature data (Rohde et al., 2013a, b) with
SST data from HadSST3 (Kennedy et al., 2011a, b). It uses
a kriging-based spatial interpolation to provide an extensive
spatial coverage for the period from 1850 to present. The
land data utilize significantly more land station data (over
40 000 stations) compared to the 10 000 land stations used
by some of the other groups (though GISTEMP and Global-
Temp have both recently updated their records to include a
larger number of land stations, including more than 20 000
sites in GHCNv4). The land component also includes the
novel homogenization technique of the Berkeley Earth tem-
perature record that detects breakpoints through neighbor
difference series comparisons, cuts land stations into frag-
mentary records at breakpoints, and combines these fragmen-
tary records into a temperature field. The ocean component
of the land–ocean product uses an interpolated variant of
HadSST v3, whose construction is described below. A ver-
sion of the Berkeley Earth interpolated dataset has been pub-
licly available for some time but has not been formally de-
scribed. Lastly, we note that HadSST v3 will be replaced with
HadSST v4 once that product becomes operational (Kennedy
et al., 2019). Aside from minor differences in the way data
are communicated and formatted, HadSST v4 should be us-
able following the same steps described here.
2 Methods
The Berkeley Earth Land/Ocean Temperature Record com-
bines the Berkeley Earth land record (Rohde et al., 2013a)
with SST data from HadSST3 (Kennedy et al., 2011a, b). The
HadSST3 data are adjusted in several ways. The primary ma-
nipulation is to replace the gridded data with an interpolated
field using a kriging-based approach. The HadSST3 data set
provides grid cell averages on a 5by 5grid and only re-
ports monthly averages for cells where data were present dur-
ing the month in question. HadSST3 often reports no data
for 40 % of ocean grid cells. As described below, the in-
terpolation produces a more complete field and reduces the
component of uncertainty associated with incomplete cover-
age. While providing a more complete field, the interpolation
does not materially change the apparent rate of warming in
the oceans.
After interpolation, the ocean temperature anomaly field is
merged with the Berkeley Earth land anomaly field using the
fraction of land–water in each grid cell (typically reported
with a 1by 1latitude–longitude resolution). As described
below, two versions are considered with respect to the role
of sea ice. The version using air temperature above sea ice is
recommended for most users, though the other version may
be useful for certain specialists and diagnostic purposes.
2.1 Interpolation method
The HadSST3 gridded fields provide several critical compo-
nents, the temperature anomaly, the number of observations,
and several estimates of the uncertainty (Kennedy et al.,
2011a, b). The grid cell uncertainties and observation counts
allow one to treat some grid cells as having greater confi-
dence than others. Unlike land surface station data, where
each monthly average represents many temperature observa-
tions, the ocean observation counts are a true measure of the
number of instantaneous SST measurements.
Analogous to Rohde et al. (2013a), the core of the inter-
polation approach is to generate a kriging-based field us-
ing an assumed distance-based correlation function. As with
Rohde et al. (2013a), a correlation-based approach is used
rather than the more common covariance-based approach
to simplify the computational considerations and should be
adequate as long as the variance changes relatively slowly
with changes in position. A review of both the HadSST data
and climate model outputs suggested that the temperature-
to-distance correlation function could be modeled effectively
via the same spherical correlation function approach used for
land surface temperatures:
dmax 21+d
2dmax , d < dmax
R(d)=0, d dmax .(1)
The empirically estimated distance parameter dmax was
found to have a value of 2680km based on the spatial vari-
ance of the HadSST monthly averages. This is similar to,
though somewhat smaller than, the 3310 km scale adopted
in the land surface temperature study (Rohde et al., 2013a).
By contrast, the local correlation parameter R0=0.47 was
Earth Syst. Sci. Data, 12, 3469–3479, 2020
R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record 3471
Figure 1. Empirically estimated correlation versus distance for monthly average sea surface temperatures. Correlation was estimated by
comparing root-mean-square differences for all possible pairs of HadSST grid cells and all months and binning the population by distance.
The black curve reflects a best fit for the spherical correlation function model. The red dashed curve shows the corresponding correlation
model derived for land-based measurements (Rohde et al., 2013a).
estimated to be much lower in the oceans (compared to 0.86
on land). This is due to two factors. Firstly, ocean observa-
tions are individual measurements whereas land observations
reflect monthly averages. Secondly, the typical monthly fluc-
tuations in the oceanic environment are much smaller than on
land, causing a reduced signal-to-noise ratio. The estimation
of R0was based on a comparison of the variance in HadSST
grid cells with a single measurement to those with >100
observations. The latter condition provides a proxy for cells
where the random portion of measurement and sampling un-
certainty could plausibly be neglected.
Figure 1 shows an empirically estimated average correla-
tion versus distance between HadSST grid cells. This shows
the empirical length scale, though a larger intercept is used
(0.75), reflecting the fact that the average HadSST grid
cell incorporates many observations. The lower value for R0
represents the typical relationship between a single measure-
ment and the monthly average.
This treatment, using a single scale length for the whole
ocean, simplifies the analysis; however, it does ignore some
of the real variations across the oceans. For example, in
regions with boundary currents, upwelling–downwelling,
or complex ocean-to-land geographies, the scale length of
monthly average temperature variations may be smaller than
suggested here. In practice, the 5 ×5gridding of HadSST
already precludes a detailed analysis of most small features.
The interpolation presented here primarily serves to improve
the representation by smoothing over noise and filling gaps,
but it will not necessarily capture the smallest features.
The distance correlation function gives rise to a kriging
T(x, t)=θt+X
jK(xj,x,t)(SST(xj, t )θt)(2)
D(x1,t)R(kx1x2k)··· R(kx1xNk)
R(kxNx1k)··· D(xN1,t)R(kxN1xNk)
D(xj,t)=1+(Neff(xj, t )1)R0
Neff(xj, t )(4)
Neff(xj, t )=max s2
(σm(xj,t))2,1.(5) Earth Syst. Sci. Data, 12, 3469–3479, 2020
3472 R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record
Here tis the current month, T(x, t) is the interpolated tem-
perature at a general location x, SST(xj, t) is the HadSST
anomaly value in the grid cell centered at location xj,
σm(xj,t) is the measurement uncertainty associated with lo-
cation xj, and smis the average measurement uncertainty of
a single measurement. Neff(xj, t ) is then an effective number
of independent measurements associated with the grid cell.
Though HadSST provides the true number of observations
per cell, N(xj, t), we found that Neff(xj, t ), which incorpo-
rates the measurement uncertainty, appeared to give superior
results than simply relying on the reported number of ob-
servations. The incorporation of Neff(xj, t ) into the determi-
nation of the kriging coefficients Khas the effect of giving
greater weight to grid cells with less uncertainty. For integer
values of Neff(xj, t ), the formulation of D(xj, t ) is mathe-
matically equivalent to having xjappear Neff(xj, t ) indepen-
dent times in the correlation matrix. Note also that any empty
HadSST grid cells at time tare omitted from the matrix for-
mulation for K.
θtis a free parameter at each time tand effectively rep-
resents the global ocean-average temperature anomaly. Its
value is found iteratively by insisting that the spatial average
of T(x, t)θt=0.
It is instructive to note that this kriging formulation has
the property that T(xj,t)SST(xj, t ) in the limit that
Neff(xj, t )→ ∞, but will ordinarily produce a temperature
estimate based on a weighted average of multiple HasSST
grid points in the case that Neff(xj, t ) is small or moder-
ate. The latter property can be useful in suppressing noise
at grid locations with high uncertainty and/or very few mea-
It is also important to recognize that though the correla-
tion function R(d) has a very long tail, this does not mean
that average necessarily extends over a large area. In general,
the kriging coefficients K(xj,x, t ) constructed in this way
will heavily favor the nearest several data points. As long as
nearby data are available, little weight will be given to distant
grid cells. However, the long tail of the correlation function
means that the kriging will attempt to fill large holes using
distant data if no nearby data are available.
An absolute value field was also created by applying a sim-
ilar interpolation to the HadSST climatology.
C(x, m)=
P(x, m)+X
(KB(xj,x,m)(SSTCLIM(xj, m)P(x,m)))
C(x, m) is the interpolated climatology for month m,
SSTCLIM(xj,m) is the reported climatology, and
KB(xj,x,m) is a set of kriging parameters, which are
the same as K(xj, x,m) except that R0and D(xj,t) are both
replaced with 1, effectively treating the SSTCLIM(xj,m) as
if it has no uncertainty. P(xm) is a background prediction
function dependent only on the month and the latitude
of x. It is described as a piecewise cubic spline with 11
knots as free parameters equally spaced in the cosine of
latitude. These free parameters are chosen to minimize
the spatial average of C(x, m)P(x, m). By construction,
C(xj,m)=SSTCLIM(xj, m) for all xjvalues, and this
construction merely provides a way of interpolating between
grid cell centers.
In addition to the above description, a physical cutoff was
applied to the absolute temperature C(x, m)+T(x, t) at a
fixed minimum temperature of 1.8 C, which is the freezing
temperature of seawater. If the interpolation would suggest
a value lower than this, T(x, t ) was adjusted accordingly to
maintain the minimum value of 1.8 C. Such adjustments
are rare.
Finally, one last interpolation is performed using an as-
sumption of temporal persistence. Unlike land temperature
anomalies, where the temporal correlation is often only a
couple weeks, ocean temperature anomalies typically have
a temporal correlation measured in months. This can be ex-
ploited to estimate ocean temperatures based on adjacent
months when no other information is available.
Analogous to Rohde et al. (2013a), a diagnostic criterion
can be constructed V(x, t)=P
K(xj,x,t). Because of the
nature of the kriging coefficients, V(x, t )1 in the presence
of dense data and V(x, t)0 if there are no HadSST data
in the neighborhood of x.
The final estimate of the SST, including a temporal persis-
tence adjustment for regions of low V(x, t ), is then
Tfinal(x, t )=T(x, t )+(1 V(x,t))
V(x, t +1)T(x, t +1) +V(x, t 1)T(x, t 1)
V(x, t +1) +V(x, t 1) θt.
Here, t1 and t+1 refer to the temperature field 1 month ear-
lier and 1 month later, respectively. This adjustment allows
for a modest reduction in uncertainty at early times when
data are temporally sparse.
As described, this analysis is agnostic about the resolu-
tion used to sample the final temperature field. In practice,
we generally use the same 15 984-element equal-area grid
as Rohde et al. (2013a) to calculate Tfinal(x , t ), though with
non-ocean elements masked out.
2.2 Ocean uncertainty
The ocean-average uncertainty in our ocean reconstruction is
estimated following essentially the same model as adopted
by HadSST3. HadSST3 estimates the total reconstruction
uncertainty as the combination of measurement uncertainty,
coverage uncertainty, and bias uncertainty (Kennedy et al.,
2011a, b). Bias uncertainty, σbias, which reflects biases cre-
ated due to variations over time in the ways that SST has
been measured, is brought forward essentially unchanged by
Earth Syst. Sci. Data, 12, 3469–3479, 2020
R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record 3473
Figure 2. Component uncertainties for the ocean average of HadSST v3 and the corresponding transformed forms of those components after
the application of the interpolation scheme described in the text. All uncertainties are expressed as appropriate for 95 % confidence intervals
on annual ocean averages.
our analysis process (Fig. 2). Due to its slowly varying na-
ture, this uncertainty remains the most important limitation
of the detection of long-term averages.
The coverage uncertainty, σcoverage, is the uncertainty in
the large-scale average arising due to incomplete sampling of
the spatial field. As with HadSST3, our estimate of the cover-
age uncertainty is constructed by sampling a known field, ap-
plying our interpolation procedure, and seeing how well we
reproduce the underlying average of the known field. Follow-
ing HadSST3, we used the SST fields provided by HadISST
v2 as our target. The HadISST fields are spatially complete,
observation-based historical reconstructions of SST and sea
ice concentration (Titchner and Rayner, 2014). To estimate
the coverage uncertainty associated with a specified HadSST
sampling field, we mask every month of the HadISST dataset
using that sampling field, interpolate the remaining data, and
measure the error in the interpolated average relative to the
true ocean average of the whole HadISST field. The de-
viations in the ocean average are then collected across all
HadISST months, and the uncertainty for that coverage mask
is reported as the root-mean-square average of the devia-
tions. Using this technique, which is directly analogous to
the HadSST3 coverage assessment technique, we estimate
that the application of our interpolation approach typically
reduces the coverage uncertainty by 20 %–40 % (Fig. 2).
Lastly, we consider the impact of our interpolation on the
measurement and sampling uncertainty. Measurement uncer-
tainty essentially captures the errors in individual observa-
tions, while sampling uncertainty reflects the fact that water
temperatures can vary on timescales shorter than a month and
spatial scales smaller than a grid box. Though interpolation
does not change the underlying uncertainty associated with
individual measurements, by adjusting the weight of individ-
ual observations in the overall average, we affect the way
that individual measurement errors propagate into the global
average. In particular, in the presence of sparse data, lim-
ited measurements may be extrapolated over a large area. In
some circumstances, this can cause the effective uncertainty
in the global average due to these uncertainties to increase.
In essence, the interpolation may trade improvements in cov-
erage uncertainty against a greater impact for measurement
uncertainty. This largely limits our ability to reduce the over-
all uncertainty by interpolation.
The impact of measurement uncertainty on a large-scale
average depends on the error correlation. If the measurement
uncertainties were uncorrelated, then the error would gener-
ally be expected to decline with the square root of the number
of measurements. In actuality, the measurement uncertainties
are frequently correlated. In most cases, single ships report
many measurements per month. Each of those measurements
can have both random errors and a potential for systematic
bias. For a single ship, we cannot expect this bias compo-
nent of a measurement error to be reduced by increasing the
number of observations. In their analysis HadSST3 models
the entire error correlation matrix to understand the effect of
measurement errors on the global average uncertainty.
For HadSST3, the error correlation matrices were not pub-
lished. As a result, it is not possible to exactly determine Earth Syst. Sci. Data, 12, 3469–3479, 2020
3474 R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record
the effect of our interpolation procedure on the measure-
ment uncertainty. However, we can make a reasonable es-
timate. Since HadSST3 releases both the per-grid-cell mea-
surement uncertainties and the global average measurement
uncertainty, we can compare the expected measurement un-
certainty treating all grid cells as independent to what is actu-
ally observed by HadSST3 using the whole error correlation
matrix (Kennedy et al., 2011b).
σuncorrelated =q6j(A(xj)σm(xj,t))2,(8)
where A(xj) is the fraction of the Earth’s oceans represented
by grid cell xjand σuncorrelated is the measurement uncer-
tainty resulting from assuming that the measurement errors
in individual grid cells are uncorrelated with other grid cells.
We find that the measurement uncertainty reported by
HadSST3 in the ocean average is typically 2.1 times larger
than σuncorrelated, with some variation over time.
We use this estimate as a benchmark to approximate the
effect of error correlation on our analysis of measurement
σinterpolated, measurement =
σHadSST, measurement
σuncorrelated q6j(K(xj,t)σm(xj, t ))2(9)
K(xj,t)=ZZ K(xj, x , t )dx.ZZ 1dx(10)
Here the double integral denotes the integral over the surface
of the ocean. Thus K(xj,t) is effectively the weight of the xj
grid point in the global average.
The total uncertainty in the ocean average is then found by
assuming the components are independent.
bias +σ2
coverage +σ2
interpolated, measurement (11)
Over nearly all time periods, we find that interpolation does
reduce the uncertainty associated with missing coverage. In
the early period, the interpolation results in an appreciable
reduction in total uncertainty. However, the total uncertainty
in the global average is little changed in the recent period.
This is because the bias and measurement uncertainties play
a dominant role in the recent period, and the impact of these
uncertainties on the global average is little changed as a result
of the interpolation. However, even if the ocean-average un-
certainty is not changed during the recent period, the interpo-
lation may still aid in the interpretation of local- to regional-
scale features.
2.3 Land and ocean combination
The combined field is constructed by merging the Berke-
ley Earth land surface temperature with the interpolated SST
field described above. Two versions are considered that dif-
fer only in their treatment of sea ice, using either the land air
temperature (LAT) or the SST field to estimate the tempera-
ture anomaly at sea ice locations. From 1850 to near present,
the sea ice locations are estimated using the ice concentration
fields in HadISST v2 (Titchner and Rayner, 2014).
To combine LAT and SST data, both data sets are ex-
pressed on the same grid. To simplify the combination at cells
that are part land and part ocean, we have taken to adding in
the spatial climatology and doing the combination in abso-
lute temperatures.
In the case where sea ice areas are represented by SST, the
combination is straightforward:
Tcombined(x, t )=L(x)TLAT(x, t)+(1L(x))TSST(x, t ),(12)
where L(x) is the fraction of the grid cell at location xthat
is land, and TLAT and TSST are respectively the LAT as esti-
mated by Rohde et al. (2013a) and the interpolated SST as
described above.
In the case where sea ice regions are treated as land,
Tcombined(x, t )=
L(x, t)TLAT(x , t )+(1 L(x, t ))TSST(x, t ),(13)
L(x, t)=L(x)+(1 L(x))I(x, t),(14)
where I(x, t) is the ice fraction at location xat time tas re-
ported by HadISST v2 (Titchner and Rayner, 2014). For this
purpose, HadISST is also regridded onto the same grid as
LAT and SST. As HadISST is frequently delayed by a few
months compared to other climate data, it is necessary to
supplement this data set when producing near-real-time es-
timates. For this purpose, the Sea Ice Index of the National
Snow and Ice Data Center (Fetterer et al., 2017) is used for
months that are not yet available in HadISST. The modern
ice distribution in both HadISST and the Sea Ice Index are
based on satellite observations; however, we found that the
Sea Ice Index tended to have systematically more partial
melting than HadISST. To maintain consistency, a distribu-
tion transform was applied to the sea ice fractions provided
in the Sea Ice Index based on comparing the 2014–2018 ice
fields in each dataset.
It is useful to note that regardless of whether one is us-
ing SST or LAT to estimate temperatures in association with
sea ice, most such estimates involve a considerable extrapo-
lation. In the case of LAT, for example, conditions over sea
ice in the Arctic will usually be extrapolated from Greenland,
Canada, Scandinavia, and Russia. Similarly, in the Antarc-
tic, coastal stations will be extrapolated outward over the ice.
By contrast, when using SST, one extrapolates from rare SST
measurements that may be far removed from the sea ice edge.
Or, in the case that analysis of the sea ice regions is excluded
entirely, averaging methods are effectively substituting the
ocean or global average temperature anomaly.
Earth Syst. Sci. Data, 12, 3469–3479, 2020
R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record 3475
It is our belief that the anomaly field generated by extrap-
olating air temperatures over sea ice locations is a more sen-
sible approach to characterizing climate change at the poles.
The air temperature changes over the sea ice can be quite
large even while the water temperatures underneath are not
changing at all. In particular, over the last decades Arctic air
has shown a very large warming trend during the winter.
Regardless of the approach used, the spatial climatology
can then be calculated and removed (differing from the orig-
inal only in cells with a mix of land and water/sea ice). Then
the long-term trend in the climate can be computed using the
spatial average of the anomaly fields.
Uncertainties for the combined record are calculated by
assuming the uncertainties in LAT and SST time series are
independent and can be combined in proportion to the rela-
tive area of land and ocean. In the case that LAT is used over
sea ice, the uncertainties for both LAT and SST have to be
slightly recalculated by assuming that the time-varying mask
L(x, t) is applied the relevant spatial averages in the uncer-
tainty estimations described in Rohde et al. (2013a) and in
the SST section above. Doing this adjustment causes a slight
increase in LAT uncertainty (due to the extrapolation over
sea ice) and a similar small decrease in SST uncertainty.
3 Data availability
The Berkeley Earth Land/Ocean temperature product will
be updated monthly on the website and is
freely available for use to all interested researchers. A con-
venience copy of the dataset available at the time this paper
was created has been registered with Zenodo and is avail-
able at (Rohde and
Hausfather, 2020).
4 Results and conclusions
The global mean anomalies obtained from the Berkeley Earth
Land/Ocean Temperature Record are quite similar to other
published records, as shown in Fig. 3. With the exception of
some short periods prior to 1880 and before and after World
War 2, all four other temperature records examined lie within
the uncertainty envelope of the Berkeley Earth record. Differ-
ences around World War 2 relate primarily to differences in
adjustments to ERSST v5 and HadSST3 sea surface tempera-
ture records during that period (Huang et al., 2017; Kennedy
et al., 2019; Cowtan et al., 2017).
Berkeley Earth has the highest trend of any temperature
record examined for the period from 1880 to 2015, largely
due to lower surface temperature estimates prior to 1900.
These differences are driven both by increased spatial cover-
age from the inclusion of additional land records and by the
spatial interpolation of both land and ocean records (which
are more limited in both the NOAA and Hadley records).
Similarly, Berkeley Earth has among the highest warming
Figure 3. Comparison of published global surface temperature
records. The top panel shows annual anomalies (relative to a 1961–
1990 baseline period), with the Berkeley Earth uncertainty as the
shaded area. The bottom panel shows trends and two-sigma trend
uncertainties (calculated using an autoregressive–moving average,
ARMA(1,1), approach to account for autocorrelation) for various
starting dates through the end of 2015 based on monthly anomalies.
rates in the recent period (1979–2015) due primarily to
greater Arctic coverage (where warming was unusually rapid
during that period). The other records that provide robust
Arctic interpolation, CW2014 and NASA GISTEMP, also
show higher trends during this period.
From 1955 to present (after the availability of data in
Antarctica), Berkeley Earth provides globally complete cov-
erage via spatial interpolation, similar to NASA’s GISTEMP
and CW2014. This contrasts with HadCRUT4 which ex-
cludes any grid cells lacking station coverage or SST mea-
surements, or NOAA GlobalTemp where interpolation is
more limited. As shown in Fig. 4, the patterns of spatial
anomalies between the different groups tend to be quite sim-
ilar, apart from differences due to spatial coverage or gridded
field resolution.
When constructing a global surface temperature record,
sea ice produces a challenging edge case. The water tem-
perature under sea ice is tightly constrained by the freez-
ing point of water and can only change with changes in sea
ice cover. Air temperatures over sea ice are less well con-
strained and can vary significantly over time. Whether areas
with sea ice coverage are estimated using sea surface temper-
atures or surface air temperatures will have a notable effect
on the record. While most groups (GISTEMP, CW2014) that
interpolate temperatures over areas with sea ice cover use air
temperatures, Berkeley Earth has provided both variants to Earth Syst. Sci. Data, 12, 3469–3479, 2020
3476 R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record
Figure 4. Global gridded temperature anomalies for December 2015 relative to a 1961–1990 baseline for each global temperature dataset.
Grid resolution is based on the highest-resolution dataset provided by each group: 1×1lat–long for Berkeley Earth, 5×5for HadCRUT4,
1×1for NASA GISTEMP, 5×5over land and 2×2over oceans for NOAA GlobalTemp, and 5×5for Cowtan and Way (CW2014).
Figure 5. (a) Two variants of the Berkeley Earth global surface temperature product estimating temperatures under sea ice based on SSTs
(red) or proximate air temperature measurements (blue), as well as the HadCRUT temperature series for comparison. (b) The same two
versions of the Berkeley Earth data set with the HadCRUT time series subtracted.
Earth Syst. Sci. Data, 12, 3469–3479, 2020
R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record 3477
Figure 6. Berkeley Earth average absolute climatology for the period from 1951 to 1980 with the air temperature at sea ice (a) and ocean
temperature under sea ice (b) variants shown.
Figure 7. Comparison of published annual uncertainty estimates (two sigma) for Berkeley Earth, HadCRUT4 (Morice et al., 2012), GIS-
TEMP (Lenssen et al., 2019), GlobalTempv5 (Vose et al., 2012), and Cowtan and Way (2014).
allow researchers to select the series that best supports their
needs. We consider the variant using air temperature above
sea ice to be a better description of global climate change,
but the ocean temperature variants may be useful for compar-
ison and for certain specialists. Both variants of the Berkeley
Earth record are shown in Fig. 5 as well as the HadCRUT
temperature series for comparison. When SSTs under sea ice
are used, the apparent warming trend in recent years is lower
than when air temperatures are used. Comparing these ver-
sions helps to reveal the contribution of sea ice areas to the
overall global warming rate.
Figure 5 also aids in understanding the difference between
Berkeley Earth and HadCRUT. The interpolated SST field
adopted here has a nearly identical trend to the HadSST field,
differing by less than 0.01 C per century. Part of the dif-
ference between Berkeley Earth’s global temperature series
and HadCRUT is due to differences in the amount of warm-
ing estimated to have occurred over land. This is the primary Earth Syst. Sci. Data, 12, 3469–3479, 2020
3478 R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record
source of difference when comparing the Berkeley Earth se-
ries with SST at sea ice to the HadCRUT series (blue line in
Fig. 5). While this difference is not insignificant, the larger
overall difference is due to the incorporation of air tempera-
ture warming in sea ice regions, especially in the Arctic (red
line in Fig. 5). Inclusion of the rapid warming above Arctic
sea ice suggests the global average has increased an addi-
tional 0.1 C during the last 100 years compared to esti-
mates that do not include the changes in this region.
In addition to monthly temperature anomalies, Berkeley
Earth produces monthly absolute temperature fields. A cli-
matology field is estimated via kriging observations, using
elevation as a factor in the kriging process over land. Both ab-
solute temperature variants with air temperature over sea ice
and water temperature under sea ice are available, as shown
in Fig. 6. Absolute temperatures are created by adding the
climatology field to monthly anomalies.
Figure 7 provides a comparison between published un-
certainties (two sigma) for each of the major global land–
ocean temperature series. The Berkeley Earth, GISTEMP,
and CW2014 records have the lowest uncertainty of the
groups providing annual values, in part due to their spatial
interpolation reducing the uncertainty associated with cover-
The Berkeley Earth Land/Ocean surface temperature
record presented here has already been used by a number of
publications (e.g., Jones, 2015; Thorne et al., 2016; Sutton et
al., 2015). It joins a number of existing land–ocean surface
temperature products that help provide a diverse examination
of the Earth’s changing climate since 1850 and can be used
for diverse applications including climate model validation,
estimating transient climate response, examining changes in
extreme events, and other research areas.
Author contributions. RR designed and implemented the dataset
construction. ZH provided feedback, graphics, and analysis. RR and
ZH jointly prepared the manuscript.
Competing interests. The authors declare that they have no con-
flict of interest.
Review statement. This paper was edited by David Carlson and
reviewed by two anonymous referees.
Cowtan, K. and Way, R. G.: Coverage bias in the HadCRUT4 tem-
perature series and its impact on recent temperature trends, Q. J.
Roy. Meteor. Soc., 140, 1935–1944, 2014.
Cowtan, K., Hausfather, Z., Hawkins, E., Jacobs, P., Mann,
M. E., Miller, S. K., Steinman, B. A., Stolpe, M. B.,
and Way, R. G.: Robust comparison of climate models
with observations using blended land air and ocean sea
surface temperatures, Geophys. Res. Lett., 42, 6526–6534,, 2015.
Cowtan, K., Rohde, R., and Hausfather, Z.: Evaluating bi-
ases in sea surface temperature records using coastal
weather stations, Q. J. Roy. Meteor. Soc., 144, 670–681,, 2017.
Fetterer, F., Knowles, K., Meier, W. N., Savoie, M., and
Windnagel, A. K.: Sea Ice Index, Version 3. Boulder, Col-
orado USA, NSIDC: National Snow and Ice Data Center,, 2017.
Freeman, E., Woodruff, S. D., Worley, S. J., Lubker, S. J., Kent, E.
C., Angel, W. E., Berry, D. I., Brohan, P., Eastman, R., Gates, L.,
Gloeden, W., Ji, Z., Lawrimore, J., Rayner, N. A., Rosenhagen,
G., and Smith, S. R.: ICOADS Release 3.0: A major update to the
historical marine climate record, Int. J. Climatol. 37, 2211–2237,, 2017.
Hansen, J., Ruedy, R., Sato, M., and Lo, K.: Global Sur-
face Temperature Change, Rev. Geophys., 48, RG4004,, 2010.
Huang, B., Thorne, P. W., Banzon, V. F., Boyer, T., Chep-
urin, G., Lawrimore, J. H., Menne, M. J., Smith, T. M.,
Vose, R. S., and Zhang, H.-M.: Extended Reconstructed
Sea Surface Temperature, Version 5 (ERSSTv5): Upgrades,
Validations, and Intercomparisons, J. Clim., 30, 8179–8205,, 2017.
Huang, B., Menne, M. J., Boyer, T., Freeman, E., Gleason, B. E.,
Lawrimore, J. H., Liu, C., Rennie, J. J., Schreck, C. J., Sun, F.,
Vose, R., Williams, C. N., Yin, X., and Zhang, H.-M.: Uncer-
tainty Estimates for Sea Surface Temperature and Land Surface
Air Temperature in NOAAGlobalTemp Version 5, J. Clim., 33,
1351–1379,, 2020.
Ishihara, K.: Calculation of Global Surface Temperature Anomalies
with COBE-SST, (Japanese) Weather Service Bulletin 73, 2006.
Ishii, M., Shouji, A., Sugimoto, S., and Matsumoto, T.: Objective
analyses of sea-surface temperature and marine meteorological
variables for the 20th century using ICOADS and the Kobe Col-
lection, Int. J. Climatol., 25, 865–879, 2005.
Jones, P.: The Reliability of Global and Hemispheric Surface Tem-
perature Records, Adv. Atmos. Sci., 33, 1–14, 2015.
Jones, G. S.: “Apples and Oranges”: on comparing simulated his-
toric near surface temperatures changes with observations, Q. J.
Roy. Meteor. Soc.,, 2020.
Kennedy, J. J., Rayner, N. A., Smith, R. O., Parker, D. E., and
Saunby, M: Reassessing biases and other uncertainties in sea
surface temperature observations measured in situ since 1850:
2. Biases and homogenization, J. Geophys. Res.-Atmos., 116,
D14104,, 2011a.
Kennedy J. J., Rayner, N. A., Smith, R. O., Saunby, M., and
Parker, D. E.: Reassessing biases and other uncertainties in sea-
surface temperature observations since 1850 part 1: measurement
and sampling errors, J. Geophys. Res.-Atmos., 116, D14103,, 2011b.
Kennedy, J. J., Rayner, N. A., Atkinson, C. P., and Killick,
R. E.: An Ensemble Data Set of Sea Surface Tempera-
ture Change From 1850: The Met Office Hadley Centre
HadSST. Data Set, J. Geophys. Res.-Atmos., 124, 7719–
7763,, 2019.
Earth Syst. Sci. Data, 12, 3469–3479, 2020
R. A. Rohde and Z. Hausfather: Berkeley Earth Land/Ocean Temperature Record 3479
Kent, E. C., Rayner, N. A., Berry, D. I., Saunby, M., Moat, B. I.,
Kennedy, J. J., and Parker, D. E.: Global analysis of night ma-
rine air temperature and its uncertainty since 1880: the Had-
NMAT2 Dataset, J. Geophys. Res.-Atmos., 118, 1281–1298,, 2013.
Lawrimore, J. H., Menne, M. J., Gleason, B. E., Williams, C. N.,
Wuertz, D. B., Vose, R. S., and Rennie, J.: An overview of
the Global Historical Climatology Network monthly mean tem-
perature data set, version 3, J. Geophys. Res., 116, D19121,, 2011.
Lenssen, N. J. L., Schmidt, G. A., Hansen, J. E., Menne, M.
J., Persin, A., Ruedy, R., and Zyss, D.: Improvements in the
GISTEMP Uncertainty Model, J. Geophysi. Res.-Atmos., 124,
6307–6326,, 2019.
Menne, M. J. and Williams, C. N.: Homogenization of temperature
series via pairwise comparisons, J. Clim., 22, 1700–1717, 2009.
Menne, M. J., Williams, C. N., Gleason, B. E., Rennie, J. J., and
Lawrimore, J. H.: The Global Historical Climatology Network
Monthly Temperature Dataset, Version 4, J. Clim., 31, 9835–
9854,, 2018.
Morice, C. P., Kennedy, J. J., Rayner, N. A., and Jones, P.
D.: Quantifying uncertainties in global and regional tempera-
ture change using an ensemble of observational estimates: The
HadCRUT4 data set, J. Geophys. Res.-Atmos., 117, D08101,, 2012.
Morice, C. P., Kennedy, J. J., Rayner, N. A., Winn, J. P., Hogan,
E., Killick, R. E., Dunn, R. J. H., Osborn, T. J., Jones, P. D.,
and Simpson, I. R.: An updated assessment of near-surface tem-
perature change from 1850: the HadCRUT5 dataset, J. Geophys.
Res., submitted, 2020.
Richardson, M., Cowtan, K., Hawkins, E., and Stolpe, M. B.: Rec-
onciled climate response estimates from climate models and
the energy budget of Earth, Nat. Clim. Change, 6, 931–935,, 2016.
Rohde, R. and Hausfather, Z.: Berkeley Earth Combined Land
and Ocean Temperature Field, Jan 1850–Nov 2019, Zenodo,, 2020.
Rohde, R., Muller, R. A., Jacobsen, R., Muller, E., Perlmuller,
S., Rosenfeld, A., Wurtele, J., Groom, D., and Wickham, C.:
A New Estimate of the Average Earth Surface Land Tempera-
ture Spanning 1753 to 2011, Geoinfor. Geostat. Overv., 1, 1–7, 2013a.
Rohde, R., Muller, R. A., Jacobsen, R., Perlmuller, S., Rosenfeld,
A., Wurtele, J., Curry, J., Wickham, C., and Mosher, S.: Berke-
ley Earth Temperature Averaging Process, Geoinfor. Geostat.
Overv., 1, 2,, 2013b.
Smith, T. M., Reynolds, R. W., Peterson, T. C., and Lawrimore, J.:
Improvements to NOAA’s Historical Merged Land–Ocean Sur-
face Temperature Analysis (1880–2006), J. Climate, 21, 2283–
2296, 2008.
Sutton, R., Suckling, E., and Hawkins, E.: What does global mean
temperature tell us about local climate?, Philos. T. R. Soc. A,
373, 2054, 2015.
Thorne, P. W., Donat, M. G., Dunn, R. J. H., Williams, C. N.,
Alexander, L. V., Caesar, J., Durre, I., Harris, I., Hausfather, Z.,
Jones, P. D., Menne, M. J., Rohde, R., Vose, R. S., Davy, R.,
Klein-Tank, A. M. G., Lawrimore, J. H., Peterson, T. C., and
Rennie, J. J.: Reassessing changes in diurnal temperature range:
Intercomparison and evaluation of existing global data set esti-
mates, J. Geophys. Res.-Atmos. 121, 5138–5158, 2016.
Titchner, H. A. and Rayner, N. A.: The Met Office Hadley Centre
sea ice and sea surface temperature data set, version 2: 1. Sea
ice concentrations, J. Geophys. Res.-Atmos., 119, 2864–2889,, 2014.
Vose, R. S., Arndt, D., Banzon, V. F., Easterling, D. R., Gleason, B.,
Huang, B., Kearns, E., Lawrimore, J. H., Menne, M. J., Peter-
son, T. C., Reynolds, R. W., Smith, T. M., Williams, C. N., and
Wuertz, D. B.: NOAA’s Merged Land–Ocean Surface Tempera-
ture Analysis, B. Am. Meteorol. Soc., 93, 1677–1685, 2012. Earth Syst. Sci. Data, 12, 3469–3479, 2020
... Under a warming climate, increasing temperatures exacerbate the evaporation of water, a process that in turn is enhanced by increasing levels of water vapor, the world's most dominant greenhouse gas. To establish the link between the growth rates of PWV and temperature, etc. and to analyze the correlation between them, we obtained global temperature anomalies and sea height variations from Berkeley Earth [53] and NASA's Goddard Institute for Space Studies (GISS) [54] for the last two decades and compared PWV anomalies with them. After data processing, the results of the comparison are shown Under a warming climate, increasing temperatures exacerbate the evaporation of water, a process that in turn is enhanced by increasing levels of water vapor, the world's most dominant greenhouse gas. ...
... After data processing, the results of the comparison are shown Under a warming climate, increasing temperatures exacerbate the evaporation of water, a process that in turn is enhanced by increasing levels of water vapor, the world's most dominant greenhouse gas. To establish the link between the growth rates of PWV and temperature, etc. and to analyze the correlation between them, we obtained global temperature anomalies and sea height variations from Berkeley Earth [53] and NASA's Goddard Institute for Space Studies (GISS) [54] for the last two decades and compared PWV anomalies with them. After data processing, the results of the comparison are shown in Figure 12, where the red points represent the global annual values of anomalies (variations), the red error bars are the yearly fluctuations in temperature anomalies and sea height variations, and the blue error bars are the yearly fluctuation of PWV anomalies. ...
Full-text available
Precipitable water vapor (PWV) is an important component in the climate system and plays a pivotal role in the global water and energy cycles. Over the years, many approaches have been devised to accurately estimate the PWV. Among them, global navigation satellite systems (GNSS) have become one of the most promising and fastest-growing PWV acquisition methods because of its high accuracy, high temporal and spatial resolution, and ability to acquire PWV in all weather and in near real time. We compared GNSS-derived PWV with a 5 min resolution globally distributed over 14,000 stations from the Nevada Geodetic Laboratory (NGL) from 1994 to 2020 with global radiosonde (RS) data, temperature anomalies, and sea height variations. Then, we examined the temporal and spatial variability of the global PWV and analyzed its climate implications. On a global scale, the average bias and root mean square error (RMSE) between GNSS PWV and RS PWV were ~0.72 ± 1.29 mm and ~2.56 ± 1.13 mm, respectively. PWV decreased with increasing latitude, and the rate of this decrease slowed down at latitudes greater than 35°, with standard deviation (STD) values reaching a maximum at latitudes less than 35°. The global average linear trend was ~0.64 ± 0.81 mm/decade and strongly correlated with temperature and sea height variations. For each 1 °C and 1 mm change, PWV increased by ~2.075 ± 0.765 mm and ~0.015 ± 0.005 mm, respectively. For the time scale, the PWV content peaked ~40 days after the maximum solar radiation of the year (the summer solstice), and the delay was ~40 days relative to the summer solstice.
... For the near-surface air temperature, we used three in-situ temperature records and one reanalysis dataset. For in-situ datasets, we used NASA's Goddard Institute for Space Studies Surface Temperature version 4 (GISTEMP) 64 , the Berkeley Earth temperature dataset (BEST) 54 , and the Met Office Hadley Centre/ Climatic Research Unit version (HadCRUT5) 55 . ...
Full-text available
In recent decades, the warming in the Arctic has been much faster than in the rest of the world, a phenomenon known as Arctic amplification. Numerous studies report that the Arctic is warming either twice, more than twice, or even three times as fast as the globe on average. Here we show, by using several observational datasets which cover the Arctic region, that during the last 43 years the Arctic has been warming nearly four times faster than the globe, which is a higher ratio than generally reported in literature. We compared the observed Arctic amplification ratio with the ratio simulated by state-of-the-art climate models, and found that the observed four-fold warming ratio over 1979–2021 is an extremely rare occasion in the climate model simulations. The observed and simulated amplification ratios are more consistent with each other if calculated over a longer period; however the comparison is obscured by observational uncertainties before 1979. Our results indicate that the recent four-fold Arctic warming ratio is either an extremely unlikely event, or the climate models systematically tend to underestimate the amplification. Over the past four decades, Arctic Amplification - the ratio of Arctic to global warming - has been much stronger than thought, and is probably underestimated in climate models, suggest analyses of observations and the CMIP5 and CMIP6 simulations.
... We employ two homogenized data sets for this analysis. The first is the Berkeley Earth Surface Temperatures (BEST) data set, which utilizes over 40,000 weather stations around the world to create 1° × 1° gridded estimates of daily temperature from 1880 to the present (R. A. Rohde & Hausfather, 2020;R. Rohde et al., 2013). ...
Full-text available
Plain Language Summary Downscaling methods and models yield climate projections on local scales relevant to entities such as cities and individual facilities. Weighted averages of future projections are needed so that the huge amount of existing climate model data is curated to a set that is compact enough to be widely‐accessible to end users. Observations are used to judge which models most faithfully represent historical climate. In California, previous work has ostensibly observed coastal cooling and inland warming over the later part of the 20th century. Upon closer inspection, we determine that this California coastal cooling finding is likely an artifact due to historical changes in observing practices at weather stations. We then evaluate and weight downscaled models using temperature records that show California coastal cooling and separately develop weights using temperature records that do not show such cooling. When we carefully consider the information in the observational data, we find that the two approaches to weighting produce about the same results. However, the weights can significantly differ when observational uncertainty is not considered. Together, these results highlight the importance, even over a heavily‐instrumented area like California, of carefully evaluating historical climate model simulations when producing future, localized climate projections for end users.
... We use 1960-2014 temperature observations from Berkeley Earth (Rohde & Hausfather, 2020), the 20th Century Reanalysis (Compo et al., 2011), and the University of Delaware (Willmott, 2000) (UDel), regridded to a common 1°-by-1° resolution and averaged together. We similarly average precipitation from Global Precipitation Climatology Centre (Schneider et al., 2011) and UDel at a 0.5°-by-0.5° ...
Full-text available
Quantifying which nations are culpable for the economic impacts of anthropogenic warming is central to informing climate litigation and restitution claims for climate damages. However, for countries seeking legal redress, the magnitude of economic losses from warming attributable to individual emitters is not known, undermining their standing for climate liability claims. Uncertainties compound at each step from emissions to global greenhouse gas (GHG) concentrations, GHG concentrations to global temperature changes, global temperature changes to country-level temperature changes, and country-level temperature changes to economic losses, providing emitters with plausible deniability for damage claims. Here we lift that veil of deniability, combining historical data with climate models of varying complexity in an integrated framework to quantify each nation’s culpability for historical temperature-driven income changes in every other country. We find that the top five emitters (the United States, China, Russia, Brazil, and India) have collectively caused US$6 trillion in income losses from warming since 1990, comparable to 14% of annual global gross domestic product; many other countries are responsible for billions in losses. Yet the distribution of warming impacts from emitters is highly unequal: high-income, high-emitting countries have benefited themselves while harming low-income, low-emitting countries, emphasizing the inequities embedded in the causes and consequences of historical warming. By linking individual emitters to country-level income losses from warming, our results provide critical insight into climate liability and national accountability for climate policy.
... Measurements of monthly topofatmosphere longwave radiative fluxes used in this study were from the Clouds and Earth's Radiation Energy System SYN1degEd4A product, which combines data from Aqua and Terra 51 . Monthly gridded surface temperature data (land and ocean) are from the Berkeley Earth Surface Temperatures dataset 52 . Monthly SLP data are from the Hadley Centre Sea Level Pressure Dataset 53 . ...
Full-text available
Precipitation efficiency (PE) relates cloud condensation to precipitation and intrinsically binds atmospheric circulation to the hydrological cycle. Due to PE’s inherent microphysical dependencies, definitions and estimates vary immensely. Consequently, PE’s sensitivity to greenhouse warming and implications for climate change are poorly understood. Here, we quantify PE’s role in climate change by defining a simple index ϵ as the ratio of surface precipitation to condensed water path. This macroscopic metric is reconcilable with microphysical PE measures and higher ϵ is associated with stronger mean Walker circulation. We further find that state-of-the-art climate models disagree on the sign and magnitude of future ϵ changes. This sign disagreement originates from models’ convective parameterizations. Critically, models with increasing ϵ under greenhouse warming, in line with cloud-resolving simulations, show greater slowdown of the large-scale Hadley and Walker circulations and a two-fold greater increase in extreme rainfall than models with decreasing ϵ. Falling raindrops play an essential but as-yet unquantified role in planetary climate change. Here the authors use the concept of precipitation efficiency to establish that raindrops play a critical role in predicting future tropical atmospheric circulation and extreme precipitation.
Current greenhouse gas emissions will continue to affect the climate even after we reach net-zero emissions. We must understand how and prepare for a cooling planet.
The mixture of Gaussian processes (MGP) is a powerful model, which is able to characterize data generated by a general stochastic process. However, conventional MGPs assume the input variable obeys certain probabilistic distribution, thus cannot effectively handle the case where the input variable lies on a general manifold or a graph. In this paper, we first clarify the relationship between the MGP prediction strategy and the attention mechanism. Based on the attention mechanism, we further design two novel mixture models of Gaussian processes, which do not rely on probabilistic assumptions on the input domain, thus overcoming the difficulty of extending MGP models to manifold or graph. Experimental results on real-world datasets demonstrate the effectiveness of the proposed methods.
Full-text available
Recent soil moisture (SM) reconstructions revealed plunging trends and enhanced SM-temperature couplings over the last two decades in dry regions. However, how SM changed and whether the land-atmosphere interaction was intensified over time in humid regions remained unknown. Here we reported the first six-century-long regional summer SM reconstruction (1360–2000 CE) in western Europe (WE) using three individual tree-ring δ 18O chronologies in England and France. A sharp wet-to-dry change occurred around 1820, earlier than 1850–1900 CE, the commonly used historical baseline of anthropogenic climate changes. Enhanced coupling of SM-temperature followed, with stronger summer sea level pressure anomalies in dry years after the 1820s. Our results reveal that the hotter-drier regime has also become more frequent in humid WE under global warming.
Concentrations of greenhouse gases reached record levels in 2021, and energy continued to accumulate in the climate system with the ocean heat content reaching a new record high. La Niña events at the start and end of 2021, however, meant that the global mean temperature in 2021 was only between 5th and 7th warmest year on record. We give a summary of the global climate with a focus on the UK and Europre.
Full-text available
Evaluating historical simulations from global climate models (GCMs) remains an important exercise for better understanding future projections of climate change and variability in rapidly warming regions, such as the Arctic. As an alternative approach for comparing climate models and observations, we set up a machine learning classification task using a shallow artificial neural network (ANN). Specifically, we train an ANN on maps of annual mean near-surface temperature in the Arctic from a multi-model large ensemble archive in order to classify which GCM produced each temperature map. After training our ANN on data from the large ensembles, we input annual mean maps of Arctic temperature from observational reanalysis and sort the prediction output according to increasing values of the ANN’s confidence for each GCM class. To attempt to understand how the ANN is classifying each temperature map with a GCM, we leverage a feature attribution method from explainable artificial intelligence. By comparing composites from the attribution method for every GCM classification, we find that the ANN is learning regional temperature patterns in the Arctic that are unique to each GCM relative to the multi-model mean ensemble. In agreement with recent studies, we show that ANNs can be useful tools for extracting regional climate signals in GCMs and observations.
Full-text available
This analysis estimates uncertainty in the NOAA global surface temperature (GST) version 5 (NOAAGlobalTemp v5) product, which consists of sea surface temperature (SST) from the Extended Reconstructed SST version 5 (ERSSTv5) and land surface air temperature (LSAT) from the Global Historical Climatology Network monthly version 4 (GHCNm v4). Total uncertainty in SST and LSAT consists of parametric and reconstruction uncertainties. The parametric uncertainty represents the dependence of SST/LSAT reconstructions on selecting 28/6 internal parameters of SST/LSAT, and is estimated by a 1000-member ensemble from 1854 to 2016. The reconstruction uncertainty represents the residual error of using a limited number of 140/65 modes for SST/LSAT. Uncertainty is quantified at the global scale as well as the local grid scale. Uncertainties in SST and LSAT at the local grid scale are larger in the earlier period (1880s–1910s) and during the two world wars due to sparse observations, then decrease in the modern period (1950s–2010s) due to increased data coverage. Uncertainties in SST and LSAT at the global scale are much smaller than those at the local grid scale due to error cancellations by averaging. Uncertainties are smaller in SST than in LSAT due to smaller SST variabilities. Comparisons show that GST and its uncertainty in NOAAGlobalTemp v5 are comparable to those in other internationally recognized GST products. The differences between NOAAGlobalTemp v5 and other GST products are within their uncertainties at the 95% confidence level.
Full-text available
One of the largest sources of uncertainty in estimates of global temperature change is that associated with the correction of systematic errors in sea surface temperature (SST) measurements. Despite recent work to quantify and reduce these errors throughout the historical record, differences between analyses remain larger than can be explained by the estimated uncertainties. We revisited the method used to estimate systematic errors and their uncertainties in version 3 of the Met Office Hadley Centre SST data set, HadSST. Using comparisons with oceanographic temperature profiles, we make estimates of biases associated with engine room measurements and insulated buckets and constrain the ranges of two of the more uncertain parameters in the bias estimation: the timing of the transition from uninsulated to insulated buckets in the middle twentieth century and the estimated fractions of different measurement methods used. Here, we present HadSST., based on release 3.0.0 and 3.0.1 of the International Comprehensive Ocean‐Atmosphere Data Set supplemented by drifting buoy measurements from the Copernicus Marine Environmental Monitoring Service. HadSST. comprises a 200‐member “ensemble” in which uncertain parameters in the SST bias scheme are varied to generate a range of adjustments. The evolution of global average SST in the new data set is similar to that in other SST data sets, and the difference between data sets is reduced during the middle twentieth century. However, the changes also highlight a discrepancy in the global‐average difference between adjusted SST and marine air temperature in the early 1990s and hence between HadSST. and, the National Oceanic and Atmospheric Administration SST data set, ERSSTv5.
Full-text available
We outline a new and improved uncertainty analysis for the Goddard Institute for Space Studies Surface Temperature product version 4 (GISTEMP v4). Historical spatial variations in surface temperature anomalies are derived from historical weather station data and ocean data from ships, buoys, and other sensors. Uncertainties arise from measurement uncertainty, changes in spatial coverage of the station record, and systematic biases due to technology shifts and land cover changes. Previously published uncertainty estimates for GISTEMP included only the effect of incomplete station coverage. Here, we update this term using currently available spatial distributions of source data, state‐of‐the‐art reanalyses, and incorporate independently derived estimates for ocean data processing, station homogenization, and other structural biases. The resulting 95% uncertainties are near 0.05 °C in the global annual mean for the last 50 years and increase going back further in time reaching 0.15 °C in 1880. In addition, we quantify the benefits and inherent uncertainty due to the GISTEMP interpolation and averaging method. We use the total uncertainties to estimate the probability for each record year in the GISTEMP to actually be the true record year (to that date) and conclude with 87% likelihood that 2016 was indeed the hottest year of the instrumental period (so far).
Full-text available
Sea surface temperatures form a vital part of global mean surface temperature records. Historical observation methods have changed substantially over time from buckets to engine room intake sensors, hull sensors and drifting buoys, rendering their use for climatological studies problematic. There are substantial uncertainties in the relative biases of different observations which may impact the global temperature record. Island and coastal weather stations can be compared to coastal sea surface temperature observations to obtain an assessment of changes in bias over time. The process is made more challenging by differences in the rate of warming between air temperatures and sea surface temperatures, and differences across coastal boundaries. A preliminary sea surface temperature reconstruction homogenized using coastal weather station data suggests significant changes to the sea surface temperature record, although there are substantial uncertainties of which only some can be quantified. A large warm excursion in versions 4 and 5 of the NOAA Extended Reconstructed Sea Surface Temperature during World War 2 is rejected, as is a cool excursion around 1910 present in all existing records. The mid-century plateau is cooler than in existing reconstructions.
We present a new version of the Met Office Hadley Centre/Climatic Research Unit global surface temperature data set, HadCRUT5. HadCRUT5 presents monthly average near-surface temperature anomalies, relative to the 1961–1990 period, on a regular 5° latitude by 5° longitude grid from 1850 to 2018. HadCRUT5 is a combination of sea-surface temperature (SST) measurements over the ocean from ships and buoys and near-surface air temperature measurements from weather stations over the land surface. These data have been sourced from updated compilations and the adjustments applied to mitigate the impact of changes in SST measurement methods have been revised. Two variants of HadCRUT5 have been produced for use in different applications. The first represents temperature anomaly data on a grid for locations where measurement data are available. The second, more spatially complete, variant uses a Gaussian process based statistical method to make better use of the available observations, extending temperature anomaly estimates into regions for which the underlying measurements are informative. Each is provided as a 200-member ensemble accompanied by additional uncertainty information. The combination of revised input data sets and statistical analysis results in greater warming of the global average over the course of the whole record. In recent years, increased warming results from an improved representation of Arctic warming and a better understanding of evolving biases in SST measurements from ships. These updates result in greater consistency with other independent global surface temperature data sets, despite their different approaches to data set construction, and further increase confidence in our understanding of changes seen. © 2020 Crown Copyright. This article is published with the permission of the Controller of HMSO and the Queen's Printer for Scotland.
Simulated historic near surface air temperature variations are often compared with observations of land air temperatures blended with sea surface temperatures. This study investigates claims that this is not a “true like‐with‐like” comparison, which may cause small biases in simulated 20th century temperature changes, with implications for different climate attribution and projection studies. A more appropriate analysis, it is claimed, should use simulated sea surface temperatures blended with land air temperatures, then an apparent discrepancy with observed trends is reduced. As the temperature of the uppermost level in a model's ocean is used to represent simulated sea surface temperatures, that models have inconsistent ways of representing land, and that simulations have differing sea ice coverages, the claim of an idealised analysis approach is challenged. An examination of Coupled Model Intercomparison Project simulations, compared with an observational dataset of near surface temperatures, suggests there is a bias in simulated historic trends when upper ocean temperatures are used instead of marine air temperatures, but this bias is small compared to other model and observational uncertainties and the impact of analysis choices. The results indicate that it is generally appropriate to use global near surface air temperature diagnostics to compare simulated historic climate change with observed temperature changes. Alternative model diagnostics are not necessarily superior to those used in standard approaches and the emphasis of model and observational discrepancies may be based on over‐confident reasoning. This article is protected by copyright. All rights reserved.
We describe a fourth version of the Global Historical Climatology Network (GHCN)-monthly (GHCNm) temperature dataset. Version 4 (v4) fulfills the goal of aligning GHCNmtemperature values with the GHCNdaily dataset and makes use of data from previous versions of GHCNm as well as data collated under the auspices of the International Surface Temperature Initiative. GHCNm v4 has many thousands of additional stations compared to version 3 (v3) both historically and with short time-delay updates. The greater number of stations as well as the use of records with incomplete data during the base period provides for greater global coverage throughout the record compared to earlier versions. Like v3, the monthly averages are screened for random errors and homogenized to address systematic errors. New to v4, uncertainties are calculated for each station series, and regional uncertainties scale directly from the station uncertainties. Correlated errors in the station series are quantified by running the homogenization algorithm as an ensemble. Additional uncertainties associated with incomplete homogenization and use of anomalies are then incorporated into the station ensemble. Further uncertainties are quantified at the regional level, the most important of which is for incomplete spatial coverage. Overall, homogenization has a smaller impact on the v4 global trend compared to v3, though adjustments lead to much greater consistency than between the unadjusted versions. The adjusted v3 global mean therefore falls within the range of uncertainty for v4 adjusted data. Likewise, annual anomaly uncertainties for the other major independent land surface air temperature datasets overlap with GHCNm v4 uncertainties.
The monthly global 2° × 2° Extended Reconstructed Sea Surface Temperature (ERSST) has been revised and updated from version 4 to version 5. This update incorporates a new release of ICOADS release 3.0 (R3.0), a decade of near-surface data from Argo floats, and a new estimate of centennial sea ice from HadISST2. A number of choices in aspects of quality control, bias adjustment, and interpolation have been substantively revised. The resulting ERSST estimates have more realistic spatiotemporal variations, better representation of high-latitude SSTs, and ship SST biases are now calculated relative to more accurate buoy measurements, while the global long-term trend remains about the same. Progressive experiments have been undertaken to highlight the effects of each change in data source and analysis technique upon the final product. The reconstructed SST is systematically decreased by 0.077°C, as the reference data source is switched from ship SST in ERSSTv4 to modern buoy SST in ERSSTv5. Furthermore, high-latitude SSTs are decreased by 0.1°-0.2°C by using sea ice concentration from HadISST2 over HadISST1. Changes arising from remaining innovations are mostly important at small space and time scales, primarily having an impact where and when input observations are sparse. Cross validations and verifications with independent modern observations show that the updates incorporated in ERSSTv5 have improved the representation of spatial variability over the global oceans, the magnitude of El Niño and La Niña events, and the decadal nature of SST changes over 1930s-40s when observation instruments changed rapidly. Both long- (1900-2015) and short-term (2000-15) SST trends in ERSSTv5 remain significant as in ERSSTv4.
Climate risks increase with mean global temperature, so knowledge about the amount of future global warming should better inform risk assessments for policymakers. Expected near-term warming is encapsulated by the transient climate response (TCR), formally defined as the warming following 70 years of 1% per year increases in atmospheric CO2 concentration, by which point atmospheric CO2 has doubled. Studies based on Earth’s historical energy budget have typically estimated lower values of TCR than climate models, suggesting that some models could overestimate future warming. However, energy-budget estimates rely on historical temperature records that are geographically incomplete and blend air temperatures over land and sea ice with water temperatures over open oceans. We show that there is no evidence that climate models overestimate TCR when their output is processed in the same way as the HadCRUT4 observation-based temperature record3, 4. Models suggest that air-temperature warming is 24% greater than observed by HadCRUT4 over 1861–2009 because slower-warming regions are preferentially sampled and water warms less than air5. Correcting for these biases and accounting for wider uncertainties in radiative forcing based on recent evidence, we infer an observation-based best estimate for TCR of 1.66 °C, with a 5–95% range of 1.0–3.3 °C, consistent with the climate models considered in the IPCC 5th Assessment Report.
We highlight improvements to the International Comprehensive Ocean-Atmosphere Data Set (ICOADS) in the latest Release 3.0 (R3.0; covering 1662–2014). ICOADS is the most widely used freely available collection of surface marine observations, providing data for the construction of gridded analyses of sea surface temperature, estimates of air–sea interaction and other meteorological variables. ICOADS observations are assimilated into all major atmospheric, oceanic and coupled reanalyses, further widening its impact. R3.0 therefore includes changes designed to enable effective exchange of information describing data quality between ICOADS, reanalysis centres, data set developers, scientists and the public. These user-driven innovations include the assignment of a unique identifier (UID) to each marine report – to enable tracing of observations, linking with reports and improved data sharing. Other revisions and extensions of the ICOADS' International Maritime Meteorological Archive common data format incorporate new near-surface oceanographic data elements and cloud parameters. Many new input data sources have been assembled, and updates and improvements to existing data sources, or removal of erroneous data, made. Coupled with enhanced ‘preliminary’ monthly data and product extensions past 2014, R3.0 provides improved support of climate assessment and monitoring, reanalyses and near-real-time applications.