CRUTEM4 and HadCRUT4
Phil Jones
Climatic Research Unit
University of East Anglia
Norwich
Summary
• Updated versions of global temperature data
(from joint work between CRU and the Met
Office Hadley Centre)
• Improvements to land data (CRUTEM4),
marine data (HadSST3) and the method of
combination (HadCRUT4)
• Urbanization (Matt has talked about this)
• Effects of ENSO, Solar, Volcanoes on the
record
CRUTEM4 and HadCRUT4
•
•
•
•
•
•
•
•
Improved coverage of land data by better access to more station data
from National Met Service (NMS) websites (Jones et al., 2012).
The 5° by 5° latitude/longitude grid box size that is used means polar
region boxes are relatively small. Greater numbers of stations from
Russia and Canada mean that more of these boxes now have at least one
station
Improvements to adjustment procedures for sea-surface temperatures
(SSTs) mean that SST data is slightly warmer for the years 1946-1960
(problem found by Thompson et al., 2008) and also warmer in the last
two decades as most SST data are now from fixed and drifting buoys
as opposed to ships (see details about HadSST3 in Kennedy et al.,
2011a,b and HadCRUT4 in Morice et al., 2012)
Jones, P.D., Lister, D.H., Osborn, T.J., Harpham, C., Salmon, M., Morice, C.P. 2012: Hemispheric and large-scale land surface air
temperature variations: An extensive revision and an update to 2010. J. Geophys. Res. 117, doi:10.1029/2011JD017139, in press
Kennedy J.J., N.A. Rayner, R.O. Smith, M. Saunby, and D.E. Parker, 2011a: Reassessing biases and other uncertainties in seasurface temperature observations measured in situ since 1850 part 1: measurement and sampling errors. Journal of Geophysical
Research Atmospheres, 116, doi:10.1029/2010JD015218.
Kennedy J.J., N.A. Rayner, R.O. Smith, M. Saunby, and D.E. Parker, 2011b: Reassessing biases and other uncertainties in seasurface temperature observations measured in situ since 1850 part 2: biases and homogenisation. Journal of Geophysical
Research Atmospheres, 116, doi:10.1029/2010JD01522.
Morice, C.P., Kennedy, J.J., Rayner, N.A. and Jones, P.D., 2012: Quantifying uncertainties in global and regional temperature
change using an ensemble of observational estimates: the HadCRUT4 dataset. J. Geophys. Res (submitted).
Thompson, D.W.J., Kennedy, J.J., Wallace, J.M. and Jones, P.D., 2008: A large discontinuity in the mid-twentieth century in
observed global-mean surface temperature. Nature 453, 646-649.
Coverage Changes – CRUTEM4
•
•
•
•
•
Recent improvements in data availability
Comparison with CRUTEM3
Removing large countries
Independent subsets of data give the same result
Errors in large-scale averages – accounting for
coverage, bias adjustment (urbanization effects),
homogeneity adjustments (at individual sites), random
errors and uncertainties in base period estimates
• Errors depend on timescale used in any plot. SH has
much larger error ranges than the NH as more SH
areas have poor coverage (mostly Antarctica and
interior southern Africa)
• Comparisons with US groups
• Comparisons with Reanalysis (ERA-Interim)
• Global average is (2*NH+SH)/3 – for the Land
2001-2010 and coverage improvements
CRUTEM3
Improvements in
coverage has enabled
more grid boxes to be
filled, not just in this
decade, but back to
the 1920s.
Little change over the
SH.
CRUTEM4
4 minus 3
Open black
squares
are new
boxes in 4
not in 3
Neither CRUTEM3 nor
CRUTEM4 do any
infilling from
neighbouring stations
(as the two US groups
do).
Instead we’re infilling
by accessing more
data. More is now
available in near-real
time than even 5 years
ago
Comparison of trends for the period 1951-2010
CRUTEM3
CRUTEM4
4 minus 3
At least 48 years
needed to calculate a
trend
Less coverage
changes, as more
improvements coming
from the 2000s than
the earlier decades
Again this is just
better real-time
access to data,
especially from
Canada and Russia
CRUTEM4 (bold) and CRUTEM3 (lighter)
Global land average ((2*NH+SH)/3)
Both series expressed as anomalies from the same 1961-90 period
Rest of plots/maps use this base period. Smoothing is decadal adaptive filter
Omitting data from large countries
NH less contiguous US (left) SH less Australia (right)
Coverage effects – subsampling the station data
Using 5 different sets of stations, each of which has a unique 20%
of the data
Hemispheric and Global averages can be produced from far fewer station numbers
Comparison of CRUTEM4 with the 2 US groups
(NASA/GISS and NCDC/NOAA)
Green shading is two sigma error estimate for the interannual timescale
BEST
• Latest BEST papers get an even larger trend than
these 3 groups, but not significantly higher
• BEST gridding uses Kriging, so is sophisticated
interpolation
• They also get much reduced errors. Trend over the
period from the 1950s to 2010s is 0.87°± 0.05°C
• I don’t understand this, but I think I know why –
either incorrect use of their very high station counts
or the trend is calculated from the decadallysmoothed data (which is based on running means!)
• Yes they do say 1950s and 2010s, so is it the trend
from 1951-2010? This is partly why I think it is based
on the smoothed series
Comparison with ERA-Interim (NH)
Land only
ERA-Interim complete coverage for NH, so warms slightly more than CRUTEM4
Comparison with ERA-Interim (SH 0-60S)
Land only
ERA-Interim is still quite different from CRUTEM4 over the Antarctic
HadCRUT4
•
•
•
•
Coverage issues
Comparisons between HadSST2 and HadSST3
Bucket models
Dramatic increase in the amount of buoy (fixed and
drifting) data in the database with respect to ships
over the last 20 years
• Again 1961-90 base period is used, but for SST this
is totally based on ships. So as now 80% of the raw
data is from buoys, there may be issues due to ships
and buoys not measuring exactly the same thing
• Buoys measure a better absolute value
Coverage improvements
HadCRUT4 vs HadCRUT3
SST issues – HadSST3
•
•
•
•
•
•
•
•
•
•
•
•
•
Principal problem is the changeover to engine intake measurements from buckets (~1940)
Countries and shipping (merchant and naval) fleets did this at different times
Bucket design also varied between different shipping fleets
The way the SST measurement was made was not put with the data until the early 1970s
Dates and bucket types have only been discovered by looking at old books of instructions to
marine observers
ERI – Engine Room Intakes
VOS – Voluntary Observing Ships
Modern SST data come in with ship call signs and locations – problem is that the shipping
fleets are becoming more reluctant to take the data – for security and trade/economic
issues (e.g. they don’t want others to know where they are – fishing fleets)
If adjustments not made for these issues, globally averaged SST would have increased much
more than it has
SSTs are vital to many other areas of atmospheric sciences. They are necessary as the
boundary values for weather forecasts and also Reanalyses.
Thompson, D.W.J., Kennedy, J.J., Wallace, J.M. and Jones, P.D., 2008: A large discontinuity in the mid-twentieth century in
observed global-mean surface temperature. Nature 453, 646-649.
Kennedy J.J., Rayner, N.A., Smith, R.O., Saunby, M. and Parker, D.E., 2011a: Reassessing biases and other uncertainties in seasurface temperature observations since 1850 part 1: measurement and sampling errors. J. Geophys. Res.116, D14103,
doi:10.1029/2010JD015218.
Kennedy J.J., Rayner, N.A., Smith, R.O., Saunby, M. and Parker, D.E., 2011b: Reassessing biases and other uncertainties in seasurface temperature observations since 1850 part 2: biases and homogenisation. J. Geophys Res. 116, D14104,
doi:10.1029/2010JD015220.
Types of books that need to be found
Thompson, D.W.J., Kennedy,
J.J., Wallace, J.M. and
Jones, P.D., 2008: A large
discontinuity in the midtwentieth century in
observed global-mean
surface temperature. Nature
453, 646-649.
This paper showed that
British Naval Ships
continued to use buckets
between 1945 and 1960 –
contrary to what was
believed in 2006. See
Kennedy et al (2011a,b)
Time series of measurement methods
(based on assumptions as of 2006)
Weight in
global
average
Buckets
of
differing
types –
some
insulated,
some not
Recent (since the 1980s) changes to SST
measurements
• Voluntary Observing Fleet (Ships) – numbers
reducing, partly as shipping companies want to
restrict access to data in real time. Also
restricted to shipping lanes. Since 2006 data
comes without the ship identifier
• Satellites – scanning radiometers (ATSR)
• Drifters (buoys) – get to more of the world’s
oceans
• Ships take the full array of weather
measurements (air temps, clouds, humidity
etc.). Drifters just take SSTs and some do air
pressure
Drifters cause significant
cooling in global average SST
Globalaverage SST
anomaly (°C)
wrt 19611990
The base
being based
on ships –
but the
drifters are
likely the
better in an
absolute
sense
HadSST3 References – recent Kennedy et
al papers
Kennedy J.J., Rayner, N.A., Smith, R.O., Saunby, M. and
Parker, D.E. (2011a). Reassessing biases and other
uncertainties in sea-surface temperature observations
since 1850 part 1: measurement and sampling errors. J.
Geophys. Res.116, D14103, doi:10.1029/2010JD015218.
Kennedy J.J., Rayner, N.A., Smith, R.O., Saunby, M. and
Parker, D.E. (2011b). Reassessing biases and other
uncertainties in sea-surface temperature observations
since 1850 part 2: biases and homogenisation. J.
Geophys Res. 116, D14104, doi:10.1029/2010JD015220.
HadSST3 versus raw observations (red)
Grey band is
error associated
with the
assumptions
made
Definitions of
the regions
given in Kennedy
et al (2011b).
Raw data (red)
shows much
greater warming
ENSEMBLES
• Both HadSST3 and HadCRUT4 are ensemble datasets
– to illustrate the error range better
• To get HadCRUT4, CRUTEM4 had also to be
transformed into 100 Ensembles
• NH, SH and Global average series are from the
average of the 100 Ensemble members
• For each grid-box this is also done, so the average of
the 100 grids only very closely approximates the
average of the 100 hemispheric time series
• These ensembles can be considered in a similar way to
the 56 member available with 20CR (Gil Compo’s
extended reanalysis to 1871). ERA-CLIM will be doing
this, so need to get used to it. Will be used this way
in D&A studies with the models
Global and regions – with trends over specified periods
More Regions – grey band shows effects of Thompson et al. (2008)
HadSST3 – with uncertainties
Biases assumed
independent of
each other
Combining land and marine
• CRUTEM4+HadSST3=HadCRUT4
• At coasts/islands the method of combination has
reverted to the method used with HadCRUT2 (i.e.
based on area with land getting a % of at least 25 and
similarly for the ocean). This ensures long island
records don’t get ignored when in poorly sampled
ocean regions.
• Also in HadCRUT4 the NH and SH average is the
average of the 12 months
• Global average is (NH+SH)/2
HadCRUT4 vs HadCRUT3 for the global
average (with error ranges)
HadCRUT4 vs other groups
Each series has its full coverage
Interesting Plots
• Comparisons with satellite (MSU)
estimates
• Urbanization
• Factoring out ENSO, Solar, Volcanoes
Surface vs Satellite
(From IPCC AR4 WG1 report, 2007, FAQ
3.1, Figure 1). Patterns of linear global
temperature trends from 1979 to 2005
estimated at the surface (left), and for the
troposphere (right) from the surface to
about 10 km altitude, from satellite records.
Grey areas indicate incomplete data. Note
the more spatially uniform warming in the
satellite tropospheric record while the
surface temperature changes more clearly
relate to land and ocean.
Surface and satellite records
of temperature AGREE!! 2010
is exceptionally warm with
the satellites too!
Global temperatures factoring out the effect of
ENSO, volcanoes , the Sun
Grant Foster and Stefan Rahmstorf 2011 Environ. Res. Lett. 6 044022
doi:10.1088/1748-9326/6/4/044022
Conclusions
• CRUTEM4/HadCRUT4 warm by similar amounts to the older
versions
• Uncorrected SST data would lead to greater warming and SST
data at odds with land temperatures
• Recent changes highly dependent on period length and the
occurrence of ENSO events