A secret history of the observed surface temperature record Phil Jones CRU, UEA

1
A secret history of the observed surface
temperature recordPhil Jones CRU, UEA

• Developing the Global Temperature Record
• Terrestrial and Marine components
• Biases
• Boulder, June 2009

2
Earliest work late 1970s and early 1980s

• Interpolation used inverse-distance weighted best
  fit planes with a base period of 1946-60
• The base period (i.e. using anomalies) is crucial
  in all the studies. It overcomes most elevation
  issues and some urbanization issues (later)
• Grid points at 10 longitude by 5 latitude
  intersections
• Station temperature data came from an NCAR DS,
  supplied on a 6250bpi tape
• Assessed data for outliers and removed those
  beyond reasonable bounds
• Programs ran on an IBM computer at Cambridge, for
  which cards had to be submitted
• Assessment of the effects of sparser coverage in
  early decades
• Compared results with all earlier studies
• Jones, P.D., Wigley, T.M.L. and Kelly, P.M.,
  1982 Variations in surface air temperatures,
  Part 1 Northern Hemisphere, 1881-1980. Monthly
  Weather Review 110, 59-70
• Kelly, P.M., Jones, P.D., Sear, C.B., Cherry,
  B.S.G. and Tavakol, R.K., 1982 Variations in
  surface air temperatures, Part 2 Arctic regions,
  1881-1980. Monthly Weather Review 110, 71-83
• Raper, S.C.B., Wigley, T.M.L., Mayes, P.R.,
  Jones, P.D. and Salinger, M.J., 1984. Variations
  in surface air temperature. Part 3 The
  Antarctic, 1957-82. Monthly Weather Review 112,
  1341-1353

3
Pre-CRU land temperature series, each adjusted to
have Brohan et al average (HadCRUT3) over their
last 30 years of overlap (from Ch1 of AR4 zero
line is 1961-90)
AR4 shouldnt have compared land only series with
HadCRUT3!
4
Mid -1980s

• Extended the spatial coverage of the station
  record, by digitizing records held in Met Office
  and other archives
• Assessed homogeneity of the records
  (subjectively)
• Base period of 1951-70
• Each station associated with its nearest grid
  point (as in 1982) and weights were inverse
  distance
• More extended analysis of the effects of fewer
  stations in the early decades (using frozen
  grids)
• Record extended back to 1851 (from 1881 earlier)
• First combined land and marine temperature record
  in 1986
• Jones, P.D., Raper, S.C.B., Bradley, R.S., Diaz,
  H.F., Kelly, P.M. and Wigley, T.M.L., 1986
  Northern Hemisphere surface air temperature
  variations 1851-1984. Journal of Climate and
  Applied Meteorology 25, 161-179
• Jones, P.D., Raper, S.C.B. and Wigley, T.M.L.,
  1986 Southern Hemisphere surface air
  temperature variations 1851-1984. Journal of
  Climate and Applied Meteorology 25, 1213-1230
• Jones, P.D., Wigley, T.M.L. and Wright, P.B.,
  1986 Global temperature variations, 1861-1984.
  Nature 322, 430-434

5
Comparison of the versions across the
years(1982, 1986, 1994, 2003 and 2006)
NH Top SH Bottom Left as was Right 1951-80
Q was all the effort worth it? It shows how
robust the series is!
6
Questions that began to be asked (1) (some
despite being discussed the papers)

• Effect of sparser coverage dont you need more
  stations?
• The frozen grid analyses had effectively answered
  this, but many only became convinced once the
  uncertainties were displayed with error ranges
• These also led to the understanding of the number
  of spatial degrees of freedom (Neff)
• Looking at numerous seasonal and annual
  temperature anomaly maps indicated that Neff must
  be much smaller than the number of observing
  sites, and that the number varies depending on
  the season and on the timescale
• The paper that derived the necessary formulae
  (Jones et al., 1997) was based on the Wigley et
  al. (1984) r-bar paper
• The fact that Neff was about 100 led to the
  cottage industry of reconstruction the past
  temperatures from natural and documentary proxies
• The fact that r-bar is higher in winter implies
  that these would be much better, if we had more
  winter-responding proxies
• Wigley, T.M.L., Briffa, K.R. and Jones, P.D.,
  1984 On the average value of correlated time
  series with applications in dendroclimatology and
  hydrometeorology. Journal of Climate and Applied
  Meteorology 23, 201-213
• Jones, P.D., Osborn, T.J. and Briffa, K.R., 1997
  Estimating sampling errors in large-scale
  temperature averages. J. Climate 10, 2548-2568

7
Questions that began to be asked (2)

• Isnt the fact that most sites have moved more
  than once in their history important?
• Wed addressed the long-term homogeneity of the
  station series in the mid-1980s
• The effort was large in person years, but its
  affect was barely noticeable (the comparison of
  the 1982 and 1986 papers showed this)
• Clearly important for individual series, but not
  that vital as the space scale increases

8
Homogenisation adjustment uncertainty
Black line is based on 763 sets of
adjustments Red hypothesised distribution Blue
the difference, so that used for stations where
adjustments have not been made
9
Similar bimodal distribution in a recent paper on
the USHCN
Menne et al (2009) in BAMS
10
USHCN all and 70 best (latter partly determined
by surface stations.org)
The 70 obviously omit large parts of the
contiguous US
11
Questions that began to be asked (3)

• Isnt the world warming because of many stations
  being located in cities?
• The urbanization issue! People always remember
  the greatest UHI that someone has shown on a
  particular day!
• They forget the similar warming between the land
  and the ocean components
• Comparisons of rural-only station datasets
• One new example here (London)
• Recognition that UHIs exist, but what matters is
  urban-related warming trends not the UHI size
• Jones, P.D., Groisman, P.Ya., Coughlan, M.,
  Plummer, N., Wang, W-C. and Karl, T.R., 1990
  Assessment of urbanization effects in time series
  of surface air temperature over land. Nature
  347, 169-172
• Jones, P.D., Lister, D.H. and Li, Q., 2008
  Urbanization effects in large-scale temperature
  records, with an emphasis on China. J. Geophys.
  Res. 113, D16122, doi10.1029/2008/JD009916.

12
Urbanization Influence

• Homogeneity testing may not remove all urban
  affected sites if all sites are similarly
  affected by urban growth
• Approach to assess residual effect is to develop
  a dataset of rural-only stations.
• Grid the rural-only stations and then compare
  with the grid with all the stations
• The issue here is assessment of the effect on
  monthly and annual average temperatures not the
  effect on a single day

13
Urbanization Examples

• Few studies have looked at urbanization studies
  across large areas of the world
• Major studies are Jones et al. (1990), Parker
  (2004, 2006) and Peterson and Owen (2005)
  references in Chapter 3 of AR4
• Basic conclusion is that any residual warming is
  an order of magnitude less than the warming that
  has occurred over the last 100 years
• Effect can be large in rapidly developing areas
  like China but even here the large-scale
  warming is 1.6 times greater (Jones et al. 2008).

14
London
UHI greater for Tn than Tx. Central London sites
always warmest at night, but warmer during day
west of London
London has an Urban Heat Island (UHI), but no
urban-related warming since at least 1900. In
other words, the centre got warmer earlier.
15
Questions that began to be asked (4)

• What causes the temperatures to change so much
  from year to year?
• El Niño and La Niña (ENSO) explain some of the
  high-frequency variability
• Volcanoes cause cooling
• Circulation change can cause warming and cooling
  (e.g. COWL patterns)
• Possible to factor out these influences
• Jones, P.D., 1989 The influence of ENSO on
  global temperatures. Climate Monitor 17, 80-89
• Thompson. D.W.J., Wallace, J.M., Jones, P.D. and
  Kennedy, J.J., 2009 Removing the signatures of
  known climate variability from global-mean
  surface temperature Methodology and Insights. J.
  Climate (tentatively accepted).

16
Questions that began to be asked (5)

• There are sites moves and/or urbanization
  influences at the nearby site (to where this talk
  is being given)
• Youve very few sites in Africa and South America
• Few seem to follow the logic of the Neff argument
  and think that one site can influence the global
  average
• Seem to accept the limited number when it comes
  to proxy data, but not when it is instrumental
  data !
• Developing these series (or regional ones) brings
  so many insights, which are clearly apparent in
  examples
• Different observation times (and methods of
  calculating mean T) in different countries dont
  matter, as long as they dont get changed
• A few outliers dont matter. I once had some
  100C anomalies for a few boxes in Siberia, but
  the effect on NH averages was a few hundredths
• Combining the wrong months station data with a
  different months absolute temperatures produces
  far greater differences (as GISS realised last
  year!)

17
Issues Today

• SST measurement
• Exposure of thermometers in Europe in summer in
  the mid-19th century
• Homogeneity of daily temperature series easiest
  to use the monthly adjusted series, and change
  the daily accordingly (not addressed here)

18
What does the corrected series look like?
HadSST2 (no corrections after 1941)
HadSST2 (no corrections after 1941)
Global-average SST anomaly (C) wrt 1961-1990
Uncorrected data
Uncorrected data
19
Other Marine (SST) problems

• The various measuring platforms (ships, buoys,
  satellites etc) have slightly different biases
• SSTs in the 1940s (Thompson et al., 2008)
• Recent SSTs (buoys slightly cooler than ships)
• Neither of these two adjusted for yet
• Issue is the same in all cases instruments
  improved (response time, better sampling etc.)
  but no overlap measurements made beforehand. The
  bias gets sorted out in retrospect, when there is
  enough data available to clearly show the problem
• 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.

20
Huge change in marine observing network in the
past 25 years
Percentage of observations coming
from DRIFTERS and SHIPS
21
What will the corrected series look like?
New Corrections
HadSST2 (no corrections after 1941)
Global-average SST anomaly (C) wrt 1961-1990
Uncorrected data
22
Early exposure issues

• Europe affected, before the development of
  Stevenson screens
• Solution has come about from modern parallel
  measurements (in Austria and Spain, with the old
  screens)
• Effect is annually 0.4C, with most series too
  warm by up to 0.7C in June
• Surprisingly (for Austria), the effect is much
  smaller using the (TxTn)/2 method of calculating
  averages than using fixed hours
• Issue important as it is the summers that
  calibrate the natural and documentary proxies

23
Kremsmünster - Austria
24
Kremsmünster - Austria
When built in the 1770s, this monastery was the
tallest in Europe for the time
25
Kremsmünster historic minus modern diurnal
cyclesNNE exposure
Work undertaken by Reinhard Böhm et al (ZAMG,
accepted Climatic Change) 8 complete years of
overlap between the window and the modern
Austrian observing site
26
Effect through different formulae used to
calculate monthly mean T
27
Different exposures NNW (left) and N (right)
28
Adjustment across all 32 sites in the Greater
Alpine Region (GAR)
GAR is 4-19E by 43-49N To apply adjustments
need to know the NW-to-NE direction each site
faced
29
Effect of changes across the GAR
Thin is raw data. Grey is after adjustment for
site moves and observation time changes. Thick
shows the effect of exposure issues (thick black
overlies grey since about 1860).
30
Conclusions

• Individual station homogeneity assessment has
  little effect on global averages, but is vital
  for local- and regional-scale homogeneity
• Biases (buckets and some urbanization) are what
  is important for homogeneity (particularly the
  way SSTs have been measured)
• Problems in the 1945-55 will be adjusted for as
  well as issues of the differences between ships
  and drifters today (Effect will to delay the
  post-WW2 cooling to the mid-1950s and will
  slightly raise temperatures over the last 10
  years)
• Exposure issues with pre-screen measurements
  addressed in Europe (Effect is to cool summer
  temperatures by about 0.5C before about 1860)
• 2001-2008 0.18C warmer than 1991-2000, which was
  0.14C warmer than 1981-90