date: Mon Jun 13 13:16:25 2005
from: Phil Jones <p.jones@uea.ac.uk>
subject: Re: global surface temperature time series
to: "Thomas C Peterson" <Thomas.C.Peterson@noaa.gov>, Kevin Trenberth <trenbert@cgd.ucar.edu>

    Tom,
       Ch 3 of the IPCC report will discuss developments since the TAR. So Ch 1 should
    probably go up to the TAR, but it could stop earlier around 1990 (with the first IPCC
   report).
    A smooth transition will likely be up to the TAR.
       The CCSP document probably needs to go into much more detail, so may not
    be entirely relevant.
        I would have thought that Ch 1 should place greater emphasis on work pre-IPCC.
    Some of the references I gave you the other week would be best for this. There is a need
    to get across the fact that IPCC didn't invent global temperatures. Groups were working
    on this before 1990 and there were two major reviews pre-1990, namely the SCOPE
    one in 1986 and the earlier DoE State of the Art report from 1982 (which Bill Clark
   edited).
    The other thing to get across is that no matter how the data are analysed the results
    are pretty much the same - even back to Murray Mitchell.
    Cheers
    Phil
   At 20:11 08/06/2005, Thomas C Peterson wrote:

     Dear Kevin & Phil,
     I'm currently writing a new section for our IPCC introductory chapter on the history of
     global surface temperatures time series.  My description will focus on early efforts and
     with some general comments about moderately recent developments - i.e., up to the TAR.
     As I recall, your chapter discusses developments since the TAR.  Which reminded me that
     I wanted to send you the revised version of the article describing the new NCDC/NOAA
     global temperature time series (attached).  Also I wanted to know if I end around the
     time of the TAR and refer vaguely to three major groups producing surface temperature
     time series - NOAA, NASA and the UK (as per the CCSP VTT document (I'll paste the May
     5th version below for reference purposes only) will that be a smooth transition into
     your chapter or would you like me to deal with the different global analyses in some
     other manner?
     Regards,
             Tom Peterson
     2. SURFACE TEMPERATURES

     2.1       Land-based temperature data
     Over land temperature data come from fixed weather observing stations with thermometers
     housed in special instrument shelters. Records of temperature from many thousands of
     such stations exist. Chapter 2 outlines the difficulties in developing reliable surface
     temperature datasets. One concern is the variety of changes that may affect temperature
     measurements at an individual station. For example, the thermometer or instrument
     shelter might change, the time of day when the thermometers are read might change, or
     the station might move. These problems are addressed through a variety of procedures
     (see Peterson et al., 1998a for a review) that are generally quite successful at
     removing the effects of such changes at individual stations (e.g., Vose et al., 2003)
     whether the changes are documented in the metadata or detected via statistical analysis
     using data from neighboring stations as well (Aguilar et al., 2003). Subtle or
     widespread impacts that might be expected from urbanization or the growth of trees
     around observing sites might still contaminate a data set. These problems are addressed
     either actively in the data processing stage (e.g., Hansen et al., 2001) or through data
     set evaluation to ensure as much as possible that the data are not biased (e.g., Jones
     et al., 1990; Peterson, 2003; Parker, 2004; Peterson and Owen, 2005). Changes in
     regional land use such as deforestation, aforestation, agricultural practices, and other
     regional changes in land use are not addressed in the development of these data sets.
     Modeling studies have suggested over decades to centuries these affects can be important
     on regional space scales (Oleson et al., 2004).

     2.2       Marine temperature data
     Data over the ocean come from moored buoys, drifting buoys, and volunteer observing
     ships. Historically, ships have provided most of the data but in recent years an
     increasing number of buoys have been used, placed primarily in data-sparse areas away
     from shipping lanes. In addition, satellite data are often used after 1981. Many of the
     ships and buoys take both air temperature observations and sea surface temperature (SST)
     observations.  Night marine air temperature (NMAT) observations have been used to avoid
     the problem that the Suns heating of the ships deck can make the thermometer reading
     warmer than the actual air temperature.  Where there are dense observations of NMAT and
     SST, over the long term they track each other very well.  However, since marine
     observations in an area may only be taken a few times per month, SST has the advantage
     over air temperature in that water temperature changes much more slowly than that of
     air.  Also, there are twice as many SST observations as NMAT from the same platforms as
     SSTs are taken during both the day and night and SST data are supplemented in data
     sparse areas by drifting buoys which do not take air temperature measurements.
     Accordingly, only having a few SST observations in a grid box for a month can still
     provide an accurate measure of the average temperature of the month.

     2.3       Global surface temperature data
     Creating global surface temperature analyses usually involves not only merging land and
     ocean data but also considering how best to represent areas where there are few or no
     observations. One approach is to only use those grid boxes with data. This conservative
     approach avoids any error associated with interpolating data. Unfortunately, the areas
     without data are not evenly or even randomly distributed around the world, leading to
     considerable uncertainties in the analysis, though it is possible to make an estimate of
     these uncertainties. Using the conservative approach, the tropical land surface areas
     would be under represented, as would the southern ocean. Therefore, techniques have been
     developed to interpolate data to some extent into surrounding data-void regions. A
     single group may produce several different such data sets for different purposes. The
     choice may depend on whether the interest is a particular local region, the entire
     globe, or use of the data set with climate models (Chapter 5). Currently, there are
     three main groups creating global analyses of surface temperature (see Table 3A).

     2.3.1  NOAA
                 The National Oceanic and Atmospheric Administration (NOAA) National Climatic
     Data Center (NCDC) integrated land and ocean data set (see Table 3A) is derived from in
     situ data. The SSTs come from the International Comprehensive Ocean-Atmosphere Data Set
     (ICOADS) SST observations release 2 (Slutz et al., 1985; Woodruff et al., 1998; Diaz et
     al., 2002). Those that pass quality control tests are averaged into monthly 2^o grid
     boxes (Smith and Reynolds, 2003). The land surface air temperature data come from the
     Global Historical Climatology Network (GHCN) (Peterson and Vose, 1997) and are averaged
     into 5^o grid boxes. A reconstruction approach is used to create complete global
     coverage by combining together the faster and slower varying components of temperature
     (van den Dool et al., 2000; Smith and Reynolds, 2005).

     2.3.2  NASA
     The NASA Goddard Institute for Space Studies (GISS) produces a global air temperature
     analysis (see Table 3A) known as GISTEMP using land surface temperature data primarily
     from GHCN and the U.S. Historical Climatology Network (USHCN; Easterling, et al., 1996).
     The NASA team modifies the GHCN/USHCN data by combining at each location the time
     records of the various sources and adjusting the non-rural stations in such a way that
     their long-term trends are consistent with those from neighboring rural stations (Hansen
     et al., 2001). These meteorological station measurements over land are combined with in
     situ sea surface temperatures and Infrared Radiation (IR) satellite measurements for
     1982 to the present (Reynolds and Smith, 1994; Smith et al., 1996) to produce a global
     temperature index (Hansen et al., 1996).

     2.3.3  UK
     The global land and ocean data set from the UK (see Table 3A) is produced as a joint
     effort by the Climatic Research Unit of the University of East Anglia and the Hadley
     Centre of the UK Meteorological (Met) Office. The land surface air temperature data are
     from Jones and Moberg (2003) of the Climatic Research Unit.  The global SST fields are
     produced by the Hadley Centre using a blend of COADS and Met Office data bank in situ
     observations (Rayner, et al., 2003). The integrated data set is known as HadCRUT2v
     (Jones and Moberg, 2003). The temperature anomalies were calculated on a 5^ox5^o grid
     box basis. Within each grid box, the temporal variability of the observations has been
     adjusted to account for the effect of changing the number of stations or SST
     observations in individual grid-box temperature time series (Jones et al., 1997, 2001).
     There is no reconstruction of data gaps because of the problems of introducing biased
     interpolated values. The global temperature and hemispheric time series have been
     created using a technique known as optimal averaging (Parker et al., 2004; Folland et
     al., 2001a) which provides estimates of uncertainty in the time series, including the
     effects of data gaps and uncertainties related to bias corrections or uncorrected
     biases.

     2.3.4 Synopsis of surface datasets
     Since the three chosen datasets utilize many of the same raw observations, there is a
     degree of interdependence. Nevertheless, there are some differences among them as to
     which observing sites are utilized. There are three ways to assess how well the changing
     network of surface observations monitor global or regional temperature (Jones, 1995).
     The first is using frozen grids where analysis using only those grid boxes with data
     present in the sparsest years are used to compare to the full data set results from
     other years (e.g., Parker et al., 1994).  The results generally indicate very small
     errors on multi-annual timescales (Jones, 1995).  The second technique is subsampling a
     spatially complete field, such as model output, only where in situ observations are
     available.  Again the errors are small (e.g., the standard errors are less than 0.06C
     for the observing period 1880 to 1990; Peterson et al., 1998b).  The third technique is
     comparing optimum averaging, which fills in the spatial field using covariance matrices,
     eigenfunctions or structure functions, with other analyses.  Again, very small
     differences are found (Smith et al., 2005).
                 The fidelity of the surface temperature record is further supported by work
     such as Peterson et al. (1999) which found that a rural subset of global land stations
     had almost the same global trend as the full network and Parker (2004) that found no
     signs of urban warming over the period. An important advantage of surface data is the
     fact that at any given time there are thousands of thermometers in use that contribute
     to a global or other large-scale average. Besides the tendency to cancel random errors,
     the large number of stations also greatly facilitates temporal homogenization since a
     given station may have several near-neighbors for buddy-checks. While there are
     fundamental differences in the global averaging procedures applied, the differing
     techniques with the same data produce almost the same results (Wuertz et al., 2005).

     2.4       Global surface temperature variations and differences between the data sets
     Examination of the three global temperature anomaly time series (T[sfc]) from 1958 to
     the present shown in Figure 3.2.4 reveals that the three time series have a very high
     level of agreement. They all show some cooling from 1958 to around 1976, followed by
     strong warming. That most of the temperature change occurs after the mid 1970s has been
     previously documented (Karl et al., 2000; Folland et al., 2001b; Seidel and Lanzante,
     2004). The variability of the time series is quite similar as are their trends. The
     signature of the El Nio-Southern Oscillation (ENSO), whose origin is in the tropics, is
     responsible for many of the prominent short-term (several year) up and down swings of
     temperature as expected (Trenberth et al., 2002). The strong El Nio of 1997-98 stands
     out as an especially large warm event within an overall upward trend.

--
Thomas C. Peterson, Ph.D.
Climate Analysis Branch
National Climatic Data Center
151 Patton Avenue
Asheville, NC 28801
Voice: +1-828-271-4287
Fax: +1-828-271-4328

   Prof. Phil Jones
   Climatic Research Unit        Telephone +44 (0) 1603 592090
   School of Environmental Sciences    Fax +44 (0) 1603 507784
   University of East Anglia
   Norwich                          Email    p.jones@uea.ac.uk
   NR4 7TJ
   UK
   ----------------------------------------------------------------------------
