Section 4.3: Surface Climate Analysis
57
The total of these problems, coupled with the greater spatial variability of
precipitation data, makes the problem of homogeneity much more difficult
to deal with than for temperature. Precipitation networks are rarely dense
enough to allow for the homogeneity checks discussed in the next section to
be undertaken. Instead of studying individual records, analyses of regional
precipitation series have been made (Bradley et al. 1987; Diaz et al. 1989;
Folland et al. 1990, 1992).
Pressure data have an advantage over the temperature and precipitation
data bases because they are routinely analysed on to regular grid networks for
weather forecasting purposes. Monthly-mean data for the Northern Hemisphere (north of 15° N) extend back to 1873 and for the Southern Hemisphere
routine analyses began in the early 1950s. Neither analyses are complete for
the entire period, particularly during the earliest years. The sources of the
data are discussed and their quality assessed in a number of papers (Northern Hemisphere: Williams and van Loon, 1976; Trenberth and Paolino, 1980
and Jones, 1987, Southern Hemisphere: van Loon, 1972; Karoly 1984; Jones,
1991). Global analyses have been produced at the major operational weather
centres since 1979. The most widely available analyses are those from the European Centre for Medium-Range Weather Forecasts (ECMWF) and the National Meteorological Center (NMC) of NOAA. However because of changes
to the model and to data assimilation schemes, the homogeneity of the analyses for many climate applications is questionable (Trenberth et a1., 1992).
4.2.6 Data Homogenization Techniques
Many methods have been proposed for testing the homogeneity of station
records relative to those at adjacent stations (Conrad and Pollak, 1962 and
WMO, 1966). Generally, tests involve the null hypothesis that a time series of
differenees (ratios for preeipitation) between neighbouring observations will
exhibit the characteristics of a random series without serial correlation (e.g.
MitchelI, 1961; Craddock, 1977, Jones et al. 1986a,c, Alexandersson, 1986,
Potter, 1981, Young, 1993). The methods work with or without metadata.
The scope of the method and the ability to explain and correct inhomogeneities is dramatically improved, however, with detailed metadata. The
most comprehensive homogeneity exercise performed produced the United
States Historic Climate Network (Karl and Williams, 1987).
4.3 Surface Climate Analysis
4.3.1 Temperature
Various research groups have synthesized the individual site temperature
series into gridded and/or hemispheric average time series. The three
main groups are the Climatic Research Unit/U.K. Meteorological Office
57
The total of these problems, coupled with the greater spatial variability of
precipitation data, makes the problem of homogeneity much more difficult
to deal with than for temperature. Precipitation networks are rarely dense
enough to allow for the homogeneity checks discussed in the next section to
be undertaken. Instead of studying individual records, analyses of regional
precipitation series have been made (Bradley et al. 1987; Diaz et al. 1989;
Folland et al. 1990, 1992).
Pressure data have an advantage over the temperature and precipitation
data bases because they are routinely analysed on to regular grid networks for
weather forecasting purposes. Monthly-mean data for the Northern Hemisphere (north of 15° N) extend back to 1873 and for the Southern Hemisphere
routine analyses began in the early 1950s. Neither analyses are complete for
the entire period, particularly during the earliest years. The sources of the
data are discussed and their quality assessed in a number of papers (Northern Hemisphere: Williams and van Loon, 1976; Trenberth and Paolino, 1980
and Jones, 1987, Southern Hemisphere: van Loon, 1972; Karoly 1984; Jones,
1991). Global analyses have been produced at the major operational weather
centres since 1979. The most widely available analyses are those from the European Centre for Medium-Range Weather Forecasts (ECMWF) and the National Meteorological Center (NMC) of NOAA. However because of changes
to the model and to data assimilation schemes, the homogeneity of the analyses for many climate applications is questionable (Trenberth et a1., 1992).
4.2.6 Data Homogenization Techniques
Many methods have been proposed for testing the homogeneity of station
records relative to those at adjacent stations (Conrad and Pollak, 1962 and
WMO, 1966). Generally, tests involve the null hypothesis that a time series of
differenees (ratios for preeipitation) between neighbouring observations will
exhibit the characteristics of a random series without serial correlation (e.g.
MitchelI, 1961; Craddock, 1977, Jones et al. 1986a,c, Alexandersson, 1986,
Potter, 1981, Young, 1993). The methods work with or without metadata.
The scope of the method and the ability to explain and correct inhomogeneities is dramatically improved, however, with detailed metadata. The
most comprehensive homogeneity exercise performed produced the United
States Historic Climate Network (Karl and Williams, 1987).
4.3 Surface Climate Analysis
4.3.1 Temperature
Various research groups have synthesized the individual site temperature
series into gridded and/or hemispheric average time series. The three
main groups are the Climatic Research Unit/U.K. Meteorological Office