BADC Help File: GMSLP Data

This file contains background information to help you use the Met Office GMSLP data held at the BADC. This document is adapted from "Development of the Global Mean Sea Level Pressure Data Set GMSLP2" by T.A. BAsnett and D.E.Parker, 1997.


  1. Introduction
  2. Brief description of "input" gridded data sets.
  3. Creation of GMSLP2.1.
  4. Conclusion

1. Introduction:

There has long been a requirement for a reliable global gridded dataset of observed mean sea level (MSL) pressure, that would be suitable for both analysing climate variability and for vaildating climate model output. One of the earliest attempts to map global MSL pressure was made by Buchan (1868), who mapped mean monthly MSL pressure for 1857-1866, based on observations derived from a representative sample of 360 stations worldwide. Notable work in this field was carried out by Lamb and Johnson (1966), who published contoured global charts of January and July MSL pressure for each year in 1750-1966. However existing gridded MSL pressure data sets generally have a restricted coverage, being either hemispheric or 'near-global' and consist mostly of analysed data derived from hand-drawn analyses in pre-computer era and more recently from routine GCM (Global Circulation Model) operational analyses or reanalyses. Also owing to the change of analysis technique, these data sets are unlikely to be homogeneous. Although there are several 'observed' MSL pressure data sets available, which are either blends of analysed and observed data or statistical reconstructions based on observations, none of these has global coverage.

GMSLP is a global, observed, monthly historical MSL pressure data set, which has been developed in collaboration with CSIRO and NIWA. The data set begins in 1871 and is composed of gridded values on a 5° latitude and 5° longitude resolution, centred on whole multiples of 5°, e.g. 55°N, 5°N. It has been constructed from a blend of existing gridded datasets provided by the UK Met. Office, NCAR, Scripps Institute of Oceanography (SIO), University of East Anglia (UEA) and CSIRO; plus observed data extracted from the UKMO Marine Data Bank, Comprehensive Ocean-Atmosphere Data set (COADS), UKMO monthly CLIMAT message archives and land station time series.

A key use of GMSLP is in the multivariate analysis of results of atmospheric model simulations of recent climate, aimed at detecting climate change and attributing it to human activities. Thus, simulations of historical surface pressure given observed sea surface temperatures and increasing greenhouse gases, sulphate aerosols, etc..., are being compared with observations and used to help interpret simulated air temperature changes. A further aim of GMSLP is to facilitate the validation of other atmospheric model results. GMSLP is also being used in the interpretation of observed global and regional climate variations. Particular examples are: variations in the El Niño Southern Oscillation, variations in south Pacific atmospheric circulation in observations and in atmospheric model simulations, and the effect of atmospheric circulation on land surface air temperature patterns. The dataset will allow an improved assessment of the effects of atmospheric circulation changes on UK climate and therefore a more thorough interpretation, for the UK area, of model simulations with increasing greenhouse gases.

2. Brief description of "input" gridded data sets.

The input gridded data sets are summarized in Table 1.

Table 1: Gridded data sources contributing to GMSLP2.

Data Source  Area covered Period Resolution
 UKMO Historical  15N-90N  1873-1995 Nodes of grid are at multiples of 5° lat.x 10° long.
 UKMO Operational  Globe  1985-1995 Nodes of grid are at multiples of 2.25° lat.x 3.75° long.
NCAR  15N-90N  1899-1995 Nodes of grid are at multiples of 5° lat.x 5° long
SIO  42.5S-72.5N  1951-1993 Nodes of grid are at multiples of 5° lat.x 10° long
CSIRO  60S-60N  1871-1989 Nodes of grid are at multiples of 10° lat.x 10° long
 Jones  15S-60S  1951-1985  Nodes of grid are at multiples of 5° lat.x 10° long
 Jones and Wigley  60S-75S  1957-1985 Nodes of grid are at multiples of 5° lat.x 10° long 

NCAR and UKMO are both Northern Hemisphere data sets based on 5° by 5° area and 5° latitude by 10° longitude grids respectively.

The CSIRO data set is a 'near global' data set, based on a 10° by 10° area grid, constructed from a blend of land observations extracted from historical station records and marine observations from COADS.

The SIO data set (Barnett et al., 1984) also has 'near global' coverage, but on a 5° latitude by 10° longitude grid. These data were derived from a variety of different sources, including MSL pressure analyses from various national weather centres and raw ships' observations.

The data sets provided by Jones and Wigley (1988) and Jones (1991) are both limited area 5° latitude by 10° longitude grids covering part of the Southern Hemisphere. They are based solely on land station observations, which were interpolated onto 'hemespheric' grids using regression techniques.

3. Creation of GMSLP2.1.

Although GMSLP2.0 consisted of a globally complete individual MSL pressure field for every month for 1871-1994, it was heavily dependent on the component gridded analyses. To develop and further improve the GMSLP data set, the UK Met. Office introduced observed data directly.These data were a blend of marine observations from the UKMOMDB and COADS and land station time series from CSIRO, NIWA and the UKMO CLIMAT archive. The observations were blended with GMSLP2.0 to create GMSLP2.1 .

Several versions of GMSLP2.1 data were created. GMSLP2.1c was the first published version but it was found to suffer from bias over the tropical oceans,owing to over-correcting marine data for standard gravity. A revised correction scheme was devised and applied to the marine data, ultimately resulting in GMSLP2.1f, which is the most recent version of the data set.

More detailed information about the creation and analysis of this dataset is available from "Development of the Global Mean Sea Level Pressure Data Set GMSLP2" by T.A. BAsnett and D.E.Parker, 1997.

4. Conclusion

GMSLP2.1f is the most recent version of the Global Mean Sea-Level Pressure dataset. GMSLP2.1 is more reliable than GMSLP1.2, as it is strongly influenced by quality controlled observed data. GMSLP2.1f, or preceeding versions, has already been used in various analyses, especially studies of ENSO and for model validation. However there remain several weaknesses, namely the necessity of using climatology in data-sparse regions, especially in the Southern Hemisphere, together with a lack of reliable observations in the earlier decades. Some initial problems with the quality-control of several key land-station time series led to the introduction of noise into early versions of the data set: adjustments have been applied but further improvements need to be made. The problem of noisy data is relatively insignificant on a global scale, but for some regional studies it has been found that band-pass filtered data are of more use.

A completely new GMSLP3 is planned. This will use a background field based on EOF reconstruction techniques, developed by Rayner et al. (1996) for the creation of the GISST data set. By using such techniques, the UK Met. Office hope to remove the need for a background field reliant on other data sets and therefore eliminate much of the bias that has affected the creation of GMSLP2.1.