Global Temperature Deviations Monthly Means
Global Temperature Deviations Monthly Means (Decades)
Global and Hemispherical Averages
- Data Set Overview
- Sponsor
- Original Archive
- Future Updates
- The Data
- Characteristics
- Source
- The Files
- Format
- Name and Directory Information
- Companion Software
- The Science
- Theoretical Basis of Data
- Processing Sequence and Algorithms
- Hemispherical and Global Averages
- Scientific Potential of Data
- Validation of Data
- Data Access and Contacts
- FTP Site
- Points of Contact
Monthly Surface air temperature anomalies for the period 1851-1996 have been calculated by the Climate Research Unit (CRU) of the University of East Anglia, Norwich, England, using data from several sources ( Jones et al., 1986 a,b&c, and 1991). It is recommended that the data from 1851-1855 be ignored in most analyses because it is quite sparse as shown in the annual table (Jones 1997, private communication). The anomalies consist of land and ocean temperature departures from the 1961-1990 reference period, and are given on a 5 x 5 degree world grid. Hemispherical and global monthly and annual means are also included starting with 1856. Please note that this data set was recently revised (Jones, 1994); in the older data set the reference period was 1951-1970, fewer stations were included, and the gridding method was a little different. This dataset is very important for climate change studies. Despite relatively poor data coverage initially and around the two World Wars, the generally cold end of the nineteenth century and substantial warming from 1920 to 1940 are clearly shown. Slight cooling of the Northern Hemisphere took place between the 1950s and 1994, but this was followed by a warming trend in the 1980s and 1990s (Parker et al., 1994; Jones, 1994).The temperature records were obtained from various archives: the World Weather Records, published by the Smithsonian Institution (1927, 1935, 1947), and the U. S. Weather Bureau (1959-1982); material collected in the meteorological archives; and sea surface temperature data derived from the United Kingdom Meteorological Office's data bank (Bottomley et al., 1990)(UKMO; Bottomley et al., 1990), and the Comprehensive Ocean Atmosphere Data Set (COADS; Woodruff et al., 1987). The data from the several sources were carefully examined and corrections were made to compensate for known measurement problems. A brief discussion of the necessary corrections is given in Jones et al. (1991) and Parker et al., (1994) along with references to more detailed descriptions. Hansen and Lebedeff (1987) and Vinnikov et al. (1990) have also formed surface temperature anomaly datasets covering essentially the same period. All three datasets draw most of their land measurements from the same data archives and on hemispherical and global scales show similar temperature trends ( Jones et al., 1991). The CRU East Anglia data set is however unique in combining land and ocean temperature anomalies for long term analysis. Thus it shows regional mid ocean temperature anomalies that are suppressed in the land measurement only datasets. The corrections for the land measurements also differ among the three data sets; thus while the general trends of the three datasets are similar there are some differences.
For the convenience of the user, the Goddard Institute for Space Studies (GISS) Global Monthly and Annual Average Temperature Deviations are given for comparison purposes. Also available on this site is the Southern Oscillation Index. It is the normalized sea surface pressure difference between Tahiti and Darwin. The method of calculation is given in Ropelewski and Jones (1987). All missing Darwin data are infilled from Djakarta. Missing Tahiti data are infilled from Apia, Suva and Santiago. Because of the missing data, some of the years before about 1920 are a little less reliable than the later values.
The production and distribution of this data set are funded by NASA's Earth Science enterprise. The data are not copyrighted; however, we request that when you publish data or results using these data please acknowledge as follows:
The authors wish to thank Phil D. Jones and the Climatic Research Unit, School of Environmental Sciences, University of East Anglia, Norwich, U.K., for the production of this data set, and the Distributed Active Archive Center (Code 902) at the Goddard Space Flight Center, Greenbelt, MD, 20771, for putting these data in their present format and distributing them. These distribution activities were sponsored by NASA's Earth Science enterprise.This data set was constructed by the Climatic Research Unit (CRU) at East Anglia University. This is also the location of the primary archive and the source for detailed information concerning this data set. The data in its original format can be obtained from the CRU.
Note: The format of the data in the East Anglia archive is slightly different than that stored at the Goddard DAAC. For more details see the section Processing Sequence and Algorithms.
This data set will be updated as new data is made available.
CharacteristicsSource
Gridded Monthly Means Means Tables Parameters Surface Temperature deviations Annual means (summary) Northern Hemisphere Means Southern Hemisphere Means Global Means % of area reporting Units Degrees Celsius Typical Range -2 degrees to 2 degrees Celsius -1.6 to 0.7 degrees Celsius Temporal Coverage January, 1851 - December, 1996 1856-1996 Temporal Resolution: monthly means Annual means Monthly & Annual means Spatial Coverage Global (gridded) Global Hemisphere Global Spatial Resolution 5 degrees x 5 degrees Hemispherical & Global This data set is derived from the World Weather Records (WWR), published by the Smithsonian Institution (1927, 1929, 1935, 1947) and the U.S. Weather Bureau (1959-1982), The United Kingdom Meteorological Office's data banks (Bottomley et al, 1990) and the Comprehensive Ocean-Atmosphere Data Set(COADS, Woodruff et al., 1987.) Additional data were added, using material collected in published and manuscript form from meteorological archives.
The Temperature Deviation data set consists of monthly means binary data files, (1991 - 1996), a collection of gif images derived from these files (each image consists of the twelve monthly means images for a given year displayed on a single screen with a color bar), and decadal files consisting of monthly means arrays identical in format to the more recent data but grouped 120 to a file (1851 - 1990), as well as a file of global and hemispherical annual means. The entire data set, including image files, requires about 37 MB of disk storage.
Compressed:
Uncompressed:
- The decade data files have been compressed using Lempel-Ziv coding. Files with a .gz ending are compressed versions of the .bin file. When decompressing the files use the -N option so that the original .bin file name ending is restored. For additional information on decompression see aareadme file in the directory:
- software/decompression/
Name and Directory Information
File Characteristics Monthly Means (1991-96) Decades 1851-1990) Global and Hemispherical Means File size in bytes 10368 1244160 < 29000 Number of files 72 14 4 Format IEEE float ASCII text Fill value -999. -99.99 Grid size 72 x 36 NA Continent mask None - data valid over land and water Orientation North to South Start Position 177.5W, 87.5N End Position 177.5E, 87.5S Naming Convention
- The file naming convention for the monthly files is
- e_anglia.tmpdev.1nmagg.[yymm].ddd
- where
- E_Anglia = data product designator
- tmpdev = parameter name (temperature deviations)
- 1 = number of levels
- n = vertical coordinate, n = not applicable
- m = temporal period, m = monthly
- a = horizontal grid resolution, a = 5 x 5 degree
- gg = spatial coverage, gg = global (land and ocean)
- yy = year
- mm = month
- ddd = file type designation, (bin=binary file, ctl=GrADS control file) OR
- The file naming convention for the decade files is
- tmpdev.1941-50.ddd
- where
- tmpdev = parameter (temperature deviations)
- 1851-1860 = years covered in file
- ddd = file type designation, (gz=compressed, bin=binary)
- NOTE: When decompressing the data files be sure to use the -N option. This will restore the original .bin filename. For additional information on decompression see the format section of this readme and the aareadme file in the directory:
- software/decompression/
The file containing global and hemispherical means is named glb_hem_avgs.
Directory Paths
/data/surf_temp_press/tmp_dev/e_anglia/yyyy
- (where yyyy is year.)
- /data/surf_temp_press/tmp_dev/e_anglia/decades1851-1990
- /data/surf_temp_press/tmp_dev/e_anglia/gifs
- /data/surf_temp_press/tmp_dev/e_anglia/global_means
Several software packages have been made available on the CIDC CD-ROM set. The Grid Analysis and Display System (GrADS) is an interactive desktop tool that is currently in use worldwide for the analysis and display of earth science data. GrADS meta-data files (.ctl) have been supplied for each of the data sets. A GrADS gui interface has been created for use with the CIDC data. See the GrADS document for information on how to use the gui interface.
Decompression software for PC and Macintosh platforms have been supplied for datasets which are compressed on the CIDC CD-ROM set. For additional information on the decompression software see the aareadme file in the directory:
- software/decompression/
Sample programs in FORTRAN, C and IDL languages have also been made available to read these data. You may also acquire this software by accessing the software/read_cidc_sftwr directory on each of the CIDC CD-ROMs
Theoretical Basis of DataThe surface air temperature and the sea surface temperature are basic weather and climate parameters. They are normally measured by thermometers. The present data set was established to examine global and regional long term trends. These are small and are not always easy to separate from operational noise. The investigators have taken great care to address the error sources. Nevertheless Jones et al. (1986a) state, "It is considered impossible to reduce all observations to the same standard (Bradley et al., 1985). Nevertheless ...the problem is considerable reduced if all records are transformed to anomaly values from a common reference period." It is easier to establish station and regional temperature shifts than to establish an historical over all global temperature standard. Therefore first station and then regional anomalies are calculated and then joined together on a global grid to study hemispherical and global mean changes. For long term changes sampling errors or deficiencies are a major concern. Additional discussion of the various problems is given in the following sections.
Processing Sequence and Algorithms
This data set was formed by the Climatic Research Unit of East Anglia University and includes many data records that were previously unavailable; they were collected by the CRU scientists from unpublished sources and included in the data bank used for analysis. Extensive quality checks and corrections have been applied to the data to ensure, as well as possible, that the resulting data set is homogeneous. For the land stations corrections have been applied for the following problems: changes in instrumentation, changes in station location, changes in observation times, changes in the methods of calculation of monthly means, and urbanization effects (Jones et al., 1986a). Sea surface temperature anomalies have been included to create as near a complete analysis as possible. These data also required considerable correction (Jones et al., 1991). All data are presented as anomalies from the mean for the 1961-1990 period. In the previous data set posted at this site, the reference period was 1950-1979. However by 1990 only 52% of the land reference stations used to establish the 1950-1979 climatology were being routinely updated. It was therefore decided to establish a new reference period (Jones 1994). The new reference period is slightly warmer than the old one and this causes a change in the anomaly values. The 1990s anomalies will in the mean be a little smaller than in the old set, while the earlier anomalies will be a little more negative.
The data were originally collected as temperatures using thermometers. Then the data was corrected as discussed above and the anomalies from the reference period were calculated. The final step was to grid the data. The station data were gridded to a regular 5 degrees of latitude by 5 degrees of longitude grid (Jones 1994; Parker et al., 1994)). A station was used in connection with only one grid square. Because of the differences in the land and ocean data sources and the different type of corrections needed separate land and ocean grids were formed. Then the two grids were merged to form a single global grid. The land grid was used over land and the ocean grid over the ocean. If both are available in a given month for one of the boxes the weighted average of the two fields is taken with the weights being the land and ocean fractions in the box. However in a mixed box the land weight is always at least 25% so that ocean island data usually gets a larger weight than the land area would imply. The ocean island land temperatures are almost certainly more reliable than the sea surface temperatures for these boxes (Jones, 1997 private communication; compare with Parker et al., 1994).
There are no important differences in the global mean temperature trends obtained from the original and the revised gridded data sets. However some regional differences do occur. These are chiefly over continental regions. In the original land station analysis (Jones et al., 1986a) the grids were 5 degrees latitude by 10 degrees longitude. These were split into 5-degree by 5-degree grids to facilitate merging with the 5-degree by 5-degree ocean maps (Parker et al., 1994). In the reanalysis the land stations were directly averaged over 5-degree by 5-degree squares. In the computation of the earlier monthly means, the station data were weighted by the inverse of the distance to the nearest grid point (Jones et al., 1986a). In the reanalysis the monthly continental grid means are unweighted averages of all valid stations within the grid square (Jones 1994). Finally over 1000 additional land stations were used in the reanalysis.
For consistency with the other data sets in the Goddard DAAC's Interdisciplinary Data Collection, the East Anglia data was reformatted at the DAAC from the original integer values (anomalies scaled by 100) into 32-bit floating point quantities (unscaled anomaly values). In addition, whereas the original data were written out such that the left edge of the global map began at 120 degrees E. longitude , the data held at the Goddard DAAC has been modified such that the columns begin at the dateline (180 degrees west longitude). A visual comparison of the two data sets was then performed to ensure that no artifacts had been inadvertently introduced into the original data as a result of this procedure.
Hemispherical and Global Averages
Estimates of global and hemispheric monthly and annual temperature variations, relative to the 1961-1990 reference period, are presented for 1856-1996. These estimates were computed using a program supplied by Phil Jones. This program uses only the reporting (non fill value) grid squares to calculate the zonal, hemispherical and global monthly means. Thus in the hemispherical means a latitude zone is weighted by the relative geographical area of only the reporting grid squares. If a zone has no reports for that month it is ignored. In this scheme the global mean is not the mean of the northern and southern hemispherical means because there are more data gaps in the S. hemisphere. Hemispherical and global means were not calculated for the years 1851-1855 because of the paucity of data. The massive data jump in 1856 comes from ship reports of sea surface temperature. Jones and Briffa (1992) state, "Since the 1850s ships have been obliged to take weather observations and measure the temperature of the sea surface. The impetus behind this collection was an American naval captain, Matthew Fontaine Maury, who persuaded the other major maritime nations to instruct their military and merchant navies to take measurements and record these in log books." The procedure was formalized in 1853 at the Brussels Maritime Conference "for devising an uniform system of meteorological observations at sea" (Woodruff et al., 1987).The percentage of reporting 5-degree by 5-degree squares on the global grid was 3% in 1851. This increased to about 80% by 1960 and has fluctuated in that neighborhood since. These are chiefly related to ship reports. The recent percentage is about 90% in the Northern Hemisphere and between 70% and 80% in the Southern Hemisphere. Three dramatic shifts in the reporting patterns are noteworthy. In 1856 when the ships started reporting the sea surface temperature, the percentage of reporting grids jumped from 3% to 15%. The January 1856 global map indicates that the first ship reports came chiefly from British and other European ships on South American and Oriental trade routes. The start of World I caused a sharp decrease (53% to 35%) in reporting grids between January 1914 and January 1955. The Southern Hemisphere was most effected. Due to World War II there was also a sharp drop from 1939 to 1941 which again was largest in the Southern Hemisphere.
There is no one accepted, best way to obtain zonal, hemispherical and global means from insufficient sampling. Temperature anomalies vary strongly within a zone as well as latitudinally. The N. and S. hemispherical anomalies sometimes shift in opposite directions. There are also seasonal variations with anomalies often larger in some seasons than in others. Parker et al. (1994) discuss a scheme to calculate annual means which includes a grid square only if it has data for at least one month in each of the four seasons. Weighting by total zonal geographic area would at times give a zone with only one reporting station a weight similar to another zone in which all the grid squares had valid data. The weights are also affected by the size of the grid squares and how the station data are combined to obtain the monthly means for the grid squares. The chief purpose of the present data set is to examine long term trends. Two or three averaging methods may yield similar long term trends but often disagree as to which of two individual years has the larger mean temperature anomaly. Parker et al. (1994) and Hansen and Lebedeff (1987) both discuss the effect of various hemispherical and global averaging schemes. They concluded that all of the schemes they examined gave roughly the same long term trends. Hansen and Lebedeff added that their tests indicated that an averaging scheme similar to that used in the present Jones data set yields the most accurate long term trends. Additional discussions can be found in Jones et al. (1986a,b,c) and Jones (1988). For comparison purposes the Goddard Interdiscipline Data Collection includes the hemispherical and global means from the Goddard Institute for Space Studies (GISS) temperature anomaly data set. The GISS data set is based chiefly on the land stations and uses a different global grid scheme and reference period (1951-1980). The general long term trends are similar.
These global and hemispheric annual variations show little trend during the nineteenth century, marked warming to 1940, relatively steady conditions to the mid-1970's, followed by a rapid warming during the 1980's. Over the period of record, globally-averaged temperatures have risen approximately 0.5 degrees C. The warmest three years of the 1856-1996 record are, in descending order, 1995, 1990 and 1991. Globally, in 1991 the mean temperature variation was 0.29 degrees C above the 1961-90 reference period mean and 0.06 degrees C cooler than in 1990. Due to sampling and other error sources this 0.06 degree drop in 1991 may not be a significant year to year difference, but this combined with the additional 0.14 degree drop in 1992 is significant. This mean global cooling in 1991-1993 is generally attributed to the Mt. Pinatubo eruption in June, 1991. Stratospheric aerosols resulting from this eruption spread all over the globe and measurably increased the Earth's albedo for a period (Hansen et al., 1992; Lacis et al., 1992; Minnis et al., 1993) But by 1995 the mean temperature had returned to the 1990 level.
This temperature deviation data can be used for many types of studies including:
- Regional temperature variations over the last 100 or so years (Parker et al. (1994))
- Global Warming (Houghton et al. (1995);Richards (1993), Hansen and Lacis (1990). Note the cautions of Woodward and Gray (1993) concerning the limitations of certain statistical regression analysis procedures.)
- Correlations between various terrestrial climate variables ( Kyle et al. (1995); Ardanuy et al. (1992);Jones (1988))
- Correlation of variations in the climate and solar variability (Hoyt and Schatten (1993))
The data set was validated by its authors. The resulting annual mean temperatures were compared to other data sets, such as those from Russia (Vinnikov et al., 1990) and the United States (Hansen and Lebedeff, 1988). Although all the averages were highly correlated, the producers of this set believe this data set is superior because of its inclusion of marine data which represent 71% of the Earth's surface. Hansen's data set did not correct for urban warming and is thus considered to contain a warming bias of greater than 0.1 degrees Celsius per century.
To assess the effects of incomplete coverage during the early years, the producers used a frozen grid approach to analyze the changing network. They found that although the interannual variability decreased over time as more stations were used, there was no bias introduced by the sparse grid in the early part of the record. Although the data are calculated, stored, and presented in this data set to two decimal points, i.e. 0.01 degrees Celsius, the individual monthly grid point anomalies are probably only accurate to +/-0.2 degrees Celsius, given the accuracy of the original data. A detailed discussion of possible errors is given in Parker et al., (1994).. An analysis of various methods used to calculate grid point means and other points is given by Gunst et al. (1993). Jones et al. are publishing an analysis of the errors in the revised data set now on this site (it will appear in the Journal of climate in 1997).
As mentioned earlier, to ensure that the data as reformatted by the Goddard DAAC did not introduce spurious artifacts into the original data, GIF images were derived from the data as it was rewritten to the binary files and visually compared to decadal images produced from this data in Parker et al. (1994).
Points of Contact
- For information about or assistance in using any DAAC data, contact
- EOS Distributed Active Archive Center (DAAC)
- Code 902
- NASA Goddard Space Flight Center
- Greenbelt, Maryland 20771
- Internet: daacuso@daac.gsfc.nasa.gov
- 301-614-5224 (voice)
- 301-614-5268 (fax)
- To inquire about or order the original East Anglia Temperature Deviations data set, contact
- Dr. P.D. Jones
- Climatic Research Unit
- School of Environmental Sciences
- University of East Anglia
- Norwich NR4 7TJ
- UNITED KINGDOM
- Internet: P.Jones@uea.ac.uk
- Telephone: (0603) 592090
Jones, P. D. 1988. The influence of ENSO on global temperatures. Climate Monitor 17(3): 80-89.
Jones, P. D., S. C. B. Raper, C. M. Goodess, B. S. G. Cherry and T. M. L. Wigley, 1986c: A gridpoint surface air temperature data set for the southern hemisphere, U. S. Dept. of Energy, Carbon Dioxide Research Division, Washington, DC, Technical report, TR027, 73pp.
Jones, P. D., T.J. Osborn, and K.R. Briffa 1997: Estimating sampling errors in large-scale temperature averages, J. Climate, 10, 2548-2568.
Smithsonian Institution, 1927, 1935, 1947: "World Weather Records", Miscellaneous Collections, Volumes 79, 90, 104, Washington, D. C.
Woodruff, S. D., R. J. Slutz, R. J. Jenne, and P. M. Steurer, 1987: A comprehensive ocean-atmosphere data set, Bull. Am. Meteorol. Soc., 68, 1239-1250.
Woodward, W.A., and H.L. Gray, 1993: Global warming and the problem of testing for trend in time series data, J. Climate, 6, 953-962.
| 
| 
| 
|
| NASA | Goddard | GDAAC | IDC Data |