This page provides homogeneity adjustments for upper air temperatures
from the global radiosonde network. The homogeneity adjustments are
required for serious analysis of the upper air climate of the past
decades and may be used also for improving the data input for climate
data assimilation efforts (often called reanalyses) such as ERA-40 or ERA-Interim.
The adjustment process has two steps:
1) Detection of shifts in existing radiosonde observation time series
2) Estimation of the size of the shifts through comparison with suitable reference series.
Breakpoints in observed radiosonde temperature time series are detected through time series analysis of
innovation statistics from ERA-40 (1958-2001) and of the operational
data assimilation process (2002 onwards). This means that background forecast time series have been used as reference for break detection. The rationale for this
approach is documented in an ERA-40 project report Haimberger (2005) and in Haimberger (2007) (J. Climate). An update employing ERA-Interim analysis feedback data is under development.
The background forecast time series serve well as a reference for break detection but once the breakpoints are known, there are several options for breaksize estimation. We have implemented two options so far:
One is to use again the background forecasts as reference. This has been done in the RAOBCORE homogenization method. The dataset is the outcome of two years of research, one year funded by
the European Commission (Marie Curie Fellowship MEIF-CT-2003-503976 and
one year funded by the Austrian Fonds zur Förderung der
wissenschaftlichen Forschung (P18120-N10). Funds for further improving the product are available at least until 2008. The dataset is maintained in cooperation with ECMWF and with the UK MetOffice. Regular updates are planned after a calendar year has been completed.
The current version of radiosonde temperature adjustments is RAOBCORE_T_1.4.
It basically consists of a single ASCII-file containing the stationIDs
and all the adjustments and adjustment dates on 16 pressure levels for
00GMT and 12GMT. The adjustments can be added to the original datasets
to yield homogenized time series. Results from RAOBCORE_T_1.4 are documented in Haimberger, Tavolato and Sperka (2008)
For easier comparison with satellite and climate model data, a gridded version of this dataset has been made available as well. It contains layer mean temperatures for the TLS/TTS/TMT/TLT layers on a 2.5/2.5 grid.
For a recent climate model to observation comparison "Consistency of Modelled and Observed Temperature Trends in the Tropical Troposphere", Peter Thorne from UK Metoffice has created gridded NetCDF versions of RAOBCORE v1.4 as well as RICH. A RAOBCORE-adjusted dataset can be downloaded as ftp://srvx6.img.univie.ac.at/pub/raobcore14_gridded.nc The files are CF compliant and can be viewed with freeware tools such as ncview.
Using the background forecasts as reference has the disadvantage that the forecasts themselves may be influenced by biases in the radiosonde temperatuers. They may also be influenced by biases from other observing systems, most notably satellites. This problem can be avoided by creating reference series from neighboring radiosonde stations for breakpoint adjustment. This works well as long as the radiosonde network is not too sparse and as long one takes care that only homogeneous pieces of the neighboring time series are used. RICH is superficially documented in Haimberger, Tavolato and Sperka (2008), a more detailed documentation is in preparation.
A gridded NetCDF formatted RICH-adjusted radiosondet dataset can be downloaded from ftp://srvx6.img.univie.ac.at/pub/rich_gridded.nc. No individual station adjustments are provided yet. Plots of MSU-equivalent time series from RICH can be found in the RAOBCORE web visualization.
 |
The radiosonde data used can be viewed with the RAOBCORE web visualization tool.
This figure shows the observation-ERA40 background time series at the station Jan Mayen (red) in 30 hPa and the corresponding SNH-Test curve (blue). The web visualization tool includes figures like this for all stations as well as trend maps, zonal mean temperature trends and vertical break profiles.
| Version History | |
|---|---|
| 10 January 2006 | RAOBCORE_T_1.0 homogeneity adjustments have been made publicly available. They cover the period 1958-2004. It is recommended to use version 1.1 |
| 13 February 2006 | RAOBCORE_T_1.1 is the version used for preparing a mauscript submitted to J. Climate. |
| 02 August 2006 | RAOBCORE_T_1.2 is the version actually used for Haimberger, 2006 J. Climate. |
| 02 August 2006 | RAOBCORE_T_1.3 refers to the NOBGC experiment in Haimberger, 2007 J. Climate. |
| 30 January 2007 | RAOBCORE_T_1.4 is an update of Haimberger, 2007 J. Climate, reaching up to December 2006 and with more conservative ERA-40 bg modification. |
