Checking the temporal homogeneity of the ERA-40 observational input and analyses using analyses feedback data

Principal Investigator

Dr. Leopold Haimberger
Institute for Meteorology and Geophysics,
University of Vienna
Althanstrasse 14,
A-1090 Wien, Austria

leopold.haimberger@univie.ac.at

Project description

A new special project, dedicated to checking temporal homogeneity of ERA-40 data is proposed. The observational input database of ERA-40 and the final analysis may contain artificial jumps or drifts not caused by climate changes but by changes in the observing system.

Concerning the upper-air observing systems only time series of radiosonde data are sufficiently long to be useful for climate change studies. However the time series of radiosonde data often contain artificial breaks since many of the biases evident in the early radiosonde data have been corrected in later years due to improvements of the instrumentation. In this special project it will be tried to detect and correct these breaks in the temperature series and in the humidity series as fast as possible.

The inhomogeneities are sometime detectable through intercomparison of data from day/night ascents of from different levels of the radiosonde ascents. They may also be detectable through comparison with a homogeneous reference dataset. Such a reference dataset has not yet been available. Now that ERS-40 is complete, the radiosonde series can be compared with the ERA-40 first guess series. The ERA-40 first guess series is assumed to be relatively homogeneous in time and is available for comparison for every radiosonde to be investigated. The differences between first guess and radiosonde observations are available in the ERA-40 analysis feedback dataset.

Preliminary studies have shown that these differences are generally very small, so that artificial breaks in the series show up quite clearly in the time series of these differences. Various relative homogeneity test developed in the last two decades will be applied to these differences. The break locations and magnitudes suggested by the homogeneity tests shall be systematically compared with digitized metadata information (e.g. instrument type) available from datasets such as CARDS. It will be attempted to correct significant breaks using both metadata information and homogeneity test statistics. This correction should be done automatically as far as possible.

Not only the observational input to ERA-40 may be temporally inhomogeneous but also the ERA-40 analyses maybe contain artificial breaks. The homogeneity of the ERA-40 analyses may be tested by applying homogeneity tests to differences between the analyses and ensembles of radiosonde series on large spatial scales. Inhomogeneities in ERA-40 should show up on large scales since the main candidates for introducing inhomogeneities into ERA-40 analyses and first guess data are satellite observing systems. Artificial breaks in the first guess and analysis data need to be detected since they may limit the applicability of ERA-40 analyses for climate change detection studies. Within the project at least an assessment of the temporal homogeneity of the radiosonde series around the globe and of ERA-40 analyses based on the feedback data information shall be provided. Such an assessment should be valuable both for climate change detection studies using ERA-40 data as well as for possible future reanalyses. In the ideal case that a one year visit of the P.I. at ECMWF within a Marie Curie Fellowship is funded, also the development of an automatic correction procedure for radiosonde time series seems feasible.

In order to reach the project goals mainly access to the MARS archive and sufficient disk space in ECFILE is required. The computer power needed for processing the feedback is relatively small. The project should be completed by the end of 2005.

Final Report

Additional information

Project period 2004-2005.