Homogenization of the global radiosonde temperature dataset

Reanalysis projects and climate data assimilation systems are expected to deliver temporally homogeneous gridded global 3D datasets of the atmospheric state quantities temperature, humidity and wind that cover several decades. These datasets, referred to as reanalyses, should be suitable for studies of climate trends and variability. However, first and second generation reanalyses did only partly meet the requirements for these purposes. The reason lies in the frequent changes of the atmospheric observing systems during the past 50 years. The radiosonde network is practically the only global upper air observing system until 1973 and even in the satellite era it has a profound influence on the quality of the reanalyses. Temperature time series of practically all radiosonde stations, are affected by several breaks due to the introduction of improved instrumentation. These breaks, which are largest in the upper troposphere and lower stratosphere, can severely distort trend estimates. The task of removing these breaks is referred to as homogenization. In this project it is proposed to correct radiosonde temperature time series by comparing them with time series of 6h-background forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40). These time series, while not perfect, can serve as reference for the correction of many radiosondes. In a one year effort at ECMWF the P.I. developed a first version of an automatic correction method based on analysis of the difference time series between ERA-40 background forecasts and radiosonde observations (bg-obs differences). While the method yields promising results, it needs to be improved before the corrected radiosonde time series can be used safely as input for reanalyses. In this project it is proposed to (i) refine and extend the statistical tools used for the present homogenization method, (ii) compare the statistically derived corrections with physically based corrections at well documented sites. It is proposed to use mainly ERA-40 bg-obs differences but also bg-obs differences from other reanalysis efforts. It is the hypothesis of this project that the production of a 60-year global homogenized radiosonde temperature dataset, ready for assimilation in a future reanalyses, is feasible with the use of this feedback information from modern global data assimilation systems. The proposed work is planned in cooperation with ECMWF and the UK MetOffice and should be completed within three years.

Reanalysis projects and climate data assimilation systems are expected to deliver temporally homogeneous gridded global 3D datasets of the atmospheric state quantities temperature, humidity and wind that cover several decades. These datasets, referred to as reanalyses, should be suitable for studies of climate trends and variability. However, first and second generation reanalyses did only partly meet the requirements for these purposes. The reason lies in the frequent changes of the atmospheric observing systems during the past 50 years. The radiosonde network is practically the only global upper air observing system until 1973 and even in the satellite era it has a profound influence on the quality of the reanalyses. Temperature time series of practically all radiosonde stations, are affected by several breaks due to the introduction of improved instrumentation. These breaks, which are largest in the upper troposphere and lower stratosphere, can severely distort trend estimates. The task of removing these breaks is referred to as homogenization. In this project it is proposed to correct radiosonde temperature time series by comparing them with time series of 6h-background forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40). These time series, while not perfect, can serve as reference for the correction of many radiosondes. In a one year effort at ECMWF the P.I. developed a first version of an automatic correction method based on analysis of the difference time series between ERA-40 background forecasts and radiosonde observations (bg-obs differences). While the method yields promising results, it needs to be improved before the corrected radiosonde time series can be used safely as input for reanalyses. In this project it is proposed to (i) refine and extend the statistical tools used for the present homogenization method, (ii) compare the statistically derived corrections with physically based corrections at well documented sites. It is proposed to use mainly ERA-40 bg-obs differences but also bg-obs differences from other reanalysis efforts. It is the hypothesis of this project that the production of a 60- year global homogenized radiosonde temperature dataset, ready for assimilation in a future reanalyses, is feasible with the use of this feedback information from modern global data assimilation systems. The proposed work is planned in cooperation with ECMWF and the UK MetOffice and should be completed within three years.