Diagnostic studies of new climate dataset: the <br />Reanalysis data. Examples of rainfall studies over tropical Africa

Abstract

This study establishes an objective diagnosis of NCEP/NCAR reanalysis data reliability over tropical Africa at intra-seasonal to interannuel space scale. The reanalysis data are an original database available over 1958-1998 period which combines observations and atmospheric model outputs and allows us to better understand global scale atmospheric dynamics. The study of data reliability allows us to identify a great number of recommendations for using them. Over tropical Africa, the homogeneous period which should be extracted goes from 1968 to 1998. Before this year, reanalysed datasets should be used with greatest caution. However, during this period, the SMWDA, a statistical method used to detect abrupt shifts in time series and to localize them, shows three major dates: 1976/77, 1983/84 and 1987/88, that could we found in a great number of atmospheric parameters. Surface parameters like rainfall show greatest number of abrupt shifts. The most reliable atmospheric parameters among 12 that we have tested are parameters for which the NCEP model has assimilated a maximum of observations (air temperature, geopotential height, winds ...). Over tropical Africa, these parameters are representative of realistic climatic fields. The parameters associated to hydrological cycle are not very reliable. A detailed study of rainfall has showed pronounced differences between reanalysis and observation dataset. The major problems identified are an existence of many shifts in time series, a strong underestimation of seasonal cycle, time series are dominated by a low-frequency variability and a frequently incoherent interannuel variability. Thus reanalyses rainfall should not be used over tropical Africa.&lt;br /&gt;Teleconnections between major sea surface temperature anomalies and tropical African rainfall variability have been investigated to determine the ability of reanalysis to help the understanding of the role of atmospheric circulation. Ocean signals w