A simple climatology of westerly jet streams in global reanalysis datasets part 1: mid-latitude upper tropospheric jets

Abstract

A simple closed contour object identification scheme has been applied to the zonal mean monthly mean zonal wind fields from nine global reanalysis data sets for 31 years of the satellite era (1979–2009) to identify objects corresponding to westerly jet streams. The results cluster naturally into six individual jet streams but only the mid-latitude upper-tropospheric jets are considered here. The time series of the jet properties from all reanalyses are decomposed into seasonal means and anomalies, and correlations between variables are evaluated, with the aim of identifying robust features which can form the basis of evaluation metrics for climate model simulations of the twentieth century. There is substantial agreement between all the reanalyses for all jet properties although there are some systematic differences with particular data sets. Some of the results from the object identification applied to the reanalyses are used in a simple example of a model evaluation score for the zonal mean jet seasonal cycle.

Appendix: The definition of seasonal metric scores

The main quantities needed to derive the metric scores for the seasonal cycle are the mean reanalysis mean (OMean) and standard deviation (OSTD) as defined in Sect. 5 and the 31 year mean seasonal cycle (MMean) from the model to be evaluated.

The mean error score is the simple difference in annual means scaled by the annual mean of OSTD to make it dimensionless, viz,

$$MScore = (\overline{MMean} - \overline{OMean} )/\overline{OSTD}$$

where the overbar denotes the mean over the year. Monthly anomaly series are defined by,

$$MAnom_{m} = MMean_{m} - \overline{MMean}$$

where m is the month index. The annual ranges are defined as the difference between the maximum and minimum monthly values,

$$MRange = \hbox{max} \left( {MAnom} \right) - { \hbox{min} }(MAnom)$$

The observation anomaly and range are defined similarly. The range score is defined by,

$$RScore = (MRange - ORange)/\overline{OSTD}.$$

The shape error is defined as a scaled RMS error using the monthly anomaly series normalized by the range,

$$SScore = \sqrt {\mathop \sum \limits_{m} \left\{ {\left( {MNorm_{m} - ONorm_{m} } \right)/\left( {\frac{{OSTD_{m} }}{ORange}} \right)} \right\}^{2} }$$

where the normalized series are defined by,

$$MNorm_{m} = {\raise0.7ex\hbox{${MAnom_{m} }$} \!\mathord{\left/ {\vphantom {{MAnom_{m} } {MRange}}}\right.\kern-0pt} \!\lower0.7ex\hbox{${MRange}$}}.$$