|
Feature Based Diagnostics From ECMWF/NCEP Analyses And AMIPII: Model Climatologies |
Last Modified
1a. Storm Tracks
The data used in this study are instantaneous fields every 6 hours. Since we are interested in the synoptic spatial scales the following pre-processing has been performed on each field before applying the tracking algorithm. Because we want to explore a wide range of fields at multiple levels, synoptic scale features can be identified consistently in these fields if we first remove the large scale, slowly varying planetary scales. So each field is spectrally decomposed using spherical harmonics and the coefficients for total wave numbers n<=5 are set to zero before converting back to grid point space. In addition the fields are spectrally truncated to T42 and some smoothing in the form of a tapering of the spectral coefficients is applied.
The tracking algorithm is applied to both the positive and negative anomalies in the filtered fields and to individual seasons for each year; December, January February (DJF); March, April, May (MAM), June, July, August (JJA), September, October, November (SON). This is performed for each hemisphere. A limited Eulerian analysis is also performed for comparison with the feature based statistics, thus mean fields, STD and 2-6 day bandpass filtered variance are computed for the spatially filtered fields. These results my be presented as well if requested.
1b. Tropical Easterly Waves
The processing for the EW activity is also performed at the T42 resolution for all models except those that are already at lower resolution. This is the only pre-processing performed (i.e. no planetray wave removal or additional smoothing) to prevent the degredation of the already weak waves. The period covered is from May to November (7 months) for the tropical region ~5S - 40N. Additionaly seasonal cycle statistics are computed but are not displayed here.
1c. Statistics
Before computing the statistics the track ensembles for each season for each year are filtered. Since we are only interested in the most coherent systems the ensembles are filtered to retain only those tracks that have lifetimes >= 2 days and which travel further than 1000Km. This may appear some what arbitrary but is used to exclude systems that are semi-stationary (which may be important in their own right and which may be explored later) or short lived secondary activity (again the nature of this activity may be explored later).
The track ensembles for each year are then combined into total ensembles for each season for the calculation of the climatological statistics presented below. Statistics are in the form of distribution densities and mean attributes. All densities have been scaled from probability density functions to number density per unit area per season, where the unit area is equivalent to a 50 spherical cap (~106 Km2). By per season we mean that the densities have been normalized by the number of contributing track ensembles for that season. The mean attribute statistics are suppressed (not plotted) in regions where the feature or track densities are low as these will have a lower statistical significance than where the densities are high (confidence maps could be computed but this would increase the number of plots significantly and with additional computational cost). This gives some of the plots a rather ragged appearance. Smoother plots can be produced but this may give the wrong impression of the mean attributes. The thresholds used are again arbitrary but can be changed easily.
The climatological results presented here will be updated as more become available, particularly from AMIPII integration's. Also, results exploring the variability in the statistics with teleconnections will be made available later together with the methodology for the calculation.
The track density statistic is now defined in its more traditional form as opposed to the results previously found here. See Storm Track Paper for further details.
Summary table of available storm track results
These results are double page gziped postscript files of a range of statistics for each model for the 850hPa Vorticity only.
This work forms part of an ongoing study of synoptic scale features in forecast analyses, model and satellite data using feature based methods. Work on the ECMWF Re-analyses is being conducted in collaboration with Brian Hoskins at Reading. Work on an Atmospheric Model Intercomparison Project (AMIP) II diagnostic sub-project is also currently being undertaken with the Program for Climate Model Diagnosis and Intercomparison (PCMDI) (http://www-pcmdi.llnl.gov/). The main contact for the AMIP II project is James Boyle at PCMDI.
Further information on the AMIP II diagnostic sub-project can be found at http://www-pcmdi.llnl.gov/old/amip/DIAGSUBS/sp3.html For any suggestions or comments on the results presented here please contact me at kih@mail.nerc-essc.ac.uk