ERA-15 Project

<Prev   TOC   Next >

10. Some General Circulation Features in ERA.
Comparisons with ECMWF Operational Analyses.

Tropical divergence.

The operational analysis and forecasting system at ECMWF has undergone continuous development and upgrading through the years, this is indeed the very reason for doing a reanalysis at all. During our monitoring and validation we have compared different aspects in ERA with the old operations and we will show an example from 1983, when the operational model was an N48 gridpoint model with a resolution of 1.875o by 1.875o and with 16 vertical levels. The parametrization at that time included a diagnostic cloud scheme for the radiation and a Kuo-type convection. The SST analyses used were the operational NMC analyses received at ECMWF in a 5o by 5o resolution. This was near the end of the ECMWF 'gridpoint era', the spectral T63 model was introduced in April 1983. Diabatic initialization had been introduced in September 1982. Furthermore, in 1983 the TOVS were used in the form of NESDIS SATEM with a resolution of 500km rather than the 250 km cloud cleared radiances used in ERA.

Figure 6: The monthly mean velocity potential at 150 hPa (blue and red contours) from the reanalyses for March 1983: Units 10-5 m2 s-1. The divergent wind at this level is illustrated by the arrows.

Figure 7: The monthly mean velocity potential at 150 hPa (blue and red contours) from the operational analyses for March 1983: Units 10-5 m2 s-1. The divergent wind at this level is illustrated by the arrows.

The two maps in figures 6 and 7 show the monthly mean velocity potential and divergent wind at 150hPa during March 1983. This was during the most intense phase of the 1982-83 El Nino / Southern Oscillation (ENSO) event. The anomalous sea surface temperatures, up to 4-5 oC warmer at the equator around 140 oW, forces the major convective area to migrate from its normal position east of New Guinea to the position in figure 6. As vividly seen in figure 7, the ECMWF operational analyses at the time were unable to catch this anomalous circulation, indeed there is very little divergent outflow to be seen at all.

The 1982-83 ENSO event shows up in many other aspects of the ERA circulation as well. The famous pressure oscillation between Papeete (Tahiti) and Darwin (northern Australia) is well captured, as are the draughts in tropical Africa and northeastern Brazil and the excessive precipitation over the central Pacific. At higher latitudes the Pacific-North American anomaly (PNA) pattern in the surface pressure which is a fingerprint of anomalous circulation over large parts of the northern hemisphere is clearly seen. Some readers may remember that the winter of 1983 was warm in northeastern Europe and cold in the southwestern parts. The ERA mean maps show how this anomaly was connected with the ENSO event through the PNA.

Tropical cyclones

All tropical cyclones in ERA were tracked, using vorticity maxima at 850hPa and surface pressure minima in the relevant areas and seasons. The tracks were compared with existing 'best track' data based on satellite imagery and airborne in situ observations. Over the 15 years ERA is able to find about 80% (with respect to vorticity maxima) of all Northern Hemisphere reported tropical disturbances and cyclones, albeit not at all with the observed intensity. The average positional error is of the order 150km, not much more than one gridpoint. The ECMWF operations reached a similar accuracy only towards the end of the 1980's.

Looking at the performance in the different ocean basins, the picture is somewhat more mixed. In the eastern Pacific the tropical cyclones are poorly analyzed both in number and in location, this is an area with very few observations. In the Atlantic and the western Pacific, which has the greatest number of tropical cyclones, they are analyzed quite well. In the northern Indian Ocean and the southern hemisphere it is difficult to draw statistically safe conclusions due to too few tropical cyclones.

The stratosphere.

The ERA assimilation system has only four levels above 100hPa, at 70, 50, 30 and 10 hPa respectively. The potential of the analyses to capture stratospheric circulation features such as the quasi-biennial oscillation (QBO) was thus not assured. To make things worse, near the equator there are very few reliable radiosondes reaching 10hPa or above, and the NESDIS TOVS data used elsewhere above 100 hPa in ERA were not used between 20 N and 20 S just because of their bad vertical resolution which, in tests, was found to smear out the strong vertical wind shear of the QBO phenomenon.

Figure 8: The evolution of the monthly anomaly of the mean zonal wind between 5 N and 5 S over the period 1979 to 1993: Units m s-1.

In spite of all these 'ifs and buts', the reanalyses do exhibit a very well developed QBO. In figure 8 a time-height diagram of the monthly mean zonal wind anomalies is shown. The quasi-biennial switch between easterlies and westerlies, and the downward migration of the anomalies is clearly seen. Similar plots localised over Singapore (not shown here) confirm that the ERA QBO is quite realistic compared to the soundings.

Figure 9: The evolution of the global monthly temperature anomaly (full line) and 12 month moving average (dashed line) at 30 hPa for the period 1979 to 1993: Units deg. C.

Another stratospheric feature that shows up beautifully in the reanalyses is the global heating in the stratosphere caused by the aerosol clouds emitted by the volcanoes El Chichon (Mexico) in 1982 and Pinatubo (Philippines) in 1991. Since the assimilating model does not know about these emissions, the global temperature anomalies seen in figure 9 is entirely due to the data, in this case primarily the TOVS temperatures and some radiosondes. In the figure one can also notice a gradual cooling of the stratosphere over the 15 years.

Surface fluxes.

Good estimates of the surface fluxes of energy, momentum and water, over land and particularly over the oceans, are of fundamental importance for the understanding and modelling of the coupled atmosphere-ocean system. Estimates of these fluxes from observations is difficult and existing attempts vary considerably in coverage and quality. Thus the user community has expressed great expectations from the reanalyses, where an assimilating model actually calculates them globally and in detail. In ERA the fluxes are extracted from twice daily forecasts up to +24 hours.

Figure 10 Estimates of the zonally averaged total precipitation rate for four forecast periods - 0 to 6 hours (cyan), 0 to 12 hours (green), 0 to 24 hours (red) and 12 to 24 hours (blue): Units mm day-1.

It is well known that model generated fluxes generally suffer from spin-up or spin-down, i.e. they increase (or decrease) with the forecast length. This is also true for the ERA fluxes. Figure 10 shows the zonally averaged total precipitation for four forecast lengths, 00->06 (cyan), 00->12(green), 00- >24(red) and 12->24(blue) hours. It is evident that the precipitation intensity increases with the forecast length during the first 24 hours in these 11-year averages. Other fluxes, such as the evaporation and the net energy also exhibit spin up problems. Thus, in the 11 year average, the global net energy exchange drifts from 7 W/m2 going from the atmosphere to the oceans in the 00->06 forecasts to 3 W/m2 going from the oceans to the atmosphere in the 12->24 forecasts.

There are also very large inter annual variations in the fluxes, some of which may be related to varying data coverage, for instance the loss of satellite wind data over the Indian Ocean in 1980. Others are likely to be indications of real inter-annual variations in the global circulation. The distinction between data coverage and data quality related variations and variations due to genuine circulation changes will be a major, and difficult task for future research based on ERA. Some assistance may be given by the upper air general circulation statistics, such as variances and covariances, that have been collected during the assimilation, but so far not evaluated.

<Prev   TOC   Next >