ERA-15 Project

<Prev   TOC   Next >

9. Observing System Characteristics

Figure 1: The variability of the number of radiosonde data used daily in the reanalysis over the period 1979 to 1993.

Figure 2: The variability of the number of GMS cloud drift wind data used daily in the reanalysis over the period 1979 to 1993.

During the reanalysis period the global observing system experienced many changes. There were periods with one polar orbiting satellite, and others with two. Figure 1 shows the variability of the number of radiosonde height data, which the analysis has accepted and used. The FGGE year stands out clearly, as does the period 1987 to 1990. The year 1980 was a low point; from then on data slowly increased until 1986. After 1990 the radiosonde network slowly deteriorated expect in the tropics. The way observations are derived has also changed. Figure 2 shows as an example the time evolution of the number of used GMS winds. The different height assignment algorithms have produced winds on different altitudes. The number of aircraft winds also has increased through the years, and the number of drifting buoys has varied considerably.

Figure 3: The evolution of the RMS of the difference between first-guess heights and radiosonde heights in the Northern Hemisphere for the period 1979 to 1993: Units m.

Figure 3 shows the time evolution of the root mean square differences between the first guess and the radiosonde heights through the re-analysis period in the Northern Hemisphere. The improvement from 1979 to 1980 is significant above 250 hPa and is mostly due to the bias correction of radiosonde heights. There are still signs of steady improvement from 1980 onwards, which especially towards the end of the period is due to the improved quality of radiosondes.

Figure 4: The evolution of the RMS of the difference between first-guess layer mean temperatures and 1D-Var retrieved layer mean temperatures in the Northern Hemisphere for the period 1979 to 1993: Units deg.C.

The TOVS CCR data set is relatively uniform throughout the re-analysis period; the same instruments, HIRS, MSU and SSU, were used with the same specifications on all satellites. The NESDIS temperature and humidity retrievals used historically in ECMWF operations vary considerably with time due to ongoing development and tuning of the statistical retrieval algorithms. As can be seen from the time-evolution of the mean RMS of 1D-Var layer mean temperatures in the Northern Hemisphere, Figure 4, the two periods with one satellite, January 1979 to August 1981 and mid 1985 until the end of 1988, shows slightly larger errors than the rest of period with two satellites.

Figure 5: The evolution of the RMS of the difference between first-guess wind vectors and upper level cloud drift wind vectors (red line), and the difference between analyzed wind vectors and upper level cloud drift wind vectors (blue line) in the band (30 degrees N - 15 degrees N for the period 1979 to 1993: Units m s-1. The green line is the number of data used over the period.

The quality of cloud drift winds has improved also as can be seen from the time evolution of the wind RMS of first guess minus observed values in Figure. The RMS values of 9 m/s in 1979 has been reduced to 6 m/s in 1993.

There is overall evidence that the global observing system improves through the period.

<Prev   TOC   Next >