VERIFICATION OF METEOSAT WINDS FROM MPEF AND MIEC Heinrich Woick EUMETSAT, Am Kavalleriesand 31, 64295 Darmstadt, Germany ABSTRACT Operational production of wind data from METEOSAT was transferred from ESA’s Meteorological Information Extraction Centre (MIEC) to EUMETSAT’s Meteorological Products Extraction Facility (MPEF) in November 1995. The European Centre for Medium Range Weather Forecast (ECMWF) supported validation of MPEF winds by providing monthly statistics of differences between winds from MPEF and from conventional radiosonde observations. Evaluation of MPEF data available from November 1995 to May 1996 shows that MPEF high level winds from the METEOSAT water vapour channel have reached a quality better than MIEC and nearly comparable quantity, whereas high level IR winds appear to be worse. MPEF medium and low level IR winds are better than MIEC but their number is much lower. 1. INTRODUCTION The evolution of the METEOSAT wind extraction scheme has been documented by various authors, for example by Schmetz et. al. (1993). Quantitative estimates of performance improvements were performed by Woick (1993), using monthly verification data of METEOSAT winds versus radiosonde wind observations based on comparison with collocated radiosonde data. The technique used the well-known empirical correlation between speed difference and the measured radiosonde winds to compensate for the effect of seasonal variations of the performance parameters. The same method is used in this paper to study the relative performance of MPEF winds from November 1995 to May 1996 and MIEC winds from the last 15 months of operations. 2. MPEF AND MIEC WIND VALIDATION Provision of validation data by MIEC was performed without major change since MIEC became operational until the end of November 1995. ECMWF validations started in October 1995, and MPEF operational wind extraction started in November 1995. This resulted in a duplication of MIEC validation data for the month of October 1995, i.e. data were available from both, MIEC and ECMWF. This duplication is useful to assess the effect of systematic differences between the two validation schemes on the result of performance comparison between MIEC and MPEF winds.ECMWF and MIEC validation procedures differed in two respects: first, the spatial collocation window used by ECMWF is narrower than the one used by MIEC. Second, ECMWF selected the radiosonde data according to an internal quality control scheme. Third, MIEC collocations were screened in order to remove gross errors in either the radiosondes or the satellite wind vectors, in cases when the vector difference exceeded 30 m/s or the direction difference exceeded 60 degrees. A ummary of the differences is shown in Table 1. s MIEC MPEF Period early 1978 - Nov 95 Nov 95 - now validation by MIEC ECMWF exceptional validation by ECMWF, for Oct 95 only Collocation time 1.5 h 1.5 h Collocation vertical 25 hPa 25 hPa Collocation horizontal 2x2 deg lat. - long. 150 km Radiosonde selection all available from GTS selection by quality Other screening vector difference ≤30 m/s direction difference ≤ 60° Table 1: MIEC and MPEF verification method differences As an immediate effect, the number of collocations between satellite winds and radiosonde winds in the ECMWF validation scheme was reduced to only one third of the number known from MIEC. Another effect concerns the performance differences for this month and needs some discussion. Since the collocation window became more narrow and better radiosonde winds were selected for the comparison, it was expected that the ECMWF validation procedure would lead to smaller differences between satellite wind and radiosonde data. On the other hand, the screening of large deviations of vector and direction difference by MIEC could let the MIEC winds look slightly better. Table 2 shows the performance differences found by the ECMWF and the MIEC validation schemes for October 1995, respectively, for the two main validation parameters, i.e. the RMS Vector Difference and the Speed Difference. Obviously, the table supports the contrary of what was expected because the ECMWF statistics shows larger values of negative speed difference and RMS ector difference than the MIEC statistics, except for the IR high level winds. v Wind Data Set of ECMWF minus MIEC ECMWF minus MIEC ECMWF minus MIEC mean October 1995 RMS Vector Difference Speed Difference Radiosonde Speed Difference Water Vapour 0.9 m/s -0,41 m/s -1,43 IR high level - 0.3 m/s -0,90 m/s -0,04 IR medium level 0.3 m/s -0,60 m/s -0,10 IR low level 0.5 m/s -0,63 m/s -0,21 Table 2: MIEC and MPEF validation differences for October 1995 Differences between ECMWF and MIEC validation data were largest for high level water vapour winds. For this data set, mean radiosonde speed in the ECMWF statistics was lower by 1.43 m/s, probably caused by different sets of radiosonde data used by ECMWF after quality screening. This causes the ECMWF validation of MPEF WV winds to look slightly better than they would be at the same wind speed, because of the empirical correlation of speed bias and RMS vector difference with mean radiosonde speed. A correction can be made by estimating the relevant regression factors from the slopes of the two regression lines in Figure 1. The resulting effect from the differences between validation schemes of ECMWF and MIEC on high level water vapour winds is such that MPEF speed bias appears 0.6 m/s more negative and RMS vector difference appears 1.2 m/s larger than for MIEC water vapour winds.For the other height levels, radiosonde speed differences are smaller and such a correction is considered unnecessary. The overall impression from Table 2 is that either, MIEC values were too optimistic or ECMWF validation is too pessimistic. There is no obvious argument to support either of these two possibilities. These differences are based on one month of data only. However, since they look consistent amongst the different height levels, they are considered as relevant for comparison between MPEF and MIEC winds performance and are taken into account in the further discussion below. 3. MIEC VALIDATION RESULTS The performance of MIEC winds during the period from September 1994 to November 1995 is illustrated in the following scatter diagrams of monthly average speed bias and RMS vector difference between METEOSAT and
View Full Document
Unlocking...