DAQUAMAP Redo

Introduction

After the first DAQUAMAP (Data Quality in MAP) run Häberli et al. (2004) found that in the south-east of France and in the north-east of Italy the interquartile range (IQR) of the differences between the reported and corrected values were large. The reason was partly the delivery of erroneous data. After the end of the first DAQUAMAP-run, corrected data and data from additional stations were delivered.

The goal of DAQUAMAP-redo for the MAP-SOP was to repeat the whole DAQUAMAP procedure within EUMETNET programme MAP-NWS based on the "frozen" dataset in the MAP data centre.

Method

A mathematical method of quality control has been developed and continually improved in the project of DAQUAMAP in order to get a more homogeneous data set for the whole Alpine region. The DAQUAMAP procedure (also called VERAQC- the Quality Control module of the Vienna Enhanced Resolution Analysis scheme) includes a piecewise functional fitting approach which is based on a variational algorithm. Like for thin-plate spline, an integral of squares of second spatial derivatives is minimised. The second derivates are obtained from overlapping finite elements using a polynomial approach (Steinacker et al. 2000).

The difference between the first DAQUAMAP run und the redo for SOP is, that the MAP data centre has changed to a MeteoSwiss server. Thus, all scripts and programs had to be modified and tested again. The same error measurements (global range check, gross error test) and statistics (median, mean, IQR, standard deviation, etc.) as in the first run were calculated for mean sea level pressure, potential temperature and equivalent potential temperature as a measure for humidity. All stations on each height zone (<750 m MSL, 751-2300 m MSL, >2300 m MSL) are included.

Results

The results of the DAQUAMAP-redo for SOP are available from the MAP database (password required): http://www.map.meteoswiss.ch/mm-doc/daquamap/Frameresults.html.

286 stations more than the first run are included in the DAQUAMAP-redo. The geographical location of the added stations is shown in Fig.1. These additional stations are located in France (208 stations), Germany (23), Hungary (19), Italy (17), Belgium (5), Switzerland (5), North Africa (2), Russia (2), Spain (2), Czech Republic (1), Ireland (1) and Slovenia (1). Altogether, deviations for about 1150 stations are calculated on the height layer 1 (<750 m MSL) and about 130 stations on the height layer 2 (750-2300 m MSL).

The mean differences of the deviations between the first and the second run for the stations below 750 msl are 0.0014 hPa (standard deviation: 0.84) for mean sea level pressure, 0.0243 K (1.74) for equivalent potential temperature and -0.0468 K (1.25) for potential temperature. The density distribution of the differences between the calculated deviations for all stations is also calculated. The largest density is about null for all three parameters. Especially for the mean sea level pressure the distribution is narrow. The largest bandwidth appears for water vapour mixing ratio.

The percentage of the IQR is still higher for water vapour mixing ratio than for potential temperature and for mean sea level pressure (see Fig. 2). The percentage of IQR for mean sea level pressure is the smallest. In contrast to the first run the IQR's are generally smaller. For topographical layer 2 the gradient of the curve isn't so large. It means that the distribution of the IQR on the higher topographical layer is larger, but it pays attention to the smaller number of stations on this height zone.

Fig. 3 shows the interquartile range of the sea level pressure-, potential temperature- and the water vapour mixing ratio deviations for each height zone. The IQR for water vapour mixing ratio is generally the largest especially around the Mediterranean. The mean sea level pressure shows that in the south-east of France and in the north-east of Italy, there is still the largest IQR. But the magnitude of the IQR is smaller for all parameters than in the first run (see Haeberli et al. 2004.). The north-south gradient with larger values to the south still appears in the second run. On the altitudinal belt 750 - 2300 m MSL the differences between the parameters are not very large. Generally, the IQR on the altitudinal belt 2 seems to be larger than on the altitudinal belt <750 m MSL for all parameters.

If the deviations are plotted in time it appears that for example for Wynau (Switzerland) the plot and the statistic from the redo and the first run are nearly equal (Fig. 4). But if a plot from north-east Italy is shown, for example Pordedone, it shows that the deviations curve and the statistic values also differ. The IQR of Pordedone is much smaller in the second run than in the first.

Conclusions

The results of DAQUAMAP-redo are in good agreement with the first run. Now, the quality assessment and information about the station characteristics are available for the "frozen" MAP surface data set.

Acknowledgements

This work was founded by the EUMETNET programme MAP-NWS.

References

Häberli C., I. Groehn, R. Steinacker, W. Pöttschacher, and M. Dorninger, 2004: Performance of the surface observation network during MAP. Meteorologische Zeitschrift, Vol. 13, No. 2, 109-121

Steinacker R., C. Häberli, and W. Pöttschacher, 2000: A transparent method for the analysis and quality evaluation of irregularly distributed and noisy observational data. Mon. Wea. Rev., 128, 2303-2316