5.1 Data processing and selection

In section 3.4 methods for data processing were presented. In some methods parameters have to be set. The values of the parameters which only depend on characteristics of the devices were already defined in that section. In this section, the remaining parameter, the clock period $t_{\text{cl}}$, and data selection criteria are defined.

The clock period which we will use in this chapter, is $t_{\text{cl}}=$ 5 min. This value is more than twice as large as the minimum clock period of 1-2 min allowed by the applied sampling times of both the wind and rain measurements. A shorter clock period would be an advantage for detailed analyses of rain data, but 5 min is the smallest unit of time generally used for analyses of wind data over years (as e.g. in [KNMI 1992]). The effect of the clock period value on the calculated values of the rain intensities is addressed later in this section.

The following quantities are used for the analyses of the measured wind, rain and driving rain data:

The wind velocities measured at several distances from the façade at P4 are not used in the analyses in the present chapter, but will be used for comparison with the CFD results in chapter 6. Note that the reference horizontal rain intensity which we will use in this chapter is the corrected rain intensity $R_{\text {h,c}}$ (defined in section 3.4.3). Later in this section, the choice of $R_{\text {h,c}}$ is explained.

Since in chapter 4 it was concluded that the driving rain gauge of the TUE-II type is very suitable, the measurements of the other types of driving rain gauges are not considered in the present chapter. We will focus on the spatial differences between the horizontal rain intensities at the two positions on the roof of the Auditorium (P2 versus P3, see figure 3.4) and between the driving rain intensities at the two positions at the west façade of the Main Building (P4/5 versus P6, figure 3.6).

The measurement data are put in a table in which every row represents a clock period and gives the corresponding values of the above mentioned quantities. A clock period is said to be ``available'' if every device, yielding one or two of the above mentioned quantities, is working properly. A device which is simply not installed (operational), does not cause a clock period to be unavailable. Tables 3.2 and 3.3 show the periods during which the devices were operational. An out-of-order due to maintenance, power failure or other malfunction when a device was operational, causes a clock period to be unavailable. A clock period is also unavailable when a device produced less than 90% of the readings which it should have produced during a clock period given its sample time.

Table: Monthly percentage of available clock periods. A clock period is ``available'' when all installed devices (at the Auditorium and at the west façade of the Main Building) were well functioning. Based on 5 min clock periods.
year month % avail. year month % avail.
1997 12 99 1998 12 99
1998 1 97 1999 1 99
1998 2 98 1999 2 98
1998 3 82 1999 3 99
1998 4 79 1999 4 98
1998 5 98 1999 5 99
1998 6 70 1999 6 99
1998 7 90 1999 7 90
1998 8 99 1999 8 89
1998 9 99 1999 9 96
1998 10 88 1999 10 99
1998 11 82 1999 11 94

The over-all period of the full-scale measurements started on 1-12-1997 and ended on 30-11-1999. Table 5.1 shows the percentage of available clock periods per month. The presence of months with availability percentages lower than 90% is explained in the following. The months of March and April 1998 have relatively low availability figures, due to the reprogramming of the PhyDAS data acquisition system. In June 1998, the data acquisition system was down during 6 days due to a power failure, which could not be fixed immediately. The driving rain collected during these days (the reservoirs were found to be almost full after this period) was not registered. During in total three days of October 1998 the data acquisition systems were out-of-order due to a probable power failure. During November 1998 the PhyDAS data acquisition system was again reprogrammed.

Figure 5.1: Histogram of ${ ({R_{\text {h}}} - \overline {R_{\text {h,c,5}}})}/{ \overline {R_{\text {h,c,5}}} }$ with ${R_{\text {h}}} = {R_{\text {h,c,10}}}$ ( $\includegraphics[width=2em]{gen/m-ononderbroken.eps}$ in figure a), ${R_{\text {h}}} = {R_{\text {h,u,10}}}$ ( $\includegraphics[width=2em]{gen/m-streep.eps}$, fig. a), ${R_{\text {h}}} = {R_{\text {h,c,60}}}$ ( $\includegraphics[width=2em]{gen/m-ononderbroken.eps}$, fig. b), ${R_{\text {h}}} = {R_{\text {h,u,60}}}$ ( $\includegraphics[width=2em]{gen/m-streep.eps}$, fig. b). Based on horizontal rain measurements at position P2 from 1-12-1997 to 30-11-1999.
% midden\{%
%%% x-axis [cc][b]
\psfrag{ratio} [cc][b]{${ ({R_{\text{h}}} - \o...
\end{tabbing} %\}

The correction of horizontal rain intensity data and the effect of clock period duration For the analysis of rain quantities, the corrected rain intensity $R_{\text {h,c}}$ is used instead of a rain intensity calculated directly from the number of tippings during a clock period ( $R_{\text{h,u}}$). These two quantities were defined in section 3.4.3. Figure 5.1 shows histograms of the relative difference:

\frac{ R_{\text{h}} - \overline{R_{\text{h,c,5}}} }{ \overline{R_{\text{h,c,5}}} },
\end{displaymath} (5.1)

where $R_{\text {h}}$ can either mean $R_{\text{h,c,10}}$ (corrected 10-min values), $R_{\text{h,u,10}}$ (uncorrected 10-min values), $R_{\text{h,c,60}}$ (corrected 1-h values) or $R_{\text{h,u,60}}$ (uncorrected 1-h values).

$\overline{R_{\text{h,c,5}}}$ is the average of the corrected 5-min horizontal rain intensities $R_{\text{h,c,5}}$ during a corresponding 10-min or 1-h clock period. A good correspondence between 5-min and 10-min (or 1-h) rain intensity values should result in a `high peak' at zero relative difference (formula 5.1) in the histogram.

Two important conclusions are drawn from the histograms in figure 5.1. The first conclusion is that the correction is indeed an improvement and yields more realistic rain intensity data. This is supported by two observations. The first observation is that figure 5.1a shows that the corrected 10-min rain intensities correspond better with the averaged corrected 5-min rain intensities than the uncorrected 10-min rain intensities. Approximately 90% of the corrected 10-min rain intensities deviate less than $\pm $5% from the corresponding corrected 5-min averages (figure 5.1a), whereas hardly 20% of the uncorrected 10-min rain intensities deviate less than $\pm $5% from the corresponding corrected 5-min averages. The second observation is that peaks occur at a relative difference of -1 in the histograms with the uncorrected rain intensities (figures 5.1a and 5.1b). A relative difference of -1 means zero $R_{\text {h}}$ and non-zero $\overline{R_{\text{h,c,5}}}$ (eq. 5.1), and occurs more often in the case of uncorrected $R_{\text{h,u}}$. Note that the errors in the uncorrected rain intensity values result from the principle of the tipping bucket of the rain gauges (see section 3.4.3).

The second conclusion is drawn from the relatively poor correspondence between the averaged 5-min rain intensities and the 1-h rain intensities. Figure 5.1b shows that only 75% of the 1-h rain intensities deviate less than $\pm $5% from the corresponding 5-min averages. Hence, hourly rain intensity data should be applied with care, because they will give a poor indication of the actual rain intensities, especially maxima. Five-minute data (or possibly 10-min) are thus preferred for our purposes.

KNMI data Data of the weather station at Eindhoven Airport were obtained from reports of the Royal Netherlands Meteorological Institute (KNMI). Thirty-year averages of monthly wind speeds based on data of 1961-1990 were presented in an official report on climatological data of 15 principal weather stations [KNMI 1992]. Every month, the KNMI issues a bulletin with (among other parameters) precipitation amount and monthly averaged wind speed [KNMI 1997-99].

On internet, the KNMI presents wind data of 51 weather stations [KNMI 2000]. The data consist of hourly averaged wind speeds and consist of averages of wind directions of the last 10 minutes of every clock hour. The actual wind speeds were averaged and subsequently corrected to obtain the so-called potential wind speed, which corresponds to a hypothetical wind that would blow a particular location from any direction at 10 m height over a terrain with $z_0=$ 0.03 m ([Wieringa 1996] and [Verkaik 2000]). The potential wind speed of every clock hour at the Airport are thus given by the KNMI. The available Eindhoven Airport data span from 1-1-1960 to 31-12-1999.

© 2002 Fabien J.R. van Mook
ISBN 90-6814-569-X
Published as issue 69 in the Bouwstenen series of the Faculty of Architecture, Planning and Building of the Eindhoven University of Technology.