4.2 The TUE-I and TUE-II gauges

The two driving rain gauges TUE-I and TUE-II are almost identical. Gauge TUE-II is equipped with a wiper, gauge TUE-I is not.

In the previous section 4.1 it was shown that on a monthly basis the TUE-I gauge registers only approximately half of the rain amount measured by the TUE-II gauge. This is also valid for 10-min periods, as is shown in figure 4.1c. This figure shows correlations of 10-min driving rain intensities measured by the TUE-I gauge with the TUE-II readings. To visualise the possible influence of wind, the data points were selected for certain wind velocities $U_y$, and put into two sets, i.e. a set with 4-5 m s$^{-1}$ and a set with 6-7 m s$^{-1}$. The measurement points were also fitted by a linear function $y=cx$. The slopes are $c=0.54$ for $U_y =$ 4-5 m s$^{-1}$, and $c=0.51$ for $U_y =$ 6-7 m s$^{-1}$. The correlation coefficients are quite satisfactory: $r^2=$ 0.96 and 0.93, respectively. Figures 4.3c and 4.4c show the data of figure 4.1c separated into measurement points with reference rain intensities $R_{\text {h,c}}$ of 1.0-2.5 mm h$^{-1}$ and 3.0-5.5 mm h$^{-1}$, respectively. Comparing these three figures, one observes that higher wind speeds and higher horizontal rain intensities (in other words: the measurement points indicated by $\includegraphics[width=0.5em]{gen/m-rondje.eps}$ in figure 4.4c) yield more data through which a `nicer' fit can be obtained, although the correlation coefficient is not better than that of figure 4.1c. The correlation between the TUE-I and TUE-II gauges is more systematic (less scatter) at higher driving rain intensities. For lower driving rain intensities (figure 4.3c) the scatter is larger, but the slopes of the correlation for $U_y =$ 4-5 m s$^{-1}$ and 6-7 m s$^{-1}$ are approximately the same.

During the summer of 1998 we interchanged the positions of the gauges TUE-I and TUE-II. Figure 4.6a shows the correlation when TUE-I was at position P4 and TUE-II at P5. Figure 4.6b shows the situation after TUE-II was installed at P4 and TUE-I at P5. The correlation before and after the exchange is approximately the same: $R_{\text{TUE-I}} / R_{\text{TUE-II}}$ equals 0.48 (with $r^2=$ 0.89) in the first case, and 0.51 (with $r^2=$ 0.88) in the latter case. This indicates that the measurement positions P4 and P5 are close enough to have the same driving rain onslaught.

Figure 4.5: Driving rain gauge TUE-Ib (i.e. the TUE-I gauge with the reduced catchment area). Left: front plate with the catchment area ( $0.250 \times 0.250 = 0.0625$ m$^{2}$). Right: back plate and the inside. Cf. figure 3.13.

Without changing something else, the catchment area of TUE-I was reduced from 0.444 m$^{2}$ to 0.0625 m$^{2}$ in February 1999. This new version is called TUE-Ib (figure 4.5). The catchment area was reduced to investigate its influence on the reading of the driving rain gauge. Initially it was thought that a smaller catchment area would suffer less from evaporation of drops remaining on its surface, and/or that drops on a smaller collector would have shorter paths to the water flux gauge. The TUE-Ib/TUE-II correlation is depicted in figure 4.6c. Note that the correlation between TUE-Ib and TUE-II is 0.27 with a poor correlation coefficient ($r^2$) of 0.48. These values have been obtained for driving rain intensities up to 1.5 mm h$^{-1}$. When higher driving rain intensities are considered (see figure 4.6d), the reading of TUE-Ib is closer to the reading of TUE-II. A second degree polynomial fits well through the measurement points ($r^2=$ 0.87), although there is no theoretical argument for choosing such a polynomial. Of course, we expect that above a certain (high) level of driving rain onslaught, the readings of TUE-Ib and TUE-II will become the same, so the correlation should somehow change into a correlation $y=cx$ with $c=1$ (at driving rain intensities higher than shown in the figure). Figure 4.7 shows the reading of gauge TUE-Ib as a function of driving rain intensity measured by the TUE-II gauge for several wind speed ranges. It reveals that TUE-Ib yields only responses at relatively high rain intensities ( $R_{\text{f, TUE-II}}\gtrsim0.2$ mm h$^{-1}$) or wind speeds ($U_y\gtrsim6$ m s$^{-1}$). Contrary to the initial expectation, a smaller catchment area for the gauge without wiper did not yield better readings. An explanation may be that the drainage path length of the TUE-Ib gauge is too long (the drainage path from the collector to the water flux gauge is approximately 30 cm and did not change compared to the original TUE-I gauge).

Probably a model for the simulation of raindrop sticking, coagulating and running off on a driving rain collector as mentioned in [Blocken et al. 2001], will give a better understanding why the reading of the TUE-Ib gauge is worse than the TUE-I gauge. The same model can be used to study the differences between the TUE-I and the TUE-II gauges. For assumed raindrop spectra, it will give the relationship between the shape and size of the collector and the amount of collected rain water remaining on its surface and running off. The effect of the development of the raindrop spectrum during a rain event on the driving gauge readings with respect to the process of running-off can be investigated too.

Altogether one can conclude that the differences in reading between the TUE-I and TUE-II gauges are quite systematic. They can be explained by the evaporation of the collected raindrops remaining stuck on the collector of the TUE-I gauge. Raindrops are less likely to remain stuck on the collector of the TUE-II gauge --and subsequently evaporate-- because of its wiper. When the collectors were unmounted for inspection and cleaning, TUE-II was always significantly cleaner and its surface was smoother, whereas the surface of the collector of the TUE-I gauge was covered with a thin layer of dirt.

Figure 4.6: Correlations of 10-min driving rain intensities $R_{\text {f}}$ of gauge TUE-I with gauge TUE-II, when (a.) TUE-I and TUE-II were at positions P4 and P5 respectively, from December 1997 to June 1998, (b.) after swapping the positions from September 1998 to February 1999, and (c.) the catchment area of TUE-I was reduced to 0.063 m$^{2}$, from February 1999 to November 1999. Figure d is the same as c, but with extended axes.

Figure 4.7: Correlations of 10-min driving rain intensities $R_{\text {f}}$ of gauge TUE-Ib with gauge TUE-II, for selected reference wind velocity components perpendicular to the façade $U_y$ of 4-6 m s$^{-1}$ ( $\includegraphics[width=0.5em]{gen/m-kruisje.eps}$), 6-8 m s$^{-1}$ ( $\includegraphics[width=0.5em]{gen/m-rondje.eps}$), 8-10 m s$^{-1}$ ( $\includegraphics[height=0.5em]{gen/m-vierkant.eps}$) and 10-12 m s$^{-1}$ ( $\includegraphics[angle=90,width=0.6em]{gen/m-ruit.eps}$). Cf. figure 4.1c. Period: 26-2-1999 to 30-11-1999.