7. Conclusions

This chapter summarises the main conclusions from the previous chapters (indicated by $\bullet$) and gives suggestions for further research (indicated by  $\triangleright$).

Theory (chapter 2)

In chapter 2 we demonstrated that driving rain is a complex phenomenon of falling raindrops in the turbulent flow of wind around a building. Every individual drop has its individual drop trajectory. The lifetime of a drop is affected by drop interaction (collision and breakup) and the environment (wind and evaporation). However, in our ``theoretical model'' the onslaught of individual raindrops on a façade depends only on the wind field nearby the building. In our ``empirical model'' we define the onslaught in terms of driving rain intensities, i.e. amounts of driving rain water onto the building envelope per time interval. The empirical model is useful for analyses of our full-scale measurements. For both models we concentrated on the second step (from site to building façade) of the general two-step approach (figure 1.1).

Site and measurement set-up (chapter 3)

The objectives of the experiments were (a) the development and testing of driving rain gauges, and (b) the acquisition of driving rain data simultaneously with relevant weather data (in real circumstances, in full scale). The measured data should also be suited for the validation of CFD simulations of wind and driving rain in the same situation.

We chose the Main Building of the TUE mainly because of the relative simplicity of its geometry and the site topography. This simplicity relates to the following aspects. The Main Building is obviously higher than nearby buildings. Its height surpasses $20\times z_0$. Its wind field is hardly affected by other buildings and obstacles for south-west to west winds. Another important aspect is that a suitable location for site reference measurements of wind and rain was easily defined at the west side of the Main Building. Moreover, the large west façade is oriented towards the prevailing direction for wind and rain.

The instrumentation is described extensively in section 3.2. Recommendations, which do not relate to the driving rain gauges, are:

$\triangleright$

The output of the disdrometer used by us has two disadvantages for our purposes. Firstly, the disdrometer gives the raindrop number concentration spectrum, which is derived from the quantity which we are interested in, namely the raindrop mass flux spectrum. Secondly, the interval of the size classes increases with the diameter, which results in larger absolute errors of the real mass flux at larger diameters. The two disadvantages can be solved by outputting every detected raindrop with its diameter and velocity. It is recommended to adapt the software of the disdrometer in this way.

$\triangleright$

The two-dimensional video-disdrometer of [Schönhuber et al. 1994] is able to register details of raindrops. It could be used to verify actual raindrop spectra with those measured by the optical disdrometer of [Löffler-Mang and Joss 2000] which we used.

The video-disdrometer can also be mounted in a façade in order to validate simulated catch ratios $\eta (D)$ (eq. 2.26) directly from measurements of $\varphi _{\text {h}}(D)$ and $\varphi_{\text{f}}(D)$.

$\triangleright$

The measurement set-up at the TUE site may be continued and extended with instruments for the measurement of outdoor air temperature, relative humidity, solar irradiation etc., so that every important climatological parameter can be included to the benefit of heat-air-moisture studies.

Driving rain gauges (chapter 4)

The conclusions of the international full-scale driving rain gauge comparison test are listed in section 4.4 (and were partially presented in [Högberg et al. 1999]).

Adding some suggestions, we repeat the most important conclusions here:

$\bullet$

The experiments resulted in the formulation of design rules for driving rain gauges. These design rules relate to the catchment area, the prevention of drops from remaining stuck on the surface of the driving rain collector, the temporal resolution of the water flux gauge and the finish of the collector surface.

$\bullet$

The TUE-II gauge, which was developed at the TUE and is equipped with a rotating wiper (figure 3.14), accurately registers driving rain intensities. It has a good resolution for shorter time intervals (0.001 mm h$^{-1}$ for 10-min periods). Its wiper keeps the surface clean and forces impinged raindrops to coagulate and drip down (hence less evaporation). Moreover, it is not sensitive to wind.

$\bullet$

The TUE-I gauge, which is similar to TUE-II except that it is not equipped with a rotating wiper (figure 3.13), registers approximately half of the monthly driving rain amount measured by the TUE-II gauge (table 4.1). This is also valid for 10-min periods (figure 4.1c).

$\bullet$

The monthly driving rain amounts of the CTH, DTU and TUE-II gauges deviate within 30% from each other (table 4.1). On much smaller time bases, such as 10-min intervals, gauge responses can deviate significantly (figures 4.1-4.4). This applies especially for small time bases of the used tipping-bucket driving rain gauge (CTH): during a 10 min period it tips only once at a driving rain intensity of 0.18 mm h$^{-1}$.

$\triangleright$

Given the results of the CTH gauge for driving rain measurements, we suggest that for short-time intervals (like 5- or 10-min intervals) one should apply a continuous measuring principle instead of the tipping-bucket principle.

Indeed, given the necessary yet cumbersome correction of the tipping-bucket gauge data (sections 3.4.3 and 5.1), the suggestion also applies to horizontal rain measurements.

$\bullet$

The effect of size and shape of the catchment area cannot clearly be deduced from the experiments. A comparison of the CTH gauge (0.032 m$^{2}$) and the TUE-I gauge (0.5 m$^{2}$) does not give a straightforward conclusion, because of the differences in measuring principle.

$\triangleright$

It is recommended for further research to compare simultaneously the readings of driving rain gauges with different catchment areas but with the same continuous measuring principle. One could, for instance, add the CTH gauge and the TUE-II gauges as references. Further measurements with the disdrometer (with which we measured during a too short period) may be useful for explaining the differences between the readings of driving rain gauges.

Another line of research are calculations with a model for raindrops sticking, coagulating and running off on a driving rain collector, as mentioned in [Blocken et al. 2001].

$\bullet$

Lower driving rain intensities (measured with the TUE-II gauge) are overestimated by the DTU gauge (figures 4.1b-4.4b). The scatter in the DTU/TUE-II correlations is larger than in the TUE-I/TUE-II correlations. This is probably due to the noise caused by the wind acting on the freely suspended collector. The applied signal-processing method (see section 3.4.6) was kept simple and ( $\triangleright$) can perhaps be improved.

$\triangleright$

The possibility of splashing and the effects of protruding rims and other projections were not investigated. With our full-scale measurement set-up, it is possible to mount at least three gauges with different protruding rims at P4/5/7. An investigation on splashing effects is less easy; perhaps one may use video cameras.

$\triangleright$

As it occurred that the reservoirs of the TUE-II and TUE-I gauges overflowed during heavy driving rain, it is recommended to investigate other types of continuous water flux measurements. We have done some laboratory experiments with a drop-counting device [Bijsterbosch 2000]. Combined with a collector of $A_{\text{catch}}=0.5$ m$^{2}$, it has a resolution of 0.0006 mm h$^{-1}$ for 10-min periods and an accuracy of approximately 5% for driving rain intensities ranging at least from 0.12 to 1.20 mm h$^{-1}$ (one drop equals 0.05 g $\pm$ 5%). We recommend to test it in full scale first.

Full-scale measurements (chapter 5)

The full-scale experiments at the TUE site resulted in a unique series of continuous measurements during 24 months of driving rain on the west façade of the Main Building and wind and rain at the well-defined site reference location. Raindrop spectra with a disdrometer were measured during 3 months too. The measurements are detailed (data at 5-minute intervals were provided) and are available for future research (at the website http://sts.bwk.tue.nl/drivingrain/).

Section 5.4 gives a summary of the measurement results. The present study introduces the following new items to our knowledge on driving rain:

-

a time series of detailed measurements of driving rain on a façade on a particular building in an urban surrounding with well-defined site reference measurements,

-

the readings of different types of driving rain gauges were compared with each other in a full-scale comparison test (see chapter 4),

-

raindrop spectra were measured with a disdrometer of [Löffler-Mang and Joss 2000]. Unfortunately, the number of measured spectra during driving rain was rather small.

$\triangleright$

Further raindrop spectrum measurements are recommanded, because we could not conclude our investigation on the influence of raindrop spectrum on driving rain quantities,

-

driving rain intensities vary much at a façade position and on small time-scales (5 min), even for narrow ranges of reference wind speed, wind direction and horizontal rain intensity (see e.g. figure 5.14 and section 5.2.5). Moreover, the correlation between the two measurement positions P4/5 and P6 is very complex and depends to a great degree on wind direction (section 5.2.6),

-

the traditional empirical model (model 1, eq. 5.3), based on [Lacy 1965] and implemented in the British Standard 8401 [BSI 1992], is improved by taking the wind direction and the position on the façade more explicitly into account (model 2, eq. 5.4). Model 1 yields less realistic estimates of (especially maxima of) 5-min driving rain intensities than model 2. However, the models overestimate the cumulative driving rain amounts after 24 months by up to 35-45%.

$\triangleright$

In order to validate models 1 and 2 further, we recommend measurements of the distribution of driving rain intensity over the façade on more than two façade positions. Probably a better line of research are driving rain measurements on other buildings; below we will suggest some conditions for such experiments.

We investigated a large part of the two-step approach (figure 1.1) in chapter 5, but we did not show every possible analysis of the measurement data, such as:

$\triangleright$

an investigation on the relation between hourly values and 5-min (10-min) values, as generally only hourly values of measured wind and rain parameters are available at a weather station,

$\triangleright$

an analysis of the temporal development of horizontal and driving rain intensities during rain spells.

CFD simulations (chapter 6)

Section 6.5 summarises our CFD simulations of wind and driving rain at the TUE site. Figures 6.16 to 6.19 depict the results of simulated driving rain intensity distributions over the west façade of the Main Building. The main conclusions are:

$\bullet$

In spite of the known limitations of the applied $K$-$\epsilon$ model (e.g. [Murakami et al. 1992]), the practically limited number of grid cells and the use of a structured grid with inevitably non-ideally shaped grid cells, the simulated wind speed at the façade is within the standard deviation of the full-scale wind speed measurements at position P4. Simulated mean pressure coefficients over the west façade compare quite well with wind tunnel and full-scale measurements of [Geurts 1997].

$\bullet$

Catch ratios $\eta (D)$ calculated with turbulent drop dispersion have higher values than without turbulent drop dispersion. Due to the (extra) turbulent velocity component, drops are more easily driven towards the façade when they come close to it. The smaller the drops are, the easier they are driven onto the façade. Moreover, the longer a drop flies closely to a façade, the higher is the probability that it is driven onto the façade. Figure 6.3 illustrates drop trajectories resulting from the two drop trajectory models; figures 6.10 to 6.14 show simulated catch ratios $\eta (D)$.

$\bullet$

A comparison between the measurements and the simulations (based on the raindrop spectrum parameterisation of [Wessels 1972]) reveals that the results calculated with turbulent drop dispersion are likely to overestimate the measured driving rain intensities. The results calculated without turbulent drop dispersion are likely to underestimate the measurements.

$\triangleright$

To validate our CFD driving rain simulations more precisely, it is needed to measure wind speed and driving rain intensities at more positions close to and on the façade.

$\bullet$

As concluded before, the measured driving rain intensities for a given façade section, reference wind speed and wind direction show large variations. The driving rain intensities simulated without turbulent drop dispersion and based on the raindrop spectrum parameterisation of [Wessels 1972], yield an almost linear relation with horizontal rain intensity: $R_{\text{f}} \approx k(A) R_{\text{h}}$, where $k$ depends only on the shape (in our case, parameter $A$) of the raindrop spectrum. However, when measured raindrop spectra are used in the simulations, these simulated driving rain intensities show scatter too. Because the number of measured raindrop spectra during driving rain is rather small, we can not yet decisively conclude whether the relation $R_{\text{f}} \approx k(A) R_{\text{h}}$ is actually valid.

$\triangleright$

It is therefore recommended to elaborate the problem of the previous item with more measurements of raindrop spectra and driving rain.

$\bullet$

Altogether, the following aspects for reliable CFD results of wind and driving rain are important:

-

experience with the CFD program, useful references, like [Bottema 1993a], and validations of simulated data with measured data will make one conscious about the possibilities and limitations of the applied models. Standards on CFD simulations of wind nearby buildings are still in development though,

-

the building geometry and the surrounding topography should be simple enough for CFD modelling. An important condition for the surrounding topography is the presence of a site reference location and an unobstructed fetch for a certain range of wind directions. These will determine the values of the displacement height $d$ and the roughness length $z_0$. The larger region around the site should also be considered for these parameters. Preferably, the considered building has distinct dimensions compared to its surroundings (e.g. $\mathcal{H} \geq 20\times z_0$),

-

the grid and the size of the computational domain depend on the expected wind flow and raindrop trajectories. For instance, the stopping distance of the smallest considered raindrop determines the grid size nearby the building envelope,

-

the modelling of turbulent drop dispersion is still an issue for research,

-

reliable results depend on realistic values of climatological parameters, namely wind speeds, wind directions, horizontal rain intensities and raindrop spectra. For a climate like that of the Netherlands, the raindrop spectrum parameterisations of [Wessels 1972] are very useful.

Further measurements and simulations

In most of the previous suggestions we addressed further studies for our own site. More general suggestions are formulated in the following items:

$\triangleright$

From the start of the present study in 1996, only few comparisons between full-scale measurements and simulations of driving rain on a particular building in a particular topography have been reported ([van Mook 1999a] and [Blocken et al. 2001]). However, the number of studies on CFD simulations without the validation with full-scale measurements is larger (section 1.3).

Important for the advance of the knowledge on driving rain are several further studies in which CFD simulations are validated with full-scale measurements. Of course, one should carefully select another situation (i.e. a particular building and its environment), given the limitations of CFD simulations. One should also pay attention to the instrumentation and the measurement method. In order to make a comparison between our situation and other situations feasible, we recommend to carry out measurements and simulations in situations which are comparable to our situation. This means:

-

The geometry of buildings is considered comparable if it has a simple shape, i.e. without many or complex protrusions or recesses. Given the Main Building, it would be interesting to investigate e.g. two-dimensional situations of two blocks of flats or two rows of terraced houses, and three-dimensional situations of a tower, a building with a large indentation in the middle of the roof or a building with a large canopy.

With regard to the surrounding topography we refer to our above-mentioned considerations for reliable CFD results.

-

The climate of the site is nearly the same as that at the TUE site, as the driving rain onslaught in moderate maritime climates are very different from the onslaught in e.g. tropical thunderstorms (cf. [Choi 1999a]).

-

A full-scale experiment is designed with the approach described in chapters 3 and 4 and with the recommendations resulting from these chapters borne in mind. We recommend especially to obtain measured data at small time bases (of the order of 1-5 min), to measure continuously during at least 6 months, not to use tipping-bucket gauges, to measure raindrop spectra, and --of course-- to use a driving rain gauge as accurate as TUE-II.

The comparison between our situation and other situations may relate to: the validation of model 1 (eq. 5.3) and model 2 (eq. 5.4), the validation of the turbulent raindrop dispersion modelling in CFD and the influence of raindrop spectra on driving rain.

$\triangleright$

Within the context of the previous item, collaboration between driving rain research projects at different locations, like by [Blocken et al. 2001], should be stimulated.

$\triangleright$

Pictures of façades which were just exposed to driving rain, may give a nice overview on the wetting of a façade, and help in acquiring qualitative data for validation of CFD simulations. Such pictures may also serve to understand the influence of façade details (e.g. projections) on driving rain onslaught and run-off. Moreover, small projections and canopies are difficult to implement in a CFD model. It is recommended to take photographs (before, during and after rain), such as done by [El-Shimi et al. 1980] and [Snape and Atkinson 1999].

$\triangleright$

The TUE-II gauges were suitable for our driving rain measurements, but they are not easily installed due to the size of their collectors and the space needed for the reservoirs and balances. A `portable' gauge is needed when one would like to do more in situ driving rain measurements. Perhaps a driving rain gauge equipped with a drop-counting device (see above) could serve for this.