1.3 Literature survey

Driving rain has been the subject of research for many years. Main surveys can be found in [Lacy 1965], [Frank 1973], [Prior 1985] and [Flori 1988]. We must note that many references in these surveys are not easily available and sometimes not available at all.

Towards a standard The oldest instrument (according to [Middleton 1969]) which determined the direction from which rain is coming, was made in 1816. It was a so-called ``vectopluviometer'' and had a horizontal opening and a vertical opening which faced into the wind by a vane. Note that it measured the free driving rain as we define it in this thesis (figure 1.1), because the vectopluviometer is not mounted on a façade. Holmgren was probably the first who performed measurements of driving rain on a façade of a building in Trondheim in 1937 [Lacy 1965]. According to [Lacy 1965] Holmgren's gauge looked very much like the left gauge of figure 1.2a. Driving rain water was collected by a shallow square tray fixed to the wall and the collected water was guided into a bottle. Similar gauges are still used today, and we will call them ``traditional''.

During the second World War Chr. Nell measured driving rain on houses in a street in Voorschoten (NL). [Basart 1946] reported the findings of Nell who compared the driving rain data measured by two different methods: by a traditional gauge and by regular weighting of a plate made of very absorbent bricks. Differences were attributed to measurement anomalies and to the variation in the raindrop size distributions, changing from one storm to another. The report of [Basart 1946] is the only one with results of a comparison test of driving rain gauges which we found.

Figure 1.2: Driving rain gauges from (a, left) BRE (UK) [Lacy 1965], (a, right) TNO (NL), (b) Künzel [Frank 1973], (c) [Flori 1990] and (d) [Osmond 1995].

a. 
b. 
 
c.   d.

Hoppestad of the Norwegian Building Research Institute [Lacy 1965] assumed that the free driving rain intensity ($R_{\text{v}}$, i.e. rain intensity through the vertical) is proportional to the product of rain intensity through the horizontal ($R_{\text {h}}$) and the local wind speed ($U$):

$$R_{\text{v}} = \alpha R_{\text{h}} U,$$ (1.1)

where $\alpha$ is assumed to be the reciprocal of the `average' falling velocity of raindrops. An assumed average falling velocity of 4.5 m s$^{-1}$ yields $\alpha=0.2$ s m$^{-1}$.

The idea behind equation 1.1 is that the motion of a rain drop is affected by the horizontal wind speed and the vertical falling velocity, and that the ratio of rain intensities approximates the ratio between the mentioned two velocities. The reader should bear in mind that $R_{\text{v}}$ is not the driving rain intensity on a building envelope, but some kind of intermediate reference, as indicated in section 1.1.

In 1955 Hoppestad was the first to publish maps of free driving rain [Flori 1988]. The maps were based on calculated annual and monthly free driving rain amounts of all the meteorological stations in Norway. The (average) coefficient $\alpha$ was obtained by measurements of $R_{\text{v}}$ at four stations. While the maps gave a first quantitive overview of vertical rain (and indirectly of driving rain), Hoppestad underlined that the calculations did not take the variability of $\alpha$, nor corrections for the local environment (topography, terrain roughness, etc.) into account.

Lacy and Shellard introduced the ``driving rain index'' in 1962 [Lacy 1965]. The index was defined as the product of the annual mean wind speed (in m s$^{-1}$) and the annual rainfall (in m, i.e. precipitation height on the horizontal). Maps with driving rain indices (d.r.i.) were produced for the United Kingdom. A meeting of the working commission on rain penetration in Madrid 1966 at the Commission Internationale du Bâtiment promoted the elaboration of driving rain index maps for different countries to investigate whether this approach to estimate free driving rain amounts could be widely used [Frank 1973]. The idea was to classify areas of potentially low (index $<3$), moderate (index 3-7) and high (index $>7$) free driving rain onsets. By the end of the 1960s, d.r.i. data had been collected for the United Kingdom, Norway, Canada, Denmark, Poland, Rumania, Spain, West Germany and East Germany.

Künzel evaluated the maps of these countries in 1968 [Frank 1973]. He observed that the driving rain index maps gave reasonable indications for driving rain amounts in the UK and parts of Scandinavia. However, the mentioned classification did not hold for other parts of Europe: in many areas, low driving rain indices corresponded with high driving rain amounts in reality. So it seemed that the meaning of the index depended on topography and climate, and was not universal as intended by Lacy and Shellard. A reason for this discrepancy was that on average in some regions the wind speed during rain differs from the wind speed during dry periods. The annual wind speed averaged over all hours (as used for the d.r.i. by Lacy and Shellard) did thus not everywhere represent well the wind speed during rainfall.

[Prior 1985] reported that in the beginning of the 1970s Caton of the Meteorological Office noted that also in the UK the ratio of the mean wind speed during rain to that of all hours varied so much, that Lacy's index could significantly under- or overestimate free driving rain amounts. Caton prepared d.r.i. maps based on hourly products of rainfall and wind speed, taking into account wind direction. Caton also investigated a ``driving rain spell index'', which represented driving rain amounts associated with rain spells having a specified average frequency of occurrence. A rain spell has a variable duration and can consist of a series of rainfalls interspersed with periods up to 96 h without rain. For the development of a standard on the assessment of driving rain exposure of buildings by the British Standards Institution, Prior was asked to continue Caton's investigations. The results were reported in [Prior 1985] and were eventually integrated in BS 8104 [BSI 1992].

So, the British standard is based on data from 1959 to 1991, of 52 weather stations throughout the UK. Hourly wind speed data were corrected and translated to corresponding values for wind speed at 10 m height on an open level country. Hourly rainfall data was not available for every weather station; a procedure for `filling in' missing data points was applied [Prior 1985]. Two indices were calculated for every location and every of the 12 wind directions: (a) an average annual driving rain index and (b) a driving rain spell index associated with a frequency of once in three years. The first index was thought relevant for the weathering of building envelopes and the latter for assessing the risk of rain penetration through masonry walls. The standard gives a method for the assessment of the annual driving rain amount --or the worst likely spell amount in any three year period respectively-- on a wall of an particular orientation, calculated from directional indices (plotted on maps) corrected for terrain roughness, topography, obstruction by nearby buildings, building type and position on the building envelope. The British standard is still the only standard on driving rain estimations; in the Comité Européen de Normalisation a European version of this standard is in preparation.

It is not clear to us on which studies the correction factors in BS 8104 were based. Perhaps they are, among others, based on [Brown 1988] who measured ``catch ratios'' on different buildings and at different façade positions in Dorset (UK). [Brown 1988] defined the catch ratio as the quotient of the driving rain amount on a position on a façade and the free driving rain amount. Other sources for measurements of catch ratios could have been [Lacy 1965] ([Lacy 1965], [Lacy 1977]) and [Frank 1973].

In the 1990s, full-scale driving rain measurements were carried out by [Flori 1990], [Henriques 1992], [Hens and Mohamed 1994], Fraunhofer-Institut für Bauphysik (see e.g. [Künzel 1994]), [Sankaran and Paterson 1995b] (free driving rain only), [Osmond 1995] ([Osmond 1996], [Osmond 1996]), [Kerr et al. 1997], [Straube 1998], [Kragh 1998], [Högberg 1998], [Högberg 1999], [Blocken and Carmeliet 2000b], and [Choi 2001] (only free driving rain and raindrop spectrum). Driving rain measurements presented in [Högberg et al. 1999], [van Mook 1999a] and [van Mook 1999b] relate to work for this thesis and are presented in more detail in chapters 4, 5 and 6.

An original approach is developed by [Snape and Atkinson 1999], who investigated driving rain intensity patterns by comparing photos of stains and other surface depositions on a façade.

Computer simulations Perhaps the first computer simulations of driving rain were done by [Sandberg 1974] and [Rodgers et al. 1974]. In both articles only raindrop trajectories were calculated by computer; the wind speed fields around a (two-dimensional) building were obtained from wind tunnel experiments. At the end of the 1980s and beginning of the 1990s, new possibilities of computational fluid dynamics and new possibilities offered by improved hardware were exploited. [Choi 1993] presented simulations in which firstly the wind flow around a high-rise building was calculated by a $K$-$\epsilon$ model, and secondly, for a given mean wind field, trajectories of drops were calculated. Considered wind speeds were 5, 10 and 20 m s$^{-1}$, and also the considered rain intensities were extreme (at least compared to the north-west European climate): 10, 30 and 50 mm h$^{-1}$. In [Choi 1994b] it becomes clear that he was interested in extreme driving rain onsets for a building in Sydney. Other articles of the same author are: [Choi 1994c] (parametric study), [Choi 1994a] ([Choi 1994a], [Choi 1999b]) (driving rain index), [Choi 1995] (gust effects on driving rain) and [Choi 1999a] (tropical thunderstorms). The method of [Choi 1993] is used by almost every other researcher.

Two-dimensional simulations of driving rain on a building of moderate height (5.5 and 11 m) in a moderate climate (reference wind speed of 10 m s$^{-1}$ and rain intensity of 1.3 mm h$^{-1}$) were presented by [Bookelmann and Wisse 1992] and [Wisse 1994].

Contrary to the method of Choi, [Sankaran and Paterson 1995a] took dispersion of raindrops due to wind turbulence into account for a simulation of driving rain on a very tall building of 183 m height (and $45\times 31$ m$^{2}$ cross section). The wind flow with its standard deviation was calculated using a standard $K$-$\epsilon$ model. For a rainstorm with 30 mm h$^{-1}$ and 17 m s$^{-1}$, the driving rain intensity on the front face simulated by taking raindrop dispersion into account, was approximately two times the driving rain intensity simulated without raindrop dispersion. [Hangan and Surry 1998] also included turbulent raindrop dispersion in their driving rain simulations; they also compared their results with wind tunnel experiments. [Lakehal et al. 1995] studied different turbulence models (based on the $K$-$\epsilon$ model) for the calculation of turbulent dispersion of raindrops and applied them to calculate driving rain intensities on façades in a two-dimensional street-canyon. They could not conclude which was the most accurate model because the full-scale measurements were insufficient. This article is also interesting for a historical review on driving rain simulations.

Other articles on driving rain simulations are [Karagiozis and Hadjisophocleous 1996], [Karagiozis et al. 1997], [Blocken et al. 1999], [Blocken and Carmeliet 2000a] and [Etyemezian et al. 2000].

Windtunnel experiments [Surry et al. 1994] and the associated report of [Inculet and Surry 1995] described wind tunnel experiments in which driving rain on a reduced-scale building was studied qualitatively. Driving rain distributions were measured by water-sensitive paper. Full-scale driving rain tests in (boundary-layer) wind tunnels have not yet been published, as far as we know. The large wind tunnels (with cross sections of 10 m and more) at the CSTB in Nantes (FR) [Gandemer and Barnaud 1995] may be useful for full-scale driving rain experiments.

Summary Overviewing the literature on driving rain, we conclude that:

• there is only one official standard on the estimation of driving rain quantities, namely BS 8104:1992,
• only few detailed measurements, for example on hourly basis or less, are available,
• references on comparisons of driving rain gauges have not been found,
• CFD simulations of driving rain are all based on the standard $K$-$\epsilon$ model. Only in a few cases turbulent dispersion of raindrops is taken into account,
• simulations on driving rain are not compared to full-scale measurements of driving rain on the same building. Only recently, such comparisons are published by [van Mook 1999a] and [Blocken et al. 2001],
• raindrop size distributions used in simulations are based on quite old measurements (e.g. [Marshall and Palmer 1948] and [Best 1950]) in which the distributions were idealised by simple logarithmic functions, although measurements show much climatic and temporal variations (e.g. [Wessels 1972], [Joss and Gori 1978], [Uijlenhoet 1999]).