Building Serv. Eng. Res. Technol. 23,3 (2002) pp.

127–141

The performance characteristics of multi-outlet

siphonic roof drainage systems
GB Wrighta MEng PhD, JA SwafŽ elda BSc PhD MCIBSE and S Arthurb BEng PhD
a
Drainage and Water Supply Research Group, b Water, Environment and Fire Research Group, School
of the Built Environment, Heriot-Watt University, Edinburgh, UK

Siphonic roof drainage systems have been in existence for approximately 30

years, and are becoming an increasingly common element of urban drainage
infrastructure. In that time, the construction sector in most developed countries
have been gradually persuaded of the beneŽ ts that these systems offer when
compared to conventional roof drainage technologies. The work reported herein
details an ongoing UK government-funded research programme to investigate the
performance characteristics of multi-outlet siphonic roof drainage systems. The
experimental aims, apparatus and procedures are described, and results are illus-
trated. In addition, ‘real’ data obtained from three installed siphonic roof drainage
systems are discussed. Conclusions are drawn regarding the performance charac-
teristics of multi-outlet systems, and plans for future work are outlined.

1 Introduction designed to ensure sufŽ cient capacity and self-

cleansing  ow velocities.
1.1 Conventional roof drainage systems
Conventional roof drainage systems generally 1.2 Siphonic roof drainage systems
consist of a network of collection gutters con- In contrast to conventional systems, the
nected, via open outlets, to vertical downpipes. siphonic approach to roof drainage aims to
The system components are sized to ensure restrict the ingress of air into the system, and
annular  ow through the downpipes with a con- hence induce the full-bore  ow conditions
tinuous central air path, and system pressures necessary for siphonic action. This is achieved
therefore remain close to atmospheric.1 Conse- by utilizing specially designed gutter outlets,
quently, the driving head for  ow within conven- such as those shown in Figure 1, in conjunction
tional roof drainage systems is limited to the gut- with smaller diameter pipework. Once all of the
ter  ow depths, which results in relatively low air has been purged from the system, siphonic
 ow velocities within the system. This inef- action occurs and the system is said to have
Ž cient use of pipework necessitates many, rela- primed. Although siphonic gutter outlets nor-
tively large diameter, downpipes (typically 150 mally incorporate vortex-reducing elements, the
mm) each of which must be connected into a turbulent gutter  ow conditions will invariably
suitable underground drainage network. Further- result in small quantities of entrained air entering
more, the dimensions/gradients of the gutters the system (up to 10%2), even when the outlets
and the underground drainage network must be are fully submerged.
At its speciŽ c design condition, the driving
head within a siphonic roof drainage system can
be equal to the gutter  ow depths plus the full
Address for correspondence: GB Wright, Drainage and Water vertical height between the outlets and the point
Supply Research Group, School of the Built Environment,
Heriot-Watt University, Edinburgh EH14 4AS, UK. E-mail: of discharge. This yields signiŽ cantly higher
g.b.wright@hw.ac.uk  ow velocities than is possible in conventional
Ó The Chartered Institution of Building Services Engineers 2002 10.1191/0143624402bt041oa
128 Multi-outlet siphonic roof drainage systems

Figure 1 Typical siphonic gutter outlets (reproduced with permission)

systems, which means that attaining self-cleans- and enters a ‘secondary’ system3). Consequently,
ing velocities is rarely a problem and more than a siphonic system will rarely, if ever, operate at
one outlet can be connected to a single down- its design condition. If a siphonic system is
pipe. As the  ow is full-bore and depressurized, exposed to a rainfall event which exceeds the
there is considerably more  exibility in pipe design criteria,  ooding may occur and the sys-
routing, allowing most of the horizontal collec- tem may fail due to gutter overtopping. The
tion pipework to be located just below roof level, more likely scenario is that a system will be
and reducing the extent of costly underground exposed to a rainfall event below the design cri-
drainage networks. In addition to these oper- teria. When this occurs, the  ow conditions will
ational beneŽ ts, the small diameter pipework differ from those in a fully primed system, their
employed in siphonic roof drainage systems can exact nature depending on the speciŽ c character-
have less of an architectural impact than conven- istics of the rainfall event. Similar conditions can
tional systems, and may even be incorporated occur if the  ow distribution between gutter out-
within the building itself. lets is not as per design, possibly as a result of
A siphonic roof drainage system will only poorly installed roof surfaces/gutters or wind-
operate efŽ ciently at its design condition, e.g., a driven rainfall. Another disadvantage of siphonic
1 in 30 year rainfall event; that is, only one rain- roof drainage systems is that the restrictive out-
fall event matches any particular system. (The lets and small diameter pipework are relatively
situation is slightly different where ‘secondary’
easily blocked by detritus in the  ow, e.g. leaves.
systems have been installed. Typically, the ‘pri-
If a regular maintenance programme is not
mary’ system will drain rainfall events up to a
adhered to, this can lead to operational problems
predetermined notational return period. Beyond
and system failure.4
that level, runoff passes over weirs in the gutter
 GB Wright et al. 129

2 Current state of siphonic roof majority of operational problems tend to occur,

drainage technology e.g., noise, vibration and  ooding.
In addition to these considerations, a number
Since their development in Scandinavia in the of more serious problems have occurred with
late 1960s, siphonic roof drainage systems have speciŽ c siphonic roof drainage installations.
gradually become accepted by the construction These include a UK manufacturing plant, where
sector in most developed countries. Their high the site layout necessitated the use of a ‘U-bend’
capacities and low architectural impact have arrangement to connect the siphonic roof drain-
made them particularly popular for large, pres- age system under an existing road, and into the
tigious developments such as airports and major surface water sewer network. Following instal-
sporting stadia, e.g., Chep Lap Kok Airport lation, the gutters were observed regularly to
(Hong Kong) and The Olympic Stadium overtop during insubstantial rainfall events.
(Sydney, Australia). Despite this, siphonic roof Inspection of the system indicated that this was
drainage systems are virtually unheard of in the due to the large air pocket that formed in the
USA, with the authors knowing of only one such upward leg of the ‘U-bend’ arrangement, i.e., the
installation. 5 The lack of acceptance of such gutters overtopped before the system pressures
technology in the USA may be due to a lack had built up to the levels necessary to purge the
of understanding of the underlying principles of system of this air. It is considered unlikely that
siphonic systems and problems involved in current design methods would be capable of pre-
changing the necessary regulatory codes. Inter- dicting this type of system failure. Another inter-
estingly, there is no speciŽ c European standard esting example of the failure of a siphonic roof
for siphonic roof drainage systems. drainage system occurred at a storage depot in
Current design practice assumes that, for the southern England. In this case, the siphonic sys-
speciŽ ed design criteria, a siphonic system Ž lls tem was subjected to a rainfall event which,
and primes rapidly with 100% water. This although substantial, was less than the total sys-
assumption allows siphonic roof drainage sys- tem capacity (design capacity plus emergency
tems to be designed utilizing steady state provision). The resulting system pressures
hydraulic theory. The steady  ow energy equ- dropped below approximately 2 8 mH20, caus-
ation is normally employed,2,6 with the elevation ing the pipes to implode. This resulted in a
difference between the outlets and the point of reduction in system capacity, which led to over-
discharge being equated to the head losses in the topping of the gutters,  ooding of the facility
system. Although this design approach neglects and an insured loss of several million pounds.
the small quantities of entrained air that always Failures such as this illustrate the importance of
enter a siphonic roof drainage system, it has been considering system pressures, as well as total
reported to yield operational characteristics simi- system capacity, during the design process.
lar to those observed in laboratory test rigs at It should be noted that, although the siphonic
the fully primed state.2,7 However, steady state roof drainage industry and its clients are under-
design methods are not applicable when a standably reticent in publicizing problems, it is
siphonic system is exposed to a rainfall event considered that the number of system failures is
below the design criteria, when the  ow may only a tiny percentage of the large number of
contain substantial quantities of air, or an event systems installed throughout the world. Further-
with time-varying rainfall intensity. As such more, there is no evidence to suggest that
events are the norm, it is clear that current design siphonic systems are more prone to failure than
methods may not be suitable for determining the conventional systems. Blockages remain the
day-to-day performance characteristics of most common cause of operational problems and
siphonic roof drainage systems. This is a major system failures, and these can be avoided with
disadvantage, as it is during these events that the a regular maintenance programme.
130 Multi-outlet siphonic roof drainage systems

3 Previous relevant research 5 Laboratory investigation

Although siphonic roof drainage systems have 5.1 Overview

been in existence for approximately 30 years, it Experimental work was undertaken using the
has only been since the mid-1990s that substan- laboratory test rig detailed in Figure 2. To ensure
tial research has been reported.7 In terms of the realistic  ow conditions, each gutter was fed via
priming of single outlet siphonic roof drainage a rear supply trough and a simulated sloping
systems, previous laboratory-based research at roof. Pressure transducers were installed in the
Heriot-Watt University, Edinburgh, has ident- base of the gutters to measure  ow depths, and
iŽ ed a number of distinct phases, including the in the crown of the connected horizontal pipe-
formation of full-bore  ow conditions and the work to measure system pressures. In addition,
movement of trapped air pockets.7 The results of magnetic induction  owmeters were used to
this work have been used as the basis of a measure the gutter in ow rates. The transducers
numerical model capable of simulating the prim- and  owmeters were connected to a PC-based
ing of single-outlet siphonic systems.8 data acquisition system, capable of sampling
Further laboratory experimental work has con- data at frequencies of up to 30 kHz. All the pipe-
Ž rmed that, at rainfall intensities less than 40% work was transparent, allowing direct obser-
of the fully primed system capacity, single-outlet vations and high-speed video capture to assist
siphonic systems act in a similar manner to con- identiŽ cation of relevant  ow conditions.
ventional roof drainage systems.7 This work also Using the equipment detailed above, labora-
conŽ rmed the unsteady nature of the  ow con- tory experiments were undertaken to determine
ditions within siphonic systems at rainfall inten- the  ow conditions arising as a result of the fol-
sities above 40% of the fully primed system lowing realistic scenarios:
capacity. Such conditions were shown to be · design criteria rainfall events (fully primed
characterized by cyclical variations in gutter system) – constant gutter in ows;
water levels and system pressures, and were · rainfall events below the design criteria;
observed to result in large quantities of air being · design criteria rainfall events (fully primed
drawn into the system, leading to noise gener- system) – varying gutter in ows;
ation and structural vibration. · rainfall events above the design criteria;
· total blockage of one of the outlets.
4 Description of research programme In addition to the above, experimental work was
also undertaken to determine the effect, upon
The main aim of the research detailed in this system performance, of different types of system
paper is to extend the existing numerical model terminations. This was considered an essential
to enable the simulation of multi-outlet siphonic element of the investigation, as it is this section
roof drainage systems. In this context, the term which provides the interface between the
multi-outlet siphonic roof drainage system refers siphonic roof drainage system and the con-
to a system where more than one gutter outlet necting drainage network.
is connected to the same downpipe. In order to With reference to Figure 2 and the experi-
achieve this aim, it was Ž rst necessary to gain a mental data detailed herein, branch 1 refers to
better understanding of the conditions occurring the pipework connecting gutter 1 to the branch
within such systems, with particular reference to junction, branch 2 refers to the pipework con-
priming and the effect of different gutter in ow necting gutter 2 to the branch junction and com-
combinations. This was accomplished through mon pipe refers to the pipework downstream of
laboratory experimental work and Ž eld obser- the branch junction. It should also be noted that,
vations. unless otherwise stated, the gutter in ow rates
 GB Wright et al. 131

Figure 2 Schematic view of the siphonic roof drainage test rig. The capacity of the system is the same as 75 mm/h
falling on a 665 m2 roof – all drained via a 69-mm diameter pipe.

were constant throughout the testing periods, i.e., and the system in ows were relatively low,
the simulated rainfall events were assumed leading to free surface, subcritical  ow
‘instantaneously’ to reach a constant intensity. within the horizontal pipework and annular
 ow within the vertical pipework.
5.2 Design criteria rainfall event (fully 2) Formation and movement of hydraulic
primed system) – constant gutter in ows jumps: As the gutter water levels increased,
Priming of the laboratory siphonic test rig was so the system in ows increased, leading to
observed to occur when the in ow to gutter 1 supercritical  ow at the upstream end of the
was set to 5.85 l/s and the in ow to gutter 2 was branches and the formation of hydraulic
set to 7.78 l/s. As the two gutters were located jumps immediately upstream of the branch
at the same elevation above the point of dis- junction (refer to Figure 3a). As the system
charge, the difference in in ows required for in ows increased further, the hydraulic jump
siphonic conditions was due solely to the differ- in branch 1 moved upstream and its height
ent branch conŽ gurations. This is highlighted by increased. Similar observations were made
inspection of Figure 2, which indicates that the with respect to the  ow conditions in branch
head losses associated with the branch 2 con- 2, although the upstream movement of the
Ž guration would be signiŽ cantly less than those hydraulic jump was less marked.
associated with the branch 1 conŽ guration. The 3) Formation and propagation of full-bore  ow:
priming procedure of the siphonic test rig was Eventually the downstream depth of the
generally observed to occur as follows: hydraulic jump in branch 1 became equal to
1) Initial gutter in ow: At the start of the simu- that of the pipe diameter, and full-bore  ow
lated rainfall event, the gutter water levels developed (refer to Figure 3b). Once full-
132 Multi-outlet siphonic roof drainage systems

Figure 3 Priming process of the siphonic roof drainage test rig

bore  ow conditions formed, they were seen 2 was signiŽ cantly less than that trapped in
to propagate downstream into the common branch 1.
pipe and, to a lesser extent, further upstream 5) Partial repressurization of  ow: As the sys-
into branch 1. Similar observations were tem in ows continued to increase, the air-
made with respect to the  ow conditions in pocket trapped in branch 1 moved down-
branch 2, although the propagation of full- stream at the local velocity of the  ow. When
bore  ow was less marked. this air pocket passed into the vertical pipe-
4) Depressurization of  ow: When full-bore work it caused a partial repressurization of
 ow conditions reached the vertical section the entire system. The smaller airpocket in
of the common pipe, the mass of water col- branch 2 also moved downstream, although
lecting in the vertical pipework caused it appeared to become mixed with the water
depressurization of the system, which at the branch junction, forming a ‘bubbly
resulted in an increase in the system in ows.  ow’ that did not have such a signiŽ cant
This led to the development of full-bore  ow effect on system pressures.
conditions at the upstream ends of both 6) Fully primed system: Once all of the initial
branches. In turn, this trapped volumes of air air pockets had left the downstream end of
between the hydraulic jumps and the the vertical downpipe, the pressures
upstream end of the branches (refer to Fig- decreased and remained relatively constant.
ure 3c). The volume of air trapped in branch The system was then fully primed, although
 GB Wright et al. 133

it was observed that small quantities of would occur at the measured gutter in ow rates
entrained air continued to enter with the if the internal roughness of the pipework was
water in ows. 0.028 mm. Although such a roughness value is
considered to be reasonable for the type of pipe-
The gutter depths and system pressures recorded
during the priming of the siphonic test rig are work employed in the laboratory test rig, the
shown in Figure 4. The time lag between press- system pressures predicted by the design pro-
ure peaks clearly illustrates that the repressuriz- gram were up to |40% lower than those actually
ation wave was generated at the downstream end measured in the laboratory. These discrepancies
of the common pipe, and propagated upstream. were considered to be due to inaccuracies in the
The 0.04 s time lag shown between transducers predicted head losses across Ž ttings and the
3 and 5, which were 2.3 m apart, yields a wave simplifying assumptions employed within the
propagation velocity of 57.5 m/s. Noting that the program.
laboratory pipework was not restrained against The recorded data and observations conŽ rm
radial or longitudinal movement, an iterative sol- that the priming process for a multi-outlet
ution of the appropriate wave speed equation9 siphonic system is similar to that which occurs
yields an air content of 5.4% for a wave propa- with a single-outlet siphonic system.7 The only
gation velocity of 57.5 m/s. Although this can signiŽ cant difference is that the increased
only be considered to be an approximation of complexity of the multi-outlet system results in
the actual air content within the  ow, it is of a more complex  ow conditions, particularly with
similar magnitude to that previously estimated respect to the formation and movement of
for single-outlet systems.7 trapped air within the system. This is evidenced
Employing a design program used by indus- by the variable nature of the pressure traces prior
try, it was predicted that siphonic conditions to the priming of the system. This was also con-

Figure 4 Measured gutter depths and system pressures for the design criteria rainfall event (gutter 1) in ow =
5.85 l/s, gutter 2 in ow = 7.78 l/s. These data illustrates that even when fully primed (40–60 s) the pressures are not
truly ‘steady’.
134 Multi-outlet siphonic roof drainage systems

Ž rmed during the experimental work, where it intermediate gutter  ow depths, which were
was observed that the movement of the air sufŽ ciently high to ensure a continuous
pocket from branch 2 occasionally varied from siphonic action but were not high enough to
the general case outlined previously. High-speed ‘swamp’ the vortices that occurred around the
video footage indicated that the air pocket from gutter outlets. These vortices led to large
branch 2 would move downstream as a single amounts of air being entrained into the water
entity if it reached the branch junction at the  ows, which in turn resulted in lower  ow
same time as the air pocket from branch 1. If it rates and higher pressures than those associa-
reached the junction after the air pocket from ted with the fully primed system (95–100%
branch 1, the full-bore  ow conditions at the water).
junction represented such a restriction that the 3) Regime 3 – system in ows above 60% of the
air pocket could not move downstream as a sin- design criteria in ows: At these levels of
gle entity. Instead, the turbulent conditions in in ow, the system pressures initially mir-
this region led to the formation and downstream rored those occurring in a fully primed sys-
movement of a section of ‘bubbly  ow’ (air and tem, although they shortly returned to the
water mixture). type of higher, oscillatory pressures associa-
ted with Regime 2. Such conditions arose as
5.3 Rainfall events below the design criteria the gutter  ow depths were only sufŽ cient to
Experimental work has indicated that, at  ow sustain full siphonic action for a short period.
rates up to approximately 15% of the design cri- After this, the gutter depths decreased to lev-
teria in ows, the laboratory test rig behaved as els that enabled large quantities of air to
a conventional roof drainage system; that is, the become entrained with the water in ows.
 ow conditions remained free surface/annular
throughout. At all other  ow rates below the In general, it was determined that, with the
design criteria in ows, the system conditions in ow to one of the gutters Ž xed at a constant
were observed to be unsteady. With reference to rate, increasing the in ow into the remaining
Figure 5, it was determined that the  ow would gutter resulted in steadier and lower system
exhibit one of the following sets of character- pressures. This was as expected, as an increase
istics: in total system in ow leads to a decrease in
1) Regime 1 – system in ows between 15% and the volume of air being drawn into the system
40% of the design criteria in ows: These lev- together with an increase in energy losses. It
els of in ow resulted in highly unsteady con- was also apparent that, for the same total sys-
ditions, characterized by cyclical periods of tem in ow, overtopping became less likely as
positive and negative pressures. Such con- the ratio of the gutter in ows (Q gutter 1:Qgutter
ditions were caused by low gutter  ow 2) approached that of the fully primed system
depths, which meant that siphonic action (7.78:5.85 < 1.33:1). This was again as
could only be sustained for short periods, i.e., expected, a more even gutter  ow distribution
once initiated, siphonic action would quickly increasing the probability of siphonic oper-
drain one or both of the gutters, creating an ation.
airpath to the atmosphere and hence breaking The disparity between the transition from
the siphon. free surface/annular to unsteady/siphonic con-
2) Regime 2 – system in ows between 40% and ditions in the multi-outlet system and the sin-
60% of the design criteria in ows: These lev- gle-outlet system 7 mentioned previously is
els of in ow resulted in oscillating, con- considered to be due to the smaller pipe
stantly negative system pressures that were diameters employed and the  ow distorting
above those associated with the fully primed effect of the branch junction in the multi-out-
system. Such conditions were caused by let system.
 GB Wright et al. 135

Figure 5 Measured common pipe pressure (T5) for three rainfall events below the design criteria. These data illustrate
the  ow regimes which can result in vibration and noise generation.

5.4 Design criteria rainfall event (fully gutter  ow depths and system pressures mirrored
primed system) – varying gutter in ows those obtained with synchronized in ow start
As many real rainfall events progressively times (refer to Figure 4).
build in intensity, experimental work was under-
taken to assess the effect of gradually increasing 5.5 Rainfall events above the design criteria
the gutter in ows up to design criteria levels. Laboratory experiments undertaken with rain-
The only signiŽ cant difference between these fall events above the design criteria indicated
results and those obtained with constant gutter that the system pressures were almost identical
in ows was that it took longer for the system to those obtained at the design condition. How-
pressures and gutter  ow depths to build up to ever, the additional system in ows above the
the those necessary purge the air from the sys- design criteria levels resulted in continuously
tem, and initiate siphonic action increasing gutter depths, which would have
Additional experimental work was also under- eventually lead to overtopping of the gutter(s).
taken to determine the effect of staggering the If the slight variations in driving head associated
gutter in ow start times, which would represent with higher gutter depths are disregarded, these
systems incorporating widely varying roof geo- observations conŽ rm that the system pressures
metries, e.g., one gutter outlet serving a steeply occurring once a siphonic system has become
pitched roof and one gutter outlet serving a shal- primed are the minimum possible, and the
lower pitched roof. As may be appreciated, the capacity is the maximum possible, for that parti-
resulting  ow conditions were very complicated, cular system.
exhibiting two or three of the unsteady  ow
regimes identiŽ ed previously. However, it was 5.6 Total blockage of one of the outlets
apparent from the data collected that, after a An example of the data obtained from labora-
short period at the design criteria in ows, the tory experiments undertaken with one of the out-
136 Multi-outlet siphonic roof drainage systems

Figure 6 Measured gutter depths and system pressures with the outlet in gutter 1 blocked/unblocked/blocked (gutter
1 in ow = 0 l/s, gutter 2 in ow = 11.3 l/s).

lets blocked is shown in Figure 6. This scenario the case in an unblocked system. The data in
seeks to represent the sudden blockage of an out- Table 1 also highlight that system pressures were
let caused by detritus in the  ow. It can be seen considerably lower when outlet 1 was blocked.
that, whilst the outlet in gutter 1 (outlet 1) was This would indicate that, if a system were
completely blocked (0–178 s), the laboratory test designed to operate at pressures below approxi-
rig acted as a single-outlet siphonic system, with mately 2 7 mH2O, a complete blockage of one of
the pressures stabilized at the relevant fully the outlets might result in the onset of cavitation
primed levels and the  ow depth in gutter 2 and/or failure of the system by pipe deformation.
approaching a steady state. Table 1 summarizes Figure 6 also indicates that, when outlet 1 was
the salient system conditions pertaining to Fig- unblocked (178–197 s), the system quickly
ure 6, when outlet 1 was blocked, and Fig- reverted to a multi-outlet mode of operation. As
ure 4, when both outlets were open. As shown, experimental restrictions meant that there could
although the total system capacity was lower be no in ow into gutter 1, an airpath to the
with outlet 1 blocked, the capacity of the open atmosphere was created, leading to the cessation
outlet in gutter 2 was actually higher than was of siphonic action, an increase in system press-

Table 1 Measured system conditions with outlet 1 unblocked and blocked

Outlet blocked Fully primed Capacity of outlet Minimum measured pressure (mH2 O)
capacity (l/s) in gutter 2 (l/s)
Transducer 1 Transducer 2 Transducer 3

None 13.63 7.78 2 0.552 2 0.595 2 1.388

Outlet 1 11.30 11.30 2 1.719 2 1.846 2 2.147
 GB Wright et al. 137

ures and a decrease in system  ow rates. This between the gutter outlets and the point of dis-
resulted in a very rapid increase in the water charge (free discharge case) or the point at which
level within gutter 2, and would have led to the downpipe enters water (submerged discharge
overtopping of this gutter if outlet 1 was not re- case), it can be deduced from Figure 7 that the
blocked (after 197 s). In real situations, the water driving head for each of the four conŽ gurations
collected in a gutter whose outlet is blocked was different. In addition, it is clear from Fig-
would prevent such dramatic pressure changes ure 7 that the head losses associated with each
occurring when the outlet became unblocked. of the four conŽ gurations varied. As a result, the
However, these data do indicate the operational gutter in ows necessary to cause priming of the
problems that could occur in a multi-outlet four different conŽ gurations were different.
siphonic system if one of the outlets is not sub- From the data shown in Table 2, it is clear that
merged and is allowing large quantities of air to the use of any conŽ guration other than a freely
enter the system. discharging vertical downpipe will result in a
lower system capacity.
5.7 Effect of different system termination Figure 8a and b shows a sample of the experi-
conŽ gurations mental data obtained using the four different ter-
To ensure the efŽ cient operation of a siphonic mination conŽ gurations. As shown, the con-
roof drainage system, it is essential that full-bore Ž gurations that discharged under water resulted
 ow conditions are broken before any connec- in signiŽ cantly longer priming times and higher
tion to the surface water sewer network. If not, gutter  ow depths than those discharging
the  ows within the siphonic system and the directly to the atmosphere. This was because
sewer network may interact, leading to unpre- greater (positive) system pressures were required
dictable conditions and potential problems. to purge the initial air pockets.
Breaking of full-bore  ow conditions can only The experimental data discussed above high-
be guaranteed by ensuring that the  ow exits the light the importance of the interface between a
siphonic system above the highest water level in siphonic roof drainage system and a surface
the surface water sewer. However, as surface water sewer. For example, consider the con-
water sewers are normally designed to a lower ditions within a siphonic roof drainage system
level of risk than roof drainage systems, it is during a severe rainfall event. Theoretically,
clear that a rainfall event which causes the prim- there should be no operational problems if the
ing of a siphonic roof drainage system may also rainfall intensity is less than the design criteria.
cause surcharging of the downstream surface However, if the downstream surface water sewer
water sewer, a scenario that could lead to the happens to surcharge, and the water level in the
type of  ow interactions discussed above. There- manhole rises above that of the siphonic system
fore, experimental work was undertaken to discharge point, the driving head of the siphonic
determine the effect of terminating a siphonic system will reduce. This could increase the time
roof drainage system under water. In addition, required to prime the system, and will certainly
data was also collected to determine the effect reduce the total capacity of the system. Such a
of a right-angled termination, which often proves scenario could lead to failure of the system by
necessary due to the site layout. The four differ- gutter overtopping.
ent system terminations that were investigated
are shown in Figure 7. It should be noted that, 6 Field observations
due to space restrictions, the length of the verti-
cal downpipe in these conŽ gurations was To complement the laboratory investigation,
reduced to 4.07 m.  ow conditions have been monitored within
As the driving head for a siphonic roof drain- three siphonic roof drainage systems installed at
age system is deŽ ned as the elevation difference the National Archives of Scotland Document
138 Multi-outlet siphonic roof drainage systems

Figure 7 Siphonic system termination conŽ gurations investigated

Table 2 Variation in design criteria gutter in ows with termination conŽ guration

System termination Fully primed capacity (l/s) Fully primed capacity (as % of type 1 in ows)
type
Gutter 1 Gutter 2 Gutter 1 Gutter 2

1 7.46 5.71 100 100

2 7.35 5.45 98.5 95.4
3 7.28 5.54 97.6 97.0
4 7.11 5.57 95.3 97.5

Figure 8 (a) Variation in common pipe pressure (T5) with system termination conŽ guration (fully primed conditions).
(b) Variation in gutter 1  ow depth (T1) with system termination conŽ guration (fully primed conditions)
 GB Wright et al. 139

Repository Building, Edinburgh. Whilst a (SIPHONET12) is capable of simulating the

detailed description of the monitoring equipment priming phase of a single-outlet siphonic roof
and protocols is given elsewhere,10 a schematic drainage system. This model utilizes a method
of the systems that are being monitored is shown of characteristics-based solution technique,
in Figure 9. Since June 2000, system pressures which has been employed successfully at Heriot-
have been recorded when the rainfall intensity Watt University in the simulation of both free
exceeds 5 mm/h, and rainfall intensities have surface and full-bore  ow conditions. However,
been recorded using a tipping-bucket rain gauge. during the development of the model it became
In addition, gutter  ow depths have been moni- clear that the method of characteristics may not
tored since September 2001 using modiŽ ed air be particularly suited to the simulation of mov-
pressure transducers. ing hydraulic jumps. In addition, numerical stab-
As anticipated, the vast majority of the ility problems were also encountered with the
recorded rainfall events have been below the transition between free surface and full-bore
design criteria of the monitored systems, and  ow conditions. As these deŽ ciencies would
much of these data have conŽ rmed the labora- become more limiting in the more complex case
tory Ž ndings. The data shown in Figure 10 rep- of multi-outlet systems, it was decided to employ
resents the most signiŽ cant rainfall event, in a new modelling approach. This will incorporate
terms of prolonged siphonic action, that has been the Lax–Wendroff Ž nite difference solution
recorded to date. This event had a maximum technique.13
rainfall intensity of 105 mm/h, which equates to
a return period of 32 years,11 and appeared to 8 Conclusions and future work
result in continuous siphonic action for a period
of approximately 500 s. The conclusions of this ongoing research pro-
An analysis of the Ž eld data collected to date gramme may be summarized as follows:
indicates that 7% of recorded events resulted in
prolonged siphonic action and 50% of recorded · The priming of a multi-outlet siphonic roof
events resulted in signiŽ cant negative system drainage system is similar, although more
pressures. complex, to that of a single-outlet system.
· Current design programs may yield inaccurate
7 Development of the numerical model system pressures, which could lead to oper-
ational problems and/or system failure.
As stated previously, the existing numerical · At rainfall intensities below the design cri-
model developed at Heriot-Watt University teria, the  ow conditions within a multi-outlet

Figure 9 Schematic of the siphonic roof drainage systems being monitored at the National Archives of Scotland
Document Repository Building, Edinburgh.
140 Multi-outlet siphonic roof drainage systems

Figure 10 Measured conditions (system 2) on 2 August 2000 (refer to Figure 9 for system and transducer layout). The
time-varying nature of this rainfall event is typical, and demonstrates the importance of using rainfall hyetographs
to assess system performance fully.

siphonic roof drainage system are unsteady, of the type of operational problems and system
and may exhibit one of three different  ow failures detailed previously.
regimes.
· The complete blockage of one of the outlets Acknowledgements
in a multi-outlet siphonic roof drainage sys-
tem may lead to system pressures falling The researchers remain grateful for the assist-
below their design levels, and could result in ance given by: Dales Fabrications Ltd (UK),
system failure by cavitation and/or pipe defor- EPSRC (UK), Full ow Ltd (UK), Geberit AB
mation. (Switzerland), HR Wallingford Ltd (UK), Pick
· To ensure efŽ cient operation of a siphonic Everard (UK), Simona Ltd (Germany), Som-
roof drainage system, consideration must be merhein AB (Sweden), The Scottish Executive
given to its interaction with the downstream (UK) and Vanderweil Engineering Inc (USA).
surface water sewer network.
The Ž nal phase of this current research pro- References
gramme involves the further development of the
1 BSI, BS EN 12056-3 Gravity drainage systems
numerical model. It is intended that the Ž nal
inside buildings. Roof drainage, layout and
model will be capable of accurately simulating calculation. UK: British Standards Institute,
the  ow conditions within multi-outlet siphonic 2000.
roof drainage systems for all realistic rainfall 2 May RWP, Escarameia M. Performance of
events. It is anticipated that such a model will siphonic drainage systems for roof gutters.
be used for diagnostic design purposes and code Report No. SR463. Wallingford: HR
formulation, which should reduce the occurrence Wallingford, 1996.
 GB Wright et al. 141

3 Bramhall M, Saul AJ. Hydraulic performance of Proceedings of Water Supply and Drainage for
siphonic rainwater outlets. Proceedings of the Buildings, CIB W62 1999, Edinburgh, 1999.
8th International Conference on Urban Storm 9 Wylie EB, Streeter VL. Fluid transients in
Drainage, Sydney, 1999. systems. Englewood Cliffs, New Jersey:
4 Bowler R, Arthur S. Siphonic roof drainage – Prentice-Hall, 1993.
design considerations. Proceedings of Water 10 Arthur S, SwafŽ eld JA. Onsite evaluation of an
Supply and Drainage for Buildings Seminar, installed siphonic roof drainage system.
CIBW62 1999, Edinburgh, 1999. Proceedings of Water Supply and Drainage for
5 Rattenbury J. Fundamentals of siphonic roof Buildings Seminar, CIBW62 2000, Rio de
drainage. http://www.pmengineer.com/CDA/ Janeiro, 2000.
ArticleInformation/features/BNP Features Item/ 11 CEH Wallingford, Flood estimation handbook
0,2732,21863,00.html, 3 January 2001. (Vol. 2: Rainfall Frequency Estimation).
6 Sommerhein P. In: Garside S ed.. UV-System Wallingford: CEH Wallingford, NERC, 1999.
technical manual. Sweden: Sommerhein AB, 12 Arthur S, SwafŽ eld JA. Numerical modelling of
1996. the priming of a siphonic roof drainage system.
7 Arthur S, SwafŽ eld JA. Siphonic roof drainage: Proceedings of CIBSE, Building Serv. Eng. Res.
the state of the art. Urban Water 2001; 3, 43– Technol. 1999; 20.
52. 13 Lax PD, Wendroff B. System of conservation
8 Arthur S, SwafŽ eld JA. Numerical modelling of laws. Commun. Pure Appl. Math. 1960; 13:
a siphonic rainwater drainage system. 217–37.