This article was downloaded by: [Erciyes University]

On: 22 December 2014, At: 09:39

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Hydraulic Research

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tjhr20

An experimental study of Roman dropshaft

hydraulics
a
H. Chanson MIAHR
a
Department of Civil Engineering , The University of Queensland , Fluid Mechanics
Hydraulics and Environmental Engineering, Brisbane, QLD4072, Australia E-mail:
Published online: 01 Feb 2010.

To cite this article: H. Chanson MIAHR (2002) An experimental study of Roman dropshaft hydraulics, Journal of Hydraulic
Research, 40:1, 3-12, DOI: 10.1080/00221680209499868

To link to this article: http://dx.doi.org/10.1080/00221680209499868

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 An experimental study of Roman dropshaft hydraulics
Une etude experimentale de l'hydraulique des puits de rupture romains
H. CHANSON, MIAHR, Senior Lecturer, Fluid Mechanics, Hydraulics and Environmental Engineering, Department of Civil Engineer-
ing, The University of Queensland, Brisbane QLD4072, Australia, Email: [email protected]

Abstract
In Roman aqueducts, series of vertical dropshafts were used to dissipate the kinetic energy of the flow: i.e., the dropshaft cascades. A re-analysis of
Roman dropshaft hydraulics is conducted with physical model tests. Three basic flow patterns are observed. The results demonstrate that the vertical
dropshafts could be very efficient energy dissipators and re-oxygenation structures, under appropriate flow conditions. The optimum operation of Roman
dropshaft is discussed and an analytical model is developed to predict these conditions. Further the performances of aqueduct dropshafts are compared
with modern dropshaft designs, and the operation of dropshaft cascades is discussed. Additional material is available upon request.

Résumé
Certains aqueducs romains étaient équipes de cascades de puits de rupture, concues pour briser l'énergie cinétique de l'écoulement. On a conduit une
série de tests sur modèles physiques pour ré-analysei l'hydraulique des puits de rupture romains. Les résultats démontrent trois régimes d'écoulement.
Dans des conditions optimales. les puits de rupture sont des systèmes de perte de charge tres efticaces. Ces conditions optimales ont été déterminées
avec des modèles physiques et analytiques. Les performances des puits romains sont aussi comparées avec celles de puits modernes. et on discute aussi
de l'opération des cascades de puits. Sur demande. Pauteur peut fournir de plus amples informations.
Downloaded by [Erciyes University] at 09:39 22 December 2014

Introduction today (e.g. JAIN and KENNEDY 1983, APELT 1984,

RAJARATNAM et al. 1997) although it must be optimised for
The hydraulic expertise of the Romans has contributed signifi­ optimum operation, and to prevent scour and erosion. A third ap­
cantly to the advance of science and engineering in the Antiquity plication is the aeration of the flow: e.g., re-oxygenation (e.g.
and up to the end of the Middle Age. The aqueducts at Rome, in ERVINE and AHMED 1982).
France and North Africa left magnificent ruins for example (e.g. Table 1 summarises the main characteristics of well-documented
ASHBY 1925, RAKOB 1974). Their construction was an enor­ Roman aqueduct dropshaft and dropshaft cascades. Although
mous task conducted by the army (e.g. FEVRIER 1991). The to­ their intended purpose(s) is today poorly understood, the cascades
tal cost and duration of the works were functions of' the difficul­ were used in steep topography predominantly : e.g., at Recret,
ties : i.e., tunnels, bridges, arcades, raised foundation, siphons. Vaugneray and Grézieu-la-Varenne, Yzeron aqueduct (Lyon.
Completion of the works could take from 3 years to over 15 Fr.); at Montjeu, Autun aqueduct (Fr.) (Table 1). Most Roman
years. The cost ranged between 23 to 69 millions of US$ per dropshafts were designed with the outflow channel located on the
kilometre'. The aqueducts were designed for small flow rates (0.2 opposite wall to the inflow (Fig. 1, 2A and 3A). An unusual de-
to 2 mVs maximum) with flat longitudinal slopes : i.e., 1 to 3
m/km in average (e.g. HODGE 1992). Short sections with steep-
gradients up to 78% were sometimes introduced (CHANSON
2000). Field observations suggest primarily three designs : steep
"smooth" chutes, stepped channel and the dropshaft cascade
(Fig. 1). The stepped chute design was common also with darn
spillways 2 . The use of dropshaft 3 and dropshaft cascade along the
main branch of aqueducts was a specific engineering feature of a
new aqueduct. From a sole hydraulic perspective, the dropshafts
might have been used for: (1) a vertical drop in invert elevation,
(2) kinetic energy dissipation, and (3) flow aeration. In the first
application, a dropshaft allows the connection between two flat
conduits, located at different elevations, along a very short dis­
tance : i.e., the shaft length L. The second application is the dissi­
pation of the kinetic energy of the flow. Such a design is still used
Fig. 1 Sketch of a dropshaft cascade in Roman aqueduct

Revision received July 26. 2000. Open for discussion till August 31. 2002.

1. That is, 1 to 3 millions of sesterces per km (LEVEAU 1991) converted into 1999 US$ using the silver weight of the sesterce as a reference.
2. The oldest stepped spillway was built around BC 1,300 in Greece (CHANSON 1997) and several stepped chutes were used prior to the Roman era
(e.g. CHANSON 1995. pp. 23-37). Roman engineers built also several significant stepped spillways : e.g., Kasserine dam, Tunisia AD 100?, Oued
Guergour dam, Tunisia AD 100?.
3. "dropshaft" in English, "puit de rupture" in French, "Toshecken" in German, "poz.o resulto" in Spanish.

JOURNAL OF HYDRAULIC RESEARCH. VOL. 40. 2002. NO.

 Table I: D r o p s h a f t c a s c a d e s in R o m a n a q u e d u c t s

Steep Section Ret'. Geometry Flow conditions Remarks

(I) (2) (3) (4) (5)
Dougga aqueduct [Ca]
Oued Melah B~3.3m AH - 4 to 5 m Located downstream of 200-m long bridge,
b - 0.35 m (tunnel) upstream of tunnel..
Vaugneray. Yzeron aqueduct [Col] AH = 21.9 m Vaugneray branch of Yzeron aqueduct.
Puit du Bourg Rectangular dropshaft: h = 2.55 m, dL. = 0.24 m (?)
b = 0 . 4 m . B = 1.14 m. L = 1.9 m d/s flow conditions :
d ~ 0.35 m,
V ~ 1.33 m/s
Recret/Grézieu-la-Varcnne, [Col] Rectangular dropshafts AH = 38 m Main branch of Yzeron aqueduct.
Yzeron aqueduct
Puit Gouttenoire Square dropshaft: h = 2.55 m, d„ = 0.197m
b = 0.55 m. B = L = 1.18 m,
P = 1.12 m
Puit-en-bas Rectangular dropshaft: h = 2.5 m, d c = 0.197 m
b = 0.55 m. B = L = 1.17 m, d/s flow conditions : d ~ 0.15 m,
D = 1.26 m, P = 1.35 m V ~ 1.9 m/s
Chabet Ilelouine. Cherchell [LP] AH = 12.28 m 4 series of steep chutes followed by circular
aqueduct dropshaft.
Downloaded by [Erciyes University] at 09:39 22 December 2014

Puit amont Circular dropshaft : h = 0.77 m, Supercritical upstream flow : Located downstream of steep smooth chute.
b = 0.94 m, 0 = L = 2.03 m, V ~ 8 m/s
P > 1.75 m
Gunugu aqueduct AH = 20 m
Moulin Romain ]LP] Circular dropshaft: h ~ 3.5 to 4 m, Upstream channel : 0.86-m wide.
b = 0.38 m. 0 = L = 0.80 m
Rusicade aqueduct [Ve] Circular dropshafts

3eaulieu aqueduct [CQ1 AH = 37 m Combination of steep chutes and dropshafts

Puit d'Olivari Dropshaft: h = 6.2 m, Rectangular or circular ? 147-m between
b ~ 0.45 to 0.6 m dropshaft.
Puit du Chateau Dropshaft : h ~ 8 m Rectangular or circular ? 167-m between
dropshaft.
Brisecou Forest, Montjeu aq- [CQ. Rectangular dropshaft: h = 4.4 m, A H ~ 140 m A series of 24 dropshafts (possible combina­
ueduct PR] b = 0.8 m, B = 3.0 m, L = 2.4 m. x ~ 770 m tion with steep chutes).
D = 1.57 m, P > 0.8 m
9 dropshafts (h = 4.4 m) 15 to 30-m between dropshaft.
15 dropshafts (h = 4 . 4 m ) 50 to 120-m between dropshaft.
Cuicul aqueduct [Al]
Grand thermes distribution line Circular (?) dropshafts: h~l to 0.4 m, AH ~ 3 m Series of 4 dropshafts on an urban distribu­
b = 0.45 m, 0 = L = 0.80 m x~85m tion line.
Köln aqueduct [Gr]
Mechernich-Lessenich Rectangular dropshaft : h = 0.35 m. [Gr, p. 97] One dropshaft installed along a
b = 0.7 to 0.75 m, B = 0.9 m, L = 1.185 m. steep section (AH ~ 84 m, 0 ~ 2.6°).
P = 0.2 m
Valdepuentes aqueduct (Aqua |Lop. Several dropshafts and dropshaft cascades.
Vetus). Cordoba Vil,
Vi2]
One (?) dropshaft AH = 3 m Hornillo
Dropshaft cascade : 34 dropshafts A H - 120 m Cerro de los Pinos. Upstream of
x ~ 400 m Valdepuentes bridge.
Three unusual 90-degree bend shafts:
spiramina.
Pozo No. 11: circular dropshaft: h = 2.9 m, Shaft equipped with steps sculpted inside (h
b = 0.6m, 0 = L = O.61 m, ~ 0.5 m).
P = 0.76 m
u/s slope : 6 - 5%
d/s slope : G ~ 5%
Pozo No. 21: circular dropshaft: h = 3.1 m. 90-degree bend shaft equipped with steps
b = 0.5 m, 0 = L = 0.55 m, sculpted inside (h ~ 0.6 m).
P = 0.9 m

JOURNAL OF HYDRAULIC RESEARCH. VOL. 40. 2002. NO.

 Dropshaft cascade: AH~200m(?) Madinat-al-Zhara. Dowstrcam of
7 dropshafts Valdepucntes bridge.
Dropshaft cascade (?) : at least two AH ~ 20 m After the road (Elev. 180-160 m R.L.).
dropshafts x ~ 340 m
One (?) dropshaft Elev. loOmR.L..
Dropshaft cascade (?) : at least three x ~ 490 m Cortijo los N.. Downstream of junction with
dropshafts Veneros branch junction.

References : [AI] A L L A I S (1933); [Cal C A R T O N (1899); [Col] C O N S E I L G E N E R A L DU R H O N E (1991); [CQ] C O Q U E T (1966); [Gr]
G R E W E (1986); [LP] LEV EAU and P A I L L E T (1976); [Lop] L O P E Z - C U E R V O (1985); [PR] P I N E T T E and R E B O U R G ( 1986); [Ve] V E R T E T
(1977); [Vil] V I L L A N U E V A (1993); [Vi2] V I L L A N U E V A (1996).
Notes : d c : critical flow depth; AH : total head loss. Sites names are listed from upstream to downstream (for a given aqueduct branch)

sign was used at a small number of shafts along the Valdepuentes

aqueduct (Spa.) : the flow direction turned 90-degrees (Fig. 2B
and3B).
Overall there is little information on the hydraulic performances
of dropshaft cascades in Roman aqueducts. In the present study,
the hydraulics of Roman dropshafts is re-visited using both physi­
Downloaded by [Erciyes University] at 09:39 22 December 2014

cal and analytical models. The results provide a new understand­

ing of the dropshaft cascade operation. The performances are also
compared with modern designs. It will be shown that the Roman
dropshaft design is an efficient low-head energy dissipator.

Experimental apparatus

Three dropshaft models were studied (Table 2, Fig. 3). The mod­
els were built in marine plywood and perspex, with a vertical
square dropshaft (B = L). The upstream channel was open while
the downstream conduit was covered and ended with a free
overfall. The discharges were deduced from the brink depth mea­
surements which were first calibrated in-situ with volume-per-
time discharge data. [A calibration curve was obtained for each
model.] Free-surface elevations were recorded with pointer
gauges in the upstream and downstream channels, while the free-
surface height in the shaft was measured with rulers. The total
head was measured with a total head tube (0 = 1 mm). Additional
information were obtained with high-speed photography and
video-camera. Further details of the experimental setup are pre­
sented in CHANSON (1998).
The upstream and downstream channels operated as free-surface
flow for all investigated flow conditions. Indeed water flowed as
open channel flow in the Roman aqueducts although the conduit
height D ranged from 1 to 2 m for ease of construction and main­
tenance. The water depths were usually less than 0.2 to 0.3 m at
maximum flow rates (e.g. BLACKMAN 1978, HODGE 1992,
CHANSON 1998). Fig. 2 Well-documented Roman drophshafts. (A) Puit-en-bas, Recret,
Yzeron aqueduct, France (Qmux ~ 0.15 m7s)
Most experiments were conducted with subcritical inflow condi­
tions. A number of tests were performed with supercritical in­
flows in the model 2 only.

JOURNAL OF HYDRAULIC RESEARCH, VOL. 40, 2002, NO. I 5

 Table 2: Experimental studies of vertical dropshafts

Study Upstream con­ Shaft Downstream con­ Flow conditions Remarks

duit duit
(1) (2) (3) (4) (5) (6)
APELT(1984) Circular Square Circular 0.0022 < Q < 0.026 mVs Energy dissipation downstream of cul­
0 = 0.152 m B = L = 0.152m 0 = 0.152 m verts.
h = 0.325 m, P = 0
Circular Square Circular 0.0022 < Q < 0.026 m 7s
0 = 0.152 m B = L = 0.203 m 0 = 0.152 m
h = 0.325 m, P = 0
RAJARATNAM et al. Circular Vertical plate N/A Q < 0.042 mVs Series 3. Sewer dropshaft models.
(1997) 0 = ().154m L = 0.5 m. h = 2.9 m
Circular Circular Circular 0.0021 < Q < 0.042 m7s Series 4.
0 = 0.154 m 0 = 0.29 m 0 = 0.29 m
h = 2.11 m,P = 0
Circular Circular Circular 0.0021 < Q < 0.042 m7s Series 5. Curved inlet radius.
0 = 0.154 m 0 = 0.29 m 0 = 0.29 m
h = 2.11 m. P = 0
Circular Circular Circular 0.0021 < Q < 0.042 m-Vs Series 6. Curved inlet radius.
0 = 0.154 m 0 = 0.152 m, P = 0 0 = 0.152 m
Downloaded by [Erciyes University] at 09:39 22 December 2014

Present Study
Model 1 (Recret.Puit- Rectangular Square Rectangular 0.0002 < Q < 0.018 nrVs Open upstream conduit. 1:4 scale model
en-bas) b = 0.144 m B = L = 0.30 m b = 0.15 m (Yzeron aqueduct).
6 = 0.17° h = 0.505 m D = 0.25 m
P = 0.365 m 6 = 0.37°
Model 2 (Valdepuentes) Rectangular Square Rectangular 0.00005 <Q <().() 19 mVs Open upstream conduit. Dimensions
b = 0.11 m B = L = 0.20 m b = 0.11 m (a) Sub-critical inflow scaled upon Valdepuentes dropshaft di­
9 = 0.17° h = 0.688 m D = 0.21 m (b) Supercritical inflow: 2 < mensions.
P = 0.201 m Fr < 15
Model 3 (Valdepuentes, Rectangular Square Rectangular 0.00008 < Q < 0.022 mVs Open upstream conduit. 1:4.5 scale
90-degree bend) b = 0.11 m B = L = 0.20 m b = 0.11 m model (Valdepuentes aqueduct).
6 = 0.17° h = 0.688 m D = 0.21 m
P = 0.201 m 90-degree bend

Table 3: Transition between flow regimes (subcritical inflow conditions)

Flow patterns and dropshaft operation

Flow conditions h/L h/D dc/L cL/L
Several flow patterns were observed as functions of the flow rate Data Theory
(1) (2) (3) (4) (5)
and dropshaft dimensions. For the models 1 and 24, three flow
Transition regime Rl to R2
regimes may occur (Fig. 4). The transition conditions between
Model 1 2.90 3.48 0.15 0.09
regimes are summarised in Table 3. The observations (Table 3, Model 2 3.34 3.18 0.10 0.08
column 4) are compared with analytical calculations of the nappe Transition regime R2 and R3
trajectory developed by CHANSON (1998) (Table 3, column 5). Model 1 2.90 3.48 0.30 0.12
At low flow rates, the free-falling nappe impacts into the shaft Model 2 3.34 3.18 0.15 0.11
pool (regime R1, Fig. 3 A and 4A). Substantial air bubble entrain-
ment takes place and the entrained air bubbles occupy most of the For larger discharges, the upper nappe of the free-falling jet im­
dropshaft pool. At the lowest flow rates, the free-falling nappe pacts into the downstream channel, flowing in between the inlet
contracts down the jet trajectory until the outer edges intersect invert and obvert (regime R2). The pool free-surface level in­
(regime Rla). When the side-edges of the jet intersect, a 'fin' or creases significantly, and lesser air bubble entrainment is ob­
'central ridge' develops with a thickness much smaller than the served in the pool. High risks of scour exist at the obvert edge and
other transverse dimension until nappe impact into the shaft pool. on the invert of the downstream channel (Fig. 4B). The jet impact
With increasing discharges, the nappe trajectory enlarges, the jet onto the downstream channel is further associated with the devel­
outer edges do not intersect before impingement and no central- opment of sidewall standing waves in the conduit, similar to those
ridge formation occurs (regime Rib). In the downstream channel, observed by CHANSON and TOOMBES (1998).
the flow is supercritical (in absence of downstream backwater At large flow rates, the free-jet impacts onto the opposite wall
effect) and shock waves develop. above the downstream conduit obvert (regime R3) (Fig. 3B and
4C). Significant water deflections take place in the shaft. During
the study, the impact angle of the nappe onto the wall was shal-
4. in the model 3, only the regimes Rl and R3 are observed.

JOURNAt. OF HYDRAULIC RESEARCH. VOL. 40. 2002. NO.

 Section A-A
Downloaded by [Erciyes University] at 09:39 22 December 2014

Valdepuentes aqueduct, spiramina

(Aqua Vetus. Cordoba)
Pozzo resalto No. 21
upstream of Valdepuentes aqueduct
Fuentes de la Teja-Madinat al Zahra

Fig. 3 Photographs of the physical models. (A) Photograph of the

model 1 (Regime Rl, Q = 0.001 mVs) (high-speed photograph)

1 Top view
for rectangular and circular conduits respectively, D is the down­
stream conduit height, Q is the flow rate, and b is the upstream
channel width (Fig. 4A). The results are consistent with the obser­
vations of APELT (1984) and RAJARATNAM et al. (1997) (Ta­
ble 1, Fig. 5A).
Fig. 2 Well-documented Roman drophshafts. (B) Pozzo resalto No. 21. The dimensionless bubble penetration depth is plotted in Figure
Valdepuentes aqueduct, Spain (Q n m ~ 0.255 rrrVs). Fuentes de 5B as a function of the dimensionless flow rate d c /h where dc is
la Teja-Madinat al Zahra, upstream of Valdepuentes bridge the critical depth in the upstream channel and h is the drop in in­
vert elevation. In flow regimes Rl and R3. substantial flow aera­
low and did not produce a formed roller as observed by tion takes place, the bubbles plunge deep down to the shaft bot­
RAJARATNAM et al. (1997). tom and the bubble cloud occupies more than half of the shaft
For all the experiments, the flow in the downstream conduit was pool volume (e.g. Fig. 3). The entrained bubbles enhance the air-
found to leave the dropshafl as a supercritical open channel flow. water interface area and the air-water gas transfer : i.e.. re-aera­
A similar observation was noted by RAJARATNAM et al. tion. The flow regime R2 is less efficient in entraining air because
(1997). the nappe interacts with the downstream conduit inlet. The en­
trained bubbles do not have enough downward momentum to
reach the shaft bottom.
Flow properties
Residual energy data are presented in Figure 5C. The data are
Pool free-surface height data are reported in Figure 5A where y presented as H r „7H, as a function of the dimensionless flow rate
is the free-surface height above downstream invert (Fig. 4A). The 3/ l n ° —
pool height rises with increasing discharges up to about y /D ~ 1 dc/h (or \/Q2/(g*D2)/hfor circular pipes), where Hres is the residual
to 1.2, then remains stable with larger flow rates, until it rises head in the downstream channel and H, is the upstream total head
again forQ' > 1.2, whereQ' = Q/\/g*t>2*D3 and Q' = Qn/g*7i2*D-'Vl6 measured above the downstream channel invert. The results are
compared with the data of RAJARATNAM et al. (1997). Low

JOURNAL o r HYDRAULIC Rl SLARCH. VOL. 40. 2002. NO

 Roman dropshaf't
regime Rl

-► x

bubble Section A-A

Downloaded by [Erciyes University] at 09:39 22 December 2014

swarm

•*- — ►
l
► b
A ▲
—► b B —► ;b
*
i
if
Bottom view
Fig. 4 Basic flow patterns, including notation. (A) Regime Rl.

Regime R2

Fig. 3 Photographs of the physical models. (B) Photograph of the

model 3 (Regime R3, Q = 0.004 nrVs) Risks of scour
and damage

residual heads, associated with high energy dissipation, are

achieved at low flow rates (Regime R l ) (Fig. 5C). Poor dissipa­
tion performances are observed in regime R2. In regime R3, the
dimensionless residual head ranges from 20 to 55% depending
upon the model geometry.
A comparison between the three model performances suggests
that the model 3 is a more efficient dissipator (Fig. 5C). A domi­
nant feature of this design (Fig. 3) is the absence of the flow re­
gime R2 characterised by lesser energy dissipation and some risks
of scour. In regime R l , the performances of the three models are
Fig. 4 Basic flow patterns, including notation.
very close (Fig. 5).
(B) Regime R2

Discussion
as the downstream channel bed to minimise construction costs. In
Compared with modern designs, Roman dropshafts exhibited an Roman dropshafts, the pool of water acts as a cushion at nappe
unusual shape : i.e., a deep wide shaft pool. Modern dropshafts do impact, preventing scour at the shaft bottom. The shaft pool facil­
not include a pool, the shaft bottom being at the same elevation itates further the entrainment (by plunging jet) of air bubbles deep

JOURNAL OF HYDRAULIC RESEARCH, VOL. 40. 2002. NO. 1

 Regime R3

Nappe
impact
onto
opposite
wall

Fig. 5 Hydraulic performances of the dropshaft models. (B)

Dimensionless bubble swarm depth D,lb/(yp+P)

formances of the Roman dropshaft design are achieved with a

flow regime R1. The study shows also that the flow regime R2 is
Downloaded by [Erciyes University] at 09:39 22 December 2014

Fig. 4 Basic flow patterns, including notation. associated with high risks of scour and erosion at the lower con­
(C) Regime R3 duit inlet and obvert. The Roman aqueducts had to be designed
with dropshafts operating in a flow regime Rl for long-lasting
down, maximising the bubble residence time and hence the air- operation. Based on analytical calculations of the nappe trajectory
water gas transfer. The design contributes successfully to an en­ and impact conditions, the optimum operation (i.e. regime R l ) of
hancement of the DO content (dissolved oxygen content). Roman Roman dropshafts operating with subcritical inflow must satisfy:
dropshafts had a wider shaft (i.e. B/b > 2) than modern designs
(i.e. B/b = 1 to 1.5). It is believed that the wider shaft was se­
Q < 0.1292 * \ / g * b Regime Rl (1)
lected for an easier construction and maintenance (e.g. Conseil
General du Rhone 1996).
The rate of energy dissipation of Roman dropshaft is compared where b is the dropshaft inflow width, L is the shaft length and h
with the performances of modern drop structures and vortex is the invert drop (Fig. 4A) (CHANSON 1998). For dropshafts
dropshafts in Figure 6. The calculations for drop structures and operating with supercritical inflow, the inflow conditions must
vortex dropshafts were based upon the works of RAND (1955) satisfy :
and VISCHER and H ACER (1995) respectively. The results (Fig.
6) suggest that Roman dropshafts operating at low flow rates (i.e.
regime Rl) were very efficient energy dissipators by modern
standards. Today the Roman dropshaft design may be considered
as a low-head low-discharge dissipator.
The experimental investigation has highlighted that the best per-

Fig. 5 Hydraulic performances of the dropshaft models. (C)

Dimensionless residual head H,„7H, as a function of the
3/~
Fig. 5 Hydraulic performances of the dropshaft models. (A) dimensionless flow rate dc/h (or \/Q2/(g*D^Vh for circular
Dimensionless shaft pool free-surface height yp/D pipes)

JOURNAL OF HYDRAULIC RESEARCH, VOL. 40. 2002, NO. I 9

 Conclusion
Roman engineers built dams, spillways and hydraulic structures
all around the Mediterranean sea. Some aqueducts were equipped
with dropshafts and dropshaft cascades (Table 1). The hydraulics
of Roman dropshaft was investigated using laboratory models.
The results are compared with an analytical model.
Three flow regimes were observed basically. Optimum dropshaft
operation occurred in the flow regime Rl, characterised by low
flows and nappe impact into the shaft pool. The dropshaft opera­
tion was more efficient in terms of energy dissipation and air bub­
ble entrainment then modern designs (Fig. 6). For intermediate
flow rates, unfavourable flow conditions may occur (regime R2).
At large discharges, the operation of Roman dropshaft is similar
Fig. 6 Rate of energy dissipation AH/H, as a function of the ratio dc/h to that of modern structures.
(or \JQ2/(g*D-)/h for circular pipes). Comparison between Ro­
The issues of re-aeration should also be considered. Most aque­
man dropshaft model (Present study) and modern drop struc­
tures : drop structures (RAND 1955) and vortex dropshaft ducts were covered along their entire length, limiting the gas
(VISCHER and HAGER 1995) transfer at the free-surface and the downstream waters were low
in dissolved oxygen (DO) content, unless re-oxygenation devices
Downloaded by [Erciyes University] at 09:39 22 December 2014

were installed. The writer suggests that dropshafts might have

Vh 1 h been introduced, in place of steep chutes, to enhance the water
. <— *— *
\/g * L 2 L quality and to re-oxygenate the water.
Regime Rl (2)
L
\ /« Q' Q
\
8 *—+ — Acknowledgments
[ V h g*b"*h-*L Vg*b2*h2*L J The writer wishes to acknowledge the following people (in alpha­
where Vb is the inflow velocity. betical order) for their help and assistance : Professor C. Apelt,
The University of Queensland, Australia; Dr D. Blackman,
Monash University, Australia; Ms Chou Y.H., Brisbane, Austra­
Operation of dropshaft cascade
lia; Dr M.R. Gourlay, The University of Queensland, Australia;
A dropshaft cascade (Fig. 1) is basically a high-head dissipator. Dr A.T. Hodge, Carleton University, Canada; Mr G. Illidge, The
For example, AH = 200 m at Madinat-al-Zhara, Valdepuentes University of Queensland, Australia; Mr P. Leveau, Université
aqueduct. History showed that this design could operate success­ d'Aix-en-Provence, France; Professor N. Rajaratnam, University
fully for long periods and required (likely) little maintenance. For of Alberta, Canada; Dr A.V. Villanueva, University of Cordoba,
example, the Valdepuentes aqueduct and its dropshaft cascades Spain.
were used by the Romans and later re-used by the Muslims : their
operation lasted for several centuries (VILLANUEVA 1993,
References/bibliographie
1996). Of interest the model 3 design (with a 90-degree bend)
may provide a very-compact dissipation structure if the drop- ALLAIS, Y. (1933). "L'Alimentation en Eau d'une Ville Ro-
shafts are installed in a (square) spiral disposition. maine d'Afrique : Cuicul (Djemila)," ('The water Supply of a
The operation of the dropshaft cascades is not only a function of Roman City in Africa : Cuicul (Djemila).') 5th Congres Inter-
the dropshaft performances (see above) but it is also related to the national d'Archéologie, Alger, pp. 93-117 (in French).
connecting conduit geometry. At Recret (Yzeron aqueduct), the APELT, C.J. (1984). "Goonyella Railway Duplication Drop
conduits had a mild slope (9 ~ 0.1 to 0.6°), and a hydraulic jump Structures and Energy Dissipators at Culvert Outlets. Model
took place between dropshafts, as sketched in Figure 1. Similarly Studies." Report CH27/84, Dept. of Civil Engineering, Uni­
the connecting conduits had a mild slope at Montjeu (Brisecou versity of Queensland, Australia, Feb., 10 pages, 11 figures &
forest), Cuicul and Köln aqueducts. At Cerro de los Pinos 37 plates.
(Valdepuentes aqueduct), the channel slope was steep (about 2.9°) ASHBY, T (1935). "The Aqueducts of Ancient Rome."
leading to a supercritical flow in the conduits. The effect of a Clarendon Press, Oxford, UK, edited by LA. RICHMOND,
supercritical inflow was investigated in the model 2. The results 342 pages.
showed that the inflow conditions affected the type of flow re­ BLACKMAN, D.R. (1978). "The Volume of Water Delivered by
gime particularly at low flow rates (e.g. Eq. (1) and (2)). How­ the Four Great Aqueducts of Rome." Papers of the British
ever, for a given flow regime, the inflow Froude number had little School at Rome, Vol. 46, pp. 52-72.
if no effect on the rate of energy dissipation, shaft aeration and CARTON (1899). "Notices sur les Fouilles Exécutées a Dougga."
pool level. [Slightly better energy dissipation was observed with Bulletin Trimestriel de Géographie et d'Archéologie de la
supercritical inflows with Fr > 10 in regime R3.] S.G.A.P.O., Vol. 13, pp. 157-159 (in French).

10 JOURNAL OF HYDRAULIC RESEARCH. VOL. 40, 2002. NO. I

 CHANSON, H. (1995). "Hydraulic Design of Stepped Cascades, L O P E Z - C U E R V O , S. (1985). "Medina Az-Zahra Ingeniera y
Channels, Weirs and Spillways." Pergamon, Oxford, UK, Formas." Publicaciones del Ministerio de Obras Publicas y
Jan., 292 pages. Urbanismo, Madrid, Spainl69 pages (in Spanish).
CHANSON, H. (1997). "A Short History of Stepped Cascades in P I N E T T E , M., and R E B O U R G , A. (1986). "Autun (Saone-et-
Australia." ANCOLD Bulletin, No. 106, Aug., pp. 101-111. Loire) Ville Gallo-Romaine." Guides Archéologiques de la
CHANSON, H. (1998). "The Hydraulics of Roman Aqueducts : France, Imprimerie Nationale, Ministère de la Culture et de la
Steep Chutes, Cascades and Dropshafts." Research Report No. Communication, France (In French).
CE156, Dept. of Civil Engineering, University of Queensland, RAJARATNAM, N., M A I N A L I , A., and H S U N G , C.Y (1997).
Australia (ISBN 0 86776 775 3). "Observations on Flow in Vertical Dropshafts in Urban Drain­
CHANSON, H. (2000). "Hydraulics of Roman Aqueducts : Steep age Systems." Jl of Environmental Engrg., ASCE, Vol. 123,
Chutes, Cascades and Dropshafts." American Jl of Archaeol- No. 5, pp. 486-491.
ogy, Vol. 104, No. 1, Jan., pp. 47-72. RAKOB, F (1974). "Das Quellenheigtum in Zaghouan und die
C H A N S O N , H., andTooMBES, L. (1998). "Supercritical Flow at Römische Wasserleitung nach Karthago." Mitt, des Deutschen
an Abrupt Drop : Flow Patterns and Aeration." Can. Jl of Civil Archaeologischen Instituts Roemische Abteilung, Vol. 81, pp.
Eng., Vol. 25, No. 5, Oct., pp. 956-966. 41-89, Plates 21-76 & Maps (in German).
C O N S E I L G É N É R A L D U R H O N E (1991). "Préinventaire des Mon­ R A N D , W. (1955). "Flow Geometry at Straight Drop Spillways."
uments et Richesses Artistiques. II L'Aqueduc Romain de Proceedings, ASCE, Vol. 81, No. 791, Sept., pp. 1-13.
l'Yzeron." ('Pre-inventory of the Monuments and Art Trea­ V E R T E T , H . (1977). "Observations sur les Aqueducs deRusicade
sures. II The Roman Aqueduct of the Yzeron.') Bosc Frères (Algérie)." ('Observations on the Aqueducts of Rusicade
Downloaded by [Erciyes University] at 09:39 22 December 2014

Pubi, Lyon, France, Henri HOURS Editor, 168 pages & 1 (Algeria).') Journées d'Etude sur les Aqueducs Romains,
Map (in French). Lyon, France, 26-28 Mai (also Les Belles Lettres Pubi, Paris,
CONSEIL G E N E R A L DU R H O N E (1996). "Préinventaire des France, 1983, pp. 349-369) (in French).
Monuments et Richesses Artistiques. IV Lyon. L'Aqueduc V I L L A N U E V A , A.V. (1993). "El Abastecimiento de Agua a la
Romain du Gier." ('Pre-inventory of the Monuments and Art Cordoba Romana. I : El Acueducto de Valdepuentes." ('The
Treasures. IV Lyon. The Roman Aqueduct of the Gier.') Bosc Water Supply of the Roman Cordoba. 1 : Aqueduct of
Frères Puhl, Lyon, France, Jean BURDY Editor, 407 pages & Valdepuentes.') Monografias No. 197, Universidad de Cor­
1 Map (in French). doba, Servicio de Publicaciones, Cordoba, Spain, 172 pages (in
C O Q U E T , M. (1966). "Les Aqueducs Romains deBeaulieu etles Spanish).
Puits de Rupture de pente." ('The Roman Aqueducts of V I L L A N U E V A , A.V. (1996). "El Abastecimiento de Agua a la
Beaulieu and the Stepped Cascades.') Cahiers Ligures de Cordoba Romana. II : Acueductos, Ciclo de Distribución y
Préhistoire et d'Archéologie, Vol. 15, pp. 283-295 (in French). Urbanismo." ('The Water Supply of the Roman Cordoba. I I :
ERVINE, D.A., and A H M E D , A.A. (1982). "A Scaling Relation­ Aqueduct, Distribution System and Urbanism.') Monografias
ship for a Two-Dimensional Vertical Dropshaft." Proc. Intl. No. 251, Universidad de Cordoba, Servicio de Publicaciones,
Conf. on Hydraulic Modelling of Civil Engineering Structures, Cordoba, Spain, 222 pages (in Spanish).
BHRA Fluid Eng., Coventry, UK, paper E l , pp. 195-214. VISCHER, D., and HAGER, W.H. (1995). "Energy Dissipators."
FEVRIER, PA. (1979). " L Armee Romaineet la Construction des IAHR Hydraulic Structures Design Manual No. 9, Hydraulic
Aqueducs." ('The Roman Army and the Construction of Aq­ Design Considerations, Balkema Pubi., Rotterdam, The Neth­
ueducts.') Dossiers de l'Archéologie, Séries Les Aqueducs erlands, 201 pages.
Romains, Vol. 38, Oct./Nov., pp. 88-93 (in French).
G R E W E , K. (1986). "Atlas der Römischen Wasserleitungen nach
Notation
Köln." ('Atlas of the Roman Hydraulic Works near Köln.')
Rheinland Verlag, Pulheim, Germany (in German). B dropshaft width (m);
HODGE, A T . (1992). "Roman Aqueducts & Water Supply." b open channel width (m);
Duckworth, London, UK, 504 pages. Cd drag coefficient;
JAIN, S.C., and K E N N E D Y , J.F (1983). "Vortex-Flow Drop D 1-conduit diameter (m);
Dtructures for the Milwaukee Metropolitan Sewerage District 2- conduit height (m);
Inline Storage." IIHR Report No. 264, Iowa Inst. of Hyd. Res., D ab bubble penetration depth (m) measured vertically from
University of Iowa, USA.. the free-surface;
LEVEAU, P. (1991). "Research on Roman Aqueducts in the past d flow depth (m) measured perpendicular to the channel
Ten Years." Future Currents in Aqueduct Studies, Leeds, UK, bed;
T HODGE ed., pp. 149-162. dc critical flow depth (m); in a rectangular channel:
L E V E A U , P., and P A I L L E T , J.L. (1976). "L'Alimentation en Eau
de Caesareade Maurétanie et 1'Aqueduc de Cherchell." ('The dc=V^g7'
Water Supply of Caesarea of Mauretanie and the Cherchell Fr Froude number; for a rectangular channel:
Aqueduct.') Librairie Edition L'Harmattan, Paris, France, 183 Fr = V/V^*d = Q/\/g*d3*b2 ;',
pages & 10 plates (in French). g gravity constant (mis');

JOURNAL OF HYDRAULIC RESEARCH, VOL. 40. 2002. NO. I II

 H total head (m); Vh brink flow velocity (m/s);
H„s residual head (m) : H res =H, - AH; x horizontal distance (m);
H, upstream total head (m); yp free-surface height (m) in a shaft pool above the down­
h invert drop (m) at a vertical dropshaft; stream conduit invert;
L dropshaft length (m);
P (shaft) pool height (m), measured from the shaft bottom Greek symbols
to the downstream conduit invert; AH head loss (m) : i.e., change in total head;
Q total volume discharge (m 3 /s) of water; 7i 71 = 3.141592653589793238462643;
Q„u, maximum flow rate ( m / s ) of the aqueduct; 8 bed (invert) slope;
Q' dimensionless discharge number; 0 diameter (m);
1-Q' = cWg*b 2 *D 3 for rectangular channels;
Abbreviations
2-Q' = QA/g*Jt2*D-s/i6 for c i r C ular conduits; Rl flow regime Rl (Fig. 4A);
q discharge per meter width (m~/s); for a rectangular chan­ R2 flow regime R2 (Fig. 4B);
nel : q = Q/b; R3 flow regime R3 (Fig. 4C).
v flow velocity (m/s);
Downloaded by [Erciyes University] at 09:39 22 December 2014

12 JOURNAL OF HYDRAULIC RESEARCH, VOL. 40. 2002. NO. 1