THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

DRAIN DESIGN FOR DRY FEET

Virendra PROAG University of Mauritius vprog@uom.ac.mu

Abstract
Many areas in Mauritius get flooded regularly due to (a) sudden rain of unexpected magnitude (b) building permits given on historically floodable land (c) inadequate drains. Rainfall frequency and intensity records can be used to estimate the magnitude of rains and the ensuing flood flows. There is a 26 % probability that a 100 year rain will occur during the next 30 years (a generation). Even if a higher return interval, e.g. 1,000 years is taken, it is found that there is 7.2 % chance (not to be neglected as being small) that a 1,000 year flood will occur during a 75 year span a mans lifetime. In a small country like Mauritius, it is difficult to give flood warnings in advance that one can abandon house and move furniture. The flooding might occur suddenly in the middle of the night, when there has been a power cut. The only solution to have dry feet would be to have adequate drainage. Is it acceptable for ones house to get flooded every 10 years? Or every 30 years? Or never at all during ones lifetime? The three alternatives will require drains of different sizes, with different costs. Once the desired safety from floods has been accepted preferably through legislation it would be easy to (1) design for the adequate drain capacity

(2) earmark the boundary of the reserved low lying areas reserved for extreme flood conditions. This paper presents the relevant criteria to adopt for drain design in Mauritius.

1.

Introduction

There are many places which receive heavy rains without anybody being aware of the fact, simply because no area gets flooded. In other places, however, there are many tell-tale signs, either during or after the heavy rains. The signs noticed afterwards indicate the levels to which the water levels rose during the peak of the storm. If the existing drains are unable to carry the flood peaks generated, then people do notice the flooding of the surroundings, sometimes with devastating results and loss of life and material damage. A sudden heavy rainfall will cause flooding if there is no drain to carry the water away. A low lying area will certainly be flooded because all water will eventually accumulate there and it is usually difficult to make drains which are at a still lower level. The area nearby is also likely to be affected in case the drain, if any, has an inadequate carrying capacity. Flooding seems to be a regular phenomenon which occurs in many countries for several reasons, namely: (a) building permits given on historically floodable land (b) inadequate drains (c) sudden rain of unexpected magnitude It is therefore important that the catchment area of the urban environment be studied for the low lying areas and the natural draining channels. These areas

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

should in the first instance be completely avoided for building purposes.

2.

Outline Planning Schemes

An outline planning scheme aims to define zones where different activities are allowed, including housing. The main parameters that are considered are environmental, but rarely do we find drainage being taken into account because of possible impacts of its non existence. If there were no floods during some 50 years in living memory (or sometimes even during the last 40 years average service time of a building permit officer), it is reasonably felt that there is no danger of any big flood occurring in the area. Very often, there are historical records which can confirm that the given area had been flooded so many years ago sometimes, 300 years or 500 years. Unfortunately, it is not always easy to go through these records or to check them. Thus, building permits are very often given on land which, according to historical records, is prone to flooding. Building lots have often been earmarked by the land promoter within drainage channels low lying contours. While these should be a constraint against giving the building permit, political pressure or an unwary building permit officer may come in the way. At other times, a whole set of houses have been built in a low lying area which could very well form a lake if a regular means of feeding it with water was available. In this case, it is usually difficult to make drains which are at a lower level. Very often, flooding occurs because the drain, if any, has an inadequate carrying capacity, or has a carrying capacity which has not been designed to take sudden heavy rainfalls Sometimes, the drains are permanently inadequate as one road engineer explained, he designed road drains to cater only for the rainfall coming from the road. The drains coming from the nearby building lots were not supposed to discharge into his road drain !!!.

While the above explains how building permits wrongly given or drains with inadequate capacity allow flooding to occur, it is judicious to examine sudden rainfall. If intense rainfall magnitudes can be estimated, this could help in designing the appropriate drains. Thus, before designing drains to carry flood flows, it is necessary to determine the magnitude of the flood flows.

3.

Determination of Flood Flows

One approach to the problem is as follows: Walk over survey of the area Obtain local historical flooding levels from the residents Collect data Analysis of collected data to estimate flood flows Estimate size of channel sections under the bridge

4.

Walk Over Survey

Site visits undertaken on the existing or nearby regions will enable meeting people, sometimes old, who recollect what they (or their grandparents) saw during flood conditions - the flood levels observed. A backflow analysis may help in crosschecking the present flood estimates. These are certainly of use, as a check, during design work.

5. 5.1

Data Collection General Approach

How often have promoters accepted a consultants offer to design drains capable of carrying flow with a return period of 2 years? Many codes of practice indicate a good guideline to design drains, etc with a return period of 50 years. This section refers to the catchment parameters which will enable determining some further factors

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

needed for the calculation of floods. In a first stage, the catchment area, slope (= elevation difference/stream length) are required to find the time of concentration. The peak flood flow is given by the relation Qp = CiA, adjusted (for the units given) to Qp = 0.278 C i A Where C = runoff coefficient i = rainfall intensity (mm/hr) A = drainage catchment area (km2) (m3/s) Qp = Design Discharge

The runoff coefficient is a function of the vegetation, urbanisation and other factors of the catchment. The rainfall intensity to be used depends on the time it takes the whole catchment to contribute to the flow in the drainage channel. These parameters are discussed below.

Figure 1 : How rainfall is shared among different components Figure 1 shows the process of rainfall, wherein rainfall (or precipitation when it includes hail, snow, etc) is the sum of the ensuing evaporation, infiltration and runoff. The lands surface always has a slope, however small it might be, which determines the direction of flow (here, the runoff). Figure 2 indicates how the ridge at the top of a valley slope will divide rainfall, which will run along slopes of either side of the ridge. The area enclosed by a given ridge determines a catchment area. Depending on the point of interest, the catchment area will vary. Point X determines a smaller catchment area than point Y, and it turn area at point Y is smaller than that governed by point Z. Eventually, the estuary governs an even bigger area.

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

Figure 2: The catchment area gets bigger downstream of the valley

So, this diagram illustrates how rain from the valley will run to a low point. Therefore, unless a drain has been specifically designed to take this rainwater, it will run into the drain besides the road, even if the engineer wrongly believed that only water from his road would run into the road drained he designed to take water, just from the road. And, if there are no road drains, the road itself will act as a welldesigned drain. The recent heavy rainfalls in Port Louis and in other places bear good testimony to this phenomenon.

Figure 3: The estuary is the lowest and final exit drainage point

10

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

Therefore, one first rule to avoid flooding is to make sure that the catchment area of the drain being designed is not underestimated. Not rocket science, but how often ignored by engineers and planners! Figure 3 gives an overall picture of a valley (with smaller valleys inside) and indicates how everything discharges into the lowest point which happens to be the estuary. In this connection, there is a parallel with traffic flow. Unless the conveying capacity QOUT is greater than the incoming flow of traffic QIN, there is going to be a traffic jam. While this results in a halt or lower speed in case of vehicles, unfortunately with water, this higher inflow leads to non-stopping flow which results in overtopping the drain and flooding the sides. Therefore, another rule to avoid flooding would be to make sure that the carrying capacity COUT of the channel drain exceeds the peak discharge Qp . COUT Qp

Fy

= frequency factor = 1.20 for a 100-year return period.

The runoff coefficient thus works out (for this return period) to be C = 1.2 (0.45 + 0.20 x 1) = 0.78

Different authors give other formulae or tabulated values, depending on soil cover. If Figure 1 is examined again, several observations may be made: The equation, Rainfall = Evaporation + Infiltration + Runoff while holding true in all cases, does not indicate that runoff or any of the other parameters are constant, though they may be taken to be taken as an average over the year and so on. For example, during a hot sunny day, imagine that some rain falls. As the rain drops touch the ground (soil or road surface) water vapour can be seen to rise in the air. Evaporation is actually occurring! If it is a light rain, the ground surface will be seen to dry up quickly. Either all rain water evaporates on the road or some of the rain is absorbed into the earth : infiltration is taking place. The end result is, however: there is NO runoff! At the other end of the scale, even under the same sunny conditions, if there is a heavy rain, there will a substantial runoff towards a low point (drain, river, pond), because the soil has reached its infiltration capacity. The soil is momentarily saturated. The ratio of this runoff to the measured rainfall is the runoff coefficient. Again, this runoff coefficient may be measured as an average over a period of time, or at every instant or over very short intervals.

Simple logic, but how often ignored!

5.2

Runoff Coefficient Value

The runoff coefficient C represents the ratio of a peak flow and rainfall rate of selected duration determined or the same average recurrence interval from frequency analysis of flood peaks and rainfalls. There are various factors affecting the runoff coefficient which must be considered. In consideration of these, the Institution of Engineers, Australia (Abbey, 1999) recommends that the runoff coefficient C be taken as
C = Fy (0.45 + 0.20 fi) Where fi case. = impervious factor, taken as 1 as a worst

11

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

Typically, it is usual to give the average over a long interval of time for the runoff coefficient. However, for those people who have experienced cyclonic conditions, the situation is different. When there are heavy rains, in fact, rain might be falling continuously/on and off, during several days. The soil is now saturated over a longer period, and this can be felt even outside cyclonic conditions. Imagine now a sudden, heavy rainfall under these conditions. This will just be runoff. There will be NO infiltration (saturated soil) and little evaporation (the air is saturated with water vapour). So now, the runoff coefficient C = 1, taken as 1 is a worst case, that needs to be considered. Although this might be difficult to swallow, it is judicious to examine the situation in the light of actual experience. Mauritius is a tropical island with tropical heavy rains, not a desert where it rains 20 mm per year !! If, on top of that (as in Port Louis), the ground surface consists of clayey soil or is mostly paved, again the runoff coefficient is going to be C = 1. This is the third rule to consider: In tropical countries, take C = 1 This factor will increase the design flow to be Table 1: Typical Rainfall Intensities (mm/h)
Duration (mins) T = 100 years Seychelles Mauritius

underestimate the loading conditions (here, the possibility that the rain will not infiltrate at all, nor evaporate, is real. It does happen.). Furthermore, the drain is expected to be effective under extreme conditions, not only when it rains slightly.

5.3 Intense Rainfalls

Normal rainfalls do not cause flooding to occur. So a serious study of flooding needs to consider intense rainfalls. The worlds greatest recorded rainfalls, according to the World Meteorological Organisation are approximated by the equation P = 422 Td0.475 Where P = the rainfall (precipitation) depth in millimetres Td = the duration in hours The equation was obtained by fitting data from observed extreme rainfalls at many locations for durations ranging from one minute to several months. This equation is an estimate of the precipitation depths that could occur under very extreme circumstances. If Td is taken as 1 hour, the rainfall is 422 millimetres. Something to think about! Fortunately, the rainfall records in Mauritius do not indicate such extremes in Mauritius, but heavy rains with 100 mm/hr over an hour or so are not uncommon (89 mm at Dubreuil on 22nd December 1979 over 1 hour , 310 mm at La Brasserie over 150 minutes on 6th February 1992 and more recently 91 mm at Line Barracks, Port Louis, between 2 and 3 p.m, on 30th March 2013. The rainfall collected over the first half hour was 50 mm which amounts to an intensity of 100 mm/hour)

30 mins 60 Duration (mins) T = 50 years 30 60

150 120

120 100

140 105

110 90

5.4

Rainfall Intensity and Frequency

considered, for sure. However, though the engineers job is to do an economical design, he should not

To introduce the subject, a 100 m sprinter runs at a speed of 36 km/h on a 10 second race, but the average speed is much lower (22 km/h) when another

12

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

runner (or the same person) undertakes a 10,000 m marathon. Similarly, though a rainfall may last several hours (long distance race), the critical condition to observe in drain design is the highest rain intensity (highest speed over a short distance race). The rainfall intensity i is the average rate of precipitation in mm/hr from a storm having a duration equal to the time of concentration. It is assumed that runoff due to a heavy rainfall will reach a peak at the time of concentration when the whole catchment is contributing to flow. Then, the duration of the design storm is equal to the time of concentration. The time of concentration tc is thus the time taken for water to travel from the catchment boundary to the point of interest (Points X, Y or Z or estuary) in Figure 2). For small, steep areas, (e.g. Mauritius, Seychelles), the Kirpich formula has been found to give reasonable estimates for tc. In this formula, tc = 0.01947 L0.77 S -0.385 where = time of concentration in minutes tc L = maximum length of travel of water (m) and S = slope of catchment = H/L in which H = difference in elevation between the most remote point on the catchment and the outlet. Mays (2004), Reddy (2008) and Rmniras (1986) give other similar formulae, applicable in different conditions. Rainfall intensity-frequency-duration curves are usually derived by the countrys Meteorological Services. The rainfall intensity (mm/ or mm/min) figures are available for different periods of time, such as 5, 10, 15, 20, 30, 60 and 120 minutes. Table 1 shows examples of rainfall intensities for Seychelles and Mauritius.

The time of concentration rarely falls exactly on the duration time for which figures have been provided by the Meteorological Services. A judicious interpolation helps. As previously indicated, a 50 year return period is a good guideline, but sometimes the designer might feel that a 100 year return period might be better. For example, a bridge (Bindra, 1975) is a structure that is expected to be in operation during a very long period. In this context, it is natural to consider events with a return period of 50 years or even more. There are so many bridges in the world which have been standing for more than 50 years. A 100-year rainfall has a 1% chance of occurring in any single year. This issue will be discussed below. 6. Is a Return Period of 50 years Acceptable? The results obtained from the above calculations can prove to be very important in the design of hydrological structures such as bridge culverts and channels to drain the area under consideration and prevent flooding. Every structure is designed for a certain design life and it must be ensured that this structure serves for its purpose without endangering any life and property. The risk that failure of such a structure occurring during its design life has been explained by Mays (2004) as follows: Let P be the probability of the occurrence of an event, 1 P = probability that the event will not occur (1 P)(1 P) = probability that the event will not occur in two successive years. (1 P)(1 P)(1 P) = probability that the event will not occur in three successive years. = probability that the event will not (1 P) N occur during a span of N successive years. Hence, the risk, R or the probability that the event will occur during a span of N years is given by, R = 1 (1 - P) N The probability P is given by P = 1/Tr. Table 2 shows, for return periods Tr and various spans of years N,

13

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

Table 2: Risk R, that a flood of a given return period will be equalled or exceeded during periods of various lengths. Return Risk R for various spans of N years Period Tr (years) 1 5 10 50 100 500 1,000 5,000 10,000 5 1.0 0.67 0.41 0.10 0.05 0.01 0.005 0.001 0.0005 10 1.0 0.89 0.65 0.18 0.10 0.020 0.010 0.002 0.001 30 1.0 1.0 0.96 0.45 0.26 0.058 0.03 0.006 0.003 50 1.0 1.0 0.995 0.64 0.40 0.095 0.049 0.010 0.005 75 1.0 1.0 1.0 0.78 0.53 0.14 0.072 0.015 0.0075 100 1.0 1.0 1.0 0.87 0.63 0.18 0.095 0.020 0.0099 200 1.0 1.0 1.0 0.98 0.87 0.33 0.18 0.039 0.020 500 1.0 1.0 1.0 1.0 0.99 0.63 0.39 0.095 0.049

the risk R that a flood with certain return period will be equalled or exceeded during periods of span N years. Rainfall frequency and intensity records can be used to estimate the magnitude of rains and the ensuing flood flows. In this respect, it is important to note that there is a 26 % probability that a 100 year rain will occur during the next 30 years (a generation). In practical terms, this means that each generation has a 1 in 4 chance of experiencing flooding, even if an exceptional (?) rainfall intensity of 100 year has been considered. Over a 75 year lifetime, the likelihood rises to 0.53, i.e., the average person has a 1 in 2 chance of experiencing flooding during his lifetime. Is the population ready to accept this? Even if a higher return interval (e.g. 1,000 years) is taken, it is found that there is 7.2 % chance (not to be neglected as being small) that a 1,000 year flood will occur during a 75 year span a mans lifetime. It can be noticed that the risk that an event is reached or exceeded for a certain span of time, decreases with an increase in return period. This result is of-

ten used in the design of huge structures. There is also an increase in cost by considering the design of a structure for a long return period. However, this should be done to be safe from calamities causing loss of life and property. Many Codes of Practice indicate that one of the reasons for choosing a return period of 50 years has been that the average lifetime of most buildings and structures is near 50 years. This may have been true at one time. There are, however, other factors which need to be considered: (1) the use of better materials has increased the lifetime of the buildings and structures. Similarly, the corresponding drains or bridge culverts will have a longer life. (2) why should any owner, demolish his building after 50 years, if it is still serviceable? The Eiffel Tower was built in 1889, to be demolished just after the Universal Paris Exhibition. It is still standing and being regularly maintained. We have not yet seen any drain being demolished to be enlarged, except when they have really been shown to be inadequate. Even if a local authority tried to do so, it very likely that adjoining structures would prevent this.

14

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

There so many cathedrals and nearby bridges built in the eighteenth century in Mauritius (thirteenth century in Europe) still standing today. Would any present day designer still consider a 50 years lifetime for such monuments? (3) the cost of demolition becomes so high that the owner is likely to push the time limit before he has to really bring down the structure.. If these factors are considered, what return period should be considered? In dam hydrology, the notion of maximum possible flood (return period of 10,000 to 50,000 years, depending on authors) has made its appearance, for exactly the same reasons the possible danger to human life. It might be argued that with only some 100 years data or, in most cases, even less, it is difficult to make predictions (or wild guesses) about 10,000 years recurrence intervals. But, if a bridge culvert or drainage channel is needed now nobody will wait to collect another 50 years of rainfall data.

= 0.030 rivers in good condition

Design constraints are usually channel or river width and slopes, but the designer should try to see if other accompanying measures need to be taken. The choice is likely to be governed by minimum headway clearances under the bridge due (1) to the possibility of branches and trees being carried away and (2) other facilities passing under or by the side of the bridge. Some river training works might be necessary just upstream or downstream of the bridge. In this context, this formula is enlightening. The same channel will have different carrying or discharge capacities if any of the variables changes. A bigger cross sectional area will increase the channel capacity, but the effect will be attenuated if the roughness changes from a smooth, cement lining to a river in bad condition.

8.

The Case for Port Louis

7. Estimation of the Peak Design Discharge

At this stage, the peak design discharge may be calculated and hence used to design the drain or bridge culvert as the case may be. Once the design flow has been established, channel hydraulics may be used to design the channel or culvert. One typical carrying capacity formula is that of Manning

The motorway was flooded at Caudan between Rogers House and the waterfront on 11th February 2013, without much damage. There was a worse incident on 30th March 2013, with loss of life. The rainfall recorded, on 30th March 2013, at Line Barracks (less than 100 mm in 1 hour) would indicate, from Table 1, a return period of some 50100 years. However, the fact that there was a similar flooding at Place dArmes/Caudan (apparently without the underground pedestrian pathways getting submerged) on 11th April 2003 (Wright A., Moonien V., 2013) confirms the values of Table 2. A 50 year flood does not occur every 50 years! It will certainly occur during a period of 500 years, but may also occur within the next 10 years !! In the light of the above discussion, it is judicious to ask whether flooding can occur again, and how soon? The motorway from Montebello towards Port Louis is lined, practically on both sides with concrete borders or walls, which are supposed to be very

where COUT = flow in channel (m3/s) A = wetted area (m2) R = hydraulic radius (m) = wetted area/wetted perimeter (m) S = channel slope n = Mannings roughness coefficient = 0.010 smooth, cement lining = 0.013 good brickwork

15

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

Photo 1: Place dArmes. The arrows show the slope direction

Photo 2 : The Caudan Esplanade (right) is at a higher level than Place dArmes

16

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

effective against cars trying to rub into them. The walls are also provided, at regular distances, with weep holes, which are expected to evacuate water into the side drains. While these weep holes can be very effective in evacuating low flows, their small size (some 30 x 10 cm to 40 x 15 cm) becomes inadequate when heavy rains and winds bring in their loads of gravel, leaves, and mud. When these weep holes are blocked, the bituminous motorway becomes a very well designed, bitumen lined channel, which was well evidenced during the heavy rains of 30th March 2013. The motorway was conveying water which was supposed to be evacuated into the side drains. At end of July 2013, the weep holes are still of the same size! Between Edith Cavell street and the Government House, the lowest points in Port Louis occur along the La Poudrire street. Rightly so, the two channels Le Pouce stream and La Butte Thonnier canal are located on both sides of this road. The ground also has a downstream slope towards the sea. This means that any rainfall will be channelled towards these two canals/channels and towards the sea. The only problem is that at the level of the Harbour Front and Place dArmes, there is an uprising obstacle (Photos 1 and 2) in the form of the motorway and the Caudan Esplanade. This now implies, that should the peak discharge flow from heavy rainfall exceed the discharge capacity of the channels, the flood waters will not go directly towards the sea, unless and until they have overtopped the motorway and the Caudan waterfront Esplanade. Of course, with a consequential ponding of the area between the Port Louis museum and the Place dArmes. Again, this is simple logic, borne out by the events of 11th April 2003 and 30th March 2013. Even assuming that the motorway constitutes a roadblock in the evacuation of rainwater from Place dArmes, historical records (Chelin 1989) show that floods have occurred several times, prior to the construction of the motorway. This implies that

the existing canals/streams are not enough or are inadequate to evacuate the water reaching Place dArmes in case of heavy rainfall. So, knowing that a rainfall of intensity 100 mm/ hr is not uncommon (see examples and values Meteorological Services Table 1), have we proposed any new canals to evacuate more water?

9.

Conclusion

This study has proposed an approach to be adopted prior to the approval of planning or zoning schemes with respect to possible flooding. Rule 1: Do not underestimate the catchment to be drained, particularly when designing roads. The area, A, is much bigger than the road itself. Rule 2: In a tropical country like Mauritius, take C = 1, to cater for extreme conditions when the soil is saturated. Rule 3: Determine the rainfall intensity, i, using the proper and adequate return period, which is commensurate with what the population expects from engineers for leading a comfortable life. Rule 4: Determine the peak discharge Qp from the equation Qp = 0.278 C i A. The size (width and height) of the channel must consider the possibility of avoiding blockage by shrubs, leaves and trees during cyclones. Rule 5: Design the drain carrying capacity

Rule 6: Check that

COUT

Qp

It has been argued that a 50 year return period is probably too low and higher return periods should be taken, given the relatively high probability of occurrence during a mans lifetime. Once the desired safety from floods has been accepted preferably through legislation it would be easy to earmark the boundary of the reserved low lying areas reserved for extreme flood conditions. This should

17

THE JOURNAL OF THE INSTITUTION OF ENGINEERS MAURITIUS

ensure that houses do not get flooded regularly. It is essential that such guidelines and low lying boundaries be properly adhered to, particularly when establishing planning zones. A discussion of the flooding occurrences in Port Louis, before and after the construction of the motorway in the 1970s, tends to highlight a possible inadequacy of the existing drainage exits into the sea.

References
Abbey, P. 1999, Storm Water Drainage Design Guidelines, report for Government of Seychelles, Ministry of Environment and Transport. Bindra S. P. 1975, Principles and Practice of Bridge Engineering. Dhanpat Rai, New Delhi. Chelin A. 1989, Maurice : Une le et son pass. Editions du CRI, Ile de la Runion. Mays L. W. 2004, Water Resources Engineering. John Wiley, USA. Moonien V. 2013, Inondations: Port Louis dj sous leau 2003. LExpress, 11th August 2013, p. 20. Parker D. J. 1998, Warnings for Torrential Rain and Floodings in Mauritius. Recommendations for the Government of Mauritius. Proag V. 1995, The Geology and Water Resources of Mauritius. Mahatma Gandhi Institute, Mauritius. Reddy P. J. R. 2008, A Textbook of Hydrology. University Science Press, New Delhi. Rmniras G. 1986, Lhydrologie de lingnieur. Eyrolles, Paris. Ven Te Chow., Maidment D.R., and Mays L. W. 1988, Applied Hydrology. McGrawHill, New York. Wilson E. M. 1990, Macmillan, England. Engineering Hydrology.

Wright A . 2013, Rien na chang pour les autorits . LExpress, 10th August 2013, p. 9.

18