Van Baars, S. (2005). Ge´otechnique 55, No. 4, 319–323

TECHNICAL NOTE The horizontal failure mechanism of the Wilnis peat dyke

S. VAN BAARS*

KEYWORDS: embankment; failure LOCATION OF WILNIS The village of Wilnis is part of the community of Ronde Venen (‘Round Peats’). It is about 30 km south-east of SAFETY OF SECONDARY DYKES Amsterdam. There is a ring canal next to the road, which More than 50% of the Netherlands is below the high-water goes from the lake near Vinkeveen, through Wilnis, towards level of the rivers or sea. To prevent these areas from the city of Mijdrecht (Fig. 1). The ring canal is an open flooding, dykes are built along the rivers and seas, with a connection between the lake and Mijdrecht, so that: total length of about 3200 km. These dykes are called primary dykes. The precipitation of the lower areas of the (a) rainfall can be pumped up into this canal and flow country drains into the many ditches crossing these flat freely towards pump stations lands. From the ditches the water is pumped up into canals (b) the large lake reduces fluctuations of the water level with water levels sometimes several metres higher than the (c) cruise boats can travel freely to the lake. land. From these canals the water is pumped up into the rivers, or in some rare occasions into the sea. The water in the canals is stopped from flooding the land by canal dykes, Most of the soil in these areas consists of peat layers. In which are called secondary dykes. There are about some areas the peat has been excavated for heating houses. 14 000 km of secondary dykes in the Netherlands. This peat extraction occurred from the Middle Ages up until Over the years many secondary dykes in the Netherlands a century ago, when coal was found in the south-east of the have caused problems. For example, in December 1874 country, but most of it took place in the 17th and 18th 70 m of peat dyke failed a few kilometres north of Wilnis centuries. by horizontal translation, after heavy rainfall and a high In Fig. 2 a circle marks the location of the dyke failure. water level in the canal (Grundmann, 1996); in July 1947 a The old part of Wilnis is south of the ring canal. The new peat dyke failed by horizontal translation in Zoetermeer housing quarter, named Veenzijde (‘Peat-side’), is the square during the warmest and driest summer of the century; in of land north of the ring canal, and is one of the excavated August 1990 a peat dyke failed by horizontal translation in peat areas. The secondary ring dyke between Veenzijde and Bleiswijk (near Rotterdam) after two very warm and dry the canal is the left-over between canal and peat-excavation, summers (Vonk, 1994); and in January 1960 a canal dyke This peat strip was heightened with a deposit of coarse peat. failed in the village of Oostzaan near Amsterdam, leaving The area south of the ring canal is not excavated and is many houses of the village flooded. Because of this the therefore as high as the crest of the failed ring dyke on the government’s Technical Advisory Board for Water Barriers other side. The ring dyke is the only barrier between the 2 (TAW) was set up in 1965 with the task of determining new housing quarter and the 10 km of water of the Vinke- which primary and secondary dykes were unsafe. They veense Plassen (‘Finch-Peat’s Lakes’, Fig. 1). reported first on the primary dykes and then, in 1993, on the secondary dykes, saying:

1730 km secondary dykes of the most important 200 polders have been surveyed. 323 km of dyke was unsafe. 167 km of dyke has been improved already, 156 km of dyke is still unsafe (date: 1 January 1 1993). (TAW, 1993a)

The Wilnis dyke was among these unsafe dykes. The water board for Wilnis, named after three important small rivers in this area, Amstel, Gooi and Vecht (AGV), proposed a large upgrading of these dykes. Many of these dykes were in bad shape, but, because of a lack of money, they could not all be improved. The Wilnis dyke was not on the list, because it is not a main dyke surrounding a polder district; it only crosses a district. These secondary dykes were planned to be upgraded later.

Manuscript received 18 May 2004; revised manuscript accepted 10 January 2005. Discussion on this paper closes on 1 November 2005, for further details see p. ii. * Faculty of Civil Engineering, Department of Geotechnical Engineering, University of Technology Delft, The Netherlands. Fig. 1. Wilnis connected with a large lake

319 320 VAN BAARS

Fig. 2. Wilnis along the ring canal (dark line from left to right)

Fig. 4. Shifted part of the canal dyke; north to the left THE WILNIS DYKE BREACH In the early hours of Tuesday 26 August 2003, at 1.30 am, a peat dyke failed along the ring canal near the centre of the village of Wilnis. The failure took place after the warmest and driest summer in 50 years. About 60 m of dyke was translated about 10 m towards the north (Fig. 3), leaving two breaches (Figs 4 and 5). Therefore the water in the canal started to run along both ends of the shifted section of dyke into the new housing quarter of the village. Fortunately, a local contractor immediately started to close off both ends of the ring canal in the east and the west, and a side canal in the south. Within a few hours the ring canal was closed off with clay, but by that time the 600 houses were already 0.5 m under water. The 2000 residents were evacuated in the early morning. Almost all the residents were able to return to their homes that evening, after the water had been pumped out of this area. The canal dyke normally has a fairly constant (maximum) water load, like all secondary dykes, which means that failure occurred not because of a Fig. 5. Shifted part of north dyke temporary higher load but because of reduced strength. Because of the failure of the north dyke, the water level in the canal dropped instantly. As a result the supporting horizontal water pressure on the south quay disappeared, causing this quay (below the line of trees in Figs 4–6) to fail by a circular slip surface, as shown by the rotated trees

Fig. 6. Circular slip surface failure of south quay (rotated trees and grass)

and pavement (Figs 6 and 7). Apart from the huge damage, the question remained as to what caused the dyke breach.

GEOTECHNICAL DATA At the time of failure there were no data available on the dyke, not even a dyke profile. The only reliable data were Fig. 3. Empty canal (left) and shifted dyke (right); north to the about the water levels in the ditch (NAP 6.00 m) and the right canal (NAP 2.15 m) (Note: NAP is roughly main sea THE HORIZONTAL FAILURE MECHANISM OF THE WILNIS PEAT DYKE 321 Because of the soft subsoil, the old houses in the south are founded on wooden piles, and the new houses in the north on precast concrete piles. The saturated unit weights shown are average parameters based on 12 samples per soil layer. The saturated fine peat samples have a water content between 500% and 900%. The amount of organic materials is between 80% and 90% of the unsaturated weight. Because of the large amount of water and organic materials, which are lighter than water, the 3 average unit weight is not more than ªsat ¼ 10.0 kN/m . The peaty clay contains between 15% and 25% organic materials. The water content is between 160% and 300%. The deep peat has a water content between 450% and 750% and an amount of organic materials between 75% and 85%. Both values are slightly less than those for fine peat: there- fore the saturated unit weight is slightly more. The weights Fig. 7. Circular slip surface failure of south quay (rotated of the fine and coarse peat samples taken from the dried-out pavement) dyke crest vary considerably between samples: between 1.5 and 8 kN/m3, depending strongly on the water content (between 100% and 500%). The average unsaturated unit 3 level). In heavy rainfall the water level in the canal can weight is about ªunsat ¼ 5 kN/m . reach the crest of the dyke (NAP 1.50 m). In 2002 a little The dyke contained a cut-off of wooden sheeting, with its water even flooded over this part of the dyke, which had to toe at NAP 6.5 m, so it reached only to the peaty clay. Its be stopped with sandbags for one or two days. position is shown in Fig. 9. This sheet pile wall is almost Two days after the dyke failure many data were collected waterproof, which makes it easer for the crest to dry out. and reported by a Dutch geotechnical organisation hired by The grass next to the sheeting had become yellow or had the water board (GeoDelft, 2003). Many piezometers were died. Water content measurements indicated a dried-out crest installed, but they were all near the empty canal. There were for the dashed area indicated roughly in Fig. 9. This area is 2 many cone penetration tests (CPT), most with pore pressure about Iunsat ¼ 12.8m . The coarse peat dried out more than meters. These all registered a hydraulic head of the deep the fine peat. Temperature-CPTs recorded a temperature of sand layer (NAP 6.00 m) equal to that of the ditch, but roughly 198 for the first metre of peat, dropping to a clearly 1–2 m higher towards the canal, which had a higher constant 128 at 3 m depth and below. water level before the dyke failure. Triaxial tests (CU) were carried out, but the horizontal The borings and CPTs near and in the dyke indicated a layered soil structure of the anisotropic peat requires hor- fairly constant soil layer profile. The top layer below the izontal direct shearing. Also, 12 direct shear tests were crest of the dyke is medium coarse peat, as shown in Fig. 8. carried out per soil layer, comprising three stress levels (10, Below NAP 4.0 m the complete dyke consists of fine 25 and 40 kPa) at four different locations. The tests showed (forest) peat. Below this peat layer there is a thin peaty clay a wide range of soil strength per location. The cohesion of layer, which lays on top of a denser peat layer. The clay the fine peat layer varied between 0.9 and 9.1kPa. An layer contains a lot of reeds, leaves and root particles and, average cohesion of c9 ¼ 5.7 kPa was found, but this has to probably because of this, has a higher permeability. Below be reduced, for the following reasons: the denser peat layer there is a very thick sand layer. (a) The strength is strain rate dependent; long-term loading gives a lower strength. Ground level (b) A slip surface can go from weak spot to weak spot, resulting in a total strength lower than the sum of the average values (progressive failure). Coarse 3 3 ãsat 5 10·0 kN/m ãunsat 5 5·0 kN/m c¢ 5 2·5 kPa ö¢ 5 25° peat (c) In design, average values are not allowed, but reduced (design) values. NAP 24·0 m (d) The cohesion was determined at higher effective stresses. At effective normal stresses, approaching zero, the cohesion of soft peat reduces to zero in direct shear Fine ã 5 10·0 kN/m3 ã 5 5·0 kN/m3 c¢ 5 2·5 kPa ö¢ 5 25° peat sat unsat tests: there is (almost) no cohesion left at low stress levels. NAP 26·5 m (e) Therefore the Dutch Dyke Standard (TAW, 1993b) . Peaty clay ã 5 11·7 kN/m3 c¢ 5 2·5 kPa ö¢ 5 25° prescribes a cohesion of c9 ¼ 1 5 kPa (long-term design sat value) below the crest of the dyke and a cohesion of NAP 27·5 m 0 kPa below the toe of the dyke, unless proven Deep differently by tests. ã 5 10·5 kN/m3 c¢ 5 5 kPa ö¢ 5 20° peat sat Taking all this into consideration, a cohesion of about c9 ¼ 2.5 kPa for long-term loading is estimated for the coarse and NAP 29·0 m fine peat at low stresses. The course and fine peat layers show an average angle of internal friction of 9 ¼ 258. Deep ã 5 20·0 kN/m3 Also, the peaty clay is very weak at low stresses. A sand sat handful of this clay can be squeezed easily through the fingers. The direct shear tests, carried out at high effective stresses, show that this clay was a little stronger than peat, but data for this clay layer found in the archive of the water Fig. 8. Soil layer profile and soil data board (GeoDelft, 2003) also show zero strength at zero 322 VAN BAARS l 5 39·7 m

3 m 2·3 2·6 m 12·8 m 14 m 5 m

NAP 21·50 NAP 22·15 NAP 22·40 NAP 5 New Amsterdam Level 3 NAP < mean sea level (ãunsat 5 5 kN/m)

timber G NAP 24·40 sheet pile wall 3 (ãsat 5 10 kN/m ) NAP 25·80 NAP 26·00NAP 25·90 P Area 5 hor NAP 26·50 2 14·02 m2 11·83 m2 39·68 m2 19·6 m 3·25 m2 11·5 m2 F 5 kPa

Pvert

43·5 kPa 11·7 m 15 10 Saturated crest

v Unsaturated crest ¢ 5

ó Uplift 0 0 5 10152025303540 25

Fig. 9. Cross-section and vertical effective stress stress. Also, the CPTs indicate an even lower strength of the (a) This is the toe level of the wooden sheeting. clay. The cone resistance of the fine peat layer is qc ¼ (b) This is the level of the bottom of the ditch. 0.3 MPa with a friction ratio FR (¼ fs/qc) of 7%. The cone (c) This is the level indicated by CPTs with reduced resistance of the peaty clay layer is 0.2 MPa with a friction friction ratio below the failed part of the dyke. ratio of 1.5%. This clearly shows that the clay is weaker (d) This is the level indicated by CPTs with reduced than the fine peat at low stresses: therefore no higher friction ratio near the former crest position. parameters are used for the peaty clay. According to four CPTs at different positions below the The forces reacting on the failed part of the dyke, and the failed part of the dyke, the friction ratio of the clay had dyke sizes, are more or less as shown in Fig. 9. The stability even dropped below 1.0% at about NAP 7.0 m. According of this dyke can be checked easily with a simple one page to five CPTs at the position where once the dyke crest had computation. been (see Fig. 3), the friction ratio jumped from 0.3% to values between 4% and 10% at about NAP 6.3m. HORIZONTAL AND VERTICAL STABILITY In Fig. 9, the total cross-section of the dyke above the HORIZONTAL SLIP SURFACE horizontal slip surface is The risk of a peat layer under a dyke has already been I 14:0 11:5 11:8 39:7 19:6 3:3 100 m2 described twice by Ward. In 1944 a dyke near Cardigan Bay tot ¼ þ þ þ þ þ ¼ in Wales, founded on very light (floating on salt water), (1) weak peat (cu ¼ 6.0 kPa) and soft silt (cu ¼ 5.7 kPa), collapsed shortly after construction during high tide (Ward, If the whole peat dyke were saturated, this part of the dyke 1948). In 1948 a dyke under construction along the river would weight (per metre): Don in England slid over a weak, thin peat layer (c9 ¼ G ¼ I ª ¼ 1000 kN (2) 4.9 kPa and 9 ¼ 188) (Ward et al., 1955). However, neither sat tot peat,sat failure had a purely horizontal failure surface, and both If the crest of the dyke dried out and became unsaturated, failures were driven by an increasing load, whereas the this would become Wilnis dyke failure had a constant horizontal load and a G ¼ G I ª ª ¼ 936 kN (3) decreasing vertical load. Horizontal sliding is one of the unsat sat unsat ðÞpeat,sat peat,unsat failure mechanisms mentioned in books of the Dutch Tech- nical Advisory Board for Water Barriers (TAW, 1993b, This is a weight reduction of only 6.4%. 1998). Therefore dykes had to be checked for this mechan- The exact pore water distribution in the dyke is unknown, ism. However, among many dyke experts at TAW the idea but a reasonable approximation will be a linear behaviour grew that the horizontal failure mechanism existed only in between canal and ditch. Based on this linear interpretation, theory, but not in reality. Apparently they did not think the resulting force due to the pore pressure underneath the about the horizontal failure of peat dykes in the past (e.g. dyke is Wilnis, Zoetermeer, Oostzaan and Bleiswijk): therefore this P ¼ 1 p þ p l ¼ 1 3 ðÞ43:5 þ 5 3 39:7 ¼ 963 kN failure mechanism was excluded in the TAW safety manual vert 2ðÞ1 2 2 (TAW, 1999). In 1999 the water board decided to follow the (4) standards of this latest safety manual (DWR, 1999). The failure mechanism of the Wilnis dyke in 2003 was a This means that the peat dyke can weigh less than the horizontal translation of the dyke. There are four types of resulting water force when the crest of the dyke dries out. A evidence for a horizontal slip surface at about NAP 6.5m: sudden rise of the water level in the canal for a few days THE HORIZONTAL FAILURE MECHANISM OF THE WILNIS PEAT DYKE 323 Table 1. Safety factors for four different scenarios

Water level Peat Phor Pvert GFmax SF (crest) Normal Sat. 93 963 1000 116.51.25 Normal Unsat. 93 963 936 86.70.93 Crest Sat. 124 963 1000 116.50.94 Crest Unsat. 124 963 936 86.70.70

will hardly affect this force, as the peat and clay layers are the cohesion and making it easier for the dyke to fail in fairly impermeable. 2003 during drought. The effective vertical stresses for a saturated crest and an The combination of a higher water level just after drought unsaturated crest are indicated in Fig. 9. Even when the is even more unsafe, but has never occurred yet. With the crest is saturated, there is an area with uplift. When the crest safety factors calculated here, it is evident that the dyke had dries out there is little pressure left. Only 11.7 m out of to collapse one day. 39.7 m has no uplift. This shows that the Wilnis dyke was very unstable. The horizontal force with a common water level (as at the NOTATION time of failure) is approximately c9 effective cohesion of soil c undrained shear strength of soil 1ª 2 1ª 2 u Phor ¼ 2 h1 2 h2 fs sleeve friction of CPT FR friction ratio (¼ fs/qc) 1 : : 2 1 : 2 G total weight (of dyke) ¼ 2 3 10 3 ðÞ6 5 2 15 2 3 10 3 0 5 Itot total cross-sectional area ¼ 93 kN (5) P resultant force due to (pore) water pressure SF safety factor (¼ Fmax/Phor) When the water level rises suddenly at the crest, this force qc cone tip resistance of CPT will increase to ªsat unit weight of saturated soil ªunsat unit weight of unsaturated soil 1 : : 2 1 : 2 Phor ¼ 2 3 10 3 ðÞ6 5 1 5 2 3 10 3 0 5 ¼ 124 kN 9 angle of internal friction of soil (6)

The active soil pressure below the bottom of the canal on REFERENCES the wooden sheeting can be neglected because there is DWR [Department of Water Control and Sewage Systems] (1999). Nota: Groot onderhoud boezemwaterkeringen. Amsterdam: Mu- hardly any effective stress. nicipality Amsterdam and Water Board AGV. The horizontal shear resistance can be described with . GeoDelft (2003). Wilnis Kadeverschuiving, Reports CO-411242-22, c9 ¼ 2 5 kPa and 9 ¼ 258: -25 and -30. Delft: GeoDelft. F ¼ l c9 þ tan 9 ðÞG P (7) Grundmann, P. (1996). An incident during the making of the polder max vert in 1874. De Proostkoerier, No. 3, 4–9. If the safety factor is described as TAW (1993a). Systematisch Kade-onderzoek: de resultaten. Delft: Technische Adviescommissie voor de Waterkeringen. F SF ¼ max (8) TAW (1993b). Technisch rapport voor het toetsen van boezemkaden, Phor pp. 32, 52.Delft: Technische Adviescommissie voor de Water- . keringen. then failure occurs when SF , 1 00. Table 1 shows the TAW (1998). Grondslagen voor waterkeren, fig. 7.2, p. 82. Delft: safety factors for the different scenarios for water level and Technische Adviescommissie voor de Waterkeringen. crest water content. TAW (1999). Leidraad Toetsen op veiligheid (Safety Manual), fig The common situation with a saturated crest and normal 3.1.1, p. 72. Delft: Technische Adviescommissie voor de Water- water level is stable. However, the safety of the peat dyke in keringen. Wilnis can drop drastically by a loss of a few percent weight Vonk, B. F. (1994). Some aspects of the engineering practice caused by drought. This was the case in August 2003, and regarding peat in small polders. In Advances in understanding this is what led to the horizontal shear failure. and modelling mechanical behaviour of peat (eds E. den Haan, R. Termaat and T. B. Edil), pp. 389–402. Rotterdam: Balkema. The scenario with the water level at crest height after Ward, W. H. (1948). A slip in a flood defence bank constructed on heavy rainfall occurred for a short period in 2002. Accord- a peat bog. Proc. 2nd Int. Conf. Soil Mech. Found. Engng, ingly, and considering the dyke failure of 1874, the corre- Rotterdam 2, 19–23. sponding safety factor was dangerously low at that time. It Ward, W. H., Penman, A. & Gibson, R. E. (1955). Stability of a is possible that the peat fibres were damaged then, reducing bank on a thin peat layer. Ge´otechnique 5, No. 2, 154–163.