A rainfall protocol for Dawlish sea cliffs is helping protect a mainline coastal railway from rock falls. Austin Weltman and Pete Wilton explain.
Following a narrow strip of the south Devon coast between high cliffs and the sea, the London to Penzance railway certainly passes some eye-catching scenery. But the cost, since its opening in the mid-1800s, is an increasing number of slope failures, some causing significant damage to the line.
On the 6.5km stretch between Dawlish and Teignmouth, the cliff toe is close to the edge of the track, with clearance often as little as 3m. In places, the cliffs overhang the track, sloping steeply at 50infinity to 70infinity and between 45m and 60m high. On the seaward side is a narrow footpath built at the top of the sea wall, with a low stone wall separating walkers and trains.
But the impression on a sunny summer afternoon strolling along the popular Ladies Mile is banished during winter storms, when waves overtop the sea wall and can encroach on to the railway. Combined with heavy rainfall, this spells stress for the line.
The cliffs are formed of Teignmouth Breccia to the south west of Dawlish and Dawlish Sandstone to the north, where there is also a local outcrop of Exmouth Breccia.The rocks have varying strength and degrees of weathering with a number of promontories of stronger, more erosion-resistant rock extending out into the sea - the railway passes through five of these in tunnels.
Gradual loss of strength has been caused by the reduction of cementing of the rock, leading to local detachment of blocks of rock or clasts of limestone/sandstone from time to time. Further erosion is caused by washouts - surface water overtopping the cliff top or erosion from local groundwater issues. More significant rock falls near the cliff face are triggered by increasing pore-water pressures associated with heavy and prolonged rainfall.
Wave action is also causing gradual retreat of the cliff face, although the sea wall and the railway have, in many respects, halted the most aggressive erosion.
Another contributing factor to instability may date back to the extensive blasting used to trim back the cliffs when the railway was built.
This would have disturbed the surrounding rock, causing cracking and opening up discontinuities.
For some years after blasting, the cliff tops were repeatedly cut back to a slightly flatter angle because of frequent, major falls, many bringing down thousands of tonnes or rock.
Slips and rock falls continued into the early 1900s and in the 1930s a 134m long rock fall shelter was built to extend Parsons Tunnel north eastward through the rock promontory at Hole Head.
More recent slips have involved mainly weathered rock, triggered by run-off water impinging the rock high in the cliff. In March 2001, for example, a translational failure occurred within the shallow weathered surface at Sprey Point. Fortunately there is a wide gap between the railway and the base of the cliff and only slurried mudstone encroached on to the track.
But increased awareness of the dangers prompted Railtrack (now Network Rail) to commission a study in 2001 to tackle the problem.
From this, a protocol was developed for regular inspection by watchmen walking along the route. The minimum time scale was weekly, increasing to full time at particularly high risk locations during adverse weather conditions.
One difficulty with the system was at night, when storms reduced visibility so watchmen could only inspect the cliffs by cab-riding. In many respects the system was reactive rather than predictive, although zones of developing instability would probably have been noticed.
Network Rail then approached CL Associates to reassess the problem between Dawlish and Teignmouth.One of the main lines of attack was to prioritise those stretches of cliff where failure was either most likely or where, because of the proximity of the track to a potential failure zone, even relatively minor quantities of falling rock could cause a derailment by encroachment on to the tracks.
Five major sites were identified and schemes of rock stabilisation devised based largely on a scoring system (Fookes and Weltman 1989) to establish a suitable method.These included rock mesh protection, inclined drains, cliff-top cutoffs with V ditches, bolted concrete beams and arrays of rock bolts, with each element designed to Network Rail standards.
Rock trap fences were erected before stabilisation began, partly to protect trains during work and also to provide long term protection to the track from falling rock.
Their effectiveness was evaluated using RockTrap, an in-house rock fall trajectory program. Given the cliff height, steep angle and the limited clearance between the track and the toe of the cliff, most needed to be 4.7m high.
Further phases of remediation have been scheduled over a period of years, funding permitting, giving a general improvement in cliff stability. However, in the interim there is a need to provide track protection in those areas at risk.
CLA carried out a risk assessment to quantify the observed association between prolonged and heavy rainfall and cliff instability.With limited detailed records available this was not an exact procedure but sufficient to conclude that an association does exist.
Even more difficult to determine was the period and intensity of rainfall that might produce a slip sufficient to block one or both of the railway lines. Erring on the cautious side, the figure of 20mm of rain in 24 hours (or 25mm in 48 hours) was set as the first 'rainfall trigger level' that would impose an emergency speed restriction (ESR-1) of 20mph (32km/h).
Further trigger levels were based on a five day rainfall episode with associated periods during which ESR-2 would apply. Continued rainfall of more than 9mm per day meant an imposed ESR would remain in place until it stopped. The steps that might typically occur over a three day period are shown in the flow chart on page 27.
A predicted 10,000 minutes of resulting train delays (including regional knock-on effects) were allowed for in the first year (from 21 October 2002) of protocol implementation.
In practice, only 3,600 minutes actually accrued.
However, balancing the low rainfall over this period against practical problems that extended periods of ESR beyond that strictly necessary, this figure may, by default, be representative of a year of more typical rainfall.
Weekly physical inspections have continued by non-geotechnical personnel. During each, a proforma survey booklet is filled out with a preprinted face elevation map of finite lengths of cliff marked up as appropriate. The completed booklets are examined by CLA which prepares monthly summary reports for Network Rail.
More detailed reviews are carried out every six months to assess trends or incipient zones of weakness. Should a rock fall occur that endangers the track, a rapid response geotechnical team is on hand to assess the likely progress of the fall; identify how to restore stability (temporarily if necessary) and to judge when the track is safe to open to line speed.
In measuring rainfall, manual reading of rain gauges was considered too uncertain for effective operation of the ESR system. Instead an automatic rain gauge at Dawlish Warren Station measures and transmits rainfall readings every 15 minutes to Network Rail Control at Swindon.
Purpose-developed software produces an onscreen alarm if rainfall trigger levels are exceeded. Network Rail Control then follows a predetermined event tree to impose ESRs for the ensuing pattern of rainfall.
Other factors are taken into account, such as observation of slip debris or rock fall near or encroaching the line by train drivers, track crews or a member of the public. A final stage has been the development of software by CLA to alert the Swindon controller automatically of the appropriate course of action based on the event tree. The software manages the protocol by identifying the time for implementation of an ESR and its duration with continually updated real-time rainfall information.
Such a system of protecting when physical intervention is not immediately possible could be appropriate on other earthworks situations, such as potential failure of a cutting or embankment. Here, pore water pressure measurement, slope creep or other physical properties may be more appropriate control parameters and could be combined for a more comprehensive protocol.
At Dawlish, considerable savings over the earlier labour intensive approach have been made in the first year it has operated.
Additionally, the protocol is predictive rather than reactive, which improves safety and reduces reliance on simple observation. While the system may become redundant as stabilisation reduces the number of failures, it is a way of managing the challenging geotechnical aspects of a short section of the railway network.
Austin Weltman is technical director and Pete Wilton an associate at CL Associates.
remedial measures against degradation in weathered and fresh rock. Proc ICE, Part 1, 86, Apr, 359-380.