Thames Tideway Tunnel: TBM performance in the London Clay Formation and overcoming man-made obstructions Open Access

The Thames Tideway Tunnel project is a deep tunnel system for combined sewer overflow (CSO) control in London, UK, with a 25 km long main tunnel extending across London between Acton Storm Tanks in the west and Abbey Mills Pumping Station in the east (Newman, 2022), where it connects with the 7 km long Lee Tunnel (Newman et al., 2016) (Figure 1).

CSOs are controlled by diverting combined sewage flows into the tunnel system, which are then transferred, through the Lee Tunnel, to Beckton Sewage Treatment Works for treatment. The tunnel system comprises the main tunnel (built in four sections), the 1 km long Frogmore connection tunnel, the 5 km long Greenwich connection tunnel and ten short connection tunnels (Figure 1).

This paper focuses on the western-most section of the tunnel (Main Tunnel A), between Acton Storm Tanks and Carnwath Road Riverside (Figure 2). It is 7 km long with an internal finished diameter of 6.5 m. The tunnel is approximately 30 m deep at Acton Storm Tanks and 42 m deep at Carnwath Road Riverside and is connected to three CSO drop shafts (by short connection tunnels) at Putney Embankment Foreshore, Barn Elms and Hammersmith Pumping Station (Figure 2). In addition, it is linked to the smaller 1 km long, 2.6 m dia. Frogmore connection tunnel, which incorporates further CSO drop shafts at Dormay Street and King George’s Park in Wandsworth.

Main Tunnel A was constructed from the launch shaft at Carnwath Road Riverside, following the River Thames for the initial 4 km towards Hammersmith, passing beneath Fulham Railway Bridge (London Underground District line) and Putney Bridge and approximately 4 m above the north–south Thames Lea Valley Water Tunnel. It continues east–west across the northern tip of the Barnes Peninsula and then northwest across the river, approximately 4 m above the Thames Water ring main. The alignment continues northwest to Acton Storm Tanks for 1.8 km, beneath the very dense urban areas of Chiswick and Bedford Park, including several major A roads as well as the London Underground Piccadilly line.

The route of Main Tunnel A took it past the well-known landmark of Hammersmith Bridge; given the condition of this structure, this posed interesting challenges (see Lewis et al. (2025) in this issue).

A detailed ground investigation (GI) was undertaken for the tunnel prior to construction. This consisted of 35 combined cable percussion and rotary core boreholes (Figure 2), drilled to beyond tunnel depth, complemented by a marine hydrographic and seismic survey for the section of alignment beneath the River Thames (Newman and Hadlow, 2021). In addition, eight boreholes were completed along the Frogmore connection tunnel.

Compulsory geological workshops were organised for all field-based supervisory and borehole logging staff during the GI. In these, guidance was given by industry experts in the recognition and correct description of the London Basin stratigraphic sequence, not least, divisions A to E (in ascending order) within the London Clay Formation (LCF).

These were first recognised by King (1981), with each one reflecting major transgressive–regressive cycles of sedimentation during deposition, typically displayed by an upward increase in grain size in the form of greater concentrations of silt and laminae and very thin beds of fine sand, except for the lowest one (A2). Furthermore, each of the divisions contain distinct geotechnical properties and are recognised through their moisture content profiles (Standing, 2020).

A preconstruction ground model was produced, based on the GI data, which indicated that Main Tunnel A would be constructed entirely within the LCF. This further identified the presence of the B and A3 and A2 subdivisions only, within different sections of the alignment (Figure 3).

Geological faults were interpolated between boreholes where the boundaries between the LCF divisions had been vertically offset, as were discrete micro fossil fauna marker horizons, identified during a simultaneous micro palaeontological analysis. In addition, a large proportion of the GI boreholes extended through the entire London Basin geological sequence to the top of the Chalk Group. This significantly increased the robustness of the preconstruction ground model and confidence in interpreting the morphology of the faults by recording the vertical displacement of key marker horizons, such as the Mid-Lambeth Group Hiatus (Page and Skipper, 2000; Newman, 2009; Newman and Hadlow, 2021) as well as distinct stratum boundaries, especially within the Lambeth Group. There was no visible evidence within the boreholes for any significantly adverse disturbance of the groundmass adjacent to the faults, such as in the form of an increase in frequency of fissures or jointing discontinuities. The positions of the faults, such as those within the West Putney Fault Zone and the Hammersmith Fault were further enhanced in the long section plots produced during the marine hydrographic and seismic survey, in which several metres of vertical displacement of strata were evident (Newman and Hadlow, 2021).

The results of the seismic survey were calibrated using seismic reflections from existing tunnels of known depth and diameter and then integrated with those from the GI boreholes using vertical seismic profiling. This involved the identification of notable reflection horizons from more than one of three seismic reflection techniques used, with further calibration of these by using hydrophones suspended within dedicated over-water boreholes to gain confirmation on their presence (further details are given by Newman and Hadlow (2021)).

Notwithstanding the effects of fault displacement, most of the tunnel intersected the contact between the B and A3, encountering the high frequency of partings and thin beds of fine sand and silt within the top several metres of the A3 (see Figures 3 and 4). A high proportion of the GI boreholes had detected groundwater within this, albeit in most cases as merely ‘seepage’. Elsewhere, the initial section of the tunnel, from Carnwath Road Riverside, intersected the contact between the A3 and the predominantly fine sandy clay of the A2.

The observations made from borehole logging were corroborated by the results of laboratory classification tests on samples of LCF, confirming the elevations of the division boundaries coinciding with changes in natural moisture content (Figure 4), as outlined by several other authors (Hight et al., 2003; King, 1981; Pantelidou and Simpson, 2007; Standing, 2020). In addition, a progressive increase in strength with depth was evident based on laboratory triaxial tests. This reflects the increasing effects of consolidation due to depth of burial and is in line with the findings of Hight et al. (2003). It is not dependent on the type of LCF division.

Tunnel construction was undertaken using an earth pressure balance (EPB) tunnel boring machine (TBM), called Rachel, with an external cutterhead dia. of 8.1 m (Figure 5). The length of the TBM shield was 12.3 m, although the total length, including back-up gantries, was 142 m, with an operational capacity to mine at 80 mm/min, equating to an advance rate of 1.7 m in approximately 21 min. Other performance characteristics included a maximum thrust force of 50.6 kN and a maximum torque of 7.9 mN/m. The tunnel primary lining support was constructed within the rear of the TBM and comprised eight 350 mm thick, 1.7 m long trapezoidal segments to form a 7.1 m internal diameter boltless ring.

Geological logging during construction of the CSO drop shafts and associated connection tunnels proved a valuable exercise in recording the ‘as-built’ ground conditions (Newman, 2022). This included logging the TBM launch adit at Carnwath Road Riverside (Figure 6(a)), the CSO drop shafts (Figure 6(b)) and the exposed face of the main tunnel during TBM cutterhead interventions (Figure 6(c)). A total of six head interventions were made during tunnelling to log and record the tunnel face conditions and confirm ground conditions as predicted in the preconstruction geological model (Figure 7). Logging confirmed the ground conditions as predicted in the preconstruction geological model and, importantly, the absence of any unforeseen geological conditions. This provided further assurances on the ground ahead of tunnelling operations.

Consideration was given to possible effects of the different LCF divisions on the operational performance of the TBM, such as whether there might be possible face instability within the zone of fine sand partings at the top of the A3 and, as a consequence, higher face pressures applied in the TBM plenum chamber. A similar analysis had been undertaken for the Main Tunnel C TBM drive between Kirtling Street and Chambers Wharf (Figure 1), which revealed contrasting performances of the TBM in cohesive and granular strata belonging to the Lambeth Group and Thanet Formation (Newman et al., 2023).

Short of stopping the EPB TBM and undertaking a cutterhead intervention, the only visual evidence of the type of ground being encountered by a TBM during its operation is inspection of spoil arisings passing through the screw conveyor and spoil conveyor belt (Newman et al., 2010). Changes in the TBM operating parameters, however, can also provide useful indications of the type of ground, together with simultaneous changes in geology being encountered during tunnel excavation. Some of these were assessed for Main Tunnel A (Figure 7).

During operation of a TBM, water and foam compounds are injected into the cutterhead to assist in the mechanical breakdown of clay material being excavated and provide sufficient lubrication for it to pass through the cutterhead and screw conveyor without clogging. The quantities required for these are generally greater within clay strata and less in granular strata, which are less cohesive and in which saturated conditions might pre-exist (Newman et al., 2023). Any adjustments in the required volumes of these can be determined from changes in TBM operating parameters being monitored. The TBM was operated in closed (pressurised) mode over the majority of the tunnel drive. The combined quantity of water and foam added to the cutterhead remained relatively constant, between 13 000 and 18 000 l per 1.7 m ring advance, with a foam expansion ratio of between 200% and 700%. This might imply that the differences between the sandier and siltier LCF divisions were too subtle to be detected by the TBM and that they had no significant effect on its operational performance. An exception to this was within the first 1 km of the tunnel, in which a greater quantity was added, generally 17 000–21 000 l. It is possible that this reflects the greater plasticity (and therefore more cohesive behaviour) of the lower section of the A3 encountered over this duration, although – more plausibly – it could be due to a period of familiarisation of both machine and ground for the TBM operating personnel.

This could also be the cause for the gradual, progressive increase in TBM advance speed from 10 to 50 mm/min over the first 1 km of tunnel, becoming relatively consistent between 50 and 62 mm/min thereafter. In the same way, levels of torque progressively increased from an average of 3.0 MN.m within the first 1 km to generally 3.5 thereafter and then 4.0–4.5 MN.m towards Acton Storm Tanks.

Initial high levels of TBM thrust, between 20 000 and 22 000 kN are, similarly, thought to be due to this familiarisation period rather than from encountering the higher shear strength A2 subdivision. Beneath the River Thames, the TBM thrust remained relatively consistent, generally averaging 16 000 kN. However, this increased progressively to 22 000 kN on passing beneath the land section of the tunnel, towards Acton. This could be reflective of the associated increase in overburden pressure as there was no change in the type of LCF division being encountered between the river and land sections, although it is more probably due to a combination of reasons such as changes in the condition of the cutterhead and maintenance of face pressures (i.e. balancing advance speed with muck transfer by means of the screw conveyor). Continuing on this theme, the tunnel was never solely within any one particular LCF division for any significant duration (i.e. more than one division was present within the tunnel profile for most of its construction), allowing for a meaningful comparison of their effects on the operational performance of the TBM. It is thought likely, however, that any significant changes in the TBM operating parameters caused by the relatively subtle variations in geotechnical parameters between the LCF divisions would probably go undetected over the large scale of tunnel face in this example.

There was no evidence from its operating parameters that the TBM was significantly affected by potentially adverse ground conditions associated with the West Putney and Hammersmith geological faults.

Volume loss in the tunnel face, as determined from surface settlement monitoring, was recorded consistently to be 0.5–0.6% throughout the duration of TBM operation and was successfully maintained below the predicted 1%. This was a moderately conservative value, adopted for the impact assessments for the EPB TBM tunnel drive. This provides further evidence that any effects from encountering the different LCF divisions were insignificant and that the implications for surface settlement, generally in the range 5–12 mm, were minimal.

The TBM operation was more notably affected by its intersection with two pre-existing boreholes rather than changes in the ground conditions.

Prior to tunnel construction, ground conditions and associated hazardous features were identified in a construction risk register, which further outlined methods for implementing emergency contingency procedures to mitigate these if encountered. The construction risk register identified the likelihood of the TBM intersecting a nominal amount of boreholes. These ranged from those for the project’s own GI, to unknown, ungrouted, uncased boreholes, with a direct hydraulic interconnection to the Upper Aquifer or the River Thames, or cast-iron-lined boreholes and/or larger diameter brick-lined abstraction wells extending to below the tunnel invert.

The first borehole (NGR 524124, 175745) was intersected by the TBM approximately 1.5 km from the launch shaft at Carnwath Road Riverside and 85 m west of Putney Bridge. Despite the borehole being part of the project’s GI, its final position was not as originally intended due to being repositioned to avoid numerous logistical factors such as mooring chains, shipping channels and so on. Importantly, the borehole had been backfilled with bentonite upon completion. Up to this point, 844 precast segmental concrete rings had been constructed, during which the TBM had been operated in ‘open mode’. The borehole was intersected during excavation for the 845th advance, at which point a sudden and significant build up in water pressure was recorded within the cutterhead, leading to the TBM switching operations to ‘closed mode’. This was undertaken to counteract potential tunnel face instability, in which an increase in earth pressure was applied to the tunnel face (Figure 8) by carefully controlled adjustment of the TBM operational parameters (e.g. thrust and auger screw rotation).

Laboratory classification tests had been undertaken on samples of the LCF during the GI and these indicated a range in liquid limit between 58% and 82% for the B and A3 divisions, which were present in the tunnel horizon at this point. The influx of water significantly changed the consistency of the material, exceeding its liquid limit and leading to flowing-type behaviour (Figure 9(a)). Severe difficulties in spoil removal from the TBM resulted from this due to flow of the spoil down the screw conveyor and back into the cutterhead chamber. Any material that was extracted from the cutterhead, similarly, flowed off the spoil conveyor belt, rendering the immediate working area hazardous and prone to slipping accidents for operating personnel (Figure 9(b)).

Changes were made to the operation of the TBM to mitigate the conditions, most notably in its slower advance speed, although this resulted in consequent delays to the construction programme. Prior to intersecting the borehole, the TBM maintained a relatively consistent advance speed of 50–55 mm/min. Excavation for the two ring advances that were most affected by the borehole (845 and 846) was undertaken cautiously over a series of shortened intervals, at advance speeds averaging at approximately 25 mm/min (Figure 10(a)). This involved a time-consuming repetitive process of slowly advancing the TBM to fill the auger screw, then stopping to close the gate and empty the spoil from the screw; repeating this process over the duration of the advance to maintain a relatively stable earth pressure in the plenum chamber and further prevent ingress of fluid spoil or groundwater through the screw.

A significant increase in time taken to complete rings 845 and 846 was recorded as 6.5 h and 15 h respectively, as compared with the average time of generally 1 h for the ring builds before and after these (Figure 10(a)).

The TBM forward thrust also increased during construction of these rings, from generally 20 000 kN to 25 000 kN (Figure 10(b)) to instigate the increased earth pressure within the cutterhead in ‘closed mode’. This resulted in simultaneous, short-lived variations in the cutterhead torque, with generally lower values between 2 and 4 MN.m, peaking at 5 MN.m and eclipsing the average value of 3 MN.m before and after intersection of the borehole (Figure 10(c)).

A second, historical borehole (approximate NGR 521975 178175) was intersected by the TBM close to the north bank of the River Thames at Chiswick, 5.1 km from the Carnwath Road Riverside launch shaft (Figure 2). Details of the borehole had been obtained from published records during the initial desk top study and revealed that it was constructed for water abstraction for a bakery in 1893. It is not known when the borehole was last used for abstraction, although by 1911 the piezometric level of the Lower Aquifer within it had fallen to −30 m OD from −12 m OD, when it was first commissioned. The borehole was 128.9 m deep, with its collar recorded at +5.48 m OD; it was constructed using a 254 mm (10 inch) dia. steel casing, inserted to a depth of −79.5 m OD, 3 m beneath the top contact of the Chalk.

The collar coordinates provided for the borehole indicated that it was situated to the east of the tunnel alignment, 20 m from the point of its actual intersection with the tunnel centreline. There was no way of validating this (e.g. by simple walkover observation and/or land surveying) as it had long since been covered by subsequent building infrastructure. It was recognised, however, as a potential hazard for intersection by the tunnel within the construction risk register, most notably for hydraulic failure. However, in view of its perceived distance from the alignment, it was not anticipated to be intersected and was therefore deemed low risk. Instead, an increased awareness of its location was made to all tunnelling personnel, with contingency procedures put on standby if required. The tunnel crown was −30 m OD (approximately 35 m below ground) within the vicinity of the borehole.

The borehole was intersected during excavation for ring 3011 (chainage 5183 m). This was noticed by an abrupt decrease in earth pressure balance for 1 h of TBM operation and the simultaneous observation of fragments of the steel casing on the spoil conveyor belt. At this point, the TBM was stopped to enable a cutterhead intervention to be undertaken to check for underground obstructions, as well as damage to component parts and the quality of groundmass within the tunnel face. On entering the cutterhead, the presence of the borehole steel casing was witnessed (Figure 11) and subsequently removed. At the same time, the contingency procedures were activated, comprising increased frequency of surface settlement monitoring and visual checks on the integrity of third-party assets, as well as the installation of additional surface monitoring points to facilitate early warning in the event of increased ground loss and settlement. Results of the monitoring proved that there had been no increase in surface settlement associated with the borehole when compared with sections of the tunnel before and after the point of its intersection.

There were no other indications of any significant changes to the TBM operating parameters on intersecting the borehole, although some adjustments were made – as part of the contingency procedures – while steering through the zone of intersection. These included reducing the advance speed from typically 55 mm/min to 37 mm/min, the cutterhead torque from 3.4 MN.m to 3.0 MN.m, the screw conveyor torque from 37 kN.m to 31 kN.m and the cutterhead penetration from 21 mm/rotation to 15 mm/rotation.

Surface settlement data and TBM parameters were assessed during shift review group meetings, with any necessary actions being implemented, such as changes to the TBM operating parameters, an increase in the frequency of surface settlement monitoring and so on. There were no disturbances of third-party assets observed on the surface or breaches of in-tunnel or surface settlement triggers. Neither was there any requirement for TBM cleaning operations and so no cause for any management action meetings or instigation of emergency procedures such as evacuation of the tunnel.

Other than a relatively small amount of delay to the construction programme, there were no significant adverse effects on the integrity of the tunnel structure and surrounding groundmass. There were, however, concerns on whether the intersected borehole had acted as an open conduit for the downward flow of tunnelling fluids into the Lower Aquifer. This was insinuated by a very slight increase in tunnelling fluid injected into the TBM cutterhead at the point of borehole intersection, from typically 16 000 l to 17 590 l. This was only for a very limited duration, however, as the advance of the TBM shield was thought to have temporarily closed the borehole shortly after its intersection. Permanent closure followed soon after, when sealed by tailskin grouting operations (i.e. the injection of grout into the anulus between the shield and surrounding ground), as indicated by a slightly higher grout-take at the point of intersection when compared with that for previous and subsequent advances. Despite the evidence indicating that the borehole had been sealed by the operations of the TBM, it was necessary to satisfy any further concerns about the impact of the small quantity of potentially contaminated tunnelling fluid that had managed to enter the borehole.

Following a detailed risk assessment, the risk of tunnelling fluid entering the Lower Aquifer was considered very low. Calculations were undertaken to assess any potential impact of the foam component within the tunnelling fluid, which was known to contain less than 1% toxic ingredients by volume and further calculated as a proportion within the tunnelling fluid as 0.03%. This had been proven as biodegradable and, if entering the Lower Aquifer, would reach safe levels of attenuation in a matter of days. Furthermore, a number of boreholes were strategically placed along the length of the scheme in the central and east sections, dedicated for longer-term monitoring of water quality in the Lower Aquifer and to detect whether tunnelling fluid might have entered this.

On 23 March 2020, a national lockdown was instructed in response to the Covid-19 pandemic. The pandemic became noticeable on this project by the increasing number of personnel reporting as sick from Covid-19 from the middle of March. All tunnelling operations were suspended to safeguard personnel from infection and to allow time to implement additional protective measures, such as safe-working conditions within the confined spaces of the TBM. Most notably, these included appropriate social distancing protocols and one-way pedestrian access and egress routes, as well as frequent and strategically placed hand-washing facilities and sanitisers.

At the start of lockdown, the TBM had reached the edge of the settlement zone of influence for the Grade II-listed Hammersmith Bridge, approximately 4 km from the Carnwath Road Riverside launch shaft. Built in 1887, Hammersmith Bridge is one of the world's oldest suspension bridges made from wood and wrought iron, with the suspension system held in place by cast-iron pedestals. It was designed by Sir Joseph Bazalgette and is part of Britain's engineering heritage and a national landmark. During the planning and design stages of the Thames Tideway Tunnel, minimising potential impacts of ground disturbance on third-party infrastructure had been of prime importance. Hammersmith Bridge was identified as critical and negotiations with the London Borough of Hammersmith and Fulham resulted in a realignment of the tunnel 10 m to the south, away from the south bridge abutment (see Lewis et al. (2025) in this issue).

It was not possible to stop the TBM here due to the increased risk of adverse settlement from the effects of relaxation of the ground that this would cause, which in addition, would squeeze in on the shield, ultimately making it impossible to restart without outside intervention.

There was no choice other than to continue advancing the TBM, albeit with a considerably reduced operating personnel, incurring significant increases in the average duration of ring builds, typically from 21 min prior to lockdown to 28 min during lockdown (Trigle, 2022). Nonetheless, under these conditions the TBM successfully cleared the zone of influence by 28 March, with negligible impact on the bridge and without tunnel face loss reaching settlement trigger levels. This was largely due to the maximum short-term settlements being only 60% of those predicted and the long-term volume loss at 0.8%, significantly less than the predicted 1.0% (further technical details are provided by Trigle (2022)).

It was only at this point that it was safe to stop the TBM, although it was critical to maintain the face pressure to avoid excessive settlements. The normal target face pressure in the crown was set at 100 kPa, which was above the minimum pressure required for face stability and below the maximum pressure, which could cause heave.

Prior to the stoppage, settlement monitoring data was fed back to the site shift review group to verify performance and allow adjustments to be made where necessary. During the stoppage period, however, a higher face pressure of 150 kPa in the crown was set to accommodate the anticipated relaxation of the mixed earth material within the plenum chamber and to reduce the face pressure with time. To ensure stability of the face, the pressure was monitored every 30 min, over a 3-day period, until closure of the site on 31 March.

The TBM was stopped with a shield articulation of 100 mm to allow for extension or retraction of articulation to help control pressures if needed. The thrust cylinder extension was stopped 200 mm short of the full extension of 2700 mm to ensure support of the previous ring within the tailskin and allow access to the accelerator ports. These were thoroughly cleaned before shutdown to avoid becoming blocked on restart. In addition, the thrust cylinders could be extended by a further 200 mm to increase the face pressure if there was a risk of the face pressure dropping below 100 kPa, which could be done with minimal personnel. Hydraulic ram pressures were set and the hydraulic circuits were isolated to avoid leaks and pressure reductions. As part of a safe stop procedure, the TBM grout system was thoroughly cleaned: all moving parts were lubricated and the main drive, main drive gearboxes and screw gearboxes were filled with oil to ensure there was no possibility of them seizing up during the stoppage and struggling to restart.

Cameras were installed to facilitate remote visual monitoring within the TBM to ensure a continued stable environment. Face pressure and ram pressures were monitored using a telemetry system that incorporated a warning alert if the face pressure dropped below 120 kPa.

The TBM stoppage lasted for 4 weeks and tunnelling operations resumed in early May 2020, although only after a full readiness review between the contractor, programme manager and client had been undertaken to ensure that strict Covid-safe measures were in place. These dictated a staggered return to work by TBM operating personnel, starting with a small team followed by a gradual increase and full production by the beginning of June 2022.

The LCF around the TBM remained stable during this time and there were no issues on restart.

Recognition of discrete divisions within the LCF was achieved largely by the provision of detailed geological logging workshops and industry expert guidance prior to commencing the GI for the tunnel. The time spent on this exercise was rewarded in the form of a highly accurate ground model for the tunnel, validated by construction geological logging and leading to significantly reduced risk of the tunnel encountering unforeseen ground conditions. It is recommended that these activities are adopted as ‘best practice’ on future tunnelling projects.

Vertical changes in the stratigraphy along the tunnel, due to geological fault displacements, were not enough to cause any significant variation in the different LCF divisions encountered by the TBM. Instead, the majority of the tunnel was constructed along the boundary between the B and A3 and so a direct comparison of the individual effects of these on tunnelling operations was not possible. Notwithstanding this, it is suspected that the variation in geotechnical properties between the different LCF divisions was not significant enough to affect the TBM on the scale of tunnelling in this example. This might not be the case on a smaller diameter TBM and would be an interesting line of investigation during future smaller scale tunnelling projects in the LCF. There were slight anomalies in the TBM operating parameters within the first 1 km of tunnelling, however, coinciding with the intersection of the stronger, sandy clay within the A2 subdivision. It is possible that these were due to lithological influences, but more probably due to a period of familiarisation of machine and ground by the TBM operating personnel.

A more significant impact on the TBM operating performance was incurred by the intersection of two existing boreholes along the tunnel alignment. The first borehole was in the River Thames, near the Putney Embankment Foreshore site, and is likely to have acted as a conduit for water flow from the river into the TBM cutterhead. It is appreciated that, as part of best practice, GI boreholes should be located at an appropriately safe distance from a proposed tunnel alignment. However, in this instance, relocation of the Tideway boreholes was unavoidable in order to satisfy the requirements of third-party stakeholders. It is further possible that the integrity of the bentonite backfill may not have been sufficient to prevent water inflow during the TBM passage and that an alternative pressurised grout might have been more robust.

The second borehole was a land-based, archive borehole and relatively poorly documented in a manner to suggest that it was positioned away from the tunnel alignment and so perceived to be less hazardous than it turned out to be.

Intersection with both boreholes caused delay to the construction programme, although their impacts on the safety of tunnelling operations was successfully mitigated by the activation of emergency contingency procedures that had been carefully planned during the production of a construction risk register. These entailed skilful, meticulous operation of the TBM, in conjunction with increased in-tunnel and surface settlement monitoring to safeguard the integrity of the tunnel structure and surrounding groundmass. Furthermore, ongoing environmental monitoring and post-construction assessments, especially in relation to aquifer safety, have provided assurances for the lack of long-term impacts of tunnelling operations on the local ecosystem.

The authors thank Tideway and the BMB joint venture for permission to produce this paper and Ryan Moor for the valuable contribution of TBM operating data.