Rehabilitation and expansion of the Central City Tunnel System in Minneapolis, MN

The city of Minneapolis contracted with CDM Smith to update and provide a conceptual design for improvements to mitigate surcharge flooding in the Central City Storm Water Tunnel System (CCSTS). As a part of the project, CDM Smith conducted a field survey and condition assessment of the existing tunnel system. Information from the survey was then used to update the existing XPSWMM model and develop system wide alternatives.

The CCSTS provides storm-water runoff drainage for nearly the entire area of the city’s downtown commercial district. The system consists of deep storm-water tunnels constructed in the St. Peter Sandstone, approximately 21 m (70 ft) below the street’s surface. The primary tunnels comprising the Central City storm water tunnel system are located below Hennepin Avenue, Nicollet Mall, Lesalle Avenue, Marquette Avenue South, 2nd Avenue South, South 5th Street, Washington Avenue South, Portland Avenue South, 2nd Street South, and Chicago Avenue South, as shown in Fig. 1. This network of tunnels conveys the runoff from a 305-acre tributary area that is generally bound by Hennepin Avenue and 1st Avenue North to the east, 12th Street to the south, 4th Avenue South and 7th Avenue South to the west, and 2nd Street South. These tunnels were constructed between 1936 and 1940, except for the Marquette Avenue South tunnel, which was constructed between 1963 and 1964.

The CCSTS operates as a gravity flow system. These tunnels were constructed within the St. Peter Sandstone layer of bedrock and emerge from the bedrock at the Mississippi River below St. Anthony Falls. The Central City and the adjacent Chicago Avenue tunnel system converge into a single outfall at the Mississippi River. The runoff discharges from the converged outfall to a side channel of the Mississippi River, called a tailrace, located near the Guthrie Theater. The Minneapolis Division of Surface Water and Sewers provided 32 historic plats detailing the plan and profile of the tunnel system.

The tunnel plans show nine different cross-section configurations. Eight configurations within the overall system generally show the same geometric “cathedral” shape with the inside dimensions varying from 1.2 to 1.8 m (4 to 6 ft) in width and 1.8 to 2.4 m (6 to 8 ft) in height. For analysis, these eight configurations were reduced to three configurations with regards to tunnel liner, cross-sectional area and support. For simplicity, the three configurations are described according to the three types of tunnel support used during construction: none required, light timber and heavy timber.

The storm water tunnels on Hennepin Avenue, LaSalle Avenue and Marquette Avenue between 4th and 7th Street South, and Nicollet Avenue between 9th and 10th Street South all have sanitary sewers that either cross, or aligned with, the storm tunnels, but are located below the invert of the storm water tunnels. These sanitary sewers are clay pipes encased in concrete and range from 304 to 610 mm (12 to 24 in.) in diameter. The separation between the top of the sanitary tunnel and the Central City storm water tunnel is minimal, ranging from immediately beneath the storm water tunnel to 840 mm (2.75 ft).

The Central City storm water tunnel, as it approaches the convergence structure, is a 2.3-m (7.5-ft) wide by 2.4- m (7.9-ft) tall cathedral shape tunnel constructed of block below the springline, and liner above. The Chicago Avenue tunnel, as it approaches the convergence structure, is a 2.4-m (8-ft) diameter circular brick structure. The outfall structure, below the convergence of the Central City and Chicago tunnels, has unique cross-section configurations: a mushroom shape at the convergence that transitions to a cathedral shape for approximately 15.2-m (50-ft) immediately upstream of the outfall structure at the Mississippi River.

Pressurization of the Central City storm water tunnel segments has been an ongoing issue for the city, leading to repeated and expensive maintenance repairs. The existing deep tunnel storm water system, was not designed for the characteristics of drainage inflow that consist of an increase in runoff volumes and shortened time-of-concentration caused by increasing impervious surfaces in the downtown commercial district. Pressurization of the tunnel during large, intensive rainfall events has caused the liner to crack, contributing to liner failure and erosion of the sandstone immediately outside of the tunnel liner at multiple locations. The maintenance and repair process, typically, requires identifying the void locations caused by erosion, filling the voids with grout, and repairing the cracks that led to the creation of these voids.

Geological setting

The general subsurface geological profile in the drainage system area is very consistent as shown on the available tunnel plat drawings. This general geological profile consists of:

• Overburden — sand, gravel, boulders and, in some locations, a thin layer of clay below the granular material.
• Weathered rock — described differently on profiles ranging from hardpan and boulders to broken limestone.
• Rock — predominately a 4.6 to 12.2 m (15 to 40 ft) thick stratum of limestone that serves as a cap rock to a very thin soapstone, overlying St. Peter Sandstone.

Ground water was not identified on the plat sheets. However, during recent work on the Nicollet Mall project, which is within the drainage area of this project area, CDM Smith drilled borings to a depth of 12.2 m (40 ft) without encountering any ground water.

The CCSTS is located entirely within the St. Peter Sandstone. This rock is unique in that it is composed of very uniform sand size grains that are 99 percent quartz. The rock strength is developed from compressive loads and it exhibits almost no cohesion. The sandstone becomes harder and denser with depth. The rock is also very friable. Turbulent water in contact with fresh surface of sandstone will cause a rapid disintegration of the rock.

The total unit weight of the rock as reported in existing data is 21.2 kg/m3 ± 0.6 kg/m3 (135 pcf ±4 pcf) . Gradation, sieve, analyses of the sandstone indicate that approximately 90 percent of the sand grains are between 140 and 60 sieve sizes (ASTM). This is indicative of a fine sand. Porosity of the rock averages 0.28. Unconfined compressive strength testing of 11 samples ranged from 4.7 MPa to 19.4 MPa (680 psi to 2,810 psi) with an average strength of 10.8 MPa (1,570 psi). Published friction angles of the St. Peter Sandstone typically range from 54° to 65°.

Hydraulic modeling

The purpose of the hydraulic modeling analysis was to determine the extent of improvements to the Washington Avenue leg of the Central City storm water tunnel in preparation for a capital improvement project that the city has scheduled for construction starting in 2020. As part of this analysis, the XPSWMM model was used to determine the equivalent hydraulic diameter needed to provide additional hydraulic capacity for the CCSTS to prevent pressurization of tunnel during a design rainfall event.

To determine whether the observed pressure surcharge in each tunnel leg was a product of downstream constraints plus tail water, or an individual tunnel leg being constrained within a segment of the tunnel, free discharge conditions were created at the points where the Hennepin Avenue, Nicollet Mall, Marquette Avenue, and 2nd Avenue tunnels discharge into the Washington Avenue tunnel. Each leg was analyzed using the 10-year, 100-year and 500-year design storms to determine the level of service of each tunnel leg discharged into the Washington Avenue tunnel. The following describes the hydraulic capacity of each of these tunnel segments when not influenced by the hydraulic grade line (HGL) of the Washington Avenue storm water tunnel:

• Hennepin Avenue: Hennepin Avenue operated without surcharge for a 10-year design rainfall, had negligible surcharge during a 100-year design rainfall and had 1.5 to 3 m (5 to 10 ft) of surcharge for a 500-year design rainfall.
• Nicollet Mall: Nicollet Avenue, including contributing flows from the LaSalle Avenue tunnel, had negligible surcharge during a 10-year design rainfall and 6 to 15 m (20 to 50 ft) of surcharge during a 100-year design rainfall.
• Marquette Avenue: The Marquette Avenue tunnel conveyed the runoff from all rain events within the crown of the pipe, including a 500-year design rainfall.
• 2nd Avenue South: The 2nd Avenue South tunnel surcharged as much as 9 m (30 ft) during a twoyear design rainfall and had significantly greater surcharge during the larger design rainfall events.

The tunnel segments were recombined to assess how the hydraulic conditions of Washington Avenue, in combination with the known deficiencies in hydraulic capacity of each tunnel leg, influenced the total flow. The most significant changes occurred in the Hennepin Avenue and Marquette Avenue legs of the system, changing from no surcharge or negligible surcharge to surcharge in all design rainfall events. However, the Chicago Avenue tunnel and the converged Central City/Chicago Avenue outfalls have sufficient capacity for all modeled rainfall events. The hydraulic analysis indicates the need for hydraulic improvements to Washington Avenue, 2nd Avenue South and Nicollet Mall tunnels. The Marquette Avenue and Hennepin Avenue tunnel legs will have improved hydraulic performance after improvement of the Washington Avenue tunnel segment, and therefore does not need further analysis. The proposed 2020 construction will focus on improvements to the Washington Avenue tunnel segment.

Preliminary design alternatives

Initial increased conveyance capacity alternatives were developed for the Nicollet Avenue, 2nd Avenue South and Washington Avenue tunnel segments found to have insufficient hydraulic capacity to convey the runoff from a design rain event. The existing condition XPSWMM model was used to compute the equivalent circular cross-sectional area for each hydraulic option. Cross-sectional areas for 10-year and 100-year rainfall events were developed to establish the incremental cost differences for mitigating system pressurization risks for the respective design rain events. Based on the XPSWMM model results of these two methods the following alternatives were developed.

Expanded tunnels. This alternative increases the size of the existing tunnel cross-sectional area. The minimum cross-section area of an equivalent circular tunnel was computed for both the 10-year rainfall event (10.8 cm or 4.27 in. of rainfall in 24 hours) and the 100-year rainfall event (19 cm or 7.47 in.) of rainfall in 24 hours), as estimated by NOAA Atlas 14, Volume 8. The actual cross-sectional shape of an expanded tunnel will likely not be circular, given the characteristics of the engineering properties of the St. Peter Sandstone, available headspace between top of tunnel and top of St. Peter Sandstone, and conflicts with the Metropolitan Council Environmental Services (MCES) interceptor. An indepth description of the shapes and cross‑sectional areas considered is presented in the construction alternative section.

Parallel tunnels. This alternative involves construction of a new parallel tunnel adjacent to the existing tunnel. The minimum cross-sectional area of a circular parallel tunnel was computed for both the 10-year design rain event and the 100-year design rain event. A parallel tunnel could either be circular or it could be another shape if any of the constraints described in the Tunnel Expansion option are encountered. Tunnel sizes for the 10-year and the 100-year rainfall events are discussed in the construction consideration section of this paper.

Final alternative. After the initial alternatives were developed for the entire system, additional refinements were made to the model to specifically address the proposed 2020 project along the Washington Avenue Tunnel alignment. For this section, a combined approach was proposed that would construct a new parallel tunnel east and west of Portland Avenue and expand portions of the existing tunnel alignment along Portland Avenue and at the junction with the Chicago Avenue tunnel system.

Geo-structural analyses

Concurrent with the hydraulic analysis, CDM Smith completed a geo-structural evaluation of the existing tunnel system with the goal of identifying and evaluating any risks associated with enlarging tunnel cross-sections to increase hydraulic capacity of the system and for repairing the existing tunnel segments.

Existing tunnel analyses. To analyze the existing tunnel system, Rocscience Phase 2 software program was utilized. After review of the existing tunnel plats, five existing tunnel configurations were identified for evaluation. The existing tunnel materials were modeled using Mohr-Coulomb failure criteria to account for the lack of reinforcing steel and an inability to resist tensile stress. Therefore, the model assumed that when a very low tensile stress was applied to the liner failure of the tunnel liner would occur. The analysis configurations consisted of the following:

• Profile analysis of the excavation of the storm water drainage tunnel at locations where it crosses directly above an existing sanitary sewer tunnel. This analysis was performed to evaluate the magnitude of change in stress on the existing underlying sewer tunnel due to excavation above it.
• Cross-sectional analysis. An analysis of different support types was performed using adjusted rock strength values depending on the existing liner support system, including no timber, light timber and heavy timber support behind the liner. The analysis consisted of applying a cyclical internal pressure to the tunnel representing loads experienced during a 100-year rainfall event, as predicted by the XPSWMM existing conditions model, developed by CDM Smith. The frequency of the cyclical loading was based on a review of five years of historic pressure data provided by the city. During this five-year period, there were six events that surcharged the tunnel at the pressure meters. These surcharges ranged from 1.2 to 11.6 m (4 to 38 ft) above the tunnel crown. For the modeling, we extrapolated this to 20 surcharge loadings, representing the occurrence of one surcharge event every five years for a period of 100 years. The applied internal pressure represented by the 100- year rain-event is predicted to be 35 psi, (0.24 MPa) or 24.6 m (80.7 ft) of water. This represents a factor of slightly greater than twice the measured event. Each of the three different existing liner conditions and locations, were modeled as follows:
• No lining support. There are several locations shown on the city’s tunnel plats where the tunnel liner is shown as concrete placed against the sandstone without initial support. Figure 3 represents a typical No Support segment. The average rock strength parameters were used, without strength reduction, since there is no initial liner support. It was assumed that the St. Peter Sandstone at these locations was in good condition, with few joints or loose materials and a strength reduction was not applied to the model.
• Light timber support. At locations where light timber support was identified, the drainage tunnel is approximately 1.6-m (6-ft) high by 1.6-m (6- ft) wide. A light wood support encompasses the upper portion of the tunnel from springline to crown and back to the springline in a trapezoidal configuration. It was assumed that the ribs were used in locations where the rock quality exhibited some joints or fractures, requiring some additional initial support. To account for this condition, the model used a reduced rock strength of the intact rock. It is assumed that the timber supports provides a seepage path for ground water outside the tunnel and leakage through the tunnel to cause erosion of the sandstone. This results in a source of sand to migrate through cracks in the lining and creates an ongoing process of deterioration of lining support by creating progressively larger areas of unsupported lining.
• Heavy timber support. Heavy timber support locations consist of wood ribs that fully surround the tunnel perimeter with wood lagging. Heavy timber supports were used where the rock quality was significantly poorer than at other segments of the tunnel, requiring this stronger initial support. To account for this condition, the model used the reduced rock strength of the intact rock. The same process of loss of strength of the initial support system was used to model the behavior of the tunnel as a function of time and cyclical loads.
• Reduction in liner strength. The purpose of a reduction in strength model was to account for degradation of the underlying timber supporting the tunnel liner related to the environmental cycles of wet and dry conditions. As the wood shrinks in volume and decreases in strength, deformation of the sandstone would follow with each cyclical loading due to a storm event. This loss of external support originally provided by the sandstone, causes a tensile loading on the unreinforced segments of the concrete liner. The tensile loading results in liner cracks. This creates a pathway for seepage of ground water from outside the tunnel liner during non-storm events, and leakage into the sandstone during a pressurized storm event to cause erosion of the sandstone. The resulting sand migration through liner cracks likely results in an ongoing process of cracking of liner by creating increasing areas of unsupported lining over time. To account for this long-term reduction in liner strength, it was assumed that the timber strength reduced by 5 percent between each cyclical loading event.
• Concrete liner loading. This model provided an assessment of the liner after each loading event was conducted. Providing there was continuous rock support against the liner, deformations were found to be minimal. However, where joints were formed due to shrinkage of the unreinforced concrete, the measured cracks were of sufficient size to allow passage of sand grains into the tunnel. This loss of ground was modeled by assuming a void behind the tunnel lining at each tunnel crack.
• Combined effects. To evaluate the locations where several factors may increase the loads, an analysis was performed taking into account a combined effect of nearby sanitary sewers, the weakened condition of the tunnel lining, and disturbance to the rock.

Expansion of the tunnel system

In addition to performing an analysis of the existing tunnel, CDM Smith performed a similar analysis on the proposed parallel and expanded tunnel configurations. During the analysis, two constraints were identified for the proposed tunnel expansion.

A review of the relationship between the existing tunnel and the caprock above the tunnel, as drawn on the tunnel plats, showed that several tunnel segments are close to the caprock and have limited space available for vertical tunnel expansion without penetrating the caprock. According to the historical data, tunnels excavations, at elevations above the limestone caprock, are significantly more challenging to support and are double to triple the construction cost than if the excavations were below the caprock. Therefore, primarily horizontal tunnel expansion, with limited vertical expansion, was evaluated. To maintain a gravity system, lowering the invert for expansion was eliminated from consideration.

Additionally, there are several adjacent sanitary and storm drain tunnels that either share a wall or are very close to one another. Because of these adjacent tunnels, it was concluded that the storm tunnel cannot be lowered, or substantially re-aligned due to the conflicts created by these nearby, and crossing, sanitary tunnels. Therefore, the adjacent tunnel expansion analysis only evaluated the option to increase the cross-sectional area of selected tunnels along their existing alignment to increase the hydraulic capacity of the storm water tunnel system.

For expansion of the existing tunnel two possible configurations were identified and evaluated. The configurations consisted of the following:

• Excavation within the existing tunnel. The increase in the tunnel cross-section would be constructed using sequential excavation method to reduce excessive stresses on the lining left in place. CDM Smith assumed a sequential excavation on both sides of the existing tunnel would require a minimum width of 2.4 m (8 ft) for equipment access. The excavation width would likely result in a flat roof that would not be stable as a function of the sandstone structure. However, the sandstone could be made stable with rock bolts fully anchored into the overlying limestone rock.
• Excavation adjacent to the existing tunnel. To evaluate this condition, sequential excavation of tunnel adjacent to the existing tunnel to a width of 2.4 m (8 ft) was assumed. Additionally, assuming a relatively flat excavation roof this excavation can be made stable with rock bolts anchored into the limestone.

Results

System rehabilitation. The Phase 2 model results indicate that the existing tunnel structures are stable where the tunnel liner is in contact with the St. Peter Sandstone. However, accumulations of sand can be an indication of eroded sandstone behind the liner, causing additional stress on the liner. It is assumed that the reason for these deposits is the combination of deterioration of the wood supports and the natural behavior of the friable sandstone that causes the fine sand to erode as the ground water moves along the outside of the tunnel liner. The basis for this assumption is that the tunnel liner consists of unreinforced concrete that ranges in thickness from 178 to 350 mm (7 to 12 in.), based on the details shown on the tunnel plats provided by the city. There is no indication of expansion joints being installed in any of the tunnels. Inspections of the tunnels indicated that spacing of vertical cracks and transverse cracks averaged about 17.4 m (57 ft) apart, measured along the tunnel axis. Considering that 90 percent of the sandstone grain size is fine sand and can pass through about 90 percent of the observed cracks, the possibility of sand grains migrating and, thus, creating void spaces, as shown in the analyses, self‑perpetuates the deterioration of the tunnel lining as a function of surcharge loading. Based on the Phase 2 analysis, an increase in cracking frequency and crack width should be anticipated in locations where poor rock conditions or timber supports are present. This is supported by the locations where cracking was observed in the inspection data.

Some rehabilitation of the existing tunnel system is required in the form of repairing the cracks and filling any voids that are present behind the lining to stop further long-term deterioration of the liner. Determination of the locations and approximate volume of voids behind the liner could be conducted by an extensive geophysical survey from inside the tunnels. The results of such a survey would then be used to develop a program for repairs to the tunnel. This rehabilitation operation would mitigate the risk of failure for the existing tunnel system. However, it would not provide any increase in the hydraulic capacity of the tunnel. Therefore, the tunnel would still be subject to surcharge and street flooding.

System expansion. The modeling performed indicates that, where there is adequate sandstone cover, the tunnel can be enlarged laterally requiring minimal vertical expansion to create a stable shape. To maintain stability of the existing tunnel, which must remain in use during construction, external braces to support the tunnel liner would be required. Rock anchors and a new shotcrete liner would be required for tunnel support. Depending on the increased size of the tunnel, excavation can be performed either by a hydraulic lance or a roadheader. These excavation procedures will be discussed in greater detail later.

As the tunnels advance closer to the river and maintain their gravity slope, the sandstone thickness above the tunnel crown increases and there is adequate sandstone cover to expand the tunnel upward and maintain a cathedral shape for stability purposes. This expansion should be limited in height to maintain about 0.7 m (2 ft) of sandstone above the crown of the expanded tunnel cross-section. A first estimate of the cathedral shape can be calculated based on the friction angle of the sandstone. The height above the springline of the tunnel is about the sum of half the existing tunnel width plus the proposed increase in the width divided by tangent of the friction angle divided by two.

The advantage of a vertical expansion in the sandstone is that it eliminates the need for the rock anchors.

There are relatively short segments of the existing tunnels that are shown to have heavy timber support. Our interpretation of using this initial support system is that the rock is in poor condition relative to the other sandstone encountered in the CCSTS. These areas may require some additional ground modification such as grouting or using a welded wire mesh to prevent fall out of rock during the excavation for the tunnel expansion.

Proposed tunnel construction

As previously stated, construction of the Washington Avenue portion of this project is anticipated to begin in 2020. The preliminary alignment consists of both a parallel tunnel and some portions of the alignment where the existing tunnel will be expanded to meet the hydraulic capacity requirements. The required hydraulic capacity of the parallel tunnels would range in size from the equivalent of a 2- to 3.6-m (6.5- to 12-ft) internal diameter circular tunnel. The changes in the proposed size of the parallel tunnel, need to maintain flow in the existing tunnel and location of several cross passages greatly increase the complexity of the proposed construction. The following tunneling methods were considered for construction of a parallel tunnel:

• Hydraulic lance. The original tunnel construction used hand-held lances that emit highly pressurized streams of water that cut through the sandstone. The benefits of the approach are the ability to excavate in small spaces and to create noncircular shapes. The disadvantages include slower pace of excavation and limited number of contractors having experience with the hydraulic lance. Hydraulic lances are advantageous as a secondary method used for areas, such as transition structures, that will have a unique shape that cannot be created by a boring machine.
• Tunnel boring machine (TBM). Advantages include large boring face and efficient boring speed. Disadvantages include need for large -diameter access shaft, longer time to set up and inability to maneuver machine through tight radius curves.
• Road header machine. Advantages include smaller area needed for equipment installation, and ability to maneuver into non-circular shapes and non straight alignments. Disadvantages include slow rate of advancement and the need for more personnel in the tunnel.

The hydraulic lance method has not been used for several years in the Minneapolis area and finding labor and equipment using this method can be a limitation for this method. Generally, a TBM is a more economical method of tunnel excavation given the proposed length of approximately 1,066 m (3,500 ft). However, the alignment requires three 90° turns where new shafts would be required. The tunnel alignment also would be required to cross six lateral connection tunnels. An additional limitation for excavating the tunnel with a TBM is that the tunnel diameter is set by the machine. The required hydraulic capacity reduces to the west and therefore a TBM would perform unnecessary excavation.

The use of a roadheader also has limitations. These limitations are based on the size of the machine versus the excavation size. Roadheader power and ability to cut rock is a function of the machine size. To excavate a tunnel that is only about 2.4 m (8 ft) in height and of the rock strength presented in the modeling report a small machine will be sufficient. A hydraulic roadheader is able to excavate the rock into any cross-sectional shape that has been determined to be the most stable, creating tunnels that are able to obtain the required equivalent hydraulic capacity as it changes along the alignment. The other advantages are: it can make very short radius turns that would eliminate the need for a shaft extending to the street.

Use of the road header also allows for construction of a non circular tunnel and is particularly apt for constructing a tunnel with a non circular (cathedral or other) shape that takes advantage of the properties of the St. Peter Sandstone. The disadvantages to the roadheader are that the shape is not circular. Because of the high quartz content tool wear can be expected to lead to higher tool wear/replacement and advancement rate is less than that of a TBM.

Conclusion

Final design, construction and rehabilitation of the Central City Tunnel system will face many challenges: Contractor will be required to maintain the existing flow in the tunnel; Project is located within a densely populated urban setting with a myriad of shallow utilities, heavyvolume traffic streets making it difficult to locate shafts and staging areas; Unique engineering properties of the St. Peter Sandstone present their own challenges where the ground behavior can be unpredictable during construction. However, the expanded tunnel system will greatly improve the performance of the tunnel system; reduce surface flooding and annual maintenance cost.

References

Central City Tunnel System Hydrologic and Hydraulic Analysis Modeling Using XPSWMM, Central City, Eleventh Ave, and Chicago Ave Tunnel Systems, June 2015.

Central City Tunnel System Feasibility Study, Central City Tunnel System Pressure-Mitigation Options, June 2015.

Engineering aspect of the St. Peter Sandstone in the Minneapolis-St. Paul area of Minnesota, Charles M. Payne, University of Arizona, 1967. GSI: A Geologically Friendly Tool for Rock Mass Strength Estimation, P. Marinos and E. Hoek. 2000.

National Oceanic and Atmospheric Agency (NOAA) Atlas 14, Volume 8, Version 2 for Minneapolis.