The phenomenon commonly referred to as “geysering” that occursin below-grade sewer systems manifests as an explosive release ofwater or an air/water mixture through vertical ventilation shafts ormanholes. Fig. 1 shows a geyser event in Minneapolis. Based onother objects such as street light fixtures in the complete images, thegeyser appears to rise at least 20 m above the land surface. Similarheights of rise are seen in images in posted videos from the ChicagoTunnel and Reservoir Project (TARP) and other systems. Geysershave also been reported in hydropower systems and were attributedto air entrained in a hydraulic jump within a portion of the pipingsystem (Nielsen and Davis 2009). However, no instrumentationwas installed in that system to document behavior during a geyseroccurrence.
During the early 1980s, geysers in the Chicago tunnel systeminitiated interest in development of analysis methods so that geyserdevelopment could be predicted during design and proceduresimplemented to avoid the occurrence. Numerical models of tran-sient mixed flow, where both free surface and pressurized flow
could occur simultaneously in nearly horizontal conduits, weredeveloped for this purpose, with the model by Cardle andSong (1988) being a notable early contribution to the literatureon the topic. Additional contributions in more recent years includeCapart et al. (1997), and Politano et al. (2007) These represent vari-ous approaches to the task of predicting the simultaneous occur-rence of mixed flow with sufficient numerical accuracy topredict the hydraulic bores that may form in a rapidly filling con-duit. In spite of the numerical sophistication associated with theimplementation of any of these models, it is not clear that they haveany link to the phenomena depicted in Fig. 1. To date, there is verylittle discussion in the literature regarding the physical processesinvolved in geysering. Guo (1989) and Guo and Song (1990)describe geyser formation as linked to inertial oscillations in theconduit flow. In Guo and Song (1991), the following discussionis presented: “If the water level rises above ground surface, the gey-ser occurs. It has been ascertained that if the dropshaft is ventilated,as most are, the cover could not be blown off by air pressure alone.That is, most blowoffs are caused by the impact force of the risingwater. Therefore, it is sufficient to study the hydrodynamics alone.”This statement does not address the fundamental question ofwhether geysers involve an air-water interaction, an issue of criticalimportance to design, since a numerical model that requires the hy-draulic grade line to rise to the ground surface in order for a geyserto occur could entirely miss an occurrence that is associated with airrelease. Several previous studies have considered air interactions infilling conduits (e.g., Zhou et al. 2002; Li and McCorquodale 1999;Izquierdo et al. 1999), but they do not directly address the issue ofgeyser formation. This paper presents information to indicate thatthe event depicted in Fig. 1 must be attributed to some interactionwith air; available laboratory observations and research on two-phase flow are used to suggest the nature of this interaction.
1Professor, Dept. of Civil and Environmental Engineering, Univ. ofMichigan, 113 EWRE, Ann Arbor, MI 48109-2125. E-mail: [email protected]
2Ph.D. Candidate, Dept. of Civil and Environmental Engineering, Univ.of Michigan, Ann Arbor, MI 48109. E-mail: [email protected]
3Assistant Professor, Dept. of Civil Engineering, Auburn Univ., 238Harbert Engineering Center, Auburn, AL 36844. E-mail: [email protected]
A number of laboratory experiments that involve the release oftrapped air pockets at a vertical riser have been performed bythe writers. However, the experiments were conducted for reasonsother than corroborating the field data and therefore do not consti-tute a reproduction of the visualized sequence of events in the videofrom which Fig. 1 has been extracted. Results of some of theselaboratory experiments have been previously reported by Vascon-celos (2005) and Wright et al. (2008). Vasconcelos (2005) per-formed a set of experiments that involved the setup indicated inFig. 2. These experiments involved isolation of a volume of airunder pressures somewhat above atmospheric in a pipe on one sideof a butterfly valve while the opposite side was filled with stagnantwater. Following a sudden opening of the valve, the air pocket in-truded into the water-filled portion of the pipe and propagated to thevertical riser that was partially surcharged with water. The air beganto rise into the vertical shaft because of its buoyancy, forcing thewater upward ahead of it while a portion of the water passed down-ward around the perimeter of the air bubble as a film flow, some-what resembling the flow at the front of a Taylor bubble such asdescribed by Davies and Taylor (1950). Depending on the exper-imental conditions, the water in the vertical shaft could be totallyincorporated into the film flow, in which case the water rise termi-
nated and the air escaped to the atmosphere. In other experiments,the film flow was insufficient to eliminate the water above the bub-ble, and the water rose until it spilled out the top of the shaft. Ineither case, the rising air bubble was accompanied by a drop inpressure within the pipeline, as indicated in Fig. 3, which is a pres-sure record for a case where the water overflowed the vertical shaft;the pressure drop to about atmospheric pressure toward the end ofthe pressure record indicates the air release out the top of the shaft.Here, the initial water level in the vertical shaft was about 0.25 mabove the pipe invert. The disturbance starting at about 7 s is as-sociated with the opening of the butterfly valve. The air pockettakes about 4 s to propagate to the vertical shaft, after which therelatively sudden pressure decrease down to atmospheric pressureis associated with air rise through the shaft.
A geyser resembling that depicted in Fig. 1 did not occur in thelaboratory experiments, although water was ejected somewhatexplosively from the top of the shaft under some test conditions.The difference is likely because of the significant difference in geo-metric scale between the laboratory and field applications. At thefield scale, the relative velocity between the rising air and the down-ward film flow can result in the phenomenon of flooding instabilityas described by Guedes de Carvalho et al. (2000). Flooding insta-bility arises because of the large interfacial shear between the risingair and the water film flow and could result in water droplets beingsheared off the falling film and lifted upward with the air flow. Atthe laboratory scale, the air-rise velocity is insufficient to result inthe flooding instability according to the criteria established byGuedes de Carvalho et al. (2000). It is estimated that the geyser
Fig. 1. Geyser event occurring at a manhole in a storm water tunnel inMinneapolis
Fig. 2. Apparatus for releasing a large air pocket into the surcharge pipeline
Fig. 3. Pressure head variation as the air pocket is released up throughthe vertical shaft
height is approximately 20 m from the video record, and the jet exitvelocity can be estimated by the simple conversion of velocity toelevation head as V2=ð2gÞ ¼ h with h the rise height of the waterjet. Using h ¼ 20 m would result in an exit velocity V of 19:8 m=s,which is much greater than required for the onset of floodinginstability.
Field Observations
System Description
The geyser image presented in Fig. 1 was extracted from a videorecorded as part of a data collection effort to investigate the recur-rence of geysering at two adjacent manholes in a storm-watercollection system that flows beneath the median of Interstate 35in Minneapolis. The video recorder was installed near a manholeat the 35th Street overpass and was accompanied by the installationof a rain gauge, as well as velocity and pressure transducerswithin the tunnel. Data were collected during the time period1996–2005, with 13 events recorded between 1999 and 2005.Individual events have involved more than one discrete geyser.Data are presented subsequently for a rainfall event that occurredon the morning of July 11, 2004, and involved a sequence of ninegeysers typically lasting 10–25 s, with intervals of approximately1 min between subsidence of one geyser and the initiation of thesubsequent one. The writers have other video records from thissame location that depict even stronger geysers and have differenttime intervals between geyser formation.
The storm-water tunnel is a 3.66 m diameter arch cross section,with the tunnel invert located 28.6 m below grade at the location ofthe 2.44 m diameter manhole. Velocities were recorded at 5 minintervals with an American Sigma area (through measured depth)velocity meter, and rainfall was recorded at 1 min intervals. Pres-sure transducers were installed at 0.47 and 2.88 m above the tunnelinvert. The data acquisition system for pressure was programmed torecord at 5 min intervals until the water depth exceeded 1.46 m.Data were collected every five seconds for water depths between1.46 and 2.93 m and every second for depths greater than 2.93 m.
Details on the horizontal tunnel geometry have not been madeavailable. Modifications to the two manholes that experiencedgeysering were made in 2009 to eliminate the phenomenon.
Results
On the morning of July 11, 2004, the pressure transducersincreased sampling frequency at approximately 5:22 a.m. Nineindependent geysers were observed in the video record of the event;although there is some variability, each geyser lasted about10–30 s, with about 75–90 s separating the onset of each one.The velocity record indicates that the velocity was relativelyconstant at about one meter per second between about 5:30 and8:50 a.m. Fig. 4 presents the pressure record for a half-hour timeperiod spanning the occurrence of the nine observed geysers. Thepressure record is from the lower pressure transducer and convertedto a pressure head relative to the tunnel invert. Superimposed on thefigure are the visual observations of the geyser occurrences indi-cated from the video record. These are indicated by the verticallines indicating the beginning and end of individual geysers; theheight of the line has no relationship to pressures indicated onthe vertical axis. The pipe crown elevation is also noted on Fig. 4.Fig. 5 provides more details on the pressure record by focusing on ashorter time interval spanning several individual geysers.
A number of observations can be made regarding the data pre-sented in Figs. 4 and 5. First, the pressure head relative to the tunnel
invert never went above about 6 m during the entire event. Since thetunnel invert is 28.6 m below grade, the discussion in Guo andSong (1991) described in the Introduction would exclude the pos-sibility of geysers because the hydraulic grade line never ap-proaches the ground surface. This clearly indicates that someother mechanism must be invoked to explain the observed geysers.Another observation is related to the estimated geyser discharge. Ifthe jet exit velocity is estimated at 19:8 m=s as discussed above,this would require an upward discharge of 93 m3=s through themanhole. This compares to a tunnel discharge of less than 9 m3=sindicated by the measured tunnel velocities during the geyserevents. It seems unlikely that a water discharge even approachingthe tunnel discharge could be supported up through the manhole,indicating that the jet of water must contain only a small percentageof water by volume and consists mostly of air. This possibility al-lows the development of the estimated large discharge velocity withrelatively low tunnel pressures. Consider that a pressure differencecan drive an air-water mixture according to the approximationΔP ∼ ρmV2 with the air/water mixture at a density ρm ≪ ρw, withρw the water density. Using this relationship, a maximum observedpressure head (relative to the tunnel crown) from Fig. 4 of about1.55 m and the previously estimated jet velocity of 19:8 m=s, therequired ρm would be about 40 kg=m3, implying a water content ofthe jet on the order of 4%, which seems plausible.
Since the data do not support the notion of a solid column ofwater being lifted to the ground surface and jetted several metersinto the air, release of a series of entrapped air pockets is consideredto be the only plausible explanation for the geyser formation.Assuming that the tunnel becomes pressurized when the pressurehead in Fig. 4 exceeds the tunnel diameter, this occurs about sixminutes prior to the commencement of the first geyser event.The last geyser event ended about two minutes before the systemtransitioned back to a free surface state, and thus surcharged con-ditions existed in this portion of the tunnel for the duration of thegeyser event. As shown in Fig. 5, the pressure tends to dropabruptly at the onset of each geyser. According to the values pre-sented, the pressure head fell to nearly the same level during eachgeyser, a little greater than the pipe crown elevation. Given theuncertainty in the pressure measurements, one cannot excludethe possibility that the pressure head falls to the pipe crown, pre-sumably during the final stages of the expulsion of a discrete airpocket from along the pipe crown. This pressure variation hasstrong similarities to the laboratory results presented in Fig. 3.Following the cessation of the geyser, the pressure tends to riserelatively quickly and assume an overall trend that seems to beoccurring in the absence of the individual geysers. This behavioris inconsistent with a “hydraulic” or water-flow-only explanation of
Fig. 4. Pressure head relative to tunnel invert recorded during geyserevent of July 11, 2004
geyser formation but would be consistent with the arrival ofdiscrete air pockets at the manhole with a pressure drop as the airreleases into the vertical shaft. Although this is not conclusiveproof, the data are far more consistent with an air-water interaction,specifically with the expulsion of discrete air pockets through themanhole. The apparent regularity of the geyser events displayed inFig. 4 is not necessarily consistent with observations during otherevents on other dates with different time intervals between individ-ual geysers. The writers believe that the geometry of the particularsystem results in the propagation of discrete air pockets from thedownstream direction. Complete horizontal geometric details of thesystem were not provided to the writers, but it is understood that thelocations where the geysers have been observed are in a relativelyflat portion of the tunnel with steeper sections both upstream anddownstream. This geometry would allow the development ofmildly surcharged conditions in the flatter section while maintain-ing free surface flow on either side, providing a source for the airthat subsequently migrates along the tunnel crown to the manholeunder observation.
Conclusions
The field measurements collected during a rainfall event inMinneapolis on July 11, 2004, included visual observations ofthe manhole at the ground level along with velocity and pressuremeasurements within the storm-water tunnel near the location ofthe manhole. The measured pressures within the pipeline were inca-pable of lifting water in the manhole to even close to the groundsurface, let alone eject it 20 m into the air. Converting the estimatedrise height of the geyser into a vertical velocity in the manholeimplies a manhole discharge that is more than an order of magni-tude larger than the water flow measured to be flowing within thetunnel at the time of geyser observations. Furthermore, the mea-sured pressures and velocities do not indicate the presence oflarge-scale inertial oscillations within the tunnel. The only plausibleexplanation for the observed geyser formation is the interaction oftrapped air with water initially standing in the manhole because ofexisting surcharge conditions. The patterns of water pressure varia-tion during air expulsion through the ventilation shaft are similar inthe field and laboratory measurements, suggesting that in spite ofpossible differences in flow behavior because of the small labora-tory scale, there are similarities in the rise of trapped air pockets inthe vertical shaft in both situations.
Several implications are associated with these findings. The pro-cess of geyser formation, at least in this application, is apparently
not directly connected with surging in the tunnel system, as sug-gested by some of the previous literature on the subject. Numericalmodels that are currently applied to simulate transients in rapidlyfilling tunnel systems do not account for the air phase. The pre-dicted results of these models should be interpreted with cautionwith respect to the issue of geyser formation. Predictions that tran-sient hydraulic grade lines remain below grade should not beinterpreted to suggest that geysers will not occur. Model capabilityto predict the location of air entrapment within a system is useful,even if the subsequent motion of the air cannot be predicted with asingle phase flow model. This information can be used judiciouslyto make design decisions about the location and capacity of airventilation required in a system.
Acknowledgments
The writers would like to acknowledge the cooperation ofDr. Christopher Ellis (from the St. Anthony Falls HydraulicsLaboratory) and Bruce Irish of the Minnesota Department ofTransportation in providing the field data used in this manuscript.
References
Capart, H., Sillen, X., and Zech, Y. (1997). “Numerical and experimentalwater transients in sewer pipes.” J. Hydraul. Res., 35(5), 659–670.
Cardle, J. A., and Song, C. S. S. (1988). “Mathematical modeling ofunsteady flow in storm sewers.” Int. J. Eng. Fluid Mech., 1(4), 495–518.
Davies, R. M., and Taylor, G. I. (1950). “The mechanics of large bubblesrising through extended liquids and through liquids in tubes.” Proc. R.Soc. London, Ser. A, 200(1062), 375–390.
Guedes de Carvalho, J. R. F., Talaia, M. A. R., and Ferreira, M. J. F. (2000).“Flooding instability of high-density gas slugs rising in vertical tubesfilled with water.” Chem. Eng. Sci., 55, 3785–3802.
Guo, Q. (1989). “Geysering in urban storm drainage systems.” Proc. XXIIIIAHR Congress, National Research Council, Ottawa, Canada,S75–S81.
Guo, Q., and Song, C. S. S. (1990). “Surging in urban storm drainagesystems.” J. Hydraul. Eng., 116(12), 1523–1537.
Guo, Q., and Song, C. S. S. (1991). “Dropshaft hydrodynamics undertransient conditions.” J. Hydraul. Eng., 117(8), 1042–1055.
Izquierdo, J., Fuertes, V. S., Cabrera, E., Iglesias, P. L., and Garcia-Serra,J. (1999). “Pipeline start-up with entrapped air.” J. Hydraul. Res., 37(5),579–590.
Li, J., and McCorquodale, A. (1999). “Modeling mixed flow in stormsewers.” J. Hydraul. Eng., 125(11), 1170–1180.
Nielsen, K. D., and Davis, A. L. (2009). “Air migration analysis of theTerror Lake tunnel.” Proc., 33rd IAHR Congress, International Asso-ciation for Hydraulic Research, Madrid, Spain, 262–268.
Politano, M., Odgaard, A. J., and Klecan, W. (2007). “Case study:Numerical evaluation of hydraulic transients in a combined seweroverflow tunnel system.” J. Hydraul. Eng., 133(10), 1103–1110.
Vasconcelos, J. G. (2005). “Dynamic approach to the description of flowregime transition in stormwater systems.” Ph.D. dissertation, Dept. ofEnvironmental Engineering, Univ. of Michigan, Ann Arbor, MI.
Vasconcelos, J. G., Wright, S. J., and Roe, P. L. (2006). “Improvedsimulation of flow regime transition in sewers: Two-component presureapproach.” J. Hydraul. Eng. 132(6), 553–562.
Wright, S. J., Creech, C. T., Lewis, J. M., and Vasconcelos, J. G. (2008).“Mechanisms of flow regime transition in rapidly filling stormwaterstorage tunnels.” Environ. Fluid Mech., 8(S16), 605–616.
Zhou, F., Hicks, F. E., and Steffler, P. M. (2002). “Transient flow in arapidly filling horizontal pipe containing trapped air.” J. Hydraul.Eng., 128(6), 625–634.
Fig. 5. Details of pressure variations during geyser
