Stability of I-Walls in New Orleans during Hurricane
Transcription
Stability of I-Walls in New Orleans during Hurricane
Stability of I-Walls in New Orleans during Hurricane Katrina J. Michael Duncan, M.ASCE1; Thomas L. Brandon, M.ASCE2; Stephen G. Wright, M.ASCE3; and Noah Vroman, M.ASCE4 Abstract: Failures of I-walls during Hurricane Katrina were responsible for many breaches in the flood protection system in New Orleans. Six breaches were examined in detail by Task Group 7 of the Interagency Performance Evaluation Taskforce. Four of these failures and breaches, which occurred before the water levels reached the top of the wall, were not caused by overtopping erosion. The failure of the I-wall at the 17th Street Canal resulted from shear through the weak foundation clay. The south failure of the London Avenue I-wall was caused by subsurface erosion, which carried massive amounts of sand inland, and removed support for the wall, leading to catastrophic instability. At the north breach on London Avenue, the failure was caused by high pore pressures, combined with a lower friction angle in the loose sand, which resulted in gross instability of the I-wall under the water pressure load from the storm surge. Looking back, with the benefit of 20-20 hindsight, these stability and erosion failures can be explained in terms of modern soil mechanics, exploration techniques, laboratory test procedures, and analysis methods. An important factor in all of the cases investigated was development of a gap behind the wall as the water rose against the wall and caused it to deflect. Formation of the gap increased the load on the wall, because the water pressures in the gap were higher than the earth pressures that had acted on the wall before the gap formed. Where the foundation soil was clay, formation of a gap eliminated the shearing resistance of the soil on the flood side of the wall, because the slip surface stopped at the gap. Where the foundation soil was sand, formation of the gap opened a direct hydraulic connection between the water in the canal and the sand beneath the levee. This hydraulic short circuit made seepage conditions worse, and erosion due to underseepage more likely. It also increased the uplift pressures on the base of the levee and marsh layer landward of the levee, reducing stability. Because gap formation has such important effects on I-wall stability, and because gaps behind I-walls were found in many locations after the storm surge receded, the presence of the gap should always be assumed in I-wall design studies. DOI: 10.1061/共ASCE兲1090-0241共2008兲134:5共681兲 CE Database subject headings: Walls; Louisiana; Hurricanes; Failures; Levees; Floods. Introduction I-walls are used to raise the level of flood protection without widening the footprint of a levee. As shown in Fig. 1, I-walls are constructed by driving steel sheet piles through the levee, often penetrating into the foundation soils. In some cases the portion of the I-wall that projects above the levee crest is encased in reinforced concrete. During Hurricane Katrina, I-wall failures resulted in breaches at many locations in New Orleans. Six of these beaches are listed 1 University Distinguished Professor, Emeritus, Dept. of Civil and Environmental Engineering, 200 Patton Hall, Virginia Tech, Blacksburg, VA 24061. E-mail: [email protected] 2 Associate Professor, Dept. of Civil and Environmental Engineering, 200 Patton Hall, Virginia Tech, Blacksburg, VA 24061 共corresponding author兲. E-mail: [email protected] 3 Brunswick-Abernathy Regents Professor, Civil Engineering Dept., The Univ. of Texas, 1 University Station C1792, Austin, TX 78712-0280. E-mail: [email protected] 4 Research Engineer, USACE ERDC, 3909 Halls Ferry Rd., Vicksburg, MS 39180-6199. E-mail: [email protected] Note. Discussion open until October 1, 2008. Separate discussions must be submitted for individual papers. To extend the closing date by one month, a written request must be filed with the ASCE Managing Editor. The manuscript for this paper was submitted for review and possible publication on May 14, 2007; approved on January 25, 2008. This paper is part of the Journal of Geotechnical and Geoenvironmental Engineering, Vol. 134, No. 5, May 1, 2008. ©ASCE, ISSN 1090-0241/ 2008/5-681–691/$25.00. in Table 1. The mechanisms of failure of these walls involved instability due to shear failure within the foundation clay at the 17th Street Canal and the Inner Harbor Navigation Canal 共IHNC兲 north breach on the east side, instability due to underseepage erosion and high uplift pressures in the sand foundation soils at London Avenue, and overtopping erosion that removed support for the walls at the IHNC southeast and northwest breaches. These breaches resulted in devastating flooding in the areas the walls were designed to protect. Most disturbing were the failures that occurred before the canal water level reached the tops of the walls. Following Hurricane Katrina, the U.S. Army Corps of Engineers formed the Interagency Performance Evaluation Taskforce 共IPET兲 to conduct a comprehensive investigation of the storm and its consequences. The writers worked on the team that investigated floodwall and levee stability. The findings of the investigation are detailed in Volume V of IPET 共2007兲 and the related Appendices. This paper summarizes the results of the IPET investigation that are related to limit equilibrium analyses, underseepage, and erosion of the 17th Street Canal and London Avenue I-walls. A key finding of the IPET studies was the fact that gaps formed at many locations on the flood side of the wall as the water level rose and the wall deflected, reducing stability of the I-walls. Fig. 2共a兲 shows an I-wall with clay beneath the levee. In this case, formation of a gap eliminates the shearing resistance of the soil on the flood side of the wall, because the slip surface stops at JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 / 681 Fig. 1. I-wall cross section the gap. In addition, the water pressure in the gap is higher than the earth pressures on the wall before the gap formed. Both of these results of gap formation lead to reduced stability of the wall. Fig. 2共b兲 shows an I-wall with sand beneath the levee. In this case, formation of the gap opens a direct hydraulic connection between the water in the canal and the sand beneath the levee. This hydraulic short circuit makes seepage conditions worse, and erosion due to underseepage more likely. It also increases the uplift pressures on the base of the levee and any impermeable layers landward of the levee, reducing stability. For these reasons, formation of gaps behind the I-walls reduces I-wall stability. Gap formation was found to be an important factor in all of the failures and breaches 共except for those due to overtopping兲 that occurred in New Orleans, and it was concluded that design studies for I-walls should always assume that a gap will form behind the wall. The following sections describe studies of the failures and breaches that occurred at the 17th Street Canal, where the foundation soil was clay, and at the London Avenue Canal, where the foundation soil was sand. 17th Street Canal I-Wall A photograph of the breach in the 17th Street Canal I-wall is shown in Fig. 3. The breach is about 450 ft long. The remaining 26,000 ft of the I-wall along the canal remained stable. A cross section at Station 10+ 00, the center of the breach, is shown in Fig. 4. The canal side of the levee was made lower than the protected side to improve stability toward the canal when the canal water level was low. A considerable number of borings had been made in the breach area and in neighboring areas before the failure. Additional borings have been drilled, cone penetration tests have been performed, and test pits have been excavated after the failure. The topography, before and after the hurricane, was established using LIDAR surveys. A compilation of all of these data is included in Appendix 1 of the IPET report 共IPET 2007兲. The cross section shown in Fig. 4 is based on the information derived from these explorations, and from laboratory tests performed on samples retrieved from the area. Several hundred unconfined compression tests and unconsolidated-undrained 共UU兲 tests have been conducted on the soils at the 17th Street Canal. Undrained shear strengths measured on samples from borings within and adjacent to the breach area are plotted against elevation in Fig. 5. The strength values shown in Fig. 5 were obtained from borings drilled at the centerline of the levee and at the toe of the levee. The data shown are from a number of different types of undrained strength tests: 1. One point Q is an unconsolidated-undrained 共UU兲 triaxial compression test, using one value of confining pressure. The point represents the strength measured in a single test; 2. Q is a set 共three or four test specimens兲 of UU triaxial tests performed using a range of confining pressures; 3. UCT is an unconfined compression test. The point represents the strength measured in a single test; 4. Crest strength interpretation is the strength profile beneath the levee crest that was used in the IPET stability analyses; and 5. Toe strength interpretation is the strength profile beneath the levee toe and beyond the toe that was used in the IPET stability analyses. Table 1. Soil Conditions and Failure Mechanisms at Investigated I-Wall Breach Locations Location Soil conditions Failure mechanism 17th Street Canal Clay levee fill/marsh/foundation clay/sand Stability failure through foundation clay London Avenue south breach Clay levee fill/marsh/dense sand Underseepage erosion of foundation sand leading to removal of support for I-wall London Avenue north breach Clay levee fill/marsh/loose sand Underseepage erosion and/or foundation instability due to high uplift pressure IHNC east bank south breach Clay levee fill/marsh/clay/sand Overtopping erosion of levee fill leading to removal of support for I-wall IHNC east bank north breach Clay levee fill/marsh/clay/sand Stability failure through foundation clay IHNC west bank north breach Clay levee fill/marsh/clay/sand Overtopping erosion of levee fill leading to removal of support for I-wall 682 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 Fig. 2. Potential I-wall failure mechanisms showing; 共a兲 foundation instability through clay; 共b兲 underseepage and erosion through sand Soil Properties The levee fill is compacted CL and CH material, with an average liquid limit of about 45 and a total unit weight, ␥t, of 109 pcf. Beneath the fill is a layer of peat-like material called “marsh” that is 5 – 10 ft thick. The marsh is composed of organic material from the cypress swamp that occupied the area, together with silt and clay deposited in the swamp. The average moist unit weight of the marsh layer is about 80 pcf. Beneath the marsh is a lacustrine CH clay layer, with an average liquid limit of about 92, a PI of 65, and ␥t = 100 pcf. A compilation of all of the laboratory data collected for the 17th Street Canal investigation can be found in Appendix 1 of the IPET report 共IPET 2007兲. The measured shear strengths of the levee fill scattered very widely, from about 120 to more than 5,000 psf. Placing the greatest emphasis on data from UU tests on 5-in.-diameter samples, which appear to be the best-quality data available, su = 900 psf is a reasonable value to represent the levee fill. Although the scatter in measured values is great, close analysis of the available data shows that the marsh deposit is stronger beneath the levee crest where it was consolidated under the weight of the levee, and weaker at the toe of the levee and beyond, where it was less compressed. The measured shear strengths of the marsh scatter very widely, from about 50 to about 920 psf. Values of su = 400 psf beneath the levee crest and su = 300 psf beneath the levee toe appear to be representative of the measured values. Considerable judgment was needed to interpret the strength test results because of the scatter. Fortunately, the factor of safety is not influenced greatly by the strength of the levee fill and the marsh materials. Field explorations after the failure showed that the rupture surface passed through the clay, beneath the levee fill and the marsh material. A photograph of the side of an exploration trench that was excavated at the toe of the slide mass is shown in Fig. 6. The dark marsh material can be seen both above the lightercolored clay and below the clay, although the clay is found only beneath the marsh in its undisplaced position. The lower part of Fig. 6 shows that the marsh-clay-marsh sequence was created where the rupture surface within the clay continued upward through the marsh. Lateral displacement of the sliding mass over the underlying undisplaced material results in the marsh-clay- marsh alignment after the failure. Age dating showed that the marsh above and below the clay was the same material 共IPET 2007兲. As can be seen in Fig. 6, the rupture surface passed through the clay and the marsh, but not through the levee fill. The clay is normally consolidated beneath the levee crest, and perhaps slightly overconsolidated beneath the toe. Although the results of laboratory strength tests performed on the clay were very scattered, much more consistent strength values were derived from the results of cone penetration tests with pore pressure measurement 共CPTU tests兲. Undrained shear strengths from four CPTU tests performed through the levee, all within 250 ft of the breach, are shown in Fig. 7. It can be seen that the four tests are in close agreement, and there is little scatter in the results. The laboratory shear strength results shown in Fig. 7 were collected from test specimens that were obtained from centerline borings. The undrained strength values shown in Fig. 7 were calculated from the CPTU test data using a method developed by Mayne 共2003, 2005兲. The writers have found that, where pore pressures measured in CPTU tests are of high quality, this method provides Fig. 3. Photograph of breach in 17th Street Canal I-wall JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 / 683 Fig. 4. Cross section of 17th Street Canal I-wall at Station 10+ 00 values of undrained strength that are consistent with laboratory tests on the best-quality test specimens, with less scatter in results. Where pore pressures measured in CPTU tests are of lesser quality, the method is not so effective. Experience with a large number of cone penetration tests performed to evaluate undrained clay strengths in the New Orleans area since completion of the IPET investigation have shown that it is often difficult to measure reliable pore pressures in CPTU tests due to problems with maintaining saturation of filters and other practical difficulties. As a result, production-level testing in the New Orleans area is now often being performed without requiring pore pressure measurements. Undrained strengths of clay are being computed by dividing the total cone tip resistance 共qc兲 by a “cone factor” Nc. It has been found that consistency with high-quality CPTU tests can be achieved using values of Nc in the range of 19–25 for the clays in the New Orleans area. Using the Nc method simplifies testing and produces a greater amount of useful data for expenditure of less effort, as compared with the more exacting CPTU tests. The undrained strength line shown in Fig. 7, which was calculated from the CPTU test data using Mayne’s 共2003兲 method, is consistent with the use of a value of Nc equal to 23. The undrained shear strength increases with depth at a rate of 11 psf/ ft of depth. Although there is a large amount of scatter in the results of the laboratory tests on the clay, there is very little scatter in the results of the CPTU tests, and these values thus provide a solid basis for establishing undrained strength profiles in the clay. In the IPET report, a total unit weight of 109 pcf was mistakenly used for the lacustrine clay. This value of total unit weight, combined with the 11 psf/ ft rate of increase of strength with depth, corresponds to a value of su / p⬘ = 0.24. We have since found that a more appropriate average value of the total unit weight would be about 100 pcf, which would result in su / p⬘ = 0.29. Owing to the fact that the clay is at the bottom of the slip circles analyzed, a change in clay unit weight has essentially no effect on the overturning moment and the computed factors of safety. Strength Model The IPET strength model, developed using the data discussed in the previous paragraphs, was as follows: Fig. 5. Laboratory undrained shear strength test results from crest and toe borings and strength interpretation for 17th Street Canal I-wall at Station 10+ 00 684 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 Fig. 6. Photograph of exploration trench at failure area of 17th Street Canal I-wall, and schematic of failure mechanism The undrained strength of the levee fill was su = 900 psf, u = 0; 2. The undrained strength of the marsh material beneath the levee crest was su = 400 psf, u = 0, decreasing to su = 300 psf, u = 0 at the toe. Beyond the toe, the strength was constant, su = 300 psf, u = 0; and 3. The undrained strength of the clay was taken as 0.24 times the effective overburden pressure at the top of the clay, and increased at a rate of 11 psf/ ft at all locations. Thus the strength was highest beneath the crest, decreased from the crest to the toe. The 0.24 strength ratio was determined from the rate of increase of strength with depth 共11 psf/ ft兲 and a total unit weight of 109 pcf. Two strength profiles are shown in Fig. 5: one for strengths 1. beneath the levee crest where consolidation stresses are higher, and a second for strengths beneath the toe and beyond, where stresses are lower. This IPET strength model involves two simplifying approximations regarding clay strength: 共1兲 By using the same rate of increase of su with depth, 11 psf/ ft, throughout the clay, it is implicitly assumed that the clay is normally consolidated throughout. While the clay beneath the levee crest is most certainly normally consolidated, it is perhaps slightly overconsolidated beneath the toe. 共2兲 By using effective vertical stress equal to simple overburden pressure at all locations, redistribution of stress within the foundation from the center toward the toe of the levee is ignored. These approximations tend to overestimate clay strength beneath the crest, and underestimate strength beneath the JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 / 685 I-Wall Stability Fig. 7. Undrained shear strength increase with depth in lacustrine clay layer interpreted from laboratory tests on crest boring specimens and cone penetration tests toe. However, a detailed study for two locations showed that these approximations have a very small effect 共about 2%兲 on calculated factors of safety. Because this difference is smaller than the reasonably expected accuracy of the strength evaluations and stability analyses, it was concluded that further refining the IPET strength model was not justified. Stability analyses were performed for a range of canal water levels, bracketing the measured height of the storm surge at the time that eye-witness accounts indicated that the failure occurred. The results of analyses with and without a gap between the wall and the canal-side levee fill are shown in Fig. 8. The slip circle shown in the figure is the one for the higher water level with a gap between the wall and the levee fill. The results shown in Fig. 8 were calculated using Spencer’s method 共Spencer 1967兲. The analyses were performed with the computer program SLIDE 共Rocoscience 2005兲, and were checked using the computer program UTEXAS4 共Wright 1999兲. Additional analyses were performed using noncircular slip surfaces with UTEXAS4. The critical noncircular surface was very similar in shape and position to the critical circle, and the factor of safety 共FS兲 for the noncircular surface was 6% lower. This 6% lower factor of safety corresponds to a water level for FS= 1.00 that is 0.8 ft lower than the FS= 1.00 water level found using circular slip surfaces. It can be seen that the computed factors of safety are lower for the higher canal water level, as would be expected, and are about 25% lower for the condition with a gap behind the wall than for no gap. Based on these results, and on the fact that gaps were observed at locations where instability did not occur and condi- Fig. 8. Critical circle determined from slope stability analysis for 17th Street Canal I-wall 686 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 Fig. 9. 共a兲 Photograph of south breach area of London Avenue Canal; 共b兲 photograph of deposited sand from south breach of London Avenue Canal tions could be examined after the hurricane, it was concluded that a gap did form behind the wall, and that gap formation was a key factor in the failure of the 17th Street I-wall. One of the most important conclusions of the IPET investigation is that gaps can form behind I-walls, and that these gaps significantly reduce wall stability. It is therefore prudent to always assume that a gap will form, and this condition should be considered in all I-wall stability analyses. London Avenue I-Wall South Breach A photograph of the south breach in the London Avenue east I-wall is shown in Fig. 9共a兲, and a photograph of sand that was washed through this breach into the neighborhood is shown in Fig. 9共b兲. The breach is about 60 ft wide. A cross section through the center of the breach area, before the failure, is shown in Fig. 10. The levee is founded on a layer of marsh that overlies a dense sand layer 共SP and SP-SM with an average D10 = 0.12 mm兲. Seepage Analyses and Uplift Pressures Owing to the high permeability of the sand in the foundation, underseepage effects are important at this location. Finite-element analyses of seepage beneath the I-wall were performed using the computer program SLIDE. All of the cases analyzed represented a condition in which deflection of the wall would open a gap behind the wall, from the levee crest down to the top of the sand. This condition is consistent with the observation that gaps formed in nearby locations where failure did not occur. The permeability of the sand, based on field pumping tests, was 1.5⫻ 10−2 cm/ s. The permeability of the marsh layer was estimated as 1 ⫻ 10−5 cm/ s based on consolidation test results, and the permeability of the levee fill and the Bay Sound clay was estimated as 1 ⫻ 10−6 cm/ s. Transient and steady finite-element seepage analyses show that: 共1兲 steady seepage through the sand was established quickly; and 共2兲 the pore pressures within the sand and the uplift pressures on the base of the marsh layer are not affected by the permeability values assigned to the marsh layer and the levee fill, provided that those materials are at least two orders of magnitude less permeable than the sand. The hydraulic boundary conditions used in the seepage analyses are shown in Fig. 10. Two canal water elevations were analyzed 共7.1 and 8.2 ft NAVD88兲, covering the range of estimated canal water levels at the time of failure. A constant-head boundary condition was imposed at the location of the drain beneath Warrington Drive. Two head values at this location were analyzed: −8.4 ft NAVD88 共the normal ground water level with pumps operating兲, or −5.1 ft NAVD88 共a higher level equal to the ground surface elevation, which might have been realized with pumps not operating兲. Reports indicate that the pumps stopped operating when the wall failed, severing the power line. A no-flow boundary condition was used at the canal center line. Computed pore pressures, or uplift pressures, at the base of the marsh layer are shown in Fig. 11 for the four cases analyzed, together with the total overburden pressure at the base of the marsh layer. It can be seen that in all four cases the computed pore pressures exceed the total overburden pressure at the base of the marsh layer beyond about 15 ft from the wall. This result indicates that the marsh layer would be heaved off the underlying sand by the high uplift water pressures. How events would proceed beyond this stage cannot be defined precisely. A likely result of upward heave of the marsh would be rupture of the marsh layer at one or more weak points, and upward flow of water and sand through the rupture. This flow would relieve the high water pressure locally, and create a new hydraulic boundary condition with high hydraulic gradients within the sand at the point of rupture. Although these hydraulic gradients cannot be evaluated precisely, they would certainly be high, and would undoubtedly be capable of eroding the sand upward into the breach in the marsh. This erosion would progress rapidly back toward the levee, resulting in rapid removal of material from the landward side of the levee, quickly leading to catastrophic instability, breach of the wall and levee, and inward rush of water through the breach. Though it is not possible to document the details of this failure sequence, because there were no eyewitnesses to its development and progression, it is consistent with the known facts, and with the great volume of eroded sand shown in Fig. 9共b兲. Slope Instability At the south breach the sand was dense 共␥t = 120 pcf兲, with standard penetration test blow counts greater than 50, which would correspond to friction angles in the range of 40–46°. Cone penetration tests performed after the breach showed high tip resistance in the sand adjacent to the breach, which correspond to similar values of ⬘. A value of ⬘ = 40° was used in analyses of JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 / 687 Fig. 10. Cross section of south breach area of London Avenue Canal the stability of the levee and I-wall, with the uplift pressures shown in Fig. 11. The calculated factors of safety are shown in Fig. 12. The marsh was treated as undrained, with su = 300 psf and u = 0, based on the available test results. A value of su = 300 psf is considered appropriate for the areas beneath the canal-side levee slope and beyond the levee toe, where the slip circles pass through the marsh. The average unit weight of the marsh is about 80 pcf at both the north and the south breaches. The levee fill was also treated as undrained, with su = 900 psf and u = 0. The slip circles do not intersect the levee fill, however, and the levee strength therefore has no influence on the calculated values of factor of safety. The average unit weight of the levee fill is about 109 pcf at both the north and the south breaches. Analyses were performed with canal water levels at 7.1 and 8.2 ft NAVD88, using pore pressures in the sand from finiteelement seepage analyses without a rupture through the marsh layer. At the bases of the slices where the calculated pore pressures exceeded the overburden pressures near the top of the sand on the inboard side, zero shear strength was assigned for the sand. As discussed earlier, it was assumed that deflection of the wall toward the land side resulted in formation of a gap through the levee fill and the marsh in back of the wall, down to the top of the sand. It was assumed that the gap would not extend into the sand, because the sand is cohesionless and would slump and fill the gap. Factors of safety against instability were calculated for the range of canal water levels estimated at the time of the breach 共7.1 and 8.2 ft NAVD88兲, and the two inland water levels 共−5.1 and −8.4 ft NAVD88兲. The calculated factors of safety ranged from FS= 1.19 to 1.56. Thus, based on the available data, a mechanism of failure involving erosion and piping is clearly indicated at the south breach, but a slope stability failure mechanism is not. An analysis was performed, with the landside water level at −8.4 ft, to determine the canal water level corresponding to a calculated factor of safety equal to 1.00. It gave a level of 9.7 ft, which is 1.5 ft higher than the highest estimated water level at the time the breach occurred. Thus, instability without removal of material by erosion and piping is unlikely at the south breach. London Avenue I-Wall North Breach Analyses of failure due to erosion and piping, and due to instability, were also examined for the London Avenue north breach on the west side of the canal, which failed about 1 h after the south breach 共IPET 2007兲. The differences between the London south breach analyses and the London north breach analyses were as follows: 1. The seepage boundary conditions were different. The canal water level at the time of the north breach was 1.1– 1.3 ft higher than at the south breach because the north breach Fig. 11. Computed pore pressures at base of marsh layer at south breach of London Avenue Canal 688 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 Fig. 12. Critical circle determined from slope stability analysis for south breach of London Avenue Canal Fig. 13. Cross section of north breach area of London Avenue Canal JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 / 689 Fig. 14. Critical circle determined from slope stability analysis for north breach of London Avenue Canal occurred later. The inland seepage boundary condition ranged from −8.4 to − 3.9 ft because Pratt Drive on the north breach is at a slightly higher elevation than Warrington Drive on the south breach. 2. The cross sections are somewhat different. On the inland side of the wall at the north breach, there is a thin layer of lacustrine clay between the marsh layer and the sand, as shown in Fig. 13. 3. The sand is less dense in the north breach area 共␥t = 115 pcf in the north area, ␥t = 120 pcf in the south area兲. Standard penetration test blow counts 共NSPT兲 in this area range from 2 to 14, with an average of about 10 blows/ ft. This range of values of NSPT corresponds to values of ⬘ in the range of 30–34°. Cone penetration tests performed after the breach, in the area adjacent to the breach, showed tip resistances that correspond to about the same values of ⬘. A value of ⬘ = 32° was used in the stability analyses for the north breach area. Finite-element seepage analyses were performed with a gap behind the I-wall, with canal water levels equal to 8.2 and 9.5 ft, and with inland water levels equal to −3.9 and −8.4 ft. As for the south breach, it was found that the calculated uplift pressures at the base of the marsh layer were larger than the overburden pressures, although the calculated uplift pressures did not exceed the total overburden pressures by as high a margin as at the south. The maximum uplift pressure at the south section was about 350 psf, compared to a maximum value in the north section of about 400 psf. Thus, while the same progressive failure mechanism of heave and rupture of the marsh, followed by erosion of the sand through the rupture is possible at the north breach, it would be expected that these events would not have progressed as rapidly or as vigorously as at the south breach. Slope stability analyses were also performed for the conditions at the north breach. The calculated factors of safety for these analyses are shown in Fig. 14. The values of FS for all four conditions analyzed are less than 1.0, indicating a high likelihood of instability at his location, even without erosion of material from the landward side. Conclusions Failures of the I-wall during Hurricane Katrina were responsible for numerous breaches in the flood protection system in New Orleans. Task Group 7 of IPET examined six of these breaches in detail. Four of these failures and breaches occurred before the water levels reached the top of the wall, and were therefore not caused by overtopping erosion. The analyses described here indicate that the failure of the I-wall at the 17th Street Canal resulted from shear through the weak foundation clay. It seems probable that the south failure of the London Avenue I-wall was caused by 690 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 subsurface erosion, which carried massive amounts of sand inland, and removed support for the wall, leading to catastrophic instability. At the north breach on London Avenue, it appears that the failure was caused by high pore pressures, combined with a lower friction angle in the loose sand, which resulted in gross instability of the I-wall. Looking back, with the benefit of 20-20 hindsight, these stability and erosion failures can be explained in terms of modern soil mechanics, exploration techniques, laboratory test procedures, and analysis methods. In all of the cases investigated, an important factor was development of a gap behind the wall. Formation of the gap increased the load on the wall, because the water pressures in the gap were higher than the earth pressures that had acted on the wall before the gap formed. Where the foundation soil was clay, formation of a gap eliminated the shearing resistance of the soil on the flood side of the wall, because the slip surface stopped at the gap. Where the foundation soil was sand, formation of the gap opened a direct hydraulic connection between the water in the canal and the sand beneath the levee. This hydraulic short circuit made seepage conditions worse, and erosion due to underseepage more likely. It also increased the uplift pressures on the base of the levee and marsh layer landward of the levee, reducing stability. Because gaps behind I-walls were found in many locations after the storm surge receded, and because gap formation has such important effects on I-wall stability, it should always be assumed in I-wall design studies that a gap will form behind the wall. In a companion paper in this volume, Brandon et al. 共2008兲 examine the effects of gaps on I-wall stability, and explain how gaps can be modeled in stability analyses. Acknowledgments The results of the investigation and analysis presented in this paper represent the efforts of many individuals involved in the IPET study initiated after Hurricane Katrina. Joe Dunbar, Reed Mosher, George Sills, and Ron Wahl, all of ERDC, provided valuable contributions to the work presented in this paper. References Brandon, T. L., Wright, S. G., and Duncan, J. M. 共2008兲. “Analysis of the stability of I-walls with gaps between the I-wall and levee fill.” J. Geotech. Engrg., 134共5兲, 692–700. Ineragency Performance Evaluation Task Force 共IPET兲. 共2007兲. “Performance evaluation of the New Orleans and southeast Louisiana hurricane protection system.” Final Rep. of the Interagency Performance Evaluation Task Force, U.S. Army Corps of Engineers, 具https:// ipet.wes.army.mil典. Mayne, P. W. 共2003兲. “Class ‘A’ footing response prediction from seismic cone tests.” Proc., 3rd Int. Symp. on the Deformation Characteristics of Geomaterials, Vol. 1, Swets & Zeitlinger, Lisse, 883–888. Mayne, P. W. 共2005兲. “Integrated ground behavior: In-situ and lab tests.” Deformation characteristics of geomaterials, Vol. 2, Taylor & Francis, London, 155–177. Rocscience, Inc. 共2005兲. Slide v5.0—2D limit equilibrium slope stability analysis, Toronto. Spencer, E. 共1967兲. “A method of analysis of the stability of embankments assuming parallel inter-slice forces.” Geotechnique, 17共1兲, 11– 26. Wright, S. G. 共1999兲, UTEXAS4—A computer program for slope stability calculations, Shinoak Software, Austin, Tex. JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2008 / 691