We focus here on a particular (and deadly) type of levee failure that is characteristic of California rivers and displays the following characteristics:

-associated with landside seepage and sand boils, often at considerable distance from toe, during floods;

-failure at up to one day after flood peak;

-boils at levee toe enlarging in the last hour before failure to large sand spouts;

-loss of levee toe by sinking or sliding followed by levee subsidence and overtopping.
flood vtech:2.5:lf106x.gif:363


California's Central Valley and delta region is a nice fit to this geographic pattern of levee failure. (Is the delta region really a "delta"? Not exactly in the original "shape" sense of the word, coined by Herodotus when he realized that the form of the Nile delta was a giant greek letter delta. But geomorphically the place is in fact a delta, that is an accumulation of flat sediments on the edge of a basin, in this case the hill-pinched upper estuary of San Francisco Bay. Which means that it shares much of the history and stratigraphy of other deltas and estuaries of the world, and has conditions that match the pattern we shall adduce here.)
flood vtech:2.6:lf124.gif:304


Interesting similarity: Genesis 1 states that Eden (Sumerian for "plain") is sited at the meeting point of four rivers. The Gihon (present-day Karun) winds through Cush (Kasibites, Iran) and the Pishon now is a dry wadi from Havilah (now Arabia.) Today the junction is a vast marsh a hundred miles inland of the Persian Gulf, inhabited by the Shiite marshland Arabs, scene of the ancient cities of Ur, Uruk, and Eridu, and more recently the site of the Gulf War.
oaks x vtech:2.7:lf61.gif:205


A classic geographic urban pattern beginning with the ancient Sumerian city of Ur (of the Chaldees, of biblical fame) and carrying on to Tokyo, San Jose California, Galveston Texas etc. involves urbanization of lower floodplains, diversion and diking of rivers, and (especially in recent times) extraction of groundwater leading to subsidence. River and tidal surfaces end up well above ground levels in cities, creating problems of both underground flooding (affecting underground space and destabilizing the ground) and surface flooding of the depressed urban basin arising from levee breaches in susceptible areas upstream.
flood prevent vtech:3:lf140.gif:353


Certain patterns of subsurface geology and seepage coincide with this history, as demonstrated in the 1940s by Fiske for the Mississippi River and elaborated by other research on levee failures and flow slides conducted by the Corps of Engineers over the years. Holocene evolution of lower valleys in California is somewhat different than in the Mississippi because of smaller watersheds, tailings deposition, and tectonically defined coasts. Nonetheless mid-Holocene river evolution still gives rise to a classic fine-grained topstratum with effective thickness of 20 ft. or so. Depending on local conditions, high upward seepage pressures with associated instability may develop at relatively shallow depths beneath the toe (and throughout the landside floodplain) during or shortly (up to 36 hours) after high water.
flood geo prev vtech:4:lf138.gif:355


The Sacramento Riverconsidered longitudinally: Pleistocene gravel channels are found 25 feet below the ground in the floodplain. The gravel was deposited in fast moving braided streams when the sea was 300 feet lower (until about 7000 years ago). When sea level stabilized, a delta formed and the rivers switched to a meandering pattern, blanketing the gravels with silt.
flood geox vtech:4.5:profile2.gif:209


Notice the similarity to the stratigraphic section of the Holocene deposits in the lower Tigris-Euphrates basin
flood vtech:4.6:icoast1.gif:369


A longitudinal profile of a 20 mile stretch of the Feather River centering on the Arboga failure at Country Club Road. Notice that at both the Country Club Road and Linda areas (where notable levee failures occurred in 1997 and 1986 respectively, each causing on the order of 1/2 billion dollars in damage) the levee is built on a silt (and tailings) layer (brown color) which is roughly as thick as the levee is high. Beneath this silt and tailings is the coarser and more pervious gravel from the Pleistocene.
flood geo vtech:5:profile1.gif:171


Claim: The failure behavior of the levees during flooding demonstrates that the upwardly fining gravel layers (Laguna Formation?) at 20-30 ft beneatht the floodplain (which is deeper than usually explored in California levee safety investigations) play a critical role in failure. High seepage pressures and ample flow delivery capacity is brought to the levee toe behavior and beyond.
flood geo vtech:6:lf7.gif:195


Supporting evidence: this is the only way that sand boils and collapse features could occur hundreds or thousands of feet on the land side of failure-prone levees. Here is one (discovered by my 11 year old son Kevin) being photographed for KPIX a few hundred feet downstream from the Arboga collapse of January 1997 (video available). Ours were the first footprints in this orchard, many weeks after the failure.

animation/enlargement

flood geo vtech:8:crew.gif:196


Implication: The repressurized gravel layer pushes the potentials toward the landside and up into the sensitive foundation area. Seepage conditions in flood time are not as often assumed in (a), but as shown in (b). The gradients increase progressively with time. When they exceed 0.5 sand boils become active and the toe foundation and lower slope become mushy and unstable, effectively forming quicksand.
flood mech vtech:8.5:lf79.gif:220


In anticipating failure D, the fatal combination occurs where the levee is built over the floodplain, and is built on a silty cap (covering old gravel), that is relatively thin. These areas are marked by low ground and high groundwater, C.

Hence diagnosis of areas with acute failure risk is based on recognition of (a) levee configuration on flood plain, (b) history of seepage symptoms, (c) details of Holocene and man-made stratigraphy.
flood geo vtech:12:lf20.gif:149


Consider point (a): If you knew about the 1955 and 1986 failures, could you suggest where a future failure might be?

Notice how the levee extends well out into the floodplain (yellow) at the failure locations. This is a legacy of the "levee wars" of the late 19th century. Land owners hoped to push the river over into the other county. The situation became so unfriendly that levees were sabotaged by dynamiting.

These same areas have been subject to well-documented, flood-induced seepage (which damages orchards) for decades.
flood geox vtech:14:usgs4.gif:359


We can see point (c) most dramatically in another flood-prone river, the Pajaro River. Note how the 1995 failure, shown on the news clip to the right, occurs at a point where the county line (1890 river channel) crosses the river and the topstratum is accordingly thin and vulnerable to seepage. Note that the floodplain at the failure area is at 50 ft above sea level.
paj1 vtech flood:14.1:compare.gif:97


Consider this: 1995 the U.S. Army Corps of Engineers planned to reinforce the levee at Country Club Drive. Nonetheless, laws required that the work could not begin on the levee repair until an artificial pond was constructed for environmental "mitigation" purposes.
corps vtech:14.6:lf53.gif:362


Here is visual comparison between the 1986 and 1997 levee failures and the location of pits dug in the floodplain near the levees. Both photographs are the same scale.

Artificial pits were present at both failure sites as shown in the photographs (same scale). Pits are 1000 to 2000 ft. away from the failure points, in line with old channels that pass beneath the levees.
flood mech4 vtech:16:linda.gif:28


General observation: areas of high risk oftenappear to correspond to areas that are otherwise suitable for preventive strategies emphasizing targeted land use controls, floodway purchase options, insurance, and controlled failure.
flood prevent vtech:17::349


Without these long-term measures flood prone lands will continue under heavy development pressure (as here, where the site of the 1955 levee failure (28 fatalities) at Yuba City is for sale to developers. The levee has not been improved.)

Note that flood insurance is not required, the site being outside the levee and therefore protected against flooding. Ironically, the history in the Yuba City-Marysville area demostrates that the return period for extremely dangerous flooding arising from levee failure is close to 15 years. (1955, 1986, 1997)

Nonetheless, vast tracts of land are scheduled for development in these high-risk areas.
flood shang vtech:38.5:sale.gif:46


It appears likely that the discrimination of high risk areas (most susceptible 10-20% of levee reaches) is feasible with good reliability. Also, the proposed model should allow for improved risk-stage relationships in this type of fluvial environrment
flood prevent vtech:90::358


Short-term or flood-fighting techniques suggested by these findings:

A stategy of pressure relief in susceptible levee reaches;

Identification of susceptible top strata (pressure, stratigraphy, defects, etc)

Improved and economical pressure relief installation and monitoring;

Legal and O&M issues;

Longer term possibilities: identification of criteria for land use strategies (floodway easements, etc)

Note that cutoffs and shallow toe drains will not be effective for the condition which we claim to exist in the Central Valley of California. Deep drains and berms are more effective.
flood vtech:100:lf79a.gif:365


Quiz: now that you have learned all this, can you tell why the pattern of levee failures of January 1997 is the way it is? Would it help if you knew that the failures here (and at the Pajaro River in 1995) all took place where the floodplain elveation was about 50 feet above sea level.
flood geo vtech:150:failmap.gif:360



A near-disastrous experience in 1963 in Thailand led me to a source of wisdom: the famous engineer P.T. Bennett. Bennett showed me how tampering with the upstream natural blanket could lead to disaster. This is what led to my interest in levee failures.

animation/enlargement

Richard L. Meehan Richard Meehan has engineering degrees from M.I.T. and Imperial college, University of London. Following service with the U.S. Army Corps of Engineers he began designing dams and levees in Southeast Asia in the early 1960s, and then in the Western U.S. and South America. He has maintained a consulting engineering practice in Palo Alto and taught at Stanford University for the past twenty five years. 701 Welch Rd. #1120 Palo Alto Ca 94304 415-323-0525
flood vtech:200:rlmsk.jpg:364


Questions or Comments?

meehan@blume.stanford.edu

Description of the 1997 Arboga failure.