LESSONS FROM NOTAELE EVENTS 45

6l: exploration exposures in borings and adits were inad- three borings wcre drillcd in 1960 to "locate" the zone and
equately identified or misunderstood. Evidently guidance evaluate any relevance to pmject (Sr-172 mcters: S:-7 I
by project-experienced, exploration personnel was not metes: and 53- 105 metes). All thrce borings yielded very
available to provide mature geological judgment and eval- low core recovery, and each terminated in raveling. broken
uation of features. For example, the low-or no-core re- ground when the hole collapsed*likely evidence of broken
covery strongly implies soft sediments. clays, cavities. and/ rock due to a "tectonic" feature. Yet in spite of the sketchy
or broken rock as associated with fault zones (assuming core and general lack of data, the geotechnical advisors
lcsses not to be mechanical); but subsequent geological de- arrived at the consensus that there was no evidence for
cisions did not reflect a serious concern. Moreover. the ad- preexisting sliding, and the slope was stable. Miilled''o lik-
its of 196l crossed a strongly sheared zone (described as ened movement at the site to that of a glacier with zero
mylonized rock) that post-slide explorationl determined to velocity at the base and increasing upward. an interpreta-
be previous slide planes. The lack of water in this adit in- tion that affected any hopes realistically to control the sub-
fluenced yet another decisionro'a that "it was not practical sequent large-scale movements of 1963.
to stabilize slide by drainage." Actually, field studies Vaiont, like Malpasset in 1959 (described earlier in this
would have indicated that the heavy surface runoffon slopes chapter), reflects the tendency of some Europeans of that
of Mt. Toc largely disappears upslope of the adit and nearby period to restrict geological exploration wherever possible
slide failure plane in "karstic terrain" (Figs. 2-28, 2-29) to a minimum input for a dam-reservoir project, in spite
and the interconnected fractures and openings of lime- of on-site geological circumstances. Furthermore, Vaiont
stones. This recharge enhances the groundwater storage and demonstrates the long-standing willingness of some engi-
level throughout the slide mass, and it explains why there neers to accept the opinion of a ry!!]-kngur!_pro&ssiong!
is no large surface flow across and downslope of the slide 9f jlqilrdsa_lla!_e4e_qCq!9:_qr_lieqo!t!g_lgamapproac!_
plane.3
One early explanation for the slope collapse, in Novem-
undlhe iIg4gpryg_og11$l
I p.y,d"_r_!"" _Chapter 4).
ber 1963," implied that the sliding was an "act-of-God" Geological Setting
and could not be predicted or evaluated. Yet the geological
setting. history of sliding, ongoing movements, and poten- Rocks of the Area. The Vaiont region is characterized
tial could be examined in quantitative terms, and were ev- by a thick section of sedimentary rocks of Middle Jurassic
ident to some investigators. to Upper Cretaceous a-qe, as shown in Figs. 2-28 and2-29.
The lack of mature geological guidance during the plan- The rock column consists mostly of limestones or dolo-
ning-design phase was likewise raised during the litigation mites and alternating limy or marly layers with some clayey
proceedings of 1969 to establish "responsibility for disas- interbeds and partings. The main units, from oldest to
ter. ' ' The Italian courts accepted the testimony of Nino A. youngest, arel
Biadene. director of construction for Societa Adriatica Di
Elettricita (SADE), that "inadequate and misleading geo- o Lias Limestone-thin-bedded with partings of marl;
logical information was a major factor in causing the trag- 80 to l0O meters thick.
edy."t: . Dogger Limestone/Dolomite-thick/massive beds.
A major geological contribution of the 1978-82 partings; partly oolitic, intensely fractured. exhibiting
reevaluationl of Vaiont is the release of details on the 1959- stress relief, locally permeable/porous from solution
60 geological mapping of canyon slopes by Guidici and action: 350 meters thick. Dam foundation.
Semanza.rs This initial field mapping (not available in . Malm Limestone-thin-bedded, marly, cherty, soft.
1963) located sketchy evidence for a large-scale ancient easily folded, weak beds: 23 to 25 meters thick.
slide plane and/or a decollement feature (over 10,000 years o Upper Malm-Lower Cretaceous-medium-bedded.
ago) along the left inner-gorge wall and within the site pro- compact, cherty. clayey interbeds, a weak unit; 40 to
posed for the reservoir (Fig. 2-30). The translated text with 56 meters thick.
map sections is given as Appendix G.r An early confir- o Lower Cretaceous-marly limestone with green clayey
mation and delineation of the suspected ancient slide plane interbeds: parts soft. intensively fractured: 40 to 60
was attempted in 1959, utilizing a geophysical survey: Pro- meters thick. General location slide planes.
fessor Pietro Caloria'rl concluded that the left canyon wall . Upper Cretaceous-a scquence of brecciated lime-
was "in-place" and stable. This opinion, based on sketchy, stone. marly limcstone, and cherty limestone; gener-
indirect data, unfortunately caused SADE to be temporar- ally thin bcds, clayey interbcds common, with thick-
ily confident that the left reservoir wall was satist'actory. nesses of 10. 4, 1.5, 100. 20, and 3@ meters.
However, geologist Semanza maintained his concern. and
further field studies provided strong indications of an an- The 1963 slide plane is located within Lower Cretaceous
cient slide plane along the left canyon wall. Consequently marly units (Fig. 2-29). A detailed description of the strat-
46 ADVANCED DAM ENGINEERING FOR DESIGN, CONSTRUCTION, ANO REHABILITATION

Fig. 2-30. (A) View of Vaiont Canyon before dam. showing spectacular
V-shaped inner gorge carved in limcstones and topographic fearures of
reservoir area. Thc crest of the dam and the top of rhe 1963 slide arc
supcrimposed on the sire. Source: Kiersch ,t Civil Engineering. ASCE.
New York. (B) View of reservoir site from crest of dam taken in 1961.
showing exposure of 1960 slide and the location ofOld slide plane (reac-
tivated by 1963 slide), as well as the rugged featurcs of thc reservoir.
Source: Gerald T. McCarthy, 196t. TAMS. New York.

B
 TESSONS FROM NOTAELE EVENTS 47

igraphic column and subdivision of individual formations south limb of the Erto syncline forms a distinct "flat-seat."
into distinctive units is based on references 8 and 14-16, an important arcal feature rclevant tu the slide arca (Fig.
and a summary given in reference 3. 2-29\.
2. The Erto asymmetrical syncline became the natural
site for the ancient Vaiont River. Carving of the outer val-
Post-slide studies, 1963-85. A great amount of geo- ley was initiated before late Tertiary time. and probably
logical exploration has been completed throughout the site
minor solutioning of the limestone also began then.
and area since October 1963, including mapping, some l0O
3. During Pleistocene time, glaciers were active in
borings, test pits, adits, testing, and model studies. The
scouring the outer valley (Fig. 2-32). Adjustment to this
extensive explorations of 1964 through the 1970s pmvided
rapid unloading affected the balance of stresses and formed
additional refinements and detailed evidence on selected
a strong set of rebound joints roughly parallel to the walls
site-specific features, as given by Miiller,a's Rossi and Se-
and the floor of the outer canyon.
manza,'o Semanza,s and Broili;6 while Selli and Trevi-
4. In the past 18,000 years, glaciers disappeared from
san,15 Carloni and Mazzanti,16 and CiabattilT provide doc-
the outer Vaiont Valley, and erosion by the Vaiont River
umentation on seismological data and dynamic evaluation
has been rapid. Today, the channel is downcut up to 984
of the slide. The pertinent points and detailed contributions
ft (300 meters) below the glacial valley. The inner canyon
ofeach investigator and others are given in reference 3 and
was carved so rapidly that inherent stresses in the rocks
not repeated here.
were not fully adjusted by stress relief, giving rise to the
present imbalance of stress as encountered in shallow ex-
Clay lnterbeds and Layers. The most significant char- cavations of channel rock for the dam foundation.
acteristic of the rock column relevant to large-scale sliding 5. The change in natural groundwater level, fmm the
is the location, continuity, and physical properties of the outer to the inner valley stream levels (Fig. 2-32). further
clayey interbeds, a feature initially reportedr and later aided the subsurface solution action on limestones and
emphasizedT'8 as critical to slope stability. A detailed dis- marly beds.
cussion on the occurrence of clayey interbeds, as to strat- 6. In prehistoric time. a large landslide occurred along
igraphic and areal locations, continuity, thickness. associ- the southem margin of the inner Vaiont Valley from near
ated small-scale structural features, physical properties, and the damsite to the Pineda sector; at that time rocks involved
mineralogical composition, is given by Hendron and Pat- with the later 1963 slide were distoned. folded. and bro-
ton.3 Ma.lor slide planes invariably have sedimentary clays ken, and the ancient river channel was dammed. Relics of
and likely some mechanical clay associated with the zone this slide mass remain today on the right canyon wall of
(Fig. 2-31). The thickness and frequency of clay interbeds the present channel near the dam (Figs. 2-27 through 2-
decrease below the bedrock-slide debris contact. 29). Ancient slides elsewhere in the valley were common,
Post-slide borings by Ente Nazionale per I'Energia Elet- such as the one near Casso (Fig. 2-27) and talus infilling
trica (ENEL) encountered clayey debris, intercalated clay, of the crevasse at the head of the 1963 slide.l
and in one hole (S6C) 4 meters of clay with rock fragments 7. A similar prehistoric slide occurred in the Piave Val-
at the base of the slide mass.3 Clearly, muttiple layers of ley to form Santa Croce Lake. and rerouted the Piave River
weak clay are associated with the major surfaces of sliding. through Belluno to Venice (see locality map, Fig. 2-27).
8. A much older period of tectonic faulting (decolle-
ment) appears to have taken place throughout the area of
Events-Features Affecting Stability of Reservoir the ancient slide, with displacement in the same weak zone
Slope of clayey interbeds.E as, forexample. in the prominent fault
that bounds the 1963 slide area on the eastern margin (Fig.
Geologic Events. The Vaiont region has been modified 2-33).
by many events throughout geological time, which have
weakened the reservoir rocks in some manner. Conse- lnherent Structural Features-Retained Stress. The
quently the site is an ideal case to demonstrate some of the young folded mountains of the Vaiont region retain a part
potential problems for reservoirs (based on refiercnce l). of the active tectonic strcsses that deformed the rock se-
l. The region underwent strong mountain building qucnce. Furthermorc, a pattern of rebound joints devel-
forces in mid-Tertiary time. The thick sedimentary rock oped bencath thc floor and walls of the outer valley as ero-
section (Fig. 2-28) was deformed into a series of east-west sion progressed (Fig. 2-32): this dc-strcssing created a zone
regional folds with accompanying thrust faults. Later, of weaker, highly fractured and "laycrcd" rock some 328
cross-folding (Fig. 2-29) tilted all beds to the east along ft (100 mcteni) thick. which is accentuated by the natural
the Erto synclincal axis (within the sector of the dam and dip of rock units. Below this, the unclisturbed rock pos-
reservoir today). An abrupt monoclinal-like flexure on the sesses relatively high natural ambient stresses.
 48 ADVANCED DAM ENGINEERING FOR DESIGN, CONSTRUCTION, AND REHABTLITATION

Fig. 2-31. (A) General view of westem part of the main slide plane taken from point above dam. Note that grass "blanket" (foreground) has rraveled
downslope like a rug. (B) General view of eastem part of the main slide plane. Note east-marginal fault plane (more derail shown in Fig. 2-33). The
clays'associated failure surface (whitish) and step-like failure blocks are clearly evident. Source: Ceorge A. Kiersch, taken in October 1963.

'.r\-

K.'\ f -/,
>lii /r,

{.r,. \,$_ ',\{ ,,[ff'*

'i ,r
)t'\
r', Young Sct

JOINTS
from Sfress Relief
Fig. 2-32' Pattern of rebound joints and fmc(urcs throughout thc outer gltcial valley (pre-inncr-gorge) and within the inncr V-shaped young eanyon of
Vaiont River (postglacial). The stress reliefjoints have contributc<J ro fracturing and weakcning of the inner-g,rrge walls. Source: iicr=Lh.' ,-
 TESSONS FROM NOTAELE EVE[{TS 49

Fig. 2-33. (A) Eastern limit of 1963 slide which largely coincides with preexisting ancestral tectonic zone (decollement feature discussed in text).
Photographed from point near southeastem limit. as shown on Fig. 2-27. The failure plane changes to an east-west direction on the right edge of the
photograph and the continuation is shown in part B of the figure. (B) Eastem part of the main slide plane. looking southwest. Note clayey material and/
or beds associated with failure surface (whitish) and associated step blocks near top that periodically move downslope. Cray patches are thin soil debris.
as in foreground. Source: Ceorge A. Kiersch. taken October 25. 1963.

Rapid carving of the inner-gorge valley caused a second understood and led to the decision that stabilizing the slope
set of rebound joints to form parallel to the present Vaiont by drainage was impractical.3
canyon walls. The two sets of rebound joints, a younger 4. Gravitational crcep of the left reservoir slope during
and an older, intersect and coalesce within the upper Part the 1960-63 period was described by MtillerrE as "move-
of the inner valley (Fig. 2-32). ment of 25-30 cm per week (on occasion) with associated
local earth tremors: this stress relief effect centered at
Recent Historica! Events. Several principal recent depths of 50 to 500 m. The total rock mass undergoing
events preceded the october 9 movement: creep was estimated as 200 mil m3."
l. A slide of 916,000 yd3 (700,000 m3) occurred on the 5. During the spring and summer of 1963, the eventual
left bank of the reservoir near the dam (Figs. 2-27 and 2- slide area showed a creep distance of I cm per week, an
308) on November 4, 1960. This movement was accom- average rate since the 1960 slide.
panied by creep over a much larger area, and a pattern of 6. Beginning September 15, some geodetic stations were
tension cracks appeared upslope from the slide for the first observed to be moving 1.2 cm /day, then 2.2 cm/day on
time. These fractures (Fig. 2-27) ultimately delineated the September 26 and 4.0 cm/da;'on October 2-3.re How-
approximate limits of the 1963 slide. ever, observers believed only individual blocks were
2. As a precaution after the 1960 slide, the reservoir was creeping, based on the 1961 interpretation (glacier), and
limited to a maximum of 680 meters elevation, and a grid did not suspect the entire slope mass was moving. An ex-
of geodetic stations on concrete pillars was installed cellent, detailed reconstruction of the individually active
throughout the potential slide area to measure any move- blocks throughout the slope, and the rclative movement of
ment. Furtherrnore, a bypass tunnel was driven in the right each, is given by Hendron and Patton.3
canyon wall as a precaution (Fig. 2-27), should the river 7. About October 8, engineers realized all observation
become dammed. stations were moving, and an enorrnous mass was unstable.
3. The potential slide area was explored by four drill They began to lower the reservoir from its elevation of 710
holes (P1, Pz, Pt, Po) to depths of 170 to 200 meters, and meters in anticipation of a slide, but heavy inflow from
one man-sized adit in 196l; two borings subsequcntly runoff and slope movement reduced the actual lowering of
served as piezometers. Reportcdly, no confirmation of a water (reference I and Fig. 2-3a). The reservoir contained
major slide plane could be detected in either the holes or about 120 million ml of water at the time of collapse (Fig.
the.adit, but a strongly sheared zone was noted (discussed 2-3s).
earlier). The lack of groundwater flow in adits was mis- 8. Engincers measured a movement of 20 cm/day early
 50 ADVANCED DAM ENGINEERING FOR DESIGN, CONSTRUCTION, AND REHAEILITATION

on October 9,re and failure occurred at 22 hrs 4l mrn

(cEr).
& .) PRECIPITAIION. l0 d.y lncr.m.ntt

!rm Causes of the Slide

I
E
o
Geologic-Hydrologic Effects. The following adverse
b) BESERVOIF LEVEL
,6 features of the geologic environs and reservoir areal he lped
cause the catastrophic landslide at Vaiont:
Nora6!a,
eao !Id.
l. Rock units in the arcuate outcrop on the north slopes
of Mt. Toc are steeply tilted, and possess an abundance of
conallcCtbo bedding slips, clay interbeds, partings, and numerous Old
ot
Io.ch 1960
rlid. /
by-9..r
+ trnnal +
slide planes and faults (Fig. 2-29). Some of the rock units
are soft, weak, and susceptible to sliding, particularly the
clayey beds.
_ c)RArEoF il t 2. The steep dip of beds changes northward to Vaiont
tlMovEMENTll
I
I .o.rraoaoile r
'rI /t
t i I canyon, where rock units abruptly flatten along the Erto

;l:l ""1
1\1 o,,'nr
!
synclinal axis. The down-dip toe of steep slopes is an es-
E

^l L t
carpment offering no resistance to sliding (Fig. 2-308). The
tendency for deformation was confined to the wide, flat-
tish, chair-shaped area (Fig. 2-29) with an eastward dip.
The distorted beds were involved with earlier movement/
slides, and the steep profile of inner canyon walls is con-
Ea* ducive to gravitational creep.
aao
3. An ancient slide plane occurred throughout the left
rtt.. M0ll.r. 11964) reservoir wall and extended upslope throughout the area of
sliding (outlined in Figs. 2-27 and 2-30A). The ancestral
Fig. 2-34. Companson of water levels within slide area (P1 and P3 lo-
cations shown in Fig. 2-27) during period l96l to the 1963 failure. The
landslide dammed the ancient Vaiont River canyon, and
fluctuating levels are correlated wilh the reservoir level, rate of measured relics remain on the right wall.
movement, and the precipitation for 1960-63. Source: Miiller.r'{ 4. Tectonic faulting was widespread before the ances-

Fig. 2-35. Dam after 1963 slide with some danrage. and the slide that filled the reservoir an<l came to rrst some 140 metcrs up on the right reservoir
slope. Note.funnel-shaped features, indicating redistribution of nratcrial at dcpth with nam:w canyon after the sliding nrass "ju rnpcd " across the reservoir:
funnels wcrc noticed some 8 to l0 days after collapsc. (B) Vicw of slide nrass at upstream end with renrainder of rcservoir water in fore*grounrt. as shown
in Fig. 2-27. The character of the intact mck nrass with all its details of the outcrop upslope funher indicatcs the speed of nrovenrent. This mass "jumped"
the rescrvoir and was thrust up onto the right wall of the canyon, some 140 meters. The view is looking west along the toc of the slide, with rhe height
of the slide all above the former water level. Source: George A. Kiench, taken October 25, 1963.
 IESSOilS FROM NOTASLE EVENTS 51

tral landslide. One apparcnt featurc. a decollement struc' the hydrostatic uplift pressure on the slope" The buoyant
ture, displaced the general area of the 1963 sliding: the flow reduced gravitational friction. thererby enhancing
prominent fault observed (Fig. 2-33) along the eastern mar- sliding of the rocks. This phenomenon was further in-
gin slide apparently is a part of the early tectonic defor- creased by two additional geological conditions within the
mation, as are low-angle faults/deformation of the right slide area: the over 2300 ft (7(X) meters) wide, flattish
abutment (Fig. 2-36). "seat" of folded beds back from the canyon. which were
5. Active solutioning of limestone by groundwater has charged with groundwater beneath the slide area (Fig. 2-
occurred at intervals since mid-Tertiary time with devel- 29); and the influence of clay interbeds and impervious
opment of some interconnected openings and the widening zones on uplift pressure. The "seat" of charged limestone
of joints, fractures, and bedding planes. "Karstic terrain" rocks is much wider than was forecasted prior to the l96d-
with some sinkholes formed throughout the upper slopes of 67 exploration.6
Mt. Toc (Fig. 2-28). The openings served as small catch- 6. A multilayer artesian system was created above and
ments and avenues for recharge of the groundwater reser- below the slide planer'e by the altemating highly permeable
voir through the left canyon wall. The interconnected masses separated by impermeable slide planes and/or
groundwater network of joints-openings-faults weakened clayey interbeds.
the physical bonding of rocks and furthermore increased 7. The two sets of strong rebound joints, combined with
inclined bedding planes, tectonic and natural fracture
planes, and ancient slide planes. are contributing factors to
an unstable rock mass throughout the left inner-canyon wall
(Fig. 2-32).
8. Heavy rains for two weeks prior to the October 9 fail-
ure supplied excessive inflow and groundwater recharge
(Fig. 2-3a), both upslope and throughout the sliding mass.
The natural subsurface water level was raised in the critical
flat, chairlike sector and up dip along the Old slide plane.
The induced water level was raised to such heights that an
additional poftion of the slide plane was affected and the
friction factor of safety reduced; recent calculations3 give
a reduction of lO to l2%.
One investigatore reasoned that the multilayer aquifers
of the slope paralleled the dip of beds and would trap a
high fluid potential beneath the fault zone and associated
clays, whereas in the broken, fractured beds overlying the
Old slide plane the fluid potential would be lowered, re-
flecting the base level of the valley.3 The subsurface bank
storage area (triangle of slope rock charged with ground-
water) that ultimately affected the Old slide plane is
sketched in Fig. 2-29). Interestingly, springs were ob-
served along the Old slide plane of the canyon wall in
1959r3 (Fie. 2-3oB).
9. Swelling of some mineral constituents of the clay in-
terbeds, seams, and lenses and the marly beds likely cre-
ated additional uplift pressure. They would create pres-
sures throughout additional parts of the subsurface and
contribute to the sliding potential. The mineral composi-
tion and characteristics of clays are given by Hendron and
Patton.3
Fig. 2-36. Inner-gorge wall of limestoncs immediately downstrcam of
Vaiont dam on the left abutment, at level ofthe swinging bridge utilizcd
10. The water levels in piezometers (P1 and P3) during
during construction and the remedial anchoring of the canyon walls. March the period 1960-63 have been compared with the rate of
16, 1962. Note the numerous thrust planes and extensively broken and movement, reservoir level, and precipitation: and a corre-
distoned character of thc limestonesi small clay intcrbeds are common lation is shown in Fig. 2-34 (after reference 5) and sup-
with some clay seams. These preexisting slide planes are reprcsentative ported with a detailed analysis.3 Permeability was aided by
of past geologic and tectonic events-deformation that affectcd the integ-
extensive fractures and the interconnected system of solu-
rity and characteristics of the Vaiont resenvoir site. Counesy: SADE, Oc-
tober 1963. tion openings in limestones. The rate of groundwater mi-
52 ADVANCED OAM ENGINEERING FOR DESIGN, CONSTRUCTION, ANO EEHABILITATION

gration and buildup of bank storage became a critical fac- design-construction of sophisticated engineered
I
tor. Prior to April 1963, the reservoir was maintained at works in a complex geological region.
680 meters elevation, and an unsaturated "area of safety" 4. The diagnosis in 1960-61 that slope movement on
existed up dip along the Old slide plane. By mid-September the left valley wall was analogous to a glacier was
the reservoir was raised 20 meters (700 meters elevation) unfortunate; the investigatorl0 reasoned that any
and then to 710 meters; the slide increased its rate of creep future slide movement would be zero at the base and
because of a less unsaturated area along the Old failure increase to a maximum at the surface. A few bore-
plane. A very delicate balance existed between stability- holes through plane(s) with deformation measure-
strength of the rock mass and the internal stresses; the in- ments would have alerted SADE to this fundamental
creased bank storage level (20 meters or more) plus the geological error, and moreover provided a rude cor-
additional subsurflace recharge from excessive runoffof the relation of reservoir levels, groundwater levels, sur-
October rains provided a sufficient increase in hydrostatic face movement, and any sudden changes influenced
pressure on the Old slide plane to destroy the natural bal- by runoff.3
ance and margin of safety. 5. Important precedent case histories, such as the Vaiont
ll. A detailed stability analysis of the slide mass and a slide, should not be accepted by the geological and
reconstruction of the many movements that occurred engineering professions without a reasonably valid
throughout the complex rock mass and slope are given by geologic, hydrogeologic, and historic reconstruction
Hendmn and Patton.3 of the significant events, as demonstrated by Hendron
12. Numerous geological features discussed herein en- and Patton.s Moreover, because of the great diversity
hanced and facilitated the gravitational sliding as stresses in geologic and hydrogeologic environments among
increased and strength decreased. The creep phenomenon pmjects, standard guidelines cannot serve every po-
was apparently active for a long time at the site, although tential landslide case.
it reportedly was observed only since the 1960 slide. In- 6. Major gaps in the type of field data assembled on the
terestingly, some ten days before collapse, the animals Vaiont slide before failure involved: the lack of sub-
grazing on slopes became nervous and moved off the area stantive water pressure data, and the lack of reliable
of subsequent sliding,l suggesting a unique detection tech- movement records along any of the failure plane(s).3
nique for evaluating the stability of a slope. Pre-slide fluid pressure measurements taken from
piezometers installed at multiple levels within and
Lessons Learned below the slide would have provided the essential data
for correlation with any slide movements and the res-
The Vaiont experience tragically demonstrated how critical ervoir levels. Reliable measurements of the depth of
and important past geological events and features can be to the failure plane and the magnitude of ongoing dis-
a reservoir and surrounding area-even though the site placements would have helped to establish the depth
otherwise is wholly satisfactory for a dam of outstanding and size of the slide mass and significantly improved
design. Some of its lessons are:l the reliability of correlation studies.
7. The post-slide studies and reevaluation of evidence
l. There is no substitute for a full understanding of re- and circumstances strongly imply that the slide area
gional-areal geological history, events, and features, could have been stabilized by drainage.3
with regard to properties of the rocks involved and
the proposed project design and operation. In retrospect, it seems imperative that the following be
2. Vaiont is a benchmark case because of its inadequate recognized:2
regional-areal geological investigations for the plan-
ning and design phases of a major reservoir project. 8. Preconstruction studies must give thorough consid-
A full and meaningful recognition of geological eration to the properties of a rock mass as such, in
events, deformation, and changes induced to the en- contrast to a substance, and particularly to its poten-
virons would have determined the importance of an- tial for deformation with the passage of time. A the-
cient slide plane(s), tilting, fracturing, faulting, and oretical assessment alone is inadequate. The most
occurrence of solution-widened openings in linre- meaningful approach is a systematic appraisal that in-
stones, which combined to greatly increase the cludes:
"permeability" of the reservoir slope and allowed the . An investigation of the regional-areal geological
effects of bank storage to trigger large-scale sliding. environs and their critical features.
3. Vaiont likewise demonstrates the essential need for . An assessment of past geological events that have
mature, field-experienced geological observations, modified features and/or properties of the reservoir
evaluation, and judgmental input into the planning- rocks.
 TESSONS FROM NOTABLE EVENTS 53

o A forecast of the effect of the impoundment on r Potential for landslidcs: Rcservoir sites in opemtion
geological featurcs in the area/reservoir margins. for many years should be reassessed: new projects
o The geological reaction to changed conditions in must be evaluated for landslide potential in a critical
the context of project time. manner.
9. Project plans should set forth a system for acquiring o A tremendous amount of potential energy is stored in
data on the interaction between geological conditions a rock mass undergoing creep along an inclined prcex-
and changes induced by project operation. with the isting failure zone (as at Vaiont). With the incrcasing
following factors recognized: displacements. the friction factor drops. and the ve-
o Rock masses, under changed environmental con- locity of the mass increases. Consequently. a sliding
ditions, can weaken within short periods of time- mass has the potential to increase from slow creep to
days, weeks, months. a fantastically high rate of movement in a brief period
o The strength ofa rock mass can decrease very rap- of seconds or minutes. The energy goes into momen-
idly once creep is underway, particularly when tum, not into deforming the interior of the sliding mass
outside forces are involved. as in the typical slide with no preexisting rupture sur-
o Evidence of active creep should be considered a face.
warning that warrants immediate technical assess-
ment because acceleration to collapse can occur
quickly. Two techniques assist the engineer today in assessing the
potential for sliding: using the most improved methods for
Some engineering implications are as follows:2 observing and measuring the changes of strain within a rock
mass, and using a forewarning system in case the phenom-
. Speed of sliding movement: Rock masses are capable enon acts quickly and the failure of a rock is imminent.
of translatory movement as fast as quick clays or at a Some warning signs are:
liquid-like speed.
r Influence ofstrain energy on a rock mass: Release and
associated movement can be critical. The interplay r Movement shown by slope indicators, all types.
between a rock mass, the buoyancy (pore pressure) r Changes in the groundwater level in the rock mass,
effect, and the lightening of a rock mass allows an reflecting either inflow or alteration of permeability
accelerated release of the inherent strain energy. This due to creep by "opening" or "closing" of fractures
creates more release fractures, the cycle is repeated, and openings.
and the net result is an increase in the amount of sub-. r Rock noise (microseismic).
surface water and a stronger buoyancy effect-both o Animals showing signs of unrest. as they detect slight
aided by the energy release phenomenon. motion underfoot and leave.

THE WALTER BOULDIN DAM FAILURE

Tuolrns M. LEps

On February 10, 1975, shortly after midnight (at approxi- celerating, and by daybreak the breaching was essenrially
mately l: l0 n.u.),
the security guard at the Walter Bouldin complete, having progressed below the level of the pow-
powerhouse (Fig. 2-37), looking out a window of the right erhouse deck, over 90 ft (27 metem) below the crest of the
side of the semi-outdoor structure, saw muddy water on the dike. Heavy flows continued throu_eh the breach for about
powerhouse deck, an area illuminated by normal outside l4 hours until inflow to thc forebay was finally eliminated.
lighting. Seeing the water flowing from thc left side of thc The breach, which occurred at the maximum height section
back of the powerhouse to the right side . he quickly tele- of the dike on thc left side of thc intake and powerhouse,
phoned this information to a powerhouse supcwisor who was ovcr 120 ft (37 mcteru) dcep and 250 fr (76 meters)
lived nearby. Shortly thereafter, the outside lighting failed, wide (Fig. 2-38). The failure outflows were fully contained
so that clear and reliable observations became virtually im- in the broad. S-mile (8-km)-long. cxcavated, tailrace chan-
possible. The progress of breaching adjacent to the pow- nel. They caused no injuries to pcople residing nearby, and
erhouse of the 165-ft (50-meter)-high dike was by then ac- little damage except to project facilities.
54 ADVANCED DAM ENGINEERTNG FOR DESIGN, CONSTRUCTION. AND REHAEIL]TATION

Fig.2-37. Walter Bouldin Dam, before failure, Courtesy: Alabama Power .Fig. 2-38. Walter Bouldin Dam, after failure, Courtesy: Alabama Power
Co. Co.

Geologic Setting grained, compacted, micaceous sand, locally silty to

clayey.
The Walter Bouldin Project is located at the approximate o Pre-Cambrian to Paleozoic (Ashland Mica Schist):
fall line contact between the crystalline rocks of the Ap- weathered, metamorphic rock, slightly to completely
palachian chain (Piedmont Province) and the younger sed- decomposed, saprolitic.
iments of the Gulf Coast Plain, about 70 miles (l13 km) o Pre-Cambrian to Paleozoic (Ashland Mica Schist):
southeast of Birmingham, Alabama. unweathered, metamorphic rock, hard, dense, with
The Piedmont Province in this area consists of a broad siliceous and garnetiferous zones.
band of irregular, northeast-trending topographic features
caused by the intensive folding and metamorphism of pre-
Cambrian to Paleozoic age rocks. In the project area, the Terrace deposits comprise the surface materials over
older crystalline rocks and structures have been truncated most of the project area. These are Quaternary deposits
by the sediments of the Coastal Plain. These younger de- consisting of flat-lying sands and gravels with aggregate
posits have a general northwesterly strike, normal to the thickness of up to 50 ft (15 meters). Interbedded silts and
older system, and a slight dip to the southwest. Deposits clays of low permeability are widespread in the upper por-
thicken toward the coast in a wedgelike manner. tions of these deposits. Younger deposits occur throughout
Four formations exist at the site, as follows: the forebay area and under most of the forebay dikes.
The embankments were generally constructed on low-
. Quaternary Terrace Deposits: unconsolidated sand and permeability silt and clay interbeds that overlie more
gravel with lesser amounts of silt and clay. permeable sand and gravel strata. The overall reservoir de-
. Cretaceous (Tuscaloosa) Formation: fine to medium- sign depended upon a natural blanket over the forebayjoin-
 LESSONS FROM NOTAELE EVENTS 55

ing with a constructed low-permeability blanket tied to the tures opened at that time werc rePorled to be in excess of
impounding dikes. A lack of uniformity in the forebay one inch (25 mm) in width, and werc not specifically sealed
blanket, however, permitted seepage water to enter the un- during the dike construction operations.
derlying sand and gravel strata, and to recharge the grav- Ashland mica schist in a weathercd to decomposed con-
els. This permitted forebay seepage to pass beneath the dition underlies the Cretaceous sediments. This rock ex-
dikes and be discharged along the toe of the embankments. hibits varying degrees of permeability and soundness ac-
At the failed section, all alluvium and Cretaceous strata cording to the extent of weathering. fracturing, and general
had been locally excavated for powerhouse construction physical characteristics. In the area of the concrete struc-
down to massive mica schist, and the adjacent dike em- tures, it was generally excavated sufficiently to expose the
bankment therefore was founded on the slope resulting from better-quality underlying mck. Being much less permeable
the deep excavation for the intake and powerhouse foun- than the overlying sediments, the upper contact of this rock
dations, thus resting transitionally on all formations, from normally provides the seepage base for the lateral move-
schist to Cretaceous, to Terrace sands and gravels. ment of groundwater.
The materials of Cretaceous age consist of fine to me- In an essentially unweathered state, Ashland mica schist
dium-grained, micaceous sand with lesser amounts of silt provides the foundation for the principal concrete struc-
and clay horizons. The sands are compact and without ce- tures of the project. Its total thickness is unknown but is
mentation. The formation is moderately permeable and believed to be appreciable. This rock exhibits the usual
highly erodible. characteristics of old metamorphics, but rcmains hard and
The Cretaceous materials at the failure site showed evi- dense, providing an impermeable foundation of excellent
dence of cement grout travels in excess of 200 ft (60 me- quality.
ters) along existing and grouting-induced fractures, trans-
verse to the dam axis and adjacent to the intake and Design
powerhouse structures. Grouting had been done after the
original construction, in an attempt to reduce foundation The Walter Bouldin Project is an off-stream development
leakage. The orientation of fractures in the Cretaceous for- adjacent to Jordan Reservoir on the Coosa River, Alabama.
mation suggests that they were related to formation un- It consists of an extensively diked forebay fed by gravity
loading, wellpoint dewatering, and stress relief. The most via an interconnecting canal from nearby Jordan Reservoir,
probable time for development of preexisting fractures a 225 MW power plant and gated intake on the perimeter
would have been during excavation for the intake and pow- at the deepest point of the forebay, two forebay dikes ex-
erhouse structures and site dewatering (Fig. 2-39). Frac- tending right and left from the intake/powerhouse complex

Fig. 2-39. Walter Bouldin Dam. failure area foundation soils Note erodibility of Crctacertus slndy silt. Counesy: Alabanra Power Co.
 56 ADVANCED DAM ENGINEERTNG FOR DESIGN. CONSTRUCTION, ANO REHABILITATION

Normrl Maximum Trilwetcr

o Highly Porviour Sands and Granel: Excavation Surface Adjacent

To Left Side OI Powerhouse
@ Crotac.oui Fine Sandy Silt

@ Schist

o 100 feet

Fig. 2-40. Walter Bouldin Dam, cmss section thmugh dam in failure reach. Courtesy: Alabama Power Co.

12320 ft (710 meters) to the right and 5120 ft (1560 meters) seepage control was added, presumably because most of
to the leftl, and a S-mile (8-kn)-long excavated tailrace the terrace sands therein had been removed during exca-
channel ro carry up to 27,000 cfs (770 m3/s) (Fig. 2-37). vation for the powerhouse. Hence, the excavated slope of
At the dike section adjacent to and on the left side of the Cretaceous fine sands, and the terrace sands remaining in
intake/powerhouse (Fig. 240) where breaching occurred, that excavation (Fig. 2-39), had no piping protection.
the embankment is up to 165 ft (50 meters) high, with its Moreover, seepage exiting from those formations was hid-
crest at elevation 265 ft (81 meters). The dike was a com- den by .iprup, and the dike-to-Cretaceous-foundation con-
pacted earthfill of nearly homogeneous c(,ss section, but tact was partially inundated by tailwater.
with a thin, compacted clay upstream zone, tied, in the
intake vicinity, to a clay blanket extending upstream to the
Construction
forebay's natural earth blanket. The upstream slope of the
dikes was protected by riprap, generally only for its upper Construction of the principal features began in July 1964,
20 ft (6 meters), but adjacent to the intake for the full height and the dikes were essentially complete by April 1967. The
of 65 ft (20 meters) above the horizontal clay blanket. The dikes were constructed by a general contractor, with con-
dike upstream slope varied between 2H:lV and?.S/J:lV, struction monitored and quality-controlled by Alabama
but adjacent to and approaching the intake was transitioned Power Company personnel. Only fairly normal construc-
frbm 2Il: lV to l.3H: I I/over a distance of 100 ft (30 me- tion problems occurred, such as intemrptions for wet
ters). The downstream slope was at l.8Il: I I/. Downstrcam weather, short strikes, and groundwater control. Some
from the dam axis, the embankment was founded on a 36- 2,711,000 cy (2,070,000 m3; of fill were placed in the
in. (9l-cm) thick "drainage layer," the latter consisting of dikes, all of the fill except rockfill, filter materiat, and
unprocessed local sands and gravels, with up to l0O% riprap having been secured from nearby borrow areas in
passing a No. 4 screen and up to 507o passing a No. 200 the forebay and from excavation for the intake and pow-
scrcen. erhouse.
Significantly, the design did not include an engineered,
subsurface, toe drain to guard against piping in either the Reservoir Operations
foundation terrace sands or, more important, the underly-
ing, highly erodible, Cretaceous fine sands. Subsequent to The forebay was initially filled in May 1967. The normal
construction, as a result of springs, boils, and so on, many maximum reservoir is at elevation 252 ft (77 meters), a
smalldiameter relief wells were placed along the down- level determined by the spillway for the interconnected Jor-
stream toe to drain the terrace sands; but, in the short fail- dan Reservoir. In part because of flow constraint in the
ure reach to the left of the intake/powerhouse, no such canal interconnection, the Bouldin pool at maximum flow
 LESSONS FROM I{OTAELE EVETTITS 57

through its turbines t27,0OO cfs (770 m,/sec)l was nor- who rcported that the dikes appearcd to bc normal. The
mally a few feet below elevation 252 ft (77 meten), but first indication of failurc was the observation at l: l0 e.u..
for efficient generation it was maintained at or above ele- February 10, 1975, 8 houn later, by the all-night sccurity
vation 248 ft (76 meters) most of the time, except during guard, who saw muddy water flowing over the powerhousc
maintenance periods for projects on the Coosa River, when deck. After his telephoned rcpon to a company supervisor.
the levels briefly averaged elevation 246 ft (75 meters). various company personnel were notified, who traveled to
In September 1972, seven inadvertent reservoir draw- the site as quickly as possible. By the time they had ar-
downs occurred. The most severe, on SePtember 28, rived, the failure was well advanced. and all lighs on the
amounted to l0 ft (3 meters) in 7 hours, the level having intake and powerhouse were out because of short circuis
dropped to elevation 238 ft (73 meters), 14 ft (4.3 meters) due to invasion of the switchyard and powerhouse by water,
below maximum normal. Upon recognition of that event' debris, and mud. Hence, they had little or no opportunity,
the resewoir was promptly refilled by reduction of the in the dark, to approach or to view the failure area, and to
powerhouse discharye. Six days later, during a routine dike observe the sequence of events. Sketchy recorded imprcs-
inspection, a small, surficial slide, estimated to be 300 cy sions seem to provide a consensus that, by about l:45 e.u.,
(230 m3) was noticed on the upstream slope immediately about a 40-ft (12-meter)-long section of the dike crest-in
to the left of the intake. The slide area was carefully re- a l5Gft (45-meter) reach to the left of the inuke, near to
paired. Subsequently, no reservoir drawdown greater than but not at the intake-had sagged perhaps as much as 25 ft
6 ft (2 meters) below normal maximum pool was permit- (7.5 meten). If this was the case, water-possibly l0 ft (3
ted. meters) dee5must then have been rushing through the sag.
The critical, but unobserved, sequence is the one that had
been occurring in the several hours prior to that time.
Prefailure Remedial Actions
If the dimly observed sag was, in fact, not immediately
From the beginning of reservoir operations, significant adjacent to the intake, it must have been remarkably sim-
leakage occurred through the foundation of the left and right ilar to the sag observed at the late stage of piping failure
dikes-importantly in the short reaches adjacent to and on of both Teton and Baldwin Hills dams, just prior to col-
both sides of the intake/powerhouse, and less importantly lapse. If it was immediately at the intake contact, however,
along the over 5000 ft ( 1500 meters) length of the left dike. its relation to a possible upstream slope slide in the area,
Following the advice of consultants, 128 small-diameter where that slope was only l.3ll: I % would be inferred. It
relief wells were installed along the downstream toe of the is clear from both the Baldwin Hills and the Teton cases
left dike, a27o-ft (82-meter)-long French drain was placed that the clearly observable stage of final piping failure is
130 ft (40 meters) downstream from the left dike beginning very short, a matter of a few hours.
160 ft (49 meten) left of the powerhouse, and grout cur- Breaching was evidently complete by 6:00 A.M., but
tains in alluvium were attempted in short reaches of both continuing severe damage and erosion occurred for the fol-
dikes, beginning on the crest 150 ft (46 meters) to the left lowing 14 hours, until Jordan Reservoir had been lowered
of the intake and 130 ft (40 meters) to the right of the in- enough to cut off flow to the Bouldin forebay.
take. The holes extended down through the terrace sands
and gravels and deeply into the Cretaceous fine sands.
Somewhat curiously, the curtains were not closed against Postfailure Finding of Cause
the intake, and on the left the ungrouted gap was the lo-
cation where the 1975 breach occurred. The grouting, The cause of failure was investigated intensively by Ala-
which consumed over 6300 cf (178 m3) of cement in the bama Power Company, and separately by the Fedenl
left curtain and 4900 cf (139 m3; in the right, effected no Power Commission. The company set up an in-house Board
significant decrease in leakage. Excessiv,e grculpresqulqs- of Inquiry on February I I, 1975. The board retained thrce
ca u sed hy{rau I i,1$gtuglg_qt{! rgy g g! e scapes 9!ve ! ns*9 !_ independent consultants of wide experience to assist and
grout to-dls.tant polnts. * advise. In their June 26, 1975, rcport to the inquiry board,
-Ail observable leakage was collected, passed over weirs they advised that it was "... highly probable that the
for measurement, and monitored. An undetermined, un- breach began with a slide of the upstrcam impewious em-
seen discharge of leakage must have constantly entered the bankment zone adjacent to the east (left) side of the intake
tailrace, below tailwater level, on both sides of the pow- This is the reach in which a drawdown slide oc-
erhouse. curred in Scptember 1972, involving slippage, probably
mostly on the rockfill+o-impervious-zone, l.0f/: l.0lz,
contact, in the about 100-ft (30-meter)-long transition rcach
Failure Sequence
where the upstrcam, outer, rockfill slope transitioned from
An inspection of the dikes was made during the late after- l.3H: llz at the intake to 2.0H: lV. The overall investiga-
noon of February 9, 1975, by a knowledgeable supervisor, tion ruled out the following possible failure causes:
 58 ADVANCED DAM ENGINEERING FOR DESIGN, CONSTRUCTION, AND REHABILTTATION

r Gmundwater seepage Filtered, internal, chimney drains were added in the

. Dispersive soil in fill or foundation totally rebuilt reaches.
o Earthquake All seepage, through and under the highest dike
o Sabotage reaches on both sides of the intake, was brought to a
o Burrowing animals single, monitoring weir at a location just above tail-
. Overtopping by reservoir operations water.

The consultants advised that foundation piping was "highly Reconstruction was completed in [98 l, and the project
unlikely. " has been in satisfactory operation since that time.
To the contrary, and as offered separately and subse-
quently by an expert witness during the 1976 FPC hearing,
Conclusion
I believe piping of foundation soil in the 150 ft (45 meters)
reach of dike on the left side of the intake to be a possible Although very detailed examination of the right-hand dike
cause of failure of such importance that it should not have suggested that fill placement against the intake was locally
been discounted. The vulnerable location of initial piping imperfect, and hence the same might be true of the left
or seepage-caused subsurface erosion is visualized as being dike, such a finding has also been made in numerous other
at the downstream toe of the left dike at about dike axis cases of autopsy of fill placement, and should not be a sur-
Sta. 47 * 40, about 60 ft (18 merers) to the left of the prise to the profession. It merely confirms that the designer
intake, in the Cretaceous fine silty sand formation between must be aware of the inevitability of defects despite good
elevations 125 ft (38 meters) and 135 ft (41 meters). In this quality control, and should provide those conservative
location, the Cretaceous+o-fill contact was invisible be- measures that prevent defects from being critical, such as
cause it was covered by riprap (Fig. 240), and it was in good filter zones, total collection and visual monitoring of
and beneath the zone of fluctuation of the tailrace level (el- all leakage, and adoption of conservative slopes and zon-
evations 125 to 135 ft). Moreover, (a) the cretaceous zone ing.
had been damaged and loosened somewhat by excavation With reference to the cause of failure, it is unfortunately
during powerhouse construction to a 1.5I1: I Z slope; (b) it true that the total washout of the breached area occurred at
may have been seriously loosened by installation of a row night when observation was impossible, and removed vir-
of construction wellpoints, which tended to create a privi- tually all conclusive evidence. Hence, there is no real proof
leged seepage path transverse to the dike axis and com- of a probable single cause. My assessment, however, is
pletely under the dike from the reservoir side to the tail- that because there was no drawdown event in the forebay
water side (the holes for which were not grouted up after during February 9 to trigger an upstream slope failure, and
construction); and (c) it was not provided with filter pro- there was no physical distress evident at the dike crest as
tection against exiting seepage. With tailwater flows of up late as 8 hours before the first notice of failure, it is most
to 27,000 cfs (770 m3/sec), and total inability to observe probable that the dike failed initially by foundation piping
the gradual formation undenvater of a "pipe" in the Cre- in the highly erodible Cretaceous Formation, followed by
taceous, it could have been just a matter of time before a collapse of the crest into the quickly enlarging "pipe."
"pipe" in that highly erodible and locally pervious for-
mation worked its way slowly upstream to unrestricted
communication with the reservoir. The discharge of fines REFERENCES
into the tailrace would have been unnoticed, because of
The Baldwin Hills Reservoir Failure
both the natural turbidity of the powerhouse discharge and
the daily scouring action of that discharge. l. Califomia Depa(ment of Water Resources, "lnvestigation of Failure,
Baldwin Hills Reservoir," Apr. 1981.
2. Leps, Thomas M., "Analysis of Failure of Baldwin Hills Reservoir."
Reconstruction Proceedings of ASCE Specialty Conference, Purdue University. June
t9'12.
The redesign and reconstruction of the damaged reaches of 3. Hamilton, Douglas H., and Meehan. Richard L., "Ground Rupture
the dikes were deliberately conservative, utilizing the best in thc Baldwin Hills," Scierrce, Vol. 172, Apr. I971,
modern practices. The principal changes from the original 4. Casagrandc, Arthur. Wilson, Stanlcy D.. and Schwantes. E. Douglas,
Jr., "Thc Baldwin Hills Reservoir Failure in Retroslrct," Proceed-
design were as follows:
tngs of ASCE Specialty Confercnce, Purdue University, June 1972,

. The upstream slopes were flattened rc 2lH: tV.

o Positive dike seepage cutoffs (compacted fill and slurry The Failure of Malpasset Dam
trench) were provided throughout, tying in to imper- l. Bellier, Jean, "The Malpasset Dam." Irayaar, Vol.50. No.3E9.
vious, weathered schist. pp. 363-383, July 1967.
 LESSONS FROM NOTABLE EVENTS 59

2. Proceedings, lntemational Workshop on Dum Failures. Purdue Uni- Valont Reservoir Disaster
IN, 1985.r
versity, West Lafayene.
3. Londe. P., "Malpasset Dam." Proceedings, lnternational lUork- l. Kicrsr'h, G. A.. "Vaiont Rcscrvoir Disastcr." Civil Eagincering.
shop on Dam Failures, Purdue Univesity, 1985.r Vol. 3.1. No. 3. Mar. t961.
4. Jaeger. Charles. "The Malpasset Repon." Water Po*,er. Feb. 1963. 2. Kicrsch, G. A.. "Thc Vaiont Rcservoir Disaster," Ltndslidcs anl
5. Goguel, Jean. "The Malpasset Repon." Wuter Pox'er, p. 58. FeF Subsidencc: Ceologic Hu*trds Conferencc Proceedings. Thc Re-
ruary 1963. sources Agency, State of Califomia. pp. 136-t.15, May 26-27, 1965.
6. Puls. Louis G., Lctter to the editot. Water Po*'er, p. 229, Junc 1963. 3. Hcndron. A. J.. and Patton, F. D.. "The Vaiont Slidc, A Gcotct'h-
7. Habib, P., "Laboratory Tests to Determine the Effect of Rock Tex,- nical Analysis Based oo New Geologic Observations of thc Failurc
ture on Permeability-a Discussion," International Workshop on Surface." U.S. Army Corps Enginecrs. Technical Rcport GL-E-S-5.
Dam Failures, Purdue University, 1985.* Waterways Experimcnt Station, Vicksburg. IrlS. Vols. I and ll. June
8. Londe, Pierre. "Malpasset Dam." Proceedings, International 1985.
Workshop on Dam Failures. Purdue University. 1985.r 4. Mtiller, L.. "The Rock Slide in the Vaiont Vallcy." Rock Mechanics
9. Wittke. W., and Leonards, G. A.. "Modified Hypothesis for Failure arul Engineering Ceolog'- tournal, lntemational S<rcicty Rrrck lrte-
of Malpasset Dam," Proceedings, lnternational Workhop on Dam chanics, Vol.2. No. 3--f, pp. 148-212. 196r.
Failures, Purdue University, 1985.r 5. Nltiller, L.. "New Considcrations on the Vaiont Slidc." Roct.tfe-
10. Post. G., and Bonazzi, D.. "l:test Thinking on the lvlalpasset Ac- chanics and Engineeing Geolog-t Joumal. lnternational Society Re=k
cident." Proceedings, lnternational Workshop on Dam Failures. Mechanics. Vol.6. No. I, pp. l-9t, 1968.
Purdue University, 1985.r 6. Broili, L., "New Knowledge on the Geomorphology of the Vaiont
ll. Bonazzi, D., and Post. G., "Four Major Dam Failures Re-exam- Slide SIip Surface." Rock Mechanics an<l Engineering Geologt
ined," International Water Power and Datn Construction, pp. .12- tournal, lnternational Society Rock Mechanics. Vol. 5, No. l. pp.
zt4. Nov. 1985. 38-88. 1967.
12. Serafim. J. L., "Four Major Dam Failures Re-examined." Inrer- 7. Frattini. M., Arredi, F.. Boni, A., Fasso. C., and Scarsella, F..
national Water Po*'er ancl Dam Construction, p.44, Nov. 1985. "Relazione sulle cause che Hanno Determinato Ia frana nel Serbaroio
13. Brown, J. Guthrie, Presidential address to the 8th Intemational Com- del Vajont (9 Ottobre 1963)." Franini Commission Repon to
mission on Large Dams, Edinburgh. Scotland, ICOLD Transactions. E.N.E.L., 92 pp.. Jan. 196{.
Vol. V, p. 8. 1964. 8. Semanza. E.. "Sintesi Degli Studi Ceologici sulla frana del Vajont
14. Proceedings, Intemational Workshop on Dam Failures. Purdue Uni- dal 1959 al 1964," Memorie del Museo Tridentino di Scienze \'<t-
versity,1985.* trrali. Trento. Italy. Vol. 16. No. l. pp. l-53. 1966-61.
15. Post. G., and Bonazzi, D. "[:test Thinking on the N'lalpasset Ac- 9. Lo. K. Y.. Lee. C. F., and Gelinas. P.. "Alternative lnterpretation
cident," Proceedings, International Workhop on Dam Failures, of theVaiontSlide," inCording, E. J. (ed.). Stabiliq'of RockSlopcs,
Purdue University, 1985.t Proceedings l3th Symposium on Rock lv{echanics. University of Il-
linois. Urbana/American Society Civil En,eineers. pp. 595-61-1.
t91 t.
The St. Francis Dam Failure 10. Miitler. L., "Die Felsgleitung im Bereich Toc." Talsperrve Vajont
15, Baugeologischer Bericht to S.A.D.E. (unpublished). 1961.
l. Bowers, Nathan A., "St. Francis Dam Catastrophe-A Great Foun-
dation Failure," Engineering Ne*'s-Record, Vol. lfi). No. 12. Mar.
Il . Miller. L., "Discussion of Differences in the Chancteristic Features
of Rock Mechanics and Mountain Masses." Proceedings 5th Inter-
22. t928.
national Conference Intemational Bureau Roc'k lvlechanics. Leipzig.
2. Commission Appointed by Govemor C. C. Young, "Causes Leading
Germany (DDR). Nov. 1963.
to the Failurc of the St. Francis Dam," Califomia State Printing Of-
12. ENR. "Vaiont Trial Defendant Blames Geologists." Engineering
fice, Sacramento, Mar.24. 1928.
Neu's-Record. Ne*'York, p. 13, Feb.21.1969.
3. Gillette, H. P., "The Cause of the St. Francis Dam Failure," Engi-
13. Giudici. F.. and Semanza. E., "Studio Geologico del Serbatoio del
neering and Contracting, Apr. 1928.
Vajont" (unpublished Report) 2l pp.,68 photogrnphs,42 pp.. dis-
4. Bennett, Ralph, "St. Francis Dam Failure-An Engineer's Study of
cussions. 2 maps and sections, 1960.
the Site," Engineering News-Record, Vol. 100. No. 13, Mar. 29.
14. Rossi. D., and Semanza. 8., "Carte Geologiche del Versante Set-
1928.
tentrionale del Monte Toc e Zone Limitrofe. Prima e Dopo il Fen-
5. Jansen, Robert 8., Dams and Public Sofery, U.S. Govemment Print-
omeno di Scivolamento del 9 Ottobre 1963." Istituto di Geologia
ing Office, Denver. CO, 1983.
dell 'Universita di Ferrara. 1965.
6. Ransome, F. L., "High Dams: The Viewpoint of the Geologist,"
15. Selli. R.. and Trevisan, L.. "Caratteri e interpretazione della frana
Transactions, American Sociery of Civil Engineers, Vol. 95. Paper
del Vaiont." ln Frona Dcl V<tiont. Annali del Museo Geologico di
No. 1766, 1931.
Bologna. Ser.2. Vol. 31. No. l. 196{.
16. Carloni, G. C. and Ivlazzanti. G.. "Rilevamento Geologico dclla
Failure of Teton Dam frana dcl Vaiont." ln Frona Dcl Vaiqrt. Annali del Museo Gcolc
gico di Bologna. Ser. 2. Vol. 12. No. l. pp. 105-118. 1961.
l. lndependent Panel to Review Cause of Teton Dam Failure, "Failurc
17. Ciabatti. M.. "La Dinamica della frana del Vaiont." Lt Frana Del
of Teton Dam." Dec. 1976. /aiont, Annali del lvtuseo Geologico di Bologna, Scr. 2. Vol. 31.
2. U.S. Dept. of lnteriorTeton Dam Failure Rcview Gmup, "Failurc of
No. t. 19(y1.
Teton Dam, A Report of Findings," Apr. 1977.
18. Miillcr, L.. "Rock Mcchanics Considerations in thc Dcsign of Rrx-h
3, U.S. Dept. of lnterior Teton Dam Failure Revicw Gruup. "Failurc of
Sltrpes." in Judd. W. (cd.), Stote of Stress in Earth's Crlsr. Rand
Teton Dam, Final Report," Jan. 1980.
Corp.. Santa Monica, CA. June 1961.
19. Folberth, P. J.. "Writtcn Comnrunication on Mcasurcnrent Slope
*ln Leonards, G. A. (ed.), Engineering Geologv, Vot. 24, Nos. l-.1, lv{ovenrcnts Vaiont of Septcrnbcr-October. I96J." Lcrcr. Folbenhi
Elsevier. Amsterdam, 1987. Electnxonsult lvlilan. Italy to C. A. Kiench, May 13. 196.1.
 HYDROLOGY
CareuNo B. CecrI-to Anr-eN D. Frlpuex
Senior Civil Engineer Chief, Research Division
Pacific Gas and Electric Company Hydrologic Engineering Center
San Francisco, Califurnia U.S. Army Corps of Engineers
Davis, California

Lleweu-vN L. Cnoss
Chief Hydrologist
Chas. T. Main, Inc.
Boston, Massachusetts

INTRODUCTION an inflow design flood, or an appropriate safety evaluation

flood for an existing dam.
An important aspect in the evaluation of dam safety is the Flood studies are also performed to determine the extent
determination of the performance of the dam under maxi-
of downstream flooding due to a hypothetical failure of the
mum flood conditions. This entails sometimes complex ilam. They provide input for the planning of emergency
procedures for the study of watershed and storm character-
action in case there is an actual dam failure.
istics, and may require extensive statistical analyses or hy-
drometeorological approaches.
Engineering hydrology has such a wide scope that the Early Methods
space limitation here does not permit a complete treatment
From historical data that documented the failure of Sadd
of the subject. References are made to several textbooks el-Kafara Dam in Egypt in 2850 s.c.r (apparently due to
and other published materials for specific applications. This
an inadequate spillway) to the present regulatory concern
chapter deals primarily with the determination of hypo-
for dam safety, people have attempted to learn more about
thetical floods for use in the design and safety evaluation floods and their characteristics.
of dams, reservoirs, spillways, and diversion structures. Hydrologic analysis for the determination of floods has
Wind-generated waves for determining freeboard allow-
progressed to its present level of four major methods: the
ances are also addressed.
regional discharge method or envelope curve method, the
statistical method, the hydrometeorological method, and the
Purpose of Flood Studies risk-based analysis method. These four methods represent
the progression from T. J. Mulvaney's rational formula (O
When a dam is built to impound natural runoff, the flow = CiA, published in the late 1800s and still in use) to the
regime of the watershed is changed. When a flood occurs, hydrometeorological merhod of calculating the prob-
the incoming flow will be held back by the dam until re- able maximum precipitation (PMP) for input to a runoff
leases through properly sized outlets and spillway are pro- model to obtain the probable maximum flood (PMF) at a
vided. specific site.
Flood studies are performed to determine: the reservoir Of these four methods, the envclope curve method has
storage necessary to accommodate a given flood, the di- probably been in use the longest, and it still has a place in
mension of the spillway needed to pass excess floodwaters, preliminary evaluations and reconnaissance level studies.
the freeboard allowance required during the occurrence of Generally, this method involves the plouing of all the his-

6o
 HYDROTOGY 61

toric flood peaks for the particular river basin, rcgion, or and the constants for use, with Q in cubic feet per second
hydrometeorologically similar basins against the catchment and ,{ in square miles, arc shown in Table 3-1.3
area. Myer, Jarvis and Creager: were early practitioners of With the advent of desktop computers, thc statistical
this technique, which is still in use today. Crippen3 has methodology became morc widely used. Many investiga-
extended this practice by developing envelope curves for tors have strived to devetop a "best fit" equation for the
17 different hydrometeorological regions in the continental array of annual peak flows for a particular site. This method
United States. has become more popular as the length of rccord for some
Figure 3-l shows a sample of envelope curves taken from U.S. Geological Survey gages approaches 80 years, which
Crippen.3 gives a credible data base. Of the several equations that
Creager's equation is: have been developed, the one most universally accepted at
present and used by U.S. federal agencies is the one rec-
Q: (3-l)
8e4/-os6) ommended in Bulletin l7B of the Water Resources Coun-
46CA(o
cil, "Guidelines for Determining Flood Flow Fre-
with Q in cubic feet per second and .{ in square miles. C quency."o This publication recommends a log Pearson
is a coefficient based on experience in a given region. A Type III method for flood frequency analyses in which an-
value of C : 100 has been used to represent the maximum nual peaks are used. This probabilistic method is most use-
value. ful for the definition of flood peaks at a frequency that falls
Matthai's curye2 is based on the equation: within the period of record of data, and is usually employed
in the design of cofferdams and diversion works rather than
Q: ll,000,{0'6r (3-2) spillway design. It also has a useful role in the determi-
nation of the rainfall flood considered to be antecedent to
where I is the drainage area in square miles. the inflow design flood (IDF) or the frequency snowmelt
The general equations developed by Crippen have the flood peak to be used in conjunction with a thunderstorm
form: for the design flood.
The third major method, the deterministic or hydrome-
Q = KrAK'zKt (3-3) teorological approach, is the most widely used in the United
States today. The evolution of this approach began in the
The boundaries of the 17 regions are based on physio- 1940s when the U.S. Army Corps of Engineers and the
graphic divisions, variations in rainfall intensity, and the U.S. Weather Bureau jointly undertook the task of deter-
experience of hydrologists who have worked with flood mining the magnitude and temporal and spatial distribution
data. The regions used by Crippen are shown in Fig. 3-2, of the probable maximum precipitation (PMP), based upon

OFAIXAGE AFEA. IN SOU^R€ XILOIETEBS

t0 I O0

2 0.000
o
o z
z o
o REGTOX tO r 0.000 o
g
o CUnvE- / a
E
r" -/2 C
^roro" a
U
"".rr;/7j1.

v-
G 6
FI 00, 000

o t
a o
)
o
r.000 -,
o
=
U
C,
E
t0
/ =
u'
(
o
a HAIIrlN -a-
o
a =
=-cuav2e!
I :o
6
!
a
,.%:: .cREAOEi CURVI
lC.tO0 )
.!
EI PL AII ATIO'{
3ttE oete Axo a{uraEn

I
6
!

o cunv€
.ooo r
r
......,1
o.t
o.t I r0 ro0 I,OO0
on^lx^o€ AREA. ll. souahE rll€!1

Fig. 3-1. Envelope cunves and peak discharges. Reprintcd with pennission of ASCE
62 ADVANCED DAM ENGINEERING FOR DESIGN, CONSTRUCTION, AND REHABTLITATION

12f t2f 1130 1090 t0 50 to 10 970 930 890 950 tlo 7?o 730 6gc

t17! 1130 togo 1050 1010 970 930 ggo 950 g10 770
200 0 200 .oo 600 LrlEs

KLouErERs
'?PJ-lll-19!,J90
Fig. 3-2. Map of coterminous United States showing ffood-region boundaries. Reprinted with permission of ASCE.

the maximized synoptic situations that caused it. Over the In summary, federal, state, and local regulatory bodies
years, a series of jointly published Hydrometeorological are regularly upgrading the dam safety requirements for the
$.eports (HMR) has given the Depth-Area-Duration (D-,{- spillways of proposed and existing dams. Thus it is incum-
D) data for various regions of the United States for the bent upon the practicing hydrologist or hydrologic engineer
-
hypothesized probable maximum precipitation (PMP). As to keep current with these changing regulations and with
time goes on, more severe storm data are collected, and the the accepted techniques for meeting these regulations.
reports are revised to reflect these additional storms.
The fourth method involves the computation of site-spe-
cific floods, usually for an existing dam and for less than Earlier Standards
the PMF where the existing spillway cannot accommodate
The design floods for many existing dams in the United
the PMF. The problem in these cases is usually to derive 'States were chosen in a subjective manner, based on the
the hydrograph that produces a peak of some particular
magnitude of property damage and probable loss of project
magnirude, for example, the zero freeboard flood, the in-
investment and human life in the event of dam failure. A
cipient failure flood, or a particular-frequency flood. It does
survey made by the United States Committee on Large
not depend on the adoption in advance of specific bases or
Dams (USCOLD) in 1970 summarized the practices ac-
criteria for dam safety evaluation floods. Two types of
cepted and used by engineers in selecting design floods for
analysis are in use: (l) those that evaluate only the hy-
dams.s
draulic effects of dam failures and (2) those that go further
The survey found that spillways are sized to serve onc
and make an economic analysis to determine the design
or more of the following objectives:
that has minimum total cost. By means of dam-break stud-
ies, the safety evaluation flood can be selected at the flood
peak level where downstream flood damages would not be o To ensure that flood hazards downstream will not be
increased by overtopping of the dam. This is a trial-and- dangerously increased by malfunctioning or failure of
enrrr solution of sizing the spillway, and is also known as the dam during severe floods.
the "incremental damage assessment" method. o To ensure that service of and investment in the project
 HYOROLOGY 63

Table 3-1. Characteristics of envelope Standard 4: Keep the dam low enough and storage im-
curves.3 poundments small enough that no serious hazard would er-
ist downstrcam in the event of brcaching. and repain to thc
Q = KtA*'/'
Region Numbcr dam would be rclatively inexpcnsive and simple to accom-
1Fig. 3-2) Kl K: K,
plish.
I 23.200 0.895 - l.082
2 28.000 0.770 - 0.897 Policies Affecting Spillway Capacity Require-
5 5.1.400 0.914 _ 1._173

4 42.600 0.91E - 1.317 ments. The following general policies (based on the ac-
5 12t.000 0.838 - l.-15.1 cepted pmctices found by the USCOLD committee) pro-
6 70.000 0.917 - 1.297 vide guidance for application of the four functional design
7 49.100 0.883 - 1.351 standards.5
8 43.800 0.95-l - 1.357 r When a high dam. capable of impounding large quan-
9 75.000 0.8-19 - 1.368 tities of water, is constructed upstream of a populated com-
rtr 62.500 l. I 16 - t.37 r
ll 40.800 0.919 - 1.352 munity, a distinct hazard to that community from possible
t2 89.900 0.935 - 1.30r failure of the dam is created unless due care is exercised in
l3 6r.500 0.873 - 1.338 every phase of engineering design, construction, and op-
l4 10.000 0.710 -0.8.1.1 eration of the project to ensure complete safety. The pre-
l5h I 16.000 1.059 - 1.572 vention of overtopping of such dams during extreme floods.
t6' 98.900 t.019 -l.3.ll
t7 80.500 1.014 -l..16l including the probable maximum flood, is of such impor-
K. = tance that the additional cost for conservatively large spill-
"For dminage arca less than 1.0 sq milc. use O = K.'{;
5.358. ways is justified (notwithstanding the low probability of
'For dminage area less than 0.17 sq mile. use Q = Kn"l: I(. =
overtopping). The policy of deliberately accepting a rec-
7.3-18.
'For dminage area less than 0.05 sq mile' use Q = K'A'' K' = ognizable major risk in the design of a high-hazard dam
9.E78. simply to reduce project cost has been generally discredited
from an ethical and public welfare standpoint. This is es-
will not be unduly impaired by malfunctioning, seri- pecially true if the results of a dam failure would imperil
ous damage, or failure of the dam during floods. lives in the downstream floodplain. Legal and financial ca-
e To regulate reservoir levels as needed to avoid unac- pabilities to compensate for economic losses associated
ceptable inundation of properties, highways, rail- with major dam failures are generally considered inade-
roads, and other facilities upstream from the dam dur- quate justification for accepting such risk, particularly when
ing moderate and extreme floods. severe hazards to life are involved. Accordingly. such high-
o To minimize overall project costs insofar as pftlctica- hazard dams should be designed to conform with security
ble within acceptable limits of safety. Standard l.
. Application of Standard 2 should be confined princi-
Functiona! Design Standards. The USCOLD com- pally to structures where relatively small differentials be-
mittee also noted that functional design standards neces- tween headwater and tailwater elevations prevail during
sary to meet minimum security requirements for down- major floods. Also. overtopping should not cause serious
stream areas at minimum cost usually conform with one of damage or loss of life downstream. Examples are run-of-
the following alternatives, the selection being governed by river hydroelectric power and/or navigation dams, diver-
circumstances associated with specific projects and down- sion dams, and similar structures. In such cases. the design
stream developments. capacity of the spillway and related features may be based
Standard 1: Design dams and spillways large enough to largely on economic considerations.
ensure that the dam will not be overtopped by floods up to . Application of Standard 3 should be limited to dams
probable maximum categories. impounding a few thousand acre-feet or less. Such dams
Standard 2: Design the dam and appurtenances so that should be designed to ensure a relatively slow rate of fail-
the structure can be overtopped without failing and, insofar ure if overtopped. and should be located where hazard to
as practicable, without suffering serious damage. life and property would be minimal. The occurrcnce of
Standard 3: Design the dam and appurtenances in such overtopping floods ntust be rclatively infrequent to make
a manner as to ensure that breaching of the structure from Standard 3 acceptablc. A slow, gradual rate of brcaching
overtopping would occur at a r€latively gradual rate, such can bc accomplished by designing the dam to overtop wherc
that the rate and magnitude of increase in flood stages the breach of a large section of retatively erosion-rcsistant
downstream would be within acceptable limits, and dam- material would bc involvcd, such as through a flat abut-
age to the dam itself would be located where it could be ment section. Control may bc obtained, in some cases' by
repaired most economically. permitting more rapid erosion of a shon section of em-
 il ADVANCED DAM ENGINEERING FOR DESIGN, CONSTRUCTION. AtrtD REHABIuTATION

bankment and less rapid lateral erosion of the rcmaining The PMF should be adopted as the IDF in those situa-
embankment. tions wherc consequences attributable to dam failurc from
o Standard 4 is applicable to small recreational lakes and overtopping are unacceptable. The determination of unac-
farm ponds. In such cases, it is often p,referable to keep ceptability applies when the arca afected is evaluated, and
freeboard allowances comparatively small to ensure that the there are indications that loss of human life, extensive
volume of water impounded will never be large enough to property and environmental damage, or serious social im-
rclease a damaging flood wave if the dam should fail be- pact may be expected as a result of dam failurc.
cause of overtopping. In some instances, adoption of Stan- A flood less than the pMF may be adopted as the IDF
dard 4 may be mandatory, despite the dam owner's desire in those situations where the consequences of dam failure
to construct a higher dam, if a higher standard is not at_ are acceptable. Acceptable consequences exist when eval-
tainable. Unless appropriate safety of downstre.rm interests uation of the arca affected shows one of the following con-
can be ensured, a higher dam is not justified simply to re- ditions:
duce the frequency of damages to the project.
o There are no permanent human habitations, or com-
Recent Developments in Design Flood Criteria mercial or industrial development, nor are such habi-
tations,or commercial or industrial development,
Since the USCOLD study, other activities have developed projected to occur within the potential hazard area in
that have addressed the present thinking in the selection of the foreseeable future, and the transient population is
an inflow design flood. In addition, the concept of risk- not expected to be affected.
based analysis as it pertains to the selection of the inflow o There are only a few permanent human habitations
design flood has gained considerable acceptance in the field within the potential hazard area that would be im-
of dam safety. Two documents are discussed here to pro- pacted by failure of the dam, and there would be no
vide guidance in the hydrologic analysis of dams, eipe- significant increase in the hazard resulting from the
cially if the analysis pertains to dam safety. occurence of floods larger than the proposed IDF up
to the PMF. For example, the impoundment storage
The ICODS Report. An ad hoc interagency comminee is small, and failure would not add an appreciable vol-
on dam safety of the Federal Coordinating Council for Sci_ ume to the outflow hydrograph; so the downstream
ence, Engineering, and Technology prepared Federal inundation would be essentially the same with or with-
Guidelines for Dam Safery,6 published June 25, 1979. To out failure of the dam. The consequences of dam fail-
provide general guidelines on procedures for selecting and ure would not be acceptable if the hazard to these hab_
accommodating inflow design floods (IDF) for the use of itations were increased appreciably by the failure flood
federal agencies, the Interagency Committee on Dam wave or level of inundation, as in the case where fail-
Safety (ICODS) formed a working group on inflow design ure of a storage reservoir would add appreciably to the
floods. The result of their work is published in a manual outflow hydrograph.
called Federal Guidelines for Selecting and Accommodat-
ing Inflow Design Floods for Dams.7 In addition, the selected magnitude of the IDF should be
The latter document gives guidelines for hazard evalua_ based on the following special considerations:
tion and for determining the area affected by a theoretical
dam failure. The impacts of dam failure to be evaluated o Dams that provide vital community services such as
include loss of human life, as well as social, environmen- municipal water supply or energy may require a high
tal, and economic impacts. The hazards should be defined degree of protection against failurc to ensure that those
sufficiently for selection of an appropriate inflow design services arc continued during and following extreme
flood. The document recommended two standards, one for flood conditions when alternate services are unavail-
proposed dams and one for existing dams.
able.
o Dams should be designed to not less than some min-
Proposed Dams. Selecting an IDF for the hydrologic imum standard to reduce the risk of loss of benefits
safety design of a dam requires balancing the likelihood of during the life of the project; to hold O&M costs to a
failure by overtopping against the consequences of dam reasonable level; to maintain public confidence in
failure. Consequences of failure include both loss of life agencies responsible for darn design, construction, and
and social, environmental, and economic impacts. The in- operation; and to be in compliance with local, state,
ability to accurately define flood probabilities for rare or other regulations applicable to the facility.
events, and to assess accurately the potential loss of life
and economic impact of failure should it occur, dictates the Existing Dams, The flood selection guidelines for pro-
use of procedures that provide some latitude in meeting posed dams should be applied to the maximum extent pos-
site-specific conditions in IDF selection. sible to existing dams. The evaluation leading to selection
 HYDROTOGY 65

Table 3-2. Selection of the IDF for existing dams. large. high-hazand dams. But the rcsults of the inventory
show widespread uncertainty as to what might be the ap
Situation in Case of Failurc Minimum IDF
propriate hydrologic criteria for safety of other classes of
a. Loss of human [ife, IDF is equivalent to the PMF. dams. From study of the inventory rcsults, the following
ertensive propeny damage. or observations can be made:
serious social impacts
anributable to dam failure
. Use of PMP estimates for evaluating spillway capac-
may rcasonably be expected if ity rcquirements for large, high-hazard dams predomi-
failure occurs. nates, although a number of state agencies have indicated
b. Spccial case of (a). Total IDF selection is based on the that their standards do not require that such dams pass the
resenvoir volume is small level beyond which the potential
full estimated PMF based on the PMP.
compared with the PMF for loss of human life from dam
volume so that the thrcat lo failure outflow does not erceed
r The influence of the practices of the principal federal
human life from floods is not potential loss of life from dam-building agencies is evident in the majority of the
increased by dam failure controlled releases through standards for large, high-hazard dams, but the practices of
above that resulting fmm spillways and other release those agencies have had less effect on culrent state stan-
maximum controlled releases. facilities. A larger IDF should be
dards for small dams in less hazardous situations.
selected if economic analysis
indicates it would be cost-
. Apparently as a result of the National Dam Inspection
effective. Program for nonfederal dams,e carried out by the Corps of
c. Loss of human life IDF selection is based on an Engineers in 1977-81, several state dam safety agencies
attributable to dam failure is economic evaluation and other have adopted the spillway capacity criteria used in those
not expected. and economic relevant factors.
inspections.
and social impacts are within o Several states have adapted the standards used by the
acceptable Iimits.
d. Situation the same as in IDF is determined as described in Soil Conservation Service for the design of the tens of
condition (a)l however, an section H.4. of ICODS manual thousands of smaller dams constructed under that agency's
IDF equivalent to the PMF (see reference 7). prcgrams.
cannot k accommodated. No . Current practices include use of arbitrary criteria (such
good altematives are available
as l50Vo of the 100-year flood. fractions of the PMF, and
because of constraints
(physical, economic. social. combinations of the PMF with probability-based floods) for
etc.). which there is no apparent scientific rationale.
o Practices of the major federal dam-building agencies
for large, high-hazard dams have been adopted by most
of the IDF and the method for accommodating it must U.S. companies owning dams and by U.S. engineering
sometimes be considered simultaneously for existing dams. firms designing dams for domestic and foreign clients. Cfhe
Table 3-2 is recommended by ICODS for use as a guide regulations of the Federal Energy Regulatory Commission
for selection of the IDF for existing dams. have required such standards for licensed hydroelectric
projects.)
The National Research Council Report. A recent re- . It appears that only three agencies (the Federal Energy
port prepared by the Committee on Safety Criteria for Regulatory Commission, the Mississippi Department of
Dams, under the auspices of the National Research Coun- Natural Resources, and the New York State Department of
cil,t provided new insights into the present thinking in the Environmental Conservation) have issued explicit stan-
selection process of inflow design floods. The committee dards for existing dams that differ from the requirements
also presented data from l0 federal organizations, 35 state of new dams. (However, other responses did not specifi-
and local agencies, 9 private firms, and 4 professional en- cally state whether different standards were applicable to
gineering societies. From this information, the committee existing dams.)
concluded that there is considerable variation in the criteria The Committee on Safety Criteria for Dams8 recom-
adopted in the United States for evaluating the ability of mends the following (reprinted with permission from
dams to withstand extreme floods, esp€cially in criteria for "safety of Dams: Flood and Earthquake Criteria." O 1985
the smaller, Iess hazardous dams. Because of this vari- by the National Academy of Sciences):
ation, the reader is advised to consult with local and state
regulatory agencies on the Proper inflow design flood for a l. To the cxtcnt prdcticablc. rcscrvoir safcty evaluations
new or existing dam. should strike a balance among such considerations as project ben-
This inventory of current practices in providing dam efits. construction costs. social costs. and public safety. including
safety during extreme floods shows considerable diversity the possible conscquences of dam failure due to major eanh'
in approach by various federal, state, and local government quakes and floo<is. (While achieving such balance is the ideal.
agencies, professional societies, and privately owned firms. currently availablc technology does not pcrmit this balancing with
There is a fair consensus on the spillway requirements for full confidence ln the results.)
 66 AOVANCED DAM ENGINEERTNG FOR DESIGN, CONSTRUCTION. AND REHAEILITATION

2. Safety evaluation standards for existing dams and proposed The design and evaluation of dams against hypothetical
dams need not necessarily be the same.
floods requires determination of the hydrologic rcsponse
3. The use of probable maximum floods (pMF), based on es-
(infiltratioa, base flow, rouring, and runoff) of warersheds
timated probable maximum precipitation (pMp), as the geneal
design standard (safety evaluation flood, SEF) forproposed high-
to intense rainfall or snowmelt, with verification based on
hr"ard dams should be continued. However, instances sometimes historical storrn and runoff data (flood hydrograph analy-
will be encountered where a lower standard may be justified if sis). It atso requires determination of hypothetical storms,
failure of a dam during floods of pMF magnitude would cause no infiltration, base flow, and channel and reservoir routing.
significant increase in potential for loss of life or properry dam- (Routing, discussed later in the chapter, involves investi-
age. gation of the adequacy of existing structures above and be-
4. For existing high-hazard dams, the adopted safery evalua- low the dam under investigation to pass safely the hypo-
tion flood (SEF) should take into account estimated flood proba- thetical floods.)
bilities, expected project performance, and incremental damages The amount of water that wilt flow down a river for a
that would result from dam failure for a range of floods up to and given amount of rainfall or snowmelt varies with many fac-
including the probable maximum floods.
tors. These factors are so numerous and their evaluation so
5. In the design of new dams and spillways when design al-
uncertain as practically to preclude mathematical solutions.
tematives of approximately equal cost are available, a selection
among these altematives should give consideration to potential Several empirical methods are utilized as aids to compute
future needs for increased safety against extreme floods and earth- ntnoff from rainfall, but these methods will not be dis-
quakes. cussed here. The advent and availability of computers have
6. The safety evaluation of dams need not considerihe simul- made it easier and more practical to model the runoff phe-
taneous occurrence of the safery evaluation flood and the safety nomena through mathematical simulation. Much of the fol-
evaluation earthquake because of the extremely low probability lowing discussion of rainfall runoff also applies to snow-
of such occurrence. melt runoff; a separate section is provided later in the
7. Periodic reviews of hazard determinations and safety deci- chapter to describe the unique considerations for snowmelt
sions for all dams should be required, especially when safety
floods.
evaluations are based on criteria less conservative than the prob-
Each drainage basin requires a separate study to deter-
able maximum flood or the maximum credible earthquake.
8. Research efforts designed to provide better bases for esti- mine the rainfall-runoff relation of that basin, as there is
mating magnitudes and frequencies of extreme floods and earth- often great variation of rainfall-runoff relations between
quakes, for estimating reactions of dams to such natural phenom- neighboring drainage areas. This variation may be caused
ena, and for establishing acceptable levels of risks should be by physical characteristics that are not readity apparent,
continued. such as snow accumulation in high-elevation bands in
9. As advances o€cur in seismology, hydrology, meteorology, mountainous regions similar to the Sierra Nevada in Cali-
and the relevant data bases, and as changes are noted in public fornia. The area of application of every rainfall-runoff re-
attinrdes toward risk, the federal agencies should periodically un- Iation must be clearly indicated, as some relationships are
dertake a review of dam safety practices and standards by an in- only valid in an area of a few square miles, whereas others
dependent body representing the professions involved in engi-
have been found applicable in areas of over 100,000 sq
neering for dams and experts from other relevant disciplines.
miles. In the Ohio River Valley, for example, a single rain-
fall-runoff relation, with minor corrections for the extrem-
In the above, the safety evaluation flood (SEF) is defined
ities of the area, has given good results over the entire area;
as the largest reasonable hypothetical water inflow for
but this is an unusual instance.
which the safety of a dam and appurtenant structures is to
The hydrologic response of the watershed to precipita-
be evaluated; it should be demonstrated that this flood level
tion is usually determined and verified from historical flood
can be accommodated through storage, spillway releases,
records with a runoff model. The model should include
releases through other outlet works, or limited and accept-
consideration of nonlinear runoff response due to high rain-
able overtopping without causing failure of the structure or
fall intensities and increased hydraulic conductivity. The
uncontrolled release of impounded water.
runoff model should be developed by analytically ..recon-
stituting" historical floods to substantiate its use for esti-
HYDROLOGIC CONSIDERATIONS mating extreme events up to the severity of the probable
maximum flood (PMF). The observed rainfall runoff of
Hydrologic and hydraulic analyses for dam design or dam scveral historical floods should be used to verify that the
safety evaluation represent a very specialized and complex reconstituted hyd rographs correspond reasonably wel I with
branch of engineering for danrs. For projects where dam flood hydrographs actually recorded at selected gaging sta-
failure could have serious consequences, these analyses tions. In most cases, reconstitution studies should be made
should be directed by an engineer trained and experienced with two or more floods and possibly ar two or more key
in this specialized field. locations, particularly where possible errors could have a
 HYDROLOGY 67

serious impact on the selection of the runoff model. If the Once the region is delineated. both climatological and hy-
absence of stream gage records. the lack of an adequate drological data should be collected, and the sites for these
time and arcal precipitation definition, or other. unex- data points located on an appropriately scaled map. The
plained causes prevent the development of a runoff model. U.S. Geological Survey topographic quadrangle maps have
other means such as synthetically developed models may been found to be very useful as a basis for regional hy-
be used. drologic studies.
The development of a design flood from the selected de- The methodology fcrr such a study would proceed as fol-
sign storm involves several tools of meteorology, hydrol- lows:
ogy, and hydrologic engineering. The various steps in such
analysis, as summarized in the National Research Council l. The region of interest would be defined on an aPPro-
publications,E'ro arc generally as follows: priately scaled base maP.
2. All of the gaging stations with records of annual peak
l. Dividing the drainage area into subareas, if neces- flows both within and around the region would be

sary. located and plotted on this map.

2. Deriving a runoff model. 3. Flood frequency curyes for each gage would be cal-
3. Determining design storms using criteria contained in culated, as well as the l0-sq-mile PMP.
various National Weather Service Hydrometeoro- 4. Using a graphical technique similar to that emplol'ed
logical Reports or other rainfall design rePofts. in the preparation of a topographic map, the isopleths
4. Arranging the design storm increments into a logical forthe average annual and the l0-.25-,50-. and I0O-
storm rainfall pattem. year floods would be ploned (as a cfs /sq mile param-
5. Estimating for each time interval the losses from eter).
rainfall due to such actions as surface detention and 5. From this series of isopleth maps for the region, the
infiltration within the watershed. various return period flood peaks can be determined
6. Deducting losses from rainfall to estimate rainfall ex- for any site in the region by interpolation between the
cess values for each time interval. appropriate isoPleths.
7. Applying rainfall excess values to a runoff model of
each subarea of the basin. A major caveat in the preparation and use of this method
8. Adding to storm runoff hydro-eraph allowances for the of regional flood transposition is that major topographic
base flow of the stream, runoff from prior storms, and barriers and their orographic impact on storm precipitation
so on, to obtain the synthesized flood hydrograph for must be considered in the determination of storm isopleths.
each subarea. The effect of these natural barriers can best be seen and
9. Routing the inflow through the reservoir storage, out- evaluated by the initial preparation of an isopleths map for
lets, and spillways to obtain estimates of stordge el- 12-. 24-, and 48-hour total precipitation from major storm
evations, discharges at the dam, tailwater elevations' events.
and so forth, that describe the passage of the flood
through the reservoir. (This is essentially a process Flood Frequency Analysis
of accounting for volumes of water in inflow, stor-
age, and outflow through the flood period. If there The frequency of occurrence of both peak streamflows and
are several reservoirs in the watenhed, the reservoir runoff volumes is the very basis for reservoir spillway, out-
routing is repeated from the upperrnost to the most lets, and storage capacity design. For dam safety investi-
downstream reservoir, in turn.) gations, the hydrologic safety of a dam is determined in
relation to its ability to pass floods of a specified frequency
ordesign standard. Thus, flood frequency analysis (forpeak
RESERVOIR INFLOW PREDICTION and volume) is often used to assess dam safety.
Historical streamflow records may be used directly to
estimate discharges at various frequencies. If adequate
Regional Flood Transposition
streamflow records exist, and the watershed has remained
As previously mentioned, one of the earlicr and morc rclatively unchangcd during thc course of that recotd, then
widely used flood peak extrapolation tcchniques was thc those observed strcamflows arc probably the bcst indicator
development of regional floods. of the potential [lootl rcsponses of the watershed in its pres-
Usually the region to be investigated has been definecl ent condition. Thc Watcr Resourccs Council's guidelinesa
by some consideration other than the hydrology, for ex- prescnt the currcntly recommcnded techniques. These
ample, a utility service area, a river basin such as the Mis- guidelines describe thc use of the log Pearson Type III dis-
souri or the Tennessee, or some political consideration. tribution and associatcd topics of high and low outliers,
 6A ADVANCEO DAM ENGINEERING FOR DESIGN. COf{STRUCNON, ANO
REHABILTTATION

generalized skaw, two-station comparisons, mixed popu_ who analyzed over 20 characteristics and discuss the rcla-
lations, confidence limits, ffood estimates fr,om prccipita- tive effects of each. Drainage basin arca and nornul annual
tion data, and equivalent accuracy for independent Lsti_ prccipitation were zunong the most significant characteris-
mates for the analysis of historical nood peaks. tics. Certainly some of the most widely available examples
_ If one must predict the magnitude-frcquency of stream_ of these tebhniques are rhe U.S. Geological Suwey (USGS)
flows under some future watershed land use ievelopment "State Regrcssion Equations" (e.g., reference 16). A sim-
or rcgulatid condition, the historical streamflow records ilar report for urban areas has also been completed.tT A
cannot be used directly. In this case, one usually must re- gencralized procedure is also used by the U.S. Federal
sort to a watershed model. The same requirement arises Highway Administration (FHWA)rE for small rural wa-
where long-term historical streamflow records exist but the tersheds, generally of less than 100 sq miles.
watershed has undergone significant land-use changes dur_
Other techniques use watershed runoffcharacteristics and
ing that time. Thus, a nonstationary streamflow series ex- precipitation intensities to predict peak runoff rates. A time
ists and cannot be used directly in the frequency analysis. of concentration and infiltration index may be determined
The nonstationary series problem can also be unraveled directly from watenhed characteristics such as length and
through the use of watershed models.ll elevation change of main channel, soil characteristics, and
If a stationary series of data is available, but not at the land cover. The X-minute rainfall intensity for a desired
specific Iocations of interest, then a regional fr,equency rainfall frequency is obtained from Tp40,re HYDRO-35,ro
analysis may be undertaken.12 The regionil analysis or the state atlases.2l A peak discharge per square mile may
"tto*i
one to transfer the parameters of the flood frequency dis_ be derived from the time-of-concentration, infiltration in-
tribution at gaged locations to other locations of interest. dex, and peak 3O-minute intensity. Adjustments can be
This is accomplished through relating frequency parame- made for antecedent precipitation. This method is subject
ters to geographic and meteorologic characteristics that are to the standard criticisms of assuming that the frequency
known at the gaged and ungaged locations.
of
the runoff is the same as the rainfail.
The Water Resources Council-Hydrology Committee,
Work Group for Peak Flow Frequency for Ungaged Areas,
has recently completed a studyr3 of the following eight Frequency Analysis of Rare Floods. Recent dam
flood frequency estimation techniques: safety investigations have concentrated on determining the
frequency of rare floods used to design the dams' spiliway
Statistical estimation of peak flow capacity. Some spillways are designed for very rare fre_
Statistical estimation by moments quency flo&s, say l0,OOGyear, or l0-a probability. The
Index flood method discharge for the lO,0OGyear flood is usually computed by
extrapolating a frequency curve developed from observed
. Transfer method
Empirical equarions streamflow. Unfortunately, historical streamflow records
Single storm are almost totally inadequate for such a purpose because of
. Multiille discrete events the small sample they represent compared with the mag-
Continuous simulation nirude of the event in question. The U.S. Interagency Hy-
drology Committee, the National Research Council, and
The first phase of the Water Resources Council (WRC) the U.S. Army Corps of Engineers are studying several
study has been completed for selected watersheds in the approaches to determining the frequency of rare floods; no
northwestem and central United States. The methods re_ technique seems viable at this time.
ported in the pilot tesrs arc: USGS Equations, FHWA,
Reich, Snowmelt, Index Flood, Rational Formula, TR55,
TR20, and HEC-I. The WRC is encouraging the widest Design Hydrographs from Historical Floods
possible review of this work beforc going on with similar It is important that the temporal distribution and magnitude
applications in the southwestern and southeastem United ofany flood used in the design ofa new project (or in the
States. A laterphase of the WRC studies will include urban spillway adequacy rest for an existing facility) be deter-
areas. The rcsults of the WRC study are discussed by New- mined and compared with historical maximum floods.
ton and Herrin.la The use of several historical storrn events is advised, re-
Statistical nood peak estimation techniques predict in_ conded for the strcam in question, for other streams with
stantaneous peak flows of prescribed frequencies through a similar hydrologic characteristics in the same basin, and
regression analysis of geographic variables affecting the for strcams in another basins that are hydrologically, top-
flood runoff. An excellent discussion of drainage basin and ographically, geologically, and vegetatively similar to the
meteorologic characteristics that can be used to explain the study stream. In the case of basins larger than 300 sq miles,
behavior of streamflows is given in Thomas and Benson,rs it may be practical for the study basin as well as selected
 HYDROTOGY 69

similar basins to be divided into subbasins for the hydro- 2. The arca under this rainfall excess hydrograph is
graph investigation of historical storms. However, this planimetercd; or the ordinates arc summed and con-
iubdivision should be done carcfully because the resulting verted to inches of runoff from the basin.
flood hydrograph might not rEPresent a realistic or rcason- 3. The discrcte ordinates arc then converted by using
able routed flood. dre total inches of runoff derived in (2) to rePrcsent
Where available, hourly flow records for the storm event one inch in the unit hydrograph. The sum of these
should be obtained and plotted as flow vesus time' Flow proportioned ordinates should equal one inch of run-
data for a long-enough period to define the base flow prior off.
to the storm and the ascending and descending legs of the
historical flood hydrograph should be obtained' On the A point most often overlooked is that, by definition, a
same sheet and with the same time scale, one should plot unit hydrograph should rePresent one unit of runoff from
the discrete and accumulated precipitation for the storm the basin, usually one inch, in a unit-duration time Perid'
being analyzed. Basin mean precipitation may be assumed The main difficulty lies in determining the unit duration for
to have the same percentage of total precipitation for the a historical storm that has complex rainfall patterns' The
incremental rain as that experienced at some nearby rain first approximation of the time for the unit hydrograph' as
gage for which hourly records are available' developed in Steps l-3, can be obtained from the accu-
With the storm hydrographs and discrete and accumu- mulated storm rainfall plot' taking the time interval for the
lated precipitation plotted, the base flow must be estimated steepest portion of this plot; that is, where AP/AIequals
to d"tir" the relationship between runoff'and the hydro- a maximum (see Fig. 3-3).
graph. The simplest method is to assume that the increase The validity of this time period can be checked by the
in the Uase flow follows a straight line from the point of use of an S-curve Process in which the unit hydrograph
inflection at the start of the rising limb of the hydrograph' ordinates are summed as a series of seParate unit hydro-
graphs, with each series being advanced by the time period
sloping upward until it intersects the recession limb at a
point prior to the inflection point of the recession' The slope derived from the rainfall mass curve. If the sums of these
of tnii Uut" flow line is a maner of judgment, based on the series of offset ordinates are ploned and give a smooth S-
fact that the area under the hydrograph and the base flow curve, the assumed time period is correct. If this curve is
must be something less than the total storm precipitation' not smooth when ploned, the process is repeated, usually
Although various investigations have emphasized the slope with a shorter time period, until the rcsulting S-curve is
of thisline, the use of judgment and a knowledge of the smooth. Once the adjusted time period produces a smooth
vegetal cover and the morphology of the basin are essen- S-curve, this time can be taken as the correct time period
tiai. Also, in temperate climates the precipitation falling on for the unit hydrograPh.
a basin in the winter, with its frozen ground and sparse In most instances, these data of streamflow for the storm
vegetation, will produce a Sreater volume of surface runoff event are not taken at the site being studied. Thus it be-
than the same sto[n in the summer with vegetation at its comes necessary to adjust this hydrograph to give one inch
peak. The season of occurrence of the design storm will of runoff from the gaged drainage area to the study basin'
gou"* whether or not the frozen ground flood is the ap- This may be done by adjusting all the ordinates of the unit
propriate one. Consider, for example, that in the north- hydrograph developed from the gaged area to the study
eastern United States the PMP is most likely to be gener- area, by using the ratio of the respective drainage areas'
ated by the synoptic situation accompanying a hurricane or
The peak unit hydrograph ordinate and the unit hydrograph
tropical storrn-events that most frequently occur in Sep- ordinates on both the ascending and descending legs, down
tember. This rules out consideration of storm events that to a flow of 80% of the peak flow, should be adjusted by
have produced floods from rainfall on frozen gncund or a the squarc root of the ratio of the respective drainage arcas'
melting snow pack. If the study basin werc large enough to have been sep-
Along with the development of several storm hydro- arated into subbasins, the subbasin unit hydrograph param'
graphs and estimation of their accompanying base flows eters would be developed in the same way using rcgional
ana tne duration of storm rainfall, it is necessary to derive data. Examples of rainfall-runoff simulation in ungaged
a unit hydrograph. Rainfall-runoff simulation with com- areas arc given by the Hydrologic Engineering Center'22
puter models, for example, HEC-I,4e may be used to au- Once the unit hydrograph and smoothed S-curve have
iomatically derive a unit hydrograph. Also a hand calcu- been developed for a unit-duration time of excess for the
lation method may be used, as follows: actual storm event, it may become necessary to develop a
unit hydrograph for a shorter time interval. This is done by
l. The discrete flow ordinates above the estimated base subtracting the ordinates of two respective S-curves offset
flow are measured graphically and plotted against by the required time period. Reference 23 discusses this
time. method in detail.
 70 ADVANCED DAM ENGINEERING FOR DESIGN, CONSTRUCTION, AND REHABILITATION

!
a
a
)
E
f
o
o

c
o
a
=q
E
o
c

o
x
o
L
o
a

Augurl t 8.t 955 Augurl 20,i 955

Fig. 3-3. Flood hydrograph analysis.

Design Hydrographs from Design Storms Reports.2H3 The pMps do not have a specific frequency
because the storm is a synthetic condition determined by
Design storms are often used to compute design hydro- maximizing the worst meteorologic conditions found pos-
graphs for reservoirs and other water resource projects sible in historical storrns.
where historical flood data are not available, or they are Watershed models are generally required: when an entire
inadequate because the observed floods are only a very hydrograph is desired: in analyzing complex areas; or when
small sample of the floods that could occur. precipitation the past or proposed future watershed response functions
records often are much longer than runoff records, provid_ are changing. Watershed models are particularly desirable
ing a better sample of possible storms. This does not sumce for analyzing the effect of various water management
necessarily for very rare storrns, for which very few, if schemes.
any, observations are available. Design storms of more
common frequencies (2- to 100-year recurrence interval)
Watershed models vary widcly in complexity.
Brandstettefl compares many different models for urban
are available in National Weather Service publica- storm runoff, and many of these models are equally good
tions.le-21'24'25 Design storms
of very rare frequency (the for nonurban cases. Some are nothing more tilan simple
probable maximum precipitation, or pMp) are determined empirical eguarions within a subbasin network routing/
by the National Weather Service, Hydrometeorotogical combining framework. Othen perform a complex account-
 HYDROLOGY 71

ing of soil moisture and water in various stages of runoff. duration relationship and isohyetal pattern that is consid-
The World Meteorological Organization has made a com- ered reasonably characteristic of the region in which rhe
parison of various continuous simulation models.rs The drainage basin is located. It is an estimate for a panicular
event-oriented and continuous simulation models will be drainage area and season of year in which snowmelt is not
described in more detail in a later section. a major consideration. Runoff characteristics and the ex-
istence of water regulation structures in the basin arc con-
HYPOTHETICAL STORM COMPUTATION sidered in the analysis. In deriving SPS rainfall estimares
applicable to seasons and areas in which melting snow may
Standard Design Storms contribute a substantial volume of runoff to the hydro-
graph, appropriate allowances for snowmelt are included
It is the general practice to obtain design flood hydrographs
with, and considered a part of, SPS rainfall quantities in
from design storms. Three general types of design storms
computing the standard project flood (SPF) hydrograph.
may be used: the frequency-based storm, the standard proj-
Where floods are predominantly the result of melting snow,
ect storrn, and the probable maximum precipitation.
the SPF estimate is based on estimates of the most critical
Frequency-Based Storm. This design storm is defined combinations of snow, temperature, and water losses con-
as the depth of rainfall with a certain return interval in sidered reasonably characteristic of the region. The ffood
years, ranging from 2 to 100 years. The depths are ob- resulting from the standard project storm is called a stan-
tained from generalized studies prepared by the National dard project flood.
Weather Service, such as those in refererices 19-21,24, The SPS is generally derived in accordance with proce-
and 25. An example is given in Fig. 34. dures of the U.S. Army Corps of Engineers. Reference .t6
is used as a guideline in developing a standard project storm
Standard Project Storm (SPS!.
The SPS should rep- patterned after a major historical storm. The SPS is about
resent the most severe flood-producing rainfall depth-area- 40 to 6OVo of the probable maximum precipitation.

30. LEGENO
2.0
/ t00-YEAn CO- MtNUTE
- ^/ PaEcrPrrAfloN nNcHES,
2.O
* KtY W€Sr. FLORTOA VALU€
BEPNESENTATIVE FOF
FLOnt oA KtYs.
27.
0 200 ,aoo c00 HtLEs

0 ?00 .00 600 (tloilEiEns

:r

Fig. 3-4. 100-year 60-minute prccipitation (inches). adjusted to panial-duration serics. Source: U.S. National Weather Service. Hydro 35. 197?.

-
72 ADVANCED DAM ENGINEERING FOR DESIGN. CONSTRUCNON, ANO REHABIUTATION

Probable Maximum Precipitation (PMPl. The proba- arc taken fmm generalized charts prcparcd by the National
ble maximum precipitation is defined as the estimated depth Weather Service, the depth thus interpolated is assumed to
for a given duration, drainage area, and time of year for be the uniform depth for a l0-sq-mile arca. Reduction of
which there is virtually no risk of exceedance. The proba- these data to areal depths is sometimes done with the aid
ble maximum precipitation for a given duration and drain- of depth-area curves such as the ones shown in Fig. 36.
age are:l approximates the maximum that is physically pos-
sible within the limits of contemporary hydrometeorological Design Storm Parameters. After the desircd frc-
knowledge and techniques. PMP estimates for the United quency or design storm criteria have been established, the
States are available in generalized studies prepared by the next step is the determination of the storm parameters,
National Weather Service.2H3 Figure 3-5 shows the re- which include stonn duration, point depth, areal adjust-
gions in the United States where generalized PMP studies ment, time distribution, and areal distribution.
are available.
The flood from a PMP is generally called the probable Duration. Durations of about 6 hours or less are adequate
maximum flood (PMF). The PMF is also defined as the for small drainage basins. For larger drainage areas, du-
hypothetical flood (peak discharge, volume, and hydro- rations of up to l0 days may be required. The critical du-
graph shape) that is considered to be the most severe rea- ration for a basin is usually at least as long as the time of
sonably possible. It is based on comprehensive hydrome- concentration, the time when all of the basin is contributing
teorological application of probable maximum precipitation to runoff. Storm depths with given frequencies are avail-
and other hydrologic factors favorable for maximum flood able for durations of from 5 minutes to l0 days for loca-
runoff (such as antecedent moisture, storm area, and ori- tions within the United 51u1"r.2o'2a'5 However, design cri-
entation) for a particular basin (see, for example, reference teria for large dams require estimates of storm depths that
4l). do not have frequency levels assigned.

Development of Design Storms Balanced Storm. A balanced storm is one that is of equal
severity for all possible critical durations ofproject design.
Precipitation gage data permit determination of the fre- The severity is expressed in terms of exceedance probabil-
quency and character of precipitation at a given site. Point ity or exceedance frequency.
precipitation data are used to estimate areal variability of In the planning of a flood control project involving stor-
rain and snow, as well as for developing design storm char- age or in the development of reservoir operation rules, it is
acteristics. not ordinarily known what the critical duration will be be-
Point rainfall data are sometimes used to derive inten- cause it depends on the amounts of reservoir space and re-
sity-duration-frequency curves. When point rainfall data Iease in relation to flood magnitude. When alternative types

1O3' MERIOIAN

HMR 36 e 49

HMR 52 & 55

HMR 39 HAWAII
HM8 54 SOUTHEAST ALASKA
TP 42 PUERTO RICO ANO VIRGIN ISLANOS
TP 47 ALASKA

Fig. 3-5. Regions covercd by generalizcd PMP studies.

 HYOROTOGY 73

E
0 q
a
o

g
o
c
c
o
a
q
6o
L
o 30 rco t'^,- 30(
,"i10.,.
",,.'.I
Fig. 3-6. Depth-arca curves. Source: U.S. National Weather Service.
Fig. 3-7. Balanced storm.

of projects are considered, critical durations will be differ-

ent, and a design flood should reflect a coinparable degree largest value, the third largest value follows the largest
of protection for the various types of projects. Accord- value, the fourth largest precedes the second largest, and
ingly, balanced hypothetical storms are useful for plan- so on (Fig. 3-7). In this manner, each critical-duration in-
ning, design, and operation purposes. tensity is embedded within all longer-duration intensities.
Time panerns of floods that occur at any particular lo- Thus, whatever antecedent watershed moisture, initial res-
cation arc widely varied. Some floods have high peak flows ervoir storage, and reservoir operation policies are used,
and are of short duration, whereas others have the reverse some part of the balanced storm will produce the resulting
characteristics; and, of course, short-duration volumes can critical{uration event.
be small in relation to longduration volumes in some floods The storm area used for the point-to-area adjustment is
and large in others. No one historical flood would ordinar- critical. This adjustment compensates for the fact that
ily be representative of the same severity of peak flow and storrns covering larger areas will have lower average inten-
runoff volumes for all durations of interest. sities. Such a relationship between storrn area and average
If a project is designed to rEgulate all floods of record, intensity can be determined for different-frequency storrns
it is likely that one flood will dictate the type of project and in a region, using historical precipitation statistics. Thus,
its general features because the largest flood for peak flows the storm area used should be the area that is tributary to
is also usually the largest-volume flood. Yet, there is in- the river location in question. This requires separate stonn
formation in the other floods that can be used to obtain a centerings (areas) for each location where floods are to be
better balance of the peak flow and volumes for various computed.
durations of a design flood. It is important that this infor-
mation be used effectively. Depth. Frequency-based storm depths can be obtained
A synthetic storm of any duration from 5 minutes to l0 from the charts given in references 19-21 , 24, and 25. An
days can be generated based on depth-duration data from example for the 60-minute, 100-year chart is given in Fig.
historical storms. A triangular precipitation distribution is 34.
constructed such that the depth specified for any duration It must be noted that the rainfall depths taken from these
occurs during the central part of the storm. This is referred charts are point rainfall values with the same probabilities.
to as a balanced storm (Fig. 3-7). Depth-duration data for The estimated depths should be adjusted for areal rcduc-
various precipitation frequencies in the United States are tion. For very large drainage areas, the proper arcal distri-
available from the National Weather 5".''ri"".le-21'2a'25 bution of rainfall should be adopted in order to avoid the
These data generally are used for partial-duration fre- adoption of equal probability of the rainfall event all over
quency series and for point locations. Adjustments must be the drainage area. In other words, a rainfall of a certain
made to convert to annual frequency series and for areal probability occurring in the upper part of a watershed need
precipitation of a desired storrn area. not necessarily occur in the lower part or some other part
Cumulative precipitation for each time interval is com- of the watershed.
puted by log-log interpolation of depths from the depth-
duration data. Incremental prccipitation is then computed Storm Transposition. Because it is highly unlikely that
and rearranged so the second largest value precedes the documented storms of large precipitation quantities will