The Role of Risk Analysis in Water Resources Engineering

Larry W. Mays, Ph.D., P.E., P.H.

Department of Civil and Environmental Engineering
Arizon a State Un iversity

Introduction is beyond our rigid controls. In the design and operation

of these systems, decision s must be made under various
Uncertainties and the conse quent related risks in water kinds of uncertainty.
resources engineering design and operation are
unavoidable. Water resources projects are always subject The sources of uncertainties in water resources
to a probability of failure in achieving their intended engineering projects are many -fold. W e will first discuss
purposes. As an example, a flood control project may not the ideas of na tural unce rtainties, mo del structure
protect an area from extreme flo ods. A w ater supp ly uncertainties, model parameter u ncertainties, data
project may n ot deliver demanded w ater. This failure uncertainties, and operational uncertainties. Natural
may be due to failure of the delivery system or may be uncertainties are associate d with the random temporal and
due to lack of supply. A water distribution system may spatial fluctuations inherent in natural pro cesses. Model
not deliver water meeting quality standards even though structural uncertainties reflect the inability of a simulation
the source q uality doe s. The ration al in the selection of model or design procedure to repre sent precisely the
the design and operation parameters and the design and system’s true physical behavior or process. Model
operation standards are continually questioned. Water parameter uncertainties reflect the varia bility in the
resource engineering design and operation procedures do determination of the param eters to be u sed in the model or
not involve any required assessment and quantification of design. Data uncertainties include a.)measurement
uncertainties and the resultant evaluation of a risk. inaccuracy and errors, b.) Inadequacy of the data gauging
network, and c.)data handling a nd transcription erro rs.
For purposes of this paper risk is defined as the Operational uncertainties are associated with construction,
probab ility of failure. Failure is de fined as the event in manufacture, deterioration, maintenance, and other human
which the system fails to function w ith respect to its factors that are no t accoun ted for in th e modeling or
desired objectives . Reliability is defined as the design procedure.
complement of risk, i.e. the probability of non-failure.
Failure can be grouped into either structural failure or The four major categories of uncertainties in water
performance failure. A good example of this is for water resources engineering are; 1.)hydrologic uncertainty;
distribution systems. A structural failure such as pipe 2.)hydr aulic uncertainty; 3.)structural uncertainty; 4.) and
breakage or pump failure can cause de mands t o not b e econo mic uncertainty. Each of these uncertain ties has
met. Also operational aspects of a water distribution various compo nent uncertainties. Hydro logic uncertain ty
system such as the inability to meet demands at required can be classified into three types: inherent, parameter, and
pressure heads is a failure without any structural failure of model uncertainties. The occurrence of various
any comp onent in the system. See Mays (1989) for m ore hydrological events suc h as stream flow or ra infall even ts
details and a survey of method s for water distribution are consider ed as stoch astic processes because of the
systems. observable natural, or inherent, randomness. Because of
the lack of perfect hydrological information about these
The objective of this paper is to discuss the role of processes or events there exist informational uncertainties
uncertain ty analysis and the resultant quantification of risk about the proce sses. These uncertainties are referred to as
for the design and operation of water resources the parameter uncertainties and the m odel uncertainties.
engineering p rojects. The model uncertainty in many cases results from the lack
of data and knowledge adequa te to select the a pprop riate
The Uncertainties probab ility model or through the use of an over simplified
model such as the rational method for storm sewer design.
Uncerta inty can be defined as the occurre nce of ev ents
that are beyond our control. The uncertainty of a water Hydra ulic uncertainty is the uncertainty in the design of
resources system is an indeterministic characteristic and hydrau lic structures an d in the analysis of the performance

8
of hydrau lic strictures. It m ainly arises from three basic Tung (1992), Yen (1986), and Yen and Tung (1993).
types: model, construction and material, and operational Tung (1996) provides an excellent review of the various
conditions of flow. T he mo del unce rtainty results from methods that can be used.
the use of a simplified or an idea lized hyd raulic mo del to
describe flow conditions, which contribute to the Load-Resistance
uncertainty in determining the design capacity of
hydraulic structures. Simplified relationships such as The load for a system can be defined as an external stress
Manning’s equation are typically used to model complex to the system and the resistance can be defined as the
flow processes that cannot be adequately described, capacit y of the system to overcome the external load.
resulting in mod el errors. Load and resistance are terms that have been use d in
structu ral engineering but definitely have a place in the
Structural uncertainty refers to the failure from structural types of risk analysis that need to be performed for water
weakness. Physical failures of hydraulic structures can be resources engineering projects. In water resources
caused by water saturation and loss of soil stability, engineering these terms have a much more general
erosion or hydraulic soil failures, wave action, hy draulic meaning as illustrated in Table 1.
overloading, structural collapse, material failure, etc. An
exam ple is the structural failure of a levee sy stem eithe r in If we use the variable R for resistance and the variable L
the levee or in the adjacent soil. The structural failure for load, then we can define a failure as when the load
could be cause d by w ater saturatio n and lo ss of soil exceeds the resistance and the consequent risk as the
stability. A flood wave can cause increased saturation of probab ility of the loading exceeding the resistance,
the levee through slumping. Levees can also fail because P(L>R). A simple example of this would be the failure of
of hydraulic soil failures and wave action. a dam due to overtopping. The risk would be the
probability that the water surface elevation in a reservo ir
Econo mic uncertainty can arise from uncertainties in exceeds the elevation of the top o f the dam. In this case
construction costs, dam age costs , projected revenue, the resistance is the elevation of the top of the dam and the
operation and maintenance costs, inflation, project life, loading is the maximum elevation of the water surface of
and other intangible cost an d benefit items. Construction, a flood w ave enter ing the rese rvoir.
damage, and operation/maintenance costs are all subject
to uncertain ties because of the fluctuation in the rate of Because many uncertain variables define both the
increase of construction materials, labor costs, resistance and loading, they are both considered as
transportation costs, econom ic losses, regional random variables. A simple example would be to use the
differences, and many o thers. There are also many other rational equation, Q>CiA, to define the design discharge
econo mic and socia l uncertainties that are related to (loading) for a storm sew er. The loading , L=Q, is a
inconvenience losses. An e xamp le of this is the failure of function of three uncertain variables: the runoff
a highway crossing caused by flooding resulting in traffic- coefficient C, the rain fall intensity i, and the drainage area
related losses. A. Because none of these three variables can be
determined with complete certainty they are considered as
Analysis of Uncertainties random variables. So in this case the loading is a random
variable consisting of three random variables. If the
The objective in the analysis of unce rtainties is to resistance is defined through the use of Manning ’s
systema tically incorporate the uncertainties into the education then the resistance is a function of Manning’s
evaluation of the loading and resistance. T he most roughness factor, the pipe diameter, and the slope (friction
comm only used method is the first order analysis of slope). The tw o main contribu tors to unc ertainty in th is
uncertainties. These methods are used to determine the equation would be the friction slope and the ro ughness
statistics of the random vari ables loading and resistance factor so that they are considered as random variables.
which are typically defined through the use of The resistance is the n also a ran dom v ariable w hich is a
determ inistic models but have uncertain parameter inp uts. function of the tw o random variables.

One of the acco mpan ying pa pers by P rofessor T ung in It is interesting to note that in the storm sewer examp le
this journal briefly describes several methods for the both the loading and the resistance are defined by
analysis of uncertainties. For the first order methods one determ inistric equations, the rational equation and
can refer to Chow, Maidment and Mays (1988), Mays and Manning’s equation. Both are considered to have

9
uncertain design parameters that result in the resistance Risk Assessment
and loading being uncertain, and consequen tly are
considered as random variables. In the storm sew er Risk assessment requ ires several phases or steps which
example as in many types of hydraulic structures, the can vary for different types of water resources engineering
loading uncertain ty is actually the hydrolo gic unce rtainty projects. These steps include:
and the resistance u ncertainty is the hydr aulic uncertainty.
Step 1: Risk or hazard identification.
Step 2: Assessment of loads and resistance.
Comp osite Risk Step 3: Perform analysis of uncertainties.
Step 4: Quantify the composite risk.
The above discussion about the hydrolo gic and hy draulic Step 5: Develop the composite risk-safety factor
uncertainties being the resistance and loadin g relationships.
uncertainties leads to the idea of a composite risk. The
probab ility of failure defined previously as the risk, A Model for Risk-Based Design
P(L>R), is actually a composite risk. If only the
hydro logic uncertain ty, in particular the inherent The risk-based design o f wate r resources p rojects
hydro logic uncertainty, w ere consid ered then this wou ld promises to be, pote ntially, the m ost significant
not be a c o m posite risk. In the conventional design applic ation of unce rtainty and risk analys is. The ris k-
processes of water resources engineering projects only the based design of water resources projects integrates the
inherent hydrologic uncertainties have been considered. procedures of econom ics, uncertainty analysis, and risk
Essentially a large return period is selected an d artificially analysis in design practice. Such procedures can consider
considered as the safety factor without any reg ard to the tradeoffs among risk, economics, and other
systema tically accounting for the various uncertainties performance measures in hydraulic structure design.
that actually exist. When risk-based design is embedded into an optimization
framework, the com bined procedure is called optimal
What is being proposed herein, and in many other places risk-based design. The optimal risk-based design
in the literature, is to systematically account for the approach is the ultimate mod el for design, analysis and
uncertainties through the development of the c ompo site operation of water resources engineering projects that we
risk-safety factor relatio nships. W hat has been briefly need to strive for in the future.
described above considers th e hydro logic and hydrau lic
uncertainties in the com posite risk ev aluation. W hat is References
needed is to consider all four of the categories of
uncertainties: hydrologic, hydraulic, structural, and Chow, V.T., D.R. M aidment, and L.W. M ays, Applied
economic in the evaluation of the composite risk. Hydrology, McGraw-Hill, New York, 1989.

Duckstein, L., E.J. Plate, and M. Benedini, “Water

Safety Factor Engineering Reliability a nd Risk : A System Frame work,”
Engineering Reliability and Risk in Water Resources, (ed.
The safety factor is defined as the ratio of the re sistance to by L. Duc kstein and E.J. Plate), Martinus Nijhoff
loading, R/L. Because the safety factor, SF=R/L, is the Publishers, Boston, 1987.
ratio of two random variables, it is also a random variable.
The risk can be written as P(SF<1) and the reliability can Mays, L.W., ed ., Relia bility Analysis of Water
be written as P (SF>1) . Using the storm sew er exam ple Distribution System s, ASCE, New York, 1989.
above, both the resistance and the loading were
considered as rando m varia bles beca use they a re both Mays, L.W. a nd Y.K . Tung, Hydrosystems Engineering
functions of random variables. Consequently because the and Management, McGraw-Hill, New York, 1992.
resistance and the loading for the storm sewer design are
random variables, the safety factor for storm sewer design Tung, Y.K., Uncertainty and Reliability Analysis, Chapter
would also be a random variable. 7 in Water Resources Handbook, (ed. by L.W. May s),
McGraw-Hill, New York, 1996.

10
Yen, B.C., ed ., Stocha stic and R isk Analy sis in Hydra ulic
Engineering, Water Resources Publications, Littleton,
Colo., 1986.

Yen, B.C. an d Y.K. T ung, ed s., Reliability and

Uncerta inty Analysis in Hyd raulic Design, ASCE, New
York, 1993.

Larry W. Mays, Ph.D., P.E., P.H., is professor and

former chair of the Civil and Environmental Engineering
Department at Arizon a State Univer sity. He w as form erly
director of the Center for Research in Water Resources at
the University of Te xas at Austin, where he held an
Engineering Foundation Endowed Professorship.

A registered professional enginee r in several states and a

registered professional hydrologist, Dr. Mays has served
as principal investigator on numerous water resource
research projects sp onsored by fede ral, state, and local
government agencies. He is a member of the American
Society of Civil Engineers and many other professional
organizations, including the Universities Council on
Water Resources for which he has served as president

Dr. Mays has published extensively in water resources

literature. Among his previous books, he is coauthor of
Applied Hydrology and Hydrosystems Engineering and
Management and editor-in-chief of the Water Resources
Handbook, all published by M cGraw -Hill. He rec ently
authored a new b ook, Optimal Control of Hydrosystems,
published by Morrel Dekker, Inc.

11
Table 1. Examples of Load and Resistance for Water Resources Projects (adapted from
Duckstein et al., 1987)

Type of Problem Load Resistance

Bridge pier Scouring Pier pile depth

Flood levee Flood stage Levee height

Dam Flood duration Hydraulic and soil
Flood exposure resistance to botling
Wind sliding, erosion

Water supply Requirements or demand Supply capacity

Flood volume control Flood volume Reservoir flood storage

Max. flood stage control Incoming flood stage Cresting capacity

Underground excavation Piezometric pressure Permeability of walls

Water quality Nutrients, sediments, Cleaning capacity,

(Streams, lakes) pollutant loading low flow augmentation

Waste management Hazards (chemical, Physical, individual,

radioactive) collective

Recreation Number of visitor-days Carrying capacity of

facility

12