Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 40
5
Risk Assessment and Uncertainty
A number of risk assessment and uncertainty concepts are used in very different ways by different analysts,
so the following definitions are offered to clarify the committee’s use of terms: hazards refer to threats to people
and the things they value (such as ecosystems, security, or mission success in space), 1 risk is the probability that a
particular event or activity will result in a specified consequence,2 and uncertainty refers to the lack of knowledge
and understanding of the structure of a risk and the connections between the stages of risk evolution. Ideally, if
certainty is high, the connections among risk stages can be characterized quantitatively. A full risk assessment
should always provide not only estimates of risk but also estimates of the associated uncertainties. 3 Uncertainties
can arise from different sources, such as data inadequacies, model parameters, or lack of scientific understand -
ing of the phenomena. Evaluating the type and source of uncertainty and the ability to reduce it through further
research and experience is an important part of any risk analysis.
RISK ASSESSMENT
Overall risk is typically assessed via a probabilistic risk assessment (PRA). In essence, a PRA attempts to
determine the overall risk associated with a particular program or a mission stage by factoring in all known risks,
and their corresponding uncertainties, if known. The threat to mission success and human life from the meteoroid
and orbital debris (MMOD) environment is one of the risks to be considered within a PRA. In its most general sense,
a PRA is a systematic approach to providing quantitative answers to the following fundamental safety questions:
• What can go wrong? (What are the scenarios?)
• How likely is it to happen? (What is the frequency of each scenario?)
• What are its consequences? (i.e., of each scenario)
What is the uncertainty associated with the state of knowledge regarding these answers?4
•
1 R. Kates, C. Hohenemser, and J.X. Kasperson, eds., Perilous Progress: Managing the Hazards of Technology, Westview, Boulder, Colo.,
1985.
2 Kates et al., Perilous Progress: Managing the Hazards of Technology , 1985.
3 National Research Council, Science and Decisions: Advancing Risk Assessment , The National Academies Press, Washington, D.C., 2009,
available at http://books.nap.edu/catalog.php?record_id=12209.
4 S. Kaplan and B.J. Garrick, On the quantitative definition of risk, Risk Analysis 1:11-37, 1981, available at http://josiah.berkeley.
edu/2007Fall/NE275/CourseReader/3.pdf.
40
OCR for page 41
41
RISK ASSESSMENT AND UNCERTAINTY
Before the Challenger accident in 1986, NASA management did not encourage or seem to understand the use
of PRA, as reflected by the accident investigator and Nobel laureate Richard Feynman’s statement, “It appears that
there are enormous differences of opinion as to the probability of a failure with loss of vehicle and of human life.
The estimates range from roughly 1 in 100 to 1 in 100,000. The higher figures come from the working engineers,
and the very low figures from management.”5 After the Columbia shuttle accident, the accident investigation board
again urged NASA to enhance its risk analyses.6 This lack of attention to probabilistic risk assessment by NASA
management had resulted in the MMOD programs finding it difficult to become part of any overall risk assessment
associated with mission design and operations, since there was no agreed upon procedure for doing so. This is less
true today: NASA management has become increasingly aware of the necessity for risk management, as reflected
in NRC studies concerning MMOD with regard to the space shuttle 7 and the International Space Station (ISS).8
The initial goals of NASA’s MMOD efforts were to characterize the risk to humans in space, beginning with
NASA’s crewed spacecraft programs, more than 50 years ago.9 The primary tool for characterizing risk has been
what could be called a Poisson Consensus Model,10 which has the purpose of consolidating theory, measurements,
and assumptions into an average event rate where Poisson statistics apply. This approach requires the integration
of various statistical distributions (such as velocity and angle of impact) by techniques that were established early
in the MMOD programs.11 The history of these consensus models predates the beginning of the space program, 12
and they have since been used and their accuracy improved over the years by the international community.
Over time, NASA’s efforts have expanded to include the characterization of risk to uncrewed spacecraft and
the addition of the orbital debris population as another source of risk. This addition quickly led to the conclusion
that risk could be reduced by minimizing the growth in the orbital debris population. In addition, just as inter -
national interest has increased in minimizing the risk to Earth from natural collisions with comets and asteroids,
the NASA MMOD programs have also expanded to minimize the risk to people and assets on the ground from
reentering orbital debris. As a result of the 2010 National Space Policy, 13 which directs NASA to consider the
issues involved in the active removal of large derelict debris from orbit, the goal of minimizing risk on the ground
is likely to have increasing priority, and trade-offs between reducing the risks to Earth and the risks to spacecraft
in orbit may be required. In addition, the risk posed by MMOD has now expanded to include the possibility of
catastrophic damage to a spacecraft resulting from colliding with a tracked object in orbit. Other changes to NASA’s
mission could occur in the future, for which NASA may need to be prepared.
The hazard from the MMOD environment represents only one component of the total risk to any system or
program. It is up to NASA’s program managers to identify systems critical to their mission and manage the risk
to those systems. The responsibilities of MMOD programs include determining the probability of failure of any
critical system as a result of being hit by either a meteoroid or orbital debris object. Failure for crewed critical
systems is defined as loss of the vehicle or loss of life. In some cases what constitutes failure is obvious, such as
the penetration of a pressurized container. In other cases a cause of failure is not as obvious; examples include
events that could lead to an electrical failure and be interpreted as such (for example, spraying high-speed ejecta
5 R.P. Feynman, Personal observations on reliability of shuttle, Appendix F in Report of the PRESIDENTIAL COMMISSION on the Space
Shuttle Challenger Accident, Volume 2, June 6, 1986, available at http://history.nasa.gov/rogersrep/v2appf.htm.
6 Columbia Accident Investigation Board, History as cause: Columbia and Challenger, Chapter 8 in Columbia Accident Investigation Board
Report, Vol. 1, NASA, August 2003, available at http://www.sociology.columbia.edu/pdf-files/vaughan5.pdf
7 National Research Council, Protecting the Space Shuttle from Meteoroids and Orbital Debris, National Academy Press, Washington, D.C.,
1997, available at http://www.nap.edu/catalog.php?record_id=5958.
8 National Research Council, Protecting the Space Shuttle from Meteoroids and Orbital Debris , 1997.
9 B.G. Cour-Palais, with the assistance of an ad hoc committee, Meteoroid Environment Model¾1969 (Near-Earth to Lunar Surface), NASA
Space Vehicle Design Criteria (Environment), NASA SP-8013, March 1969, available at http://www.spaceflightnews.net/special/sp8000/
archive/00000012/01/sp8013.pdf.
10 M. Drouin, G. Parry, J. Lehner, G. Martinez-Guridi, J. LaChance, and T. Wheeler, Guidance on the Treatment of Uncertainties Associated
with PRAs in Risk-Informed Decisions Making, NUREG-1855, Vol. 1, U.S. Nuclear Regulatory Commission, March 2009.
11 D. Kessler, A Guide to Using Meteoroid-Environmental Models for Experiment and Spacecraft Design Applications, NASA TND-6596,
NASA, March 1972.
12 A.C. Lovell, Meteor Astronomy, Oxford University Press, Oxford, U.K., 1954.
13 National Space Policy of the United States of America , June 28, 2010, available at http://www.whitehouse.gov/sites/default/files/na -
tional_space_policy_6-28-10.pdf, accessed July 6, 2011.
OCR for page 42
42 LIMITING FUTURE COLLISION RISK TO SPACECRAFT
over a circuit board, severing an electrical wire, or creating a plasma). In these more complex failures, additional
hypervelocity testing may be required to determine failure mechanisms.
A solution to reducing failure rates from such collisions can be to add shielding, which sometimes must be
customized to minimize the added weight to the spacecraft. Where uncertainty exists, safety factors are sometimes
added. In other cases, the risk can be reduced by redundant systems or changes in operations, such as orienting
the spacecraft in a direction that minimizes the risk (as has been done for the space shuttle).
It is the responsibility of the MMOD programs, usually through the Hypervelocity Impact Technology Facil -
ity, to coordinate with the program managers and offer the best solutions for their mission. The MMOD programs
have put together a handbook to aid in selecting spacecraft protection options. 14 As stated in the handbook, the
definition of “failure” has a significant influence on the resulting risk. For example, failure may be defined as a
penetration of a critical item that could lead to either loss of the function of the item, or loss of the crew. Such a
definition could lead to a significant amount of hypervelocity testing and shielding development, as was the case
for the critical items identified for the ISS.15 Alternatively, the failure criterion could be as simple as the depth
of the pits on a window pane that might lead to loss of the window during launch, which was one of the critical
items identified for the space shuttle.16 In this case a sufficient amount of hypervelocity testing had already been
conducted to identify the frequency with which such pits would occur, and all that was then necessary was to
operationally plan to examine the windows after each flight and have enough spare windows on hand as replace -
ments when craters were found that exceeded the critical depth.
The focus on collisions and risk from penetration does not, however, fully cover all of the risks involving
orbital debris and interplanetary meteoroids. Other risks are discussed more fully in other chapters. In some cases
the difficulty in assessing risk is not the result of poor analysis but is the result of lack of data; for example, as is
pointed out in Chapter 4, the risk analysis to be performed is sound but suffers from the lack of measurements in
the interplanetary environment.
Finding: NASA’s MMOD risk assessment processes have evolved beyond focusing primarily on the
damage to spacecraft from collisions with debris that are too small to track, to incorporating a more
complete range of risks. More remains to be accomplished, however, including the need in some cases
for more measurements as parameters for risk analyses. As gaps are filled, NASA’s MMOD efforts can
progress toward ever more integrative risk assessment in which all sources and types of risk are modeled
and assessed.
Recommendation: Although NASA should continue to allocate priority attention and resources to col-
lision risks and conjunction analysis, it should also work toward a broad integrative risk analysis to
obtain a probabilistic risk assessment of the overall risks present in the MMOD domain in which all
sources of risk can be put in context.
UNCERTAINTY
Communication of Information About Uncertainty
NASA’s work on reducing the threat to spacecraft posed by orbital debris and meteoroids faces increasingly
challenging problems stemming from the complexity of physical changes in space, changing spacecraft designs,
increased international use of space and contributions to debris, and private and public sector initiatives in space.
An intrinsic challenge also exists in creating models that fully capture the uncertainties and the phenomena being
modeled. Examining the sources of the uncertainties, how to reduce uncertainties, identifying those that cannot
14E. Christiansen, J. Arnold, A. Davis, D. Lear, J.-C. Liou, F. Lyons, T. Prior, M. Ratliff, S. Ryan, F. Giovane, B. Corsaro, and G. Studor,
Handbook for Designing MMOD Protection, NASA TM-2009-214785, NASA, June 2009.
15 E. Christiansen, K. Nagy, D. Lear, and T. Prior, Space station MMOD shielding, Acta Astronautica 65(7-8):921-929, 2009.
16 K. Edelstein, Orbital Impacts and the Space Shuttle Windshield, NASA-TM-110594, NASA, Washington, D.C., 1995.
OCR for page 43
43
RISK ASSESSMENT AND UNCERTAINTY
be significantly reduced or removed over the near term, and estimating the time and effort required to signifi -
cantly reduce extant uncertainties are also issues that need to be addressed. It must also be understood that natural
variabilities in the MMOD environment will prevent uncertainties from being removed entirely, no matter how
sophisticated and detailed testing programs and modeling efforts become. No less a problem is how to then com -
municate uncertainties to decision makers or the public. Integrating uncertainty analysis into decision making
related to debris impact risks has progressed and can make additional progress going forward. Since many decisions
are made at the mission level, how to effectively communicate uncertainties to people with little formal training
in managing uncertainties is a matter of considerable importance. The state of knowledge of how to characterize,
catalog, and communicate the range of uncertainties is an evolving area.
A description of uncertainty is critical to guiding future research efforts, as well as communicating to those
who may be affected by the risk. There is a considerable literature on uncertainty and risk more generally. 17 A
recent NRC report also captures some of the science of risk and uncertainty.18 The principles of uncertainty are
summarized in Box 5.1 of the present report, and it is recommended that the MMOD programs increase their
efforts to adhere to those principles.
Finding: The calculation and communication of information about uncertainty are critical to properly
assessing operational alternatives based on calculated risks posed by orbital debris.
Uncertainty in MMOD Modeling
All of the MMOD models contain uncertainties (see, for example, the discussions of uncertainties pertain -
ing to the orbital debris and meteoroid models in Chapters 3 and 4 and BUMPER in Chapter 6). The building of
these models requires an examination of uncertainties that result from data, which may result from a number of
measurements, tied together with a number of assumptions. In the past, uncertainties that were judged to be large
enough to significantly affect spacecraft designs or operations were identified and brought to the attention of the
appropriate program manager. This happened in July 1987 when NASA headquarters’ senior staff was briefed on
the uncertainty in the orbital debris flux due to a lack of measurements of debris in the size range of interest to the
then-planned space station. This briefing led to the current Haystack observation program. In July 1993, a briefing
to the Shuttle Program Office about the uncertainty of possible damage to the space shuttle during a predicted
Perseid meteor storm led to a delay in the STS-51 launch and the beginning of the current meteoroid program at
Marshall Space Flight Center. Consequently, it appears that the MMOD programs effectively deal with uncertainty
when that uncertainty is either small enough to be ignored or so large that it is obvious that more data are required.
However, this approach may not be sufficient going forward. Inadequate consideration of MMOD uncertainties
is becoming more important as the program expands, more data are obtained, and safety requirements become
tighter. An increased awareness of uncertainty will also be required to adequately respond to various findings in
other chapters of this report.
As discussed elsewhere in this report, there is an opportunity to reduce uncertainties in the environment
See, for example M.G. Morgan and M. Herion, Uncertainty: A Guide Toward Dealing with Uncertainties in Quantitative Risk and Policy
17
Analysis, Cambridge University Press, Cambridge, U.K., 1990; National Research Council, Understanding Risk: Informing Decisions in a
Democratic Society (P.C. Stern and H.V. Fineberg, eds.), National Academy Press, Washington, D.C., 1996; National Science and Technology
Council, Grand Challenges for Disaster Reduction, Washington, D.C., 2005; National Research Council, Science and Decisions: Advancing
Risk Assessment, The National Academies Press, Washington, D.C., 2009; National Research Council, Science and Judgment in Risk Assess-
ment, National Academy Press, Washington, D.C., 1994; A.M. Finkel, Confronting Uncertainty in Risk Management: A Guide for Decision
Makers, Center for Risk Management, Washington, D.C., 1990; R.E. Kasperson, Coping with deep uncertainty: Challenges for environmental
assessment and decision making, pp. 337-348 in Uncertainty: Multi-disciplinary Perspectives on Risk (G. Banner and M. Smithson, eds.),
Earthscan, London, U.K., 2008.
18 National Research Council, Science and Decisions: Advancing Risk Assessment, The National Academies Press, Washington, D.C., 2009.
Science and Decisions, known as the “Silver Book,” replaced the “Red Book” (National Research Council, Risk Assessment in the Federal
Government, National Academy Press, Washington, D.C., 1983).
OCR for page 44
44 LIMITING FUTURE COLLISION RISK TO SPACECRAFT
BOX 5.1
Recommended Principles for Analysis of Uncertainty and Variability
1. Risk assessments should provide a quantitative, or at least qualitative, description of uncertainty and
variability consistent with available data. The information required to conduct detailed uncertainty analyses
may not be available in many situations.
2. In addition to characterizing the full population at risk, attention should be directed to vulnerable
individuals and subpopulations that may be particularly susceptible or more highly exposed.
3. The depth, extent, and detail of the uncertainty and variability analyses should be commensurate
with the importance and nature of the decision to be informed by the risk assessment and with what is
valued in a decision. This may best be achieved by early engagement of assessors, managers, and stake-
holders in the nature and objectives of the risk assessment and the terms of reference (which must be
clearly defined).
4. The risk assessment should compile or otherwise characterize the types, sources, extent, and mag-
nitude of variability and of substantial uncertainty associated with the assessment. To the extent feasible,
there should be homologous treatment of uncertainty among the different components of a risk assessment
and among different policy options being compared.
5. To maximize public understanding of and participation in risk-related decision making, a risk as-
sessment should explain the basis and the results of the uncertainty analysis with sufficient clarity to be
understood by the public and decision makers. The uncertainty assessment should not be a significant
source of delay in the release of a risk assessment.
6. Uncertainty and variability should be kept conceptually separate in the risk characterization.
measurements, shielding, and modeling programs. Box 5.2 summarizes an example seen in an early publication
of the Haystack data which makes use of confidence bars.19
However, there is an additional uncertainty in debris size that cannot be quantified with confidence bars
and is likely to be more important; this uncertainty is the result of three assumptions: (1) the distribution of the
fragment shapes and composition of test samples used to determine debris size from the RCS is assumed to be
representative of the distribution of shapes and composition of debris in orbit (see Chapter 2); (2) those shapes
and compositions are then assumed to be adequately described by a single parameter known as a “characteristic
length”; and (3) these assumptions carry through to the hypervelocity testing, and the program BUMPER, in which
the additional assumption is made that the “characteristic length” of a given piece of debris can be approximated
with an aluminum sphere of the same diameter (see Chapter 6). The importance of these last two assumptions can
be seen by comparing the results of two studies: under the current set of assumptions, shielding is over-designed; 20
but if the debris size had been defined as the mass having a given RCS, and that mass were approximated with an
aluminum sphere of the same mass, the shielding would be under-designed.21
However, the assumption that any particular size sphere is an approximation to any assumed or measured dis -
tribution of shapes and composition is not supported by an analysis that includes integrating over the distributions
of shape and composition.22 Finally, China’s anti-satellite test (see Box 1.2 in Chapter 1) gives reason to question
that the assumed distribution of shapes and composition is correct. The area-to-mass ratio of the fragments from
19 E.G. Stansbery, G. Bohannon, C. Pitts, T. Tracy, and J. Stanley, Radar observations of small space debris, Advances in Space Research
13(8):43-48, 1993.
20 J. Williamsen, Review of Space Shuttle Meteoroid/Orbital Debris Critical Risk Assessment Practices , Report No. P-3838, Institute for
Defense Analyses, Alexandria, Va., November 2003.
21 B. Cour-Palais, The shape effect of non-spherical projectiles in hypervelocity impacts, International Journal of Impact Engineering
26:129-143, 2001.
22 When this type of analysis is performed to relate characteristic size to RCS, it is heavily weighted toward the more numerous smaller
objects. Consequently, if such an analysis were applied to relate impact damage to some characteristic size, it might easily be weighted toward
smaller, but higher density iron or aluminum oxide debris objects.
OCR for page 45
45
RISK ASSESSMENT AND UNCERTAINTY
BOX 5.2
The Haystack Data
Stansbery et al. illustrate with 99 percent confidence bars that when the number of objects passing
through the Haystack field of view is large, as it is for the smaller debris, there is very little uncertainty in
the average flux for that diameter debris compared to the uncertainty in flux for larger diameters.1 However,
“diameter” is not directly measured; radar cross section (RCS) is measured and, consequently, there is an
uncertainty in the debris diameter corresponding to a given RCS. In considering Bohannon and Caampued,
one would expect the diameter uncertainty to also increase with decreasing flux, given that the distribution
of possible RCSs for a given diameter is used in a statistical technique to relate RCS to diameter.2,3 The
lack of increasing uncertainty shown in debris diameter indicates there may be a problem with describing
the uncertainty in debris diameter. An examination of Bohannon and Caampued4 reveals the probable
cause: only an approximate uncertainty is given, which is acknowledged as “depending on the RCS statis-
tics.” After 18 years of observations, those statistics would have reduced both uncertainties considerably.
This was illustrated 8 years later at the Third European Conference on Space Debris, when the statistical
uncertainty for the smaller, more hazardous size debris had all but disappeared, and assumptions about
shape were being tested using radar polarization measurements.5 Shape and mass still remain a significant
cause of uncertainty due to incomplete testing and analysis of all assumptions.
1 E.G. Stansbery, G. Bohannon, C. Pitts, T. Tracy, and J. Stanley, Radar observations of small space debris,
Advances in Space Research 13(8):43-48, 1993.
2 G. Bohannon and T. Caampued, Debris Size Estimation from Radar Cross Section Data Using Quadratic and
Non-Parametric Classifiers, XonTech, Inc. Report No. 930301-BE2198, Van Nuys, Calif., June 1993.
3 G. Bohannon and N. Young, Debris Size Estimation Using Average RCS Measurements, XonTech, Inc. Report
No. 930781-BE2247, Van Nuys, Calif., September 1993.
4 G. Bohannon and T. Caampued, Debris Size Estimation from Radar Cross Section Data Using Quadratic and
Non-Parametric Classifiers, XonTech, Inc. Report No. 930301-BE2198, Van Nuys, Calif., June 1993;
5 M.J. Matney and E. Stansbery, What are radar observations telling us about the low-Earth orbital debris envi-
ronment, in Proceeding of the Third European Conference on Space Debris, SP-473, European Space Agency, Paris,
France, October 2001.
China’s Fengyun-1C satellite is different from fragments from other known events. 23 Consequently, this assump-
tion about the relationship between shape and size needs to be reexamined, possibly leading to new ground tests
to obtain representative samples and new RCS calibrations from those samples (for additional discussion on RCS
calibrations, see Chapter 2).
The committee noticed a significant gap in identifying uncertainty in the more recent measurements and
models, not only in those models describing the environment, but in models like BUMPER describing the risk
to the environment (additional details on BUMPER can be found in Chapter 6). The committee asked for, but
did not receive, uncertainty analysis, nor did it receive a comparison with model predictions and measurements,
especially for those models used to predict the long-term MMOD environment. The uncertainty in these model
results are not only from the statistical nature of what is being measured, where that uncertainty can be quanti -
fied and integrated into an overall risk assessment, but from assumptions that go into the models. Although the
uncertainty in these assumptions cannot be assigned a probability of being correct, they can be altered within the
bounds of “reasonable assumptions” to determine the sensitivity of the assumptions to the predicted risk, resulting
in a range of possible risks.
Consequences of not following these principles have been identified by other NRC studies (see also, for
23J.-C. Liou and N. Johnson, Characterization of the catalog Fengyun-1C fragments and their long-term effect on the LEO environment,
Advances in Space Research 43(9):1407-1415, 2009.
OCR for page 46
46 LIMITING FUTURE COLLISION RISK TO SPACECRAFT
example, U.S. EPA 200424). These pitfalls also apply to the MMOD programs and include (1) not allowing for
optimal weighting of the probabilities and consequential errors; (2) not permitting a reliable comparison of alter-
native decisions; (3) failing to communicate the range of control options that would be comparable with different
assessments of the true state of nature; and (4) precluding the opportunity for identifying research initiatives.
Examples of these pitfalls are characteristic of the CARA/COLA programs where there is a significant lack of
uncertainty analysis associated with those programs (see Chapter 9).
In general, NASA MMOD programs have embraced some of the principles identified in Box 5.1. However,
as both the agency and the MMOD programs mature, it becomes increasingly important to better characterize risk
and uncertainty in all aspects of the MMOD problems being addressed. Other issues to be addressed include the
sources of the uncertainties, how to reduce uncertainties, identifying those that cannot be significantly reduced or
removed over the near term, and estimating the time and effort required to reduce significantly extant uncertainties.
Of course, natural variability in the MMOD environment will prevent uncertainties from being removed entirely,
no matter how sophisticated and detailed testing programs and modeling efforts become.
It is equally important that uncertainty information continue to be communicated to decision makers or pro -
gram leaders because they are the ones who will determine how to handle this information within the NASA frame-
work of mission planning and operations. While the calculation and communication of uncertainty information to
decision makers, including those who plan space missions, has improved at NASA, it is also apparent that a fully
integrated cataloging and assessment of MMOD-related uncertainties does not routinely occur in mission-planning
and decision-making activities¾as noted above, this type of information is typically conveyed to management when
the uncertainties are either small enough to be ignored, or large enough to be obvious so that either more data or
some sort of corrective action is required. Since many of these decisions appear to be made at the program level,
effective communication of uncertainty information both to the public and to the proper management levels is an
issue of considerable importance that needs constant reevaluation and oversight.
Recommendation: NASA’s meteoroid and orbital debris programs should increase their efforts to re-
duce the uncertainty and variability in models through acquisition of measurements (and where neces-
sary, to do testing and analysis) for continually improving assessment of risk and characterization of
uncertainty. Together with its MMOD efforts, NASA should continue to advance the agency’s efforts to
present information on uncertainty in risk analyses. Special attention should be given to maximizing
public understanding of uncertainty analysis through peer-reviewed papers and other publications.
Environmental Protection Agency, An Examination of EPA Risk Assessment Principles and Practices, EPA/100/B-04/001, Washington,
24
D.C., March 2004.
