Risk Assessment

1
Risk Assessment
2
Risk Analysis

• What are the risks imposed by human activities
  and natural phenomena? (Risk Assessment)
• Are these risks acceptable? (Risk Valuation)
• Can these risks be reduced? (Option Generation)
• How can the options be evaluated? (Cost/Benefit)

3
Risk Assessment

• Risk assessment asks three questions
• What can go wrong ?
• How likely is it to happen?
• What are the consequences?

4
Quantifying Risk Assessment?

• What are the risks from driving a car?
• There are 300,000 accidents per year, 1 in 100 of
  which result in death, there are 62,000,000
  people

5
High
10-2
Risk Criteria
10-3
10-4
Probability
10-5
10-6
Negligible
10-7
1
10
100
1000
10000
Number of fatalities
6
http//www.mvo.ms/SAC/SAC7_Main.pdf
7
(No Transcript)
8
ACCEPTABLE RISK There is a considerable body of
data that implicitly suggests acceptable risk.
Fell (1994, quoting Reid 1989) summarizes risk
statistics for persons voluntarily or
involuntarily exposed to various hazards,
expressed as probability of death per person per
year. They range from 0.00014 10-3 for
structural failure through 0.009 10 3 for air
travel 0.3 10-3 for road accidents 1.9 10-3 for
parachuting 2.8 10-3 for deep-sea fishing as an
occupation. For comparison, the average
30-year-old male has, statistically, a chance of
10-3 of dying this year. It has been inferred
that people, by their actions, implicitly accept
a voluntary risk up to 10-3 and tolerate
involuntary but recognized risks up to perhaps
10-5 The tolerance for risks they suddenly
discover or do not understand is lower yet. It is
also well documented that risks that may affect a
large number of people simultaneously (i.e., air
crashes) are less tolerable than risks of
individual accidents.
Fell, R. (1994). Landslide risk assessment and
allowable risk. Can.Geotech. J., 31(2), 261272.
9
Chief Medical Officers Risk Scale Negligible
Below one per million. This would be of little
concern for ordinary living if the issue was an
environmental one, or the consequence of a health
care intervention. If the word safe is to be
used it must be seen to mean negligible, but
should not import no, or zero, risk. Minimal
between one in a million and one in 100,000, and
that the conduct of normal life is not generally
affected as long as reasonable precautions are
taken. Very low between one in 100,000 and one
in 10,000 Low a risk of between one in 10,000
and one in 1,000. This would fit into many
clinical procedures and environmental hazards.
Many risks fall into this very broad
category. Moderate between one in 1,000 and one
in 100. It would cover a wide range of
procedures, treatment and environmental events.
10
High fairly regular events that would occur at a
rate greater than one in 100. They may also be
described as frequent, significant or
serious. It may be appropriate further to
subdivide this category. Unknown when the level
of risk is unknown or unquantifiable. This is not
uncommon in the early stages of an environmental
concern or the beginning of a newly recognised
disease process (such as the beginning of the HIV
epidemic). Reference On the State of Public
Health the Annual Report of the Chief
Medical Officer of the Department of Health for
the Year 1995. London HMSO, 1996.
11
Risk Assessment

• Hazard identification uses toxicology (cell,
  tissue and animal tests) and epidemiology
  (population data and field samples)
• Exposure assessment includes determination of
  sources, environmental concentrations, exposure,
  dose, and uncertainties

12
Framework for Risk Assessment
Transport and transformation
Source/ Inventory

Emission
Exposure Events
Risk Mitigation
Dose/Response
Biokinetics
Uptake
Risk Characterization
Response
Dose
13
Stages in Risk Assessment
14
A Multimedia, Multiple Pathway Exposure Model
15
Environmental Fate and Transport
16
Inter-media Transfers

• Air and soil to food

17
Mechanism of Action

• Whether a compound reaches a target tissue
  depends on
• Absorption through the GI tract, lung, or skin
• Distribution in the body
• Biotransformation
• Excretion

18
Risk From Radioactive Materials and Toxic
Chemicals in the Environment

• Risk is a function of exposure toxicity
• How much of the toxic material is the individual
  going to be exposed to?
• What amount of toxic material is likely to cause
  an adverse health effect?
• Location and strength of source (Qij)
• Model the spread of the plume (Xi)
• Model the exposure to human or other species
  (Eij)
• Model the dose response relationship (DRi)

19
Unprecedented Uncertainty

• Complicated
• Aleatory natural variation in the physical world
• Epistemic lack of knowledge about the physical
  world

• Complex
• Chaotic small changes in input may lead to large
  changes in output
• Indeterminacy many possible outcomes for given
  input (bifurcations)

20
Environmental Quality Data

• Need to understand assumptions and limitations of
  statistical procedures and how they affect
  conclusions
• Some problems
• Brief data sets
• Numerous parameters
• High variability
• Censored data (magnitude of a constituent is only
  known to be less than (or more than) a certain
  value
• Incompatible data due to different sampling
  methods

Following slides based on Statistical Procedures
for Analysis of Environmental Monitoring Data
Risk Assessment. Mc Bean E A Rovers F A.
Prentice Hall(1998)
21
Variability of Data

• Cost-benefit analysis
• Ground Water Quality Example
• Expensive methodologies limit data spatially and
  in time phenomena may be long term. Need to
  use windowing.

22
Uncertainty of Impacts from exposure to
environmental Contaminants

• How much makes us ill?
• Toxicology
• Epidemiology
• Toxicology
• most information from natural experiments due to
  accidental exposure
• Animal experiments

23
Toxicological Animal experiments

• Environmental regs set lifetime cancer risk no
  greater than 10-4 10-6. To find one cancer in
  animals at that level of exposure would, on
  average, mean testing 100,000 or more.
• Extrapolate from animals to humans?
• Instead use high doses and extrapolate. But
  different methods of extrapolation yield low dose
  estimates which vary by as much as 40 times

24
Epidemiological studies

• Cancer risk evaluated in specific working
  populations extrapolate to low dose exposure in
  general public?
• No interspecies extrapolation an advantage
• Toxicological impact falls off rapidly (not
  linearly) at lower concentrations
• NOAEL no observed adverse effect level the
  threshold for a particular hazard
• RfD reference dose. The NOAEL is adjusted to
  reflect uncertainties

25
RfD and CSF

• The reference dose(RfD) is the maximum allowable
  daily exposure to a non-carcinogenic contaminant
  for a particular body weight that is unlikely to
  cause adverse systemic effects during a lifetime
  for a particular exposure pathway (units of
  mg/kg-body weight/day)
• The cancer slope factor (CSF) or potency value is
  a measure of the change in risk with a change in
  dose of particular carcinogenic contaminant.
  (There is no NOAEL)

26
Estimation of Risk via Migration
Source
Receptor
Migration Pathway
27
Release Pathways- relative importance?

• Fugitive dust generation to impact air
• Fugitive dust generation and depsoition on other
  soils
• Leaching of source areas to contaminate ground
  water
• Human uptake by dermal contact, inhalation or
  ingestion
• Surface runoff following rain to rivers
• Groundwater seepage to rivers
• Aquatic life uptake

28
NonCarcinogenic Risk
29
Hazard index for pathway j
Total exposure Hazard index
30
Carcinogenic Risk
31
Statistical Hypothesis Testing

• Hypothesis There is no significant difference in
  the quality at sampling positions A B.
• Reject There is evidence of a difference
  confidence?
• Fail to Reject there is insufficient evidence to
  conclude that there is a difference

32
Monitoring in the Vicinity of a Landfill
B
A
Landfill waste
leachate
Direction of groundwater movement
33
Statistics A Reminder
Summary Statistics from Excel
34
The t-test for significant difference
The t-test can be used on samples which have
equal variances so we need to test for equal
variances first
35
Are the variances significantly different?
FINV(0.05,11,11)
Since the ratio of the variances is less than the
F-test value we can say that there is no evidence
for a difference in the variances at the two
locations and so the t-test can be used
36
Pooled variance and t-statistic
Tinv(1.0,22)
tlt t-critical so no evidence of a difference
37
Probability Distributions