Download m03 - The University of Hong Kong

Jf
cr
COMPARATIVE RISK
INDICATORS
&
POSSIBLE CLIMATE CHANGE
TRENDS IN HONG KONG AND
IMPLICATIONS FOR THE
SLOPE SAFETY SYSTEM
GEO REPORT No. 128
H.W. Sun
&
N.C. Evans
© The Government of the Hong Kong Special Administrative Region
First published, October 2002
Prepared by:
Geotechnical Engineering Office,
Civil Engineering Department,
Civil Engineering Building,,
101 Princess Margaret Road,
Hornantin, Kowloon,
Hong Kong.
PREFACE
In keeping with our policy of releasing information
which may be of general interest to the geotechnical profession
and the public, we make available selected internal reports in a
series of publications termed the GEO Report series. A charge
is made to cover the cost of printing.
The Geotechnical Engineering Office also publishes
guidance documents as GEO Publications. These publications
and the GEO Reports may be obtained from the Government's
Information Services Department. Information on how to
purchase these documents is given on the last page of this
report.
R.K.S. Chan
Head, Geotechnical Engineering Office
October 2002
BOOKS REGISTRATION ORDINANCE
2003 "01036
Chapter 142 \\\
-
4
EXPLANATORY NOTE
This GEO Report comprises a Special Project Report and a Technical Note on two
separate research and development projects carried out by the Special Projects Division in
2001.
The reports are presented in two separate sections in this GEO Report.
Their titles
are as follows:
Section
Title
1
Comparative Risk Indicators
2
Possible Climate Change Trends in Hong Kong and
Implications for the Slope Safety System
Page No.
5
31
SECTION 1:
COMPARATIVE
RISK INDICATORS
H.W. Sun & N.C. Evans
This report was originally produced in January 2001
as GEO Special Project Report No. SPR 1/2001
- 6 -
FOREWORD
This Report reviews the use of risk indicators as part of
the risk assessment process.
It examines the possible
implications of using different types of risk indicator for the
various types of risk assessment carried out by the GEO. The
study derives from a recommendation by the Slope Safety
Technical Review Board that the use of nominal risk indicators
be considered.
The Report was first circulated within GEO as a draft
Discussion Note, in November 2000. A wide spectrum of
comments and suggestions were received, and the report was
revised and expanded to a Special Project Report.
The study was carried out, and the report written, by Dr
H.W. Sun and Mr N.C. Evans of Special Projects Division.
Many colleagues in GEO provided valuable comments on the
original draft.
P.L.R. Pang
Chief Geotechnical Engineer/Special Projects
ABSTRACT
This Report reviews the use of risk indicators as part of the risk assessment process.
It examines the possible implications of using different risk indicators for various types of risk
assessment.
Risk assessment involves a combination of analytical methods and consideration of
social factors, and is best regarded as a framework within which informed decisions can be
made. Quantitative Risk Assessment (QRA), comprising the calculation of risk numerically
and comparison with various criteria, can form part of this process but it is not necessarily the
most important part.
Risk assessments using calculated risk are usually based on probability of death.
However, studies on the public perception of risk have highlighted its multi-dimensional and
subjective social value aspects.
Risk, especially calculated risk, is a measure of the effect of a hazard on a susceptible
population. Very few risks are distributed evenly. This can be referred to as risk
"clustering". The implications of clustering need to be considered when comparing different
types of risk.
Risk acceptance criteria are controversial. Decisions made based solely on these
criteria are placing very high reliance on both the reliability of the risk analysis and the
suitability of the risk indicator for the required purpose. There is no clear consensus on how
such criteria could be applied in different situations. The issue of cumulative risk does not
appear to be satisfactorily addressed anywhere.
Risk assessment should be iterative and consultative. Attempts should be made to
integrate social perceptions with calculation in the risk assessment process.
When
calculating risk, the risk indicator chosen, and the way in which it is calculated, should be
determined by the nature of the decision which is to be made. Decisions should not,
however, be made solely on the basis of calculated risk or cost/benefit analyses.
Social perceptions of risk do not necessarily agree with objectively assessed risk. In
Hong Kong, social perception of risk appears to be governed by a reasonably well-informed
opinion of the probability of an individual's exposure to that risk.
CONTENTS
Page
No.
Title page
5
FOREWORD
6
ABSTRACT
7
CONTENTS
8
1.
INTRODUCTION
9
2.
"CALCULATED RISK" AND "PERCEIVED RISK"
9
2.1
General
9
2.2
Uncertainty
10
2.3
Multi-dimensional Nature of Risk and Risk Measures
11
2.4
Hazards and Susceptible Populations
12
2.5
Social Values
13
2.6
Risk "Acceptability" and Cost/Benefit Analysis
15
3.
COMPARATIVE RISKS
17
4.
WAYS FORWARD
17
4.1
Risk Assessment Process
17
4.2
Decisions
18
4.3
Consultation
22
5.
CONCLUSIONS
22
6.
REFERENCES
23
LIST OF FIGURES
24
1. INTRODUCTION
This Report reviews how risk indicators can be developed and presented in different
ways, and discusses this in the context of various types of risk in Hong Kong, including risk
from landslides. Some suggestions are proposed for possible ways forward in the use of risk
assessment approaches and techniques within the Slope Safety System. The intention of the
Report is to draw attention to the various issues concerned.
Risk assessment in the broad context involves a combination of analytical methods and
assessment of social values, and is not totally amenable to the application of the physical
scientific methods or design processes with which engineers tend to be familiar. It is, rather,
a framework within which an iterative process can take place with the intention of making
informed decisions. Quantitative Risk Assessment (QRA), comprising the calculation of
numerical risk and comparison with various criteria, can form part of this process, but it is not
necessarily the most important part.
Slovic (1999) summarises the current position of risk management as follows. "Risk
management has become increasingly politicised and contentious. Polarised views,
controversy and conflict have become pervasive. Research has begun to provide a new
perspective . . . by demonstrating the complexity of the concept "risk" and the inadequacies
of the traditional view of risk assessment as a purely scientific enterprise . . . Risk assessment
is inherently subjective and represents a blending of science and judgement with important
psychological, social, cultural and political factors. ".
2. "CALCULATED RISK" AND "PERCEIVED RISK'3
2.1 General
There are methodological and disciplinary divisions between risk researchers,
scientists, engineers and lay people in their understanding and perceptions of risk. Risk
assessments using "calculated risk" are usually based on one single measure of risk. On the
other hand, studies on the public perception of risk have highlighted the multi-dimensional
aspects and the subjective social value elements of risk. It is important for regulators and
decision-makers to understand and consider these aspects of risk so that the needs of the
public can be properly addressed.
It is possible to obtain a "calculated risk" using a number of different analytical
techniques. There are problems with this approach. Calculation of risk is itself subject to
many uncertainties and to (sometimes hidden) value-laden judgements concerning the relative
importance and applicability of different risk indicators. "Calculated risk" usually takes no
account of the nature of the risk and the demographics of the affected population. Finally,
there can be communication problems and lack of understanding when discussing "calculated
risk" with the various parties involved in decision-making. These problems are discussed in
more detail below.
10 -
2.2 Uncertainty
Stern & Fineberg (1996) discuss uncertainty in risk assessment in some detail. Their
views are summarised here as follows. Risk characterisations often give misleading
information about uncertainty in several ways. They may give the impression of more
scientific certainty or unanimity than exists (or the opposite); they may suggest that
uncertainty is a matter of measurement when in fact it is a subject of judgement and
disagreement; and they may give the impression that certain risks do not exist when in fact
they have not been considered. Many risk characterisations still present point estimates of
risk, representing these as upper-bound estimates and providing little or no analysis of the
extent of overestimation. In spite of the obvious shortcomings of point estimates,
alternatives (such as probability distributions) have not yet gained widespread acceptance by
regulatory agencies and decision-makers. The problem of summarising uncertainty may
have no technical solution.
It is therefore important to declare the approach adopted in the characterisation of risk,
and what is known of the associated uncertainty, when the risk results are communicated to
the decision-makers or the concerned parties. Some of the early applications of QRA to
landslide and boulder fall hazards in Hong Kong adopted a point-value "best-estimate"
approach, without due regard for the uncertainties involved.
Uncertainty commonly surrounds the likelihood, magnitude, distribution and
implications of risks. Uncertainties may be due to random variations and chance outcomes
in the physical world. These are known as "aleatory" uncertainties and might, when
assessing landslide hazards, include rainfall distribution, progression of weathering, or
accidental occurrences such as burst water pipes. Uncertainties also arise from a lack of
knowledge. These are referred to as "epistemic" uncertainties, and, in the context of
landslide hazard, might include inadequate ground or stability analysis models, or the
incorrect or inappropriate application of analytical techniques. Sometimes, analysts may not
know which model of a risk-generating process is applicable - this is known as indeterminacy.
An example of indeterminacy arose from recent studies of landslide hazard and risk
from natural terrain (HAP, 2000). During these studies, two models of possible future
landslide frequency were developed, one based on regional trends relating to slope
characteristics (which may not always apply at individual sites), and the other on site-specific
observed landslide frequency (which may be significantly affected by occasional extreme but
localised rainfall events). The two models differed significantly in their prediction of
possible future landslide frequency, and there is no way of telling which will be more accurate
for any particular site.
When uncertainty is present but unrecognised it can be called simply ignorance.
Where uncertainty is recognisable and quantifiable the language of probability can be used.
Objective or frequency-based probability measures can describe aleatory uncertainty
associated with randomness. This approach was used to manipulate statistics of landslide
frequency, magnitude and behaviour during the recent studies of natural terrain discussed
above (HAP, 2000). The approach was found to be workable for those types of landslides
for which a substantial database of past events was available, and resulted in the calculation of
risk parameters with associated bands of uncertainty rather than point values.
11
Subjective probability measures (based on expert opinion) can describe epistemic
uncertainties associated with lack of knowledge. This approach has potential for addressing
hazards for which there are insufficient data of previous events to construct mathematical
probability distributions. This may be relevant when considering high magnitude/low
frequency events such as large natural terrain landslides or debris flows. Expert opinion
may also have a useful role to play if the probability of failure of man-made slopes is
examined and there are insufficient data to carry out a statistical analysis of historical failure
frequency. However, all such assessments need to be kept under constant review to ensure
that they reflect the current state-of-the-art with respect to both available data and
understanding of the processes involved, and the associated uncertainties should always be
recognised.
Sometimes uncertainty is recognised but cannot be measured, quantified or expressed
in statistical terms. In cases such as these, all that can really be done is to examine various
possible hazard scenarios and subjectively rank them in terms of probability and consequence.
Morgenstern (1995; 2000) also discusses the different types of uncertainty with respect
to geotechnical engineering and risk analysis, and emphasises the important role of qualitative
analyses in the risk assessment process.
2.3 Multi-dimensional Nature of Risk and Risk Measures
Risks are measured in tenns of the degree of harm that they may cause to society,
organisations or individuals. Harm can be assessed in different ways. Measures of harm
can include deaths, injuries, other health hazards, lost of production time, economic loss,
inconvenience to individuals, socio-political disruption, environmental damage, etc. The
choice of a particular measurement unit (the most usual is number of deaths) implies that we
place particular value on that indicator. Social value and perception of risk are discussed
further in Section 2.5.
Stern & Fineberg (1996) discuss the difficulties in choosing risk measures. They
consider that the choice of measure can make a big difference, especially when one risk is
compared to another. To properly characterise risk it may be necessary to use more than one
measure. All measures are value-laden. Even a simple measure such as number of deaths
embodies its own values, making no distinction between age of individuals or nature of death.
Risk characterisation often focuses on a single outcome, most often human fatalities, but risk
is multi-dimensional and even a single outcome can have multiple attributes. The general
problem is how to characterise what is known about a risk when there is no clear way to
combine its many attributes into a single measure.
One option is to reduce different risk attributes to monetary terms, enabling a total risk
"cost" to be derived. Such an approach would require monetary value to be placed on
certain non-monetary outcomes, such as death, injury, inconvenience, distress or
environmental damage. While techniques exist for doing this, they are inevitably subjective
and controversial. Cost/benefit analysis is considered further in Section 2.6.
12
2.4 Hazards and Susceptible Populations
Risk, especially calculated risk, is a measure of the effect of a hazard on a susceptible
population. When calculating risk it is necessary to be very clear about how the population
through which the risk is distributed has been defined. For instance, in a population such as
that in Hong Kong, it may be reasonable to assume that risks from certain medical conditions
are distributed relatively evenly through the entire population. In such a case as this it is, at
first sight, eminently reasonable to calculate a "global" risk in terms of susceptibility or
mortality rates divided by the entire population. If the risk applies to all equally, this
"global" risk will be a fair reflection of the "individual" risk.
In reality, very few risks are distributed evenly. Even diseases or medical conditions
to which all are exposed will tend to affect subsets of the population in different ways
(depending on age, sex, habits, wealth, etc). To take an extreme example, a small number of
individuals may be highly exposed to an unusual risk, possibly as a result of their occupation.
If the associated mortality rates, or other measures, are divided amongst the entire population,
it will mask the fact that the risk is actually concentrated on a few individuals whose level of
risk may be extreme. This can be referred to as risk "clustering".
Risks can be clustered geographically, such as for a Potentially Hazardous Installation
(PHI) with a clearly defined radius of possible impact (QRA was originally developed for
such industrial installations with clustered risk); societally, with individuals voluntarily
engaging in hazardous occupations or activities or involuntarily exposed to above-average
levels of risk due to poor housing, lack of education, restricted career choice, etc; or
temporally, i.e. some hazards materialise only occasionally, but could affect anyone in a given
population - the extreme example would be a meteorite strike! Conversely, risks within a
given population can be geographically diffuse, societally ubiquitous and temporally
relatively even. Possible examples are traffic accidents and medical conditions such as
cancer.
Between these extremes there is, of course, a continuum leading to an almost infinite
range of hazard and risk characteristics. It is very important that these factors are considered
when calculating risk, communicating the results, and comparing risks from different hazards.
If landslides and boulder falls are considered, the position with respect to clustering is
not straightforward. Different types of slope failure have different clustering characteristics.
For instance, the risk from large channelised debris flows is geographically clustered in
streamcourse mouths, temporally clustered during rainstorms and, possibly, societally
clustered in that those living in potential run-out zones in flimsy structures are most at risk.
Conversely, risk from failed man-made slopes may be more diffuse both geographically and
societally. Temporal clustering might also be more diffuse as these events do not always
occur during rainstorms (leaking water services, construction, etc, also trigger these events).
Risk from open-slope natural terrain landslides and boulder falls might fall between the two in
terms of clustering. Figure 1 shows a possible way of illustrating graphically the relative
clustering of different types of risk.
When calculating risk from landslides and boulder falls it is necessary to consider
whether or not it is appropriate to average the risk over the whole population, or to calculate
the risk to the population which is affected. There is no definitive answer to this, as the most
-
13 -
appropriate parameter will depend, to some extent, on the particular requirements of a given
risk assessment, i.e. what decisions are being made and why the risk assessment is being
carried out.
In terms of clustering characteristics, large channelised debris flows appear to share
some properties with PHIs (primarily the concentration of the hazard location and the
population at risk). Open slope natural terrain landslides and man-made slope failures seem
to be clustered somewhere between construction accidents and fire/traffic accidents with
respect to geographic distribution and population at risk. The implications of these different
types of clustering need to be considered when comparing different types of risk, or when
faced with the task of choosing the most appropriate type of risk analysis.
2.5 Social Values
The above discussion relates to the assessment of calculated risk. However, as
mentioned earlier, there is a further dimension to risk assessment which relates to social
values. Social theory states that people tend to emphasise qualitative rather than quantitative
aspects of hazard and risk. Their individual views are coloured by personal experience,
belief, the media, and local culture, and they may be naive or irrational. Even if this is the
case, such views cannot be ignored. Removing an irrational but feared risk can improve an
individual's quality of life.
Individuals tend to place a higher priority on hazards which they regard as catastrophic.
Stern & Fineberg (1996) discuss this as follows. People tend to perceive a risk as higher
when it evokes perceptions of dread, uncontrollability and catastrophe, and they want to see
strict regulations to control such risks. Specialists in risk analysis instead tend to see
riskiness as synonymous, especially for policy purposes, with expected annual mortality,
consistent with the ways that risks tend to be characterised in QRA.
Baron et al (2000) report the findings of a survey on risk perception which compared
"experts" and "non-experts". This work found that the two groups did not differ much in
what determined their worries or their desire for action, but they did differ in their beliefs
about particular risks. "Experts" tended to be more concerned about mundane but
statistically frequent events such as traffic or domestic accidents, while "non-experts" were
more concerned about statistically smaller risks from events such as cancer. This may be a
result of inflated worry due to anticipated "dreadful" consequences, possibly exacerbated by
media coverage which tends to focus on "newsworthy" rather than everyday events.
Citing statistics of "actual risks" often does little to change peoples attitudes and
perceptions. Non-specialists factor in complex, qualitative considerations including
judgements about uncertainty, dread, catastrophic potential, controllability, equity and risk to
future generations. Stern & Fineberg quote the US National Research Council (1989:52) as
follows:
"Those quantitative risk analyses that convert all types of human health hazard to a
single metric carry an implicit value-based assumption that all deaths or shortenings of life
are equivalent in terms of the importance of ignoring them. The risk perception research
shows not only that the equating of risks with different attributes is value laden, but also that
14
the values adopted by this practice differ from those held by most people. For most people,
deaths and injuries are not equal - some kinds of circumstances of harm are more to be
avoided than others. One need not conclude that quantitative risk analysis should "weight
the risks to conform to majority values. But the research does suggest that it is
presumptuous for technical experts to act as if they know, without careful thought and
analysis, the proper weights to equate one type of hazard with another. When lay and expert
values differ, reducing different kinds of hazard to a common metric (such as number of
fatalities per year) and presenting comparisons only on that metric have great potential to
produce misunderstanding and conflict and to engender mistrust of expertise."
A pilot study based on a small (34 people) sample was carried out by Hong Kong
University (HKU, 1998) specifically to investigate the perception of landslide risk in Hong
Kong. The study investigated the perceived characteristics of fifteen hazards in Hong Kong.
Figure 2 shows the perceived riskiness of the 15 hazards from their focus group survey. The
authors compiled a chart similar to that of Stern & Fineberg (1996) in which hazards are rated
into four quadrants based on known/unknown and non-dread/dread. Cancer and nuclear
power station accidents produce the highest dread. At the other extreme, bicycles and
sunbathing are least dreaded. Cancer and active smoking rank as least known.
The three characteristics which have highest correlation (correlation factors between
0.7 and 0.9) with the riskiness ratings are personal exposure, voluntariness or risk and
exposure of Hong Kong people. Other factors with less significant correlation (correlation
factors between 0.6 and 0.7) are dread, control over risk, ease of reduction and equity.
Landslides occupy a similar position to fire, traffic accidents and construction
accidents with respect to perceived knowledge. Landslides are very close also to fire in
degree of dread (higher than traffic or construction accidents). This would seem to suggest
that, from a societal point of view, the comparison of landslide risk with risks from fire,
traffic accidents and construction accidents may be justified. Note however that
construction accidents tend mainly to affect workers, making this a partly voluntary risk
(although, as discussed earlier, social factors may force individuals to accept work which
carries risk, and whether such risk can then be regarded as voluntary is arguable). As
discussed earlier, any such comparison should also consider the clustering of the particular
hazard and any subset of the population with above (or below) average risk.
The HKU study also found differences in hazard perception between different groups
of people - age, religion, etc, may be factors. So the characteristics of a given population (or
subset) could be a factor in determining their concerns. The study also investigated the
tolerability of risk. An interesting finding was that the pilot study sample considered
landslide risks to themselves could be increased slightly and would still be tolerable, whereas
risk to Hong Kong society as a whole should be decreased. By comparison, the sample felt
that risks from traffic accidents and occupational accidents should be decreased for both
themselves and for Hong Kong society as a whole. Of course, the sample size is small, and
different results could be obtained if a subset of the population living close to a large and
apparently dangerous slope was studied.
Another interesting finding was that people thought about death, traffic jams and
accidents, transport inconvenience and loss (both economic and time) when they considered
"landslides". This suggests that measuring landslide (and other) risks simply by the number
15
of deaths may not always be adequate (see also Section 2.3).
The HKU study group felt that fire and flooding were similar to landslides in terms of
severity of consequence and the large amount of money required to prevent their occurrence.
More people considered themselves to be at risk from fire, hence this was considered riskier
than landslides. The study group preferred to see money spent on preventing a single
catastrophic accident rather than on a chronic risk producing the same number of deaths.
Their response was not entirely rational as the same group also considered the chronic risk to
be worse as more people were exposed to it.
This seems to indicate that the study group considered the prevention of one landslide
with big consequences to be preferable to the prevention of lots of small ones with less severe
consequences.
Figure 3 shows the ranking of perceived riskiness from the HKU survey as related to
an assessment of the risk clustering. This very interesting plot seems to suggest that public
perception of risk is very much related to risk clustering, i.e. the more diffUse a risk (and
therefore the more chance of it affecting the individual) the greater the public concern.
2.6 Risk "Acceptability" and Cost/Benefit Analysis
One of the outcomes of analytical (quantitative) risk assessment is the definition of
"acceptance criteria", usually based on a statistical evaluation of risk-to-life. Such criteria
tend to be phrased either as Individual Risk (IR), i.e. probability of death for an individual per
year, or Societal Risk. Societal risk is usually represented as a plot on a graph of Frequency
(of incident per year) against Number of deaths per incident. This is known as an F-N curve.
Various attempts have been made to define whether a given F-N curve represents an
"unacceptable" risk. A common refinement is the definition of a zone of intermediate risk
referred to as ALARP (a statement that risks in this zone should be reduced to a level which is
As Low As Reasonably Practicable).
The position of acceptance criteria within a decision-making process can be
controversial. Decisions made based solely on these criteria are placing very high reliance
on both the reliability of the risk analysis and the suitability of the risk indicator for the
required purpose. There is no clear consensus on how such criteria could be applied in
different situations. The issue of cumulative risk (either from multiple developments or
individuals exposed to a single hazard, or to a single individual or development exposed to
multiple risks) does not appear to be satisfactorily addressed anywhere. The following
discussion raises some of the issues involved.
The Health and Safety Executive (USE) in the UK and the NSW Planning Department
(NSWPD) discuss the possible use of criteria for land-use planning in the vicinity of major
industrial hazards, as cited by ERM(1998) and Fell & Hartford (1997).
For members of the public living close to a nuclear power station or a major hazard
facility (note the definition of the susceptible population), the HSE defines individual risk
criteria of 10"4 & 10~6 per year respectively for the upper and lower bounds of ALARP,
Above the upper bound, risk is considered to be substantial and below, negligible. Note that
- 16
these are not probabilities of death but the probability of receiving a "dangerous dose" which
would result in death. The probability of death is about 10"1 for a person receiving the
"dangerous dose".
The HSE do not recommend "formal" societal risk criteria in a decision making
process. To quote their report ''there is at present no clear consensus on criteria for societal risk, and it is not even
clear how best to describe such risk. The F-N curve is a difficult concept, and it is not
apparent how to compare two F-N curves for two different situations..."
"a societal risk is (clearly) below a criterion F-N line if the whole F-N curve is below
the line, but it is not obvious when (only) part of the curve crosses the line. "
tf
another difficulty for the present purpose is that developments are considered one at
a time, and the contribution to the total national societal risk from any one development is
very small Even at local level, the additional societal risk from a small development in a
built-up area may seem small... However, over the years the small additions to societal risks
will accumulate, and eventually it will appear that there has been a considerable increase in
the number of people at risk from major hazards."
The HSE consider societal risk in a simplified way in view of these difficulties. Their
advice is based on criteria for individual risk with societal risk allowed for by using more
stringent criteria for larger developments. The HSE consider a risk "substantial" and will
recommend against housing developments if there is individual risk of 10"5 per year for more
than 25 people receiving a dangerous dose, or if there is an individual risk of 10"6per year for
more than 75 people receiving a dangerous dose. Allowing for differences in temporal
occupancy of the different categories of development, the HSE derived equivalencies in
consideration of societal risk.
The NSWPD has derived detailed individual fatality risk criteria for various land uses,
e.g. medical facilities, schools, residential, commercial, sports, industrial, etc. For
residential development, the individual fatality risk limit is 10~6 per year. The variation of
the criteria considers different vulnerabilities of people in different types of facilities. It also
has a higher level of acceptable risk for industrial areas, reflecting the voluntary aspect of risk
acceptance criteria. They suggested a qualitative approach be used for assessing societal risk,
because of difficulties in applying F-N curves.
A corollary of calculating risk analytically is that it is theoretically possible to carry
out costfaenefit analyses for different types of mitigation. This can be done by comparing
the cost of a particular measure to the reduction in risk achieved. This can be particularly
pertinent for projects falling in the ALAKP region. Theoretically, if a cost is assigned to an
individual life, it is possible to calculate how much money could or should be spent to reduce
a risk to an acceptable level. This is another way of analysing whether a project is viable or
not. The drawbacks to this approach are the same as the drawbacks with all such analytical
techniques, i.e. the uncertainties in the analysis have to be faced and the appropriateness of
the risk indicator has to be considered. A further factor is that such a process can be
considered to be extremely "cold-blooded" and might offend those concerned. Stein &
Fineberg (1996) quote Arrow et al, (1996) as follows:
-
17 -
"Benefit-cost analysis is neither necessary nor sufficient for designing sensible public
policy. If properly done, it can be very helpful to agencies in the decision-making process
there may^ be factors other than benefits and costs that agencies will want to weigh in
decisions''. Care should be taken to assure that quantitative factors do not dominate
important qualitative factors in decision-making. "
Costfaenefit analysis can, however, help to demonstrate how effective different types
of risk mitigation measure might be, i.e. which method provides most risk mitigation for a
given sum. Again, such analyses should be carried out carefully with due consideration
given to all the factors involved.
3. COMPARATIVE RISKS
Figure 4 shows some statistics on fatalities in Hong Kong. Statistics on multiple
fatality events in Hong Kong are plotted as F-N curves in Figure 5. If we are to rank risks in
terms of simple human fatalities per unit of population, these diagrams may be acceptable.
However, as discussed above, risks are both multi-dimensional and clustered (to different
degrees) and it can be misleading to compare one calculated risk with another unless they are
similar in these respects.
As an example, Figure 4 shows that the annual fatalities from trench collapses and
landslides in Hong Kong have been, over the last ten years, very similar. However, there are
many differences between these two hazards. The population at risk from trench collapses is
orders of magnitude lower than that at risk from landslides, suggesting that the individual risk
within the affected population is much higher. This is not the only difference. It could be
argued that the potential for casualties from landslides (in the absence of government
regulatory procedures and mitigation/improvement works) is considerable, whereas that from
trench collapses is limited. The voluntary/involuntary risk factor also has to be considered.
So two hazards which, at first sight, have very similar basic casualty statistics, cannot be
treated in the same way.
Figure 6 compares some Hong Kong average annual fatality numbers with the
perceived riskiness ratings as found in the Hong Kong University survey. There is no
obvious correlation between these two measurements,, except that the riskiness rating for
hazards causing annual average fatalities of more than ten are relatively high. The HKU
report concluded that subjective risk or risk perception is not solely determined by the number
of deaths. As discussed in Section 3, perceived riskiness appears to be very much related to
the clustering of the hazard and the resulting probability that an individual will be exposed to
that hazard.
4. WAYS FORWARD
4.1 Risk Assessment Process
Stern & Fineberg (1996) define good risk characterisation as that which meets the
needs of decision participants, and draw the following conclusions*
18
Risk assessments should be accurate, balanced and informative. The authors state
get the science right and get the right science". The heavier the reliance on underlying
assumptions, the greater the need for wide participation in the risk decision process. While
various statistical and analytical techniques, and techniques for estimating and representing
uncertainty, can be used to summarise what is known about a particular risk, these are not
usually integrated with broadly-based deliberation or made user-friendly.
u
Characterising uncertainty in calculated risk assessments in a way that is both accurate
and understandable may not be technically possible. A solution to dealing with uncertainty
may lie in the process leading to a risk decision, whereby the participants are provided with
enough understanding to appreciate where scientists and engineers agree and where they do
not.
The risk characterisation process should be tailored to the needs of the decision which
is to be made. Organisations responsible for characterising risk should anticipate the valuebased judgements that are likely to become contentious and consider putting them on the
agenda for the analytic/deliberative process.
Risk characterisation is more than a synthesis of information developed by analytical
techniques. Analysis has inherent limitations in the face of the multi-dimensional and valueladen nature of many risk decisions. Analytical and deliberative processes should be
blended in a way that clarifies the concerns of the interested and affected parties.
As part of an open, iterative and broadly-based deliberative process, uncertainty
analysis should inform all the parties of what is known, what is not known, and the weight of
evidence for what is only partially understood. Describing the uncertainty does not in and of
itself represent or imply an advancement in that state; it does, however, help clarify what can
be known and perhaps help identify directions for future research and data collection efforts.
If point estimates of risk are likely to contain significant errors, then explicit evaluation
of uncertainty is needed. However, just as scientific judgements concerning point estimates
are often tenuous and susceptible to overconfidence, so too are characterisations of the
uncertainty in these estimates. Uncertainty analysis should avoid the temptation to view the
evaluation and simulation results that some techniques of uncertainty analysis generate as the
equivalent of field and laboratory studies and data. Also, formal uncertainty analysis may
not help if the uncertainty in the fundamental understanding of the basic processes that drive
the risk, or of whether the risk is even present at all, is so large that a quantitative estimate can
only lead to obfuscation.
It follows that before an appropriate risk assessment methodology can be developed, it
is necessary to define both the decisions which have to be made which could be assisted by
risk assessment, and the appropriate level and nature of consultation. An understanding of
the perceptions and needs of the stakeholders is also a prerequisite to a successful outcome.
4.2 Decisions
With respect to the types of decision within GEO's remit which could be facilitated by
an assessment of the risks involved, it is perhaps appropriate to consider these at various
19 -
levels, from the regional to the site specific.
could be as follows:
A classification of the various types of decision
(a) Regional scale: What resources should be made available to
reduce deaths, injuries and economic loss from landslides,
excavation collapses or excessive deformations, boulder
falls and foundation problems?
This is not a simple or easily-managed decision. In reality,
GEO bids for resources for individual themes or projects,
based on its own in-house estimates of where resources
should be concentrated. Within GEO, the potential exists
to create a process whereby the risks from these various
hazards are discussed, and this could be a valuable
management tool when it comes to prioritising bids for
resources. Any comprehensive discussion of these risks
would have to consider both the available numerical data
(taking account of risk clustering and uncertainties as
discussed earlier) and social factors, ideally within an
iterative and consultative framework. Some work has been
done already on the calculation of global risks from
different types of landslide hazard, but further consideration
is needed on how best to present and interpret these data.
The only risk indicators which have been considered to date
relate to the number of anticipated fatalities. This may not
be adequate to reach an informed decision with respect to
allocation of, or bids for, resources.
(b) Regional scale: How can GEO best measure
effectiveness of its policies and procedures?
the
This is an area which has received much attention in recent
years. Global risks from the failure of various categories
of man-made slopes have been calculated and a trend of
decreasing risk-to-life has been demonstrated. The risk
indicator used has invariably been annual fatality rate within
the population as a whole, which could be appropriate if the
data are restricted in use to the evaluation of trends with
time. However, it might be appropriate to consider
whether the multi-dimensional nature of landslide risk
should be accounted for when carrying out these types of
assessment, i.e. should economic factors and/or social
perceptions be integrated into the assessments of annual risk?
In addition, care must be taken if using these data to either
draw conclusions about acceptability, or to compare with
other types of risk. As discussed earlier, the nature of the
risk itself and the way in which it may be clustered can
cause misleading conclusions to be drawn.
20 -
(c) Area scale: How should resources be allocated within the
Landslip Preventive Measures programme?
The GEO uses the New Priority Classification System, or
NPCS (Wong, 1998) as a first step for prioritising follow-up
Landslip Preventive Measures action on features in the New
Slope Catalogue. The NPCS is applied primarily to preGCO man-made slopes and retaining walls. Separate
classification systems were developed for soil cut slopes,
rock cut slopes, fill slopes and retaining walls. Under each
system a Total Score is calculated for each feature,
reflecting the relative risk of a failure. The Total Score is
the product of an Instability Score and a Consequence Score.
The Instability Score is calculated from a number of key
parameters that affect the likelihood of failure. The
Consequence Score reflects the likely consequence of
failure. The higher the Total Score, the higher is the
priority for follow-up action.
The Consequence Score reflects only the potential for loss
of life in the event of a failure. It does not consider other
possible dimensions of risk, such as transport inconvenience
and economic loss (identified by HKU as additional social
concerns relating to landsliding). There is potential here
for further development.
The NPCS system provides an excellent example of how
relative risk can be assessed in a formal manner by using a
specific method designed to address a specific problem (i.e.
which slopes deserve priority?). If multi-dimensional risk
could be incorporated into such a system in a way which
reflects societal concerns as well as assessed risk-to-life, the
methodology could be very powerful., and could possibly be
applied to other types of landslide hazard ranking.
(d) Site-specific scale: Is the risk to existing development from
a given slope (or area of natural terrain) acceptable? If not,
what should be done?
Some attempts to carry out this type of analysis in Hong
Kong have met problems.
Firstly, the inherent
uncertainties must be considered when comparing results
with "acceptance criteria"; secondly, application of
"acceptance criteria" is not straightforward (for instance, if
F-N curves are used, what proportion of the F-N curve
should be below the "acceptable" line?); and finally, should
societal perceptions also be addressed? If intending to
carry our cost-benefit analysis to see what, if anything,
should be done to mitigate risk, the same problems apply.
- 21
The great uncertainties in calculated risk (and the singleparameter nature of the risk indicator) cast doubt on whether
design decisions should be made solely on this basis.
It could also be argued that a more productive way of
looking at risk to existing developments would be a
development of the NPCS approach in which various
categories of slope (and possibly natural terrain) are
prioritised for investigation using a multi-dimensional risk
index.
(e) Site-specific scale: Is the risk associated with a proposed
new development potentially affected by a man-made slope
or area of natural terrain acceptable? If not, what should
be done?
This is a variant of (d) above. In the case of an isolated
development potentially affected by a slope or area of
natural terrain it is possible to carry out an analytical risk
assessment and compare this with "acceptance" criteria.
The same problems of uncertainty and concentration on a
single risk indicator apply, but the process can have value,
particularly for those sites where calculated risks are
significantly above or below "acceptance" thresholds.
However, whether decisions as to absolute acceptability can
be made on this basis is somewhat debatable, particularly in
marginal situations or where there is high uncertainty (and
this will, in fact, cover most cases). Where a development
is proposed in an area in which there are already existing
developments (or other proposed developments) a further
difficulty arises: does one assess just the risk in the one
development, or the cumulative risk in a given area? This
question was also raised by the HSE in the UK, with no
conclusions being reached. These various problems lead to
the conclusion that the quantitative part of a risk assessment
of this nature should form part of an integrated discussion of
the risks associated with different types of site layout or
mitigation. If used in this way, analytical methods could
be useful in showing how the relative level of risk rises or
falls as different design approaches are considered.
Approaches such as this could supplement the conventional
engineering approach of providing an adequate factor of
safety against slope instability, which has its own
limitations.
22
4.3 Consultation
The discussion in Section 4.2 above refers often to an iterative and consultative
process in which the information available on risk is assessed by the parties concerned with,
or affected by, the decision. Determining who should be involved in a given decision is not
straightforward, particularly when, as is the case for GEO, there are existing statutory and
administrative procedures which must be followed. However, if consideration is to be given
to a review of who should be involved in various decisions, Stern & Fineberg (1996) offer the
following list of questions, the answers to which can assist in identifying the interested and
affected parties:
Who has information and expertise that might be helpful?
Who has been involved in similar risk situations before?
Who has wanted to be involved in similar decisions before?
Who may be affected by the risk characterisation?
Who may be affected but not know they are affected?
Who may be reasonably angered if they are not included?
5. CONCLUSIONS
Risk assessment should be an iterative and consultative process. Risk calculation can
form part of this process, but it is not necessarily the most important part. Attempts should
be made to integrate social perceptions with calculation in the risk assessment process.
Calculation of risk is subject to many uncertainties and these must always be
recognised and addressed, using sensitivity analyses where appropriate. Decisions should
not be made solely on the basis of calculated risk or cos^enefit analyses.
Risk is multi-dimensional. Limiting a risk assessment to one indicator (probability of
death) may not always be appropriate.
Social perceptions of risk do not necessarily agree with objectively assessed risk. In
Hong Kong, social perception of risk appears to be governed by a reasonably well-informed
opinion of the probability of an individual's exposure to that risk.
When calculating risk, the risk indicator(s) chosen, and the way in which it is
calculated, should be determined by the nature of the decision which is to be made.
With respect to landslide risk, different types of risk assessment and/or calculation will
be required for different types of hazard and for different types of decision.
Calculated risk which relies heavily on statistical uncertainty analysis cannot be a
substitute for data on, and understanding of, the occurrence, nature and behaviour of landslide
and boulder fall hazards.
When comparing calculated landslide risk with risk from other hazards, the nature of
the hazards and the population exposed to them must be considered.
23 -
6. REFERENCES
Arrow, K.J., Cropper, MX., Eads, G.C., Hahn, R.W., Lave, L.B., Noll. R.G., Portney, P.R.,
Russell, M., Schmalensee, R., Smith, V.K. & Stavins, R.N. (1996). Benefit-Cost
Analysis in Environmental Health, and Safety Regulation: A Statement of Principles.
Washington D.C.: American Enterprise Institute, The Annapolis Centre, and
Resources for the Future.
Baron, J., Hershey, J.C. & Kunreuther, H. (2000). Determinants of priority for risk
reduction: the role of worry. Risk Analysis. Vol. 20, No. 4, pp 413-427.
ERM-Hong Kong Ltd (1998). Landslides and Boulder Falls from Natural Terrain: Interim
Risk Guidelines. GEO Report No. 75, 183 p.
Fell, R. & Hartford, D. (1997). Landslide Risk Management. In Landslide Risk
Assessment D. Cruden & R. Fell (eds), Balkema, Rotterdam, pp 51-109.
HAP (2000). Natural Terrain Hazard and Risk Area Studiesr Mount Johnston North and
Tung Chung East? Final Report. Halcrow Asia Partnership Ltd. Prepared for the
Geotechnical Engineering Office.
HKU (1998). Public Perception and Tolerability of Landslide Risk. Main Report. Social
Sciences Research Centre, The University of Hong Kong, prepared for the
Geotechnical Engineering Office, 88 p.
Kwong, J.S.M. (2000). Study on Trench Excavation. Geotechnical Engineering Office,
Hong Kong. (Draft paper in preparation).
Morgenstern, N.R. (1995). Managing risk in geotechnical engineering. The 3rd Casagrande
Lecture. Proceedings of the 10th Pan-American Conference on Soil Mechanics and
Foundation Engineering, Guadalajara, Vol. 4, pp 102-126.
Morgenstenx, N.R. (2000). Performance in geotechnical practice. The Inaugural Lumb
Lecture. Transactions, Hong Kong Institution of Engineers, Vol. 7, No. 2, pp 1-15.
Slovic, P. (1999). Trust, emotion, sex, politics and science: surveying the risk-assessment
battlefied. Risk Analysis, Vol. 19, No. 4, pp 689-701.
Stern, P.C. & Fineberg, H.V. eds (1996). Understanding Risk - Informing Decisions in a
Democratic Society. National Academy Press, Washington D.C., 249 p.
Transport Department (annual). Road Traffic Accident Statistics (from 1983 to 1999).
Road Safety & Standards Division, Transport Department, Hong Kong.
Wong, C.K.L. (1998). The New Priority Classification Systems for Slopes and Retaining
Walls. GEO Report No. 68, Geotechnical Engineering Office, Hong Kong, 117 p.
24
LIST OF FIGURES
Figure
No.
Page
No.
1
Subj ective Assessment of Risk Clustering
25
2
Perceived Risk in Hong Kong
26
3
Public Perception of Risk Related to Risk Clustering
27
4
Statistics on Fatalities in Hong Kong (1972-1999)
28
5
Multiple Fatality Events in Hong Kong
29
6
Annual Fatalities and Perceived Riskiness
30
25
Diffused Risk
Geographical Space/Population
-^Clustered
Diffused —
Diffused
Construction Accidents
Fire Accidents
Household Accidents
Traffic Accidents
Man-made Slope
Failures
Boulder Falls from
Natural Terrain
Clustered
Clustered Risk
Legend:
PHI
CDF
Potential hazardous installation
Channelised debris flows
Figure 1 - Subjective Assessment of Risk Clustering
I
•
g
e
:.
I
t
^n
3
K>
i
hd
o
2.
fD
PC/3
Fire
\.
::
,
::
»
.>>
-
C
I
S gII
r
Cancer
Motor Vehicles
C/3
c/a
1
c
f.
o
£ 3
Landslides
Construction Accidents
X
?
I
Nuclear Electric Power
/
Handguns
3
Unhygenic Foods
i.,
:
??
Commerical Aviation
I
Mean Rating of Riskiness
0 = No risk at all
Non-nuclear Electric Power
Pesticides
Bicycles
Sharks
Sunbathing
Active Smoking
27 -
Diffused Risk
Geographical Space/Population
Diffused Diffused
-> Clustered
Construction Accidents
Fire Acci dents
Household Aiccidents
Traffic Accidents
D
Man-made Slope
Failures
Oper Hillside
Lands lides and Nuclear rower
Boulder Falls from
Cation
Natural Terrain
Clustered
Clustered Risk
Legend:
PHI
CDF
Potential hazardous installation
Channelised debris flows
Ranking in perceived riskiness (after HKU, 1998)
Figure 3 - Public Perception of Risk Related to Risk Clustering
28
100000
10000
.2
;tt
-=
1000
a
—
0
i
100
LO
1970
1975
1980
1985
Year
1990
1995
2000
Legend:
Total number of deaths(1)
Road traffic accidents(2)
Construction site accidents(1)
Landslides (other than boulder falls)
Trench collapse(5)
Notes:
Cancer(1)
Fire(3)
Lightning(4)
Boulder falls
(1) Figures are obtained from Hong Kong Annual Digest of Statistics - 1999
Edition on the internet web site of the Census and Statistics Department.
(2) Figures are obtained from Road Traffic Accident Statistics (1983-1999) by
Transport Department.
(3) Figures are obtained from Fire Services Department direct.
(4) Figures are obtained from Census and Statistics Department direct.
(5) Figures are obtained from Kwong (2000).
Figure 4 - Statistics on Fatalities in Hong Kong (1972-1999)
1000
I
~
0.1
z—
0
0.01
—e
0.001
0.0001
10
Number of Fatalities (N)
100
Legend:
Statistics - Traffic accidents 1983-1999(1)
Statistics - Fire accidents 1986-1999(1)
Statistics - Historical landslide incidents 1917-1995(2)
Statistics - Historical boulder falls 1926-1995(2)
Analytical - Failures of pre-GCO man-made slopes in 1977(3)
Analytical - Failures of pre-GCO man-made slopes in 2000(3)
Analytical - Natural terrain landslides 1994 (base line case)(4)
Analytical - Natural terrain landslides 1994 (base line + 20 m case)(4)
Notes:
(1)
(2)
(3)
(4)
(5)
Sources of data are same as those in Figure 4.
Data are extracted from ERM (1998).
Data are extracted from Cheung & Shiu (2000).
Data are extracted from Sun & Evans (1999).
The analytical F-N curves presented above were derived to study relative changes in risk
with time, and do not necessarily reflect actual fatalities.
Figure 5 - Multiple Fatality Events in Hong Kong
30
100
90
Ig
c/3
«<D
80
m
^ i
1|
|J
S, a
U
1
£
•
I!
(-! .og
o
O
•
8^
S.I«
g>
•
•
*0
• 1
70
^r
0>
|
1 60
CW
CO
.8
"^3
l~i
50
i
1I 40
!
:
n-q
^
3
0
03
O
•
3
^
d>
0
W
i-l d.
Nuclear electric•
powei
Unhygenic foods
•
Handguns
PH
Commerical aviation
30
P^cti<~iH/-c
Conventional electric X)wer
Sharks
20
Bicycles
Sunbathing
10
^
0
]
10
100
1000
10000
100000
Average Annual Fatalities
Legend:
To me
To a random person in HK
Notes:
(1) Sources of fatalily data are same as those in Figure 4.
(2) Ranges of annual fatalily data for caculations of average annual fatalities
are same as those shown in Figure 4.
(3) Perceived riskiness ratings are obtained from HKU (1998).
Figure 6 - Annual Fatalities and Perceived Riskiness
31
SECTION 2:
POSSIBLE CLIMATE
CHANGE TRENDS IN
HONG KONG AND
IMPLICATIONS FOR THE
SLOPE SAFETY SYSTEM
N.C. Evans
This report was originally produced in September 2001
as GEO Technical Note No. TN 5/2001
32 -
FOREWORD
This Report discusses possible climate change trends in
Hong Kong, with particular reference to rainfall, and examines
what the implications might be in terms of landslide risk and
slope safety. The conclusions drawn are based on analyses of
Hong Kong climatological parameters since the early 1950s,, and
the current scientific consensus on future small-scale and largescale climate change.
The climatological data were supplied by Dr W.L. Chang
of the Hong Kong Observatory. Dr Chang also provided many
valuable comments on data analysis and the urban heat-island
phenomenon. His assistance is gratefully acknowledged. The
rainfall analyses were carried out by Mr N.C Evans of Special
Projects Division, who also wrote the Report.
P.L.R. Pang
Chief Geotechnical Engineer/Special Projects
33
ABSTRACT
The Slope Safety Technical Review Board recommended in 2000 that GEO examine
rainfall data to see whether any possible climate change related trends were visible and, if so,
what the implications might be in terms of landslide risk and slope safety.
The current scientific consensus on climate change is that "global warming" is a real
phenomenon, and that noticeable regional climate changes in the next 50 to 100 years are
probable. In addition, there is an increasing acceptance that development often results in
local climate change.
Analysis of 50 years of annual rainfall data from Hong Kong shows that average
annual rainfall has increased slightly over this period. However, this increase has not been
uniform, and has been concentrated in the central (developed) parts of Hong Kong,
particularly the northern parts of Hong Kong Island, Kwai Chung, Tsing Yi, Shatin, Tai Po
and Tsuen Wan, where annual increases of several hundred millimetres appear to have
occurred.
It seems probable that these changes are related to the development of an 'urban heatisland'. The fifty-year trend in other climatological parameters appears to support this
hypothesis.
These apparently local changes might continue together with Hong Kong's on-going
development. In addition, it is possible that global climate change in the next 50 to 100
years might be significant enough to have a local impact.
Past records should not be automatically assumed to be a sound basis for predicting
future events. It would be prudent to review regularly those rainfall parameters used for
slope stability analysis, slope design, landslide risk assessment and landslide emergency
preparedness planning.
34
CONTENTS
Page
No.
Title Page
31
FOREWORD
32
ABSTRACT
33
CONTENTS
34
1.
INTRODUCTION
35
2.
GLOBAL CLIMATE CHANGE
35
2.1
Intergovernmental Panel on Climate Change (IPCC)
35
2.2
Sources of Global Climate Change
35
2.3
Consensus on Recent Global Climate Change
36
2.4
Consensus on Future Global Climate Change
36
3.
ANALYSIS OF LOCAL CLIMATE RECORDS
37
3.1
Rainfall
37
3.2
Possible Urban Heat Island Effects
38
4.
CONCLUSIONS
39
5.
RECOMMENDATIONS
40
6.
REFERENCES
40
LIST OF FIGURES
41
- 35
1. INTRODUCTION
One of the recommendations coming out of the 2000 SSTRB visit was that GEO
should examine rainfall records to see whether any climate-change related trends were visible
in Hong Kong, and whether there were implications for landslide risk indicators. This
Report summarises the work carried out to date, and reaches some initial conclusions.
Background information is presented first to allow the Hong Kong data to be viewed in
context.
2. GLOBAL CLIMATE CHANGE
2.1 Intergovernmental Panel on Climate Change (IPCC)
The Intergovernmental Panel on Climate Change (IPCC) is a United Nations body
which brings together hundreds of scientists and researchers from many countries. The
IPCC has published a number of important documents over the last decade, and has recently
released a new series of reports entitled "Climate Change 2001" (IPCC, 2001a & b).
Although not all scientists agree with the IPCC, the reports it produces are the closest
documents available to a consensus from the world scientific community. The IPCC reports
can be viewed at and downloaded from the IPCC web site (www.ipcc.ch/).
The IPCC uses the term 'climate change' to refer to any change in climate over time,
whether due to natural variability or as a result of human activity. In the past the term has
sometimes been used to refer solely to human-induced changes, and this has caused some
confusion. The main conclusions of the latest IPCC reports are summarised below.
2.2 Sources of Global Climate Change
Changes in climate occur as a result of both internal variability within the climate
system and external factors (both natural and anthropogenic). The influence of external
factors can be broadly compared using the concept of radiative forcing, which is a measure of
the influence of the factor on the balance of incoming and outgoing energy in the Earthatmosphere system. A positive radiative forcing, such as that produced by increasing
concentrations of greenhouse gases, tends to warm the surface. A negative radiative forcing,
which can arise from an increase in some types of aerosols (microscopic airborne particles)
tends to cool the surface.
Positive forcing (warming) factors which are assessed to have affected the climate
since AD 1750 include increasing concentrations of the "greenhouse" gases, increases in
tropospheric ozone, aviation-induced contrails (very minor) and an increase in solar radiation
(natural). Negative forcing (cooling) factors over the same period include increases in
stratospheric ozone, increases in aerosols (sulphates, organic carbon and smoke from biomass
burning) and changes in reflectance (albedo) due to changes in land-use. Temporary
negative forcing (lasting only a few years) results from episodic explosive volcanic activity.
The combined change in radiative forcing of the two major natural factors (solar
variation and volcanic aerosols) is estimated to have been negative (i.e. a net cooling effect)
for the past two and possibly the past four decades.
36 -
2.3 Consensus on Recent Global Climate Change
•
The global-average surface temperature increased by about 0.6°C during the 20th
century (discounting the effects of localised temperature rises due to urbanisation the so-called "urban heat island" phenomena, discussed later). This wanning has
been unusual and is unlikely to be entirely natural in origin.
It seems likely that the 1990s was the wannest decade, and 1998 the warmest year,
in the global instrumental record (since 1861). It also appears likely that the
temperature increase in the 20th century has been the largest of any century in the
past 1,000 years.
•
Global sea-level rose between 0.1 and 0.2 in during the 20th century.
•
Natural forcing may have contributed to the observed warming in the first half of
the 20th century, but cannot account for the warming observed in the second half.
Most of the observed wanning over the last 50 years is likely to have been due to
the increase in greenhouse gas concentration.
2.4 Consensus on Future Global Climate Change
*
The IPCC has examined 35 different scenarios for future emissions and has run a
number of different climate models. The following conclusions relate to the main
predicted global effects and those that are potentially of particular significance to
Hong Kong.
*
Global average surface temperature is projected to increase by between 1.4 and
5.8°C between 1990 and 2100. This rate of warming is probably without
precedent in the last 10,000 years.
*
Warming in the northern regions of North America, and north and central Asia, is
likely to exceed the global average by more than 40%. By contrast, in south and
southeast Asia the warming is expected to be less than the global average.
Global mean sea level is predicted to rise by between 0.09 and 0.88 m between
1990 and 2100.
*
On a global scale, more intense precipitation events are very likely over many
areas, particularly the mid to high latitudes of the northern hemisphere. It cannot
be assumed that future hydrological regimes will be the same as those in the past.
However, precipitation is very variable in both time and space, and the main
characteristic of the long-term global total record is its marked year-to-year
variability. If smaller regions are examined, the year-to-year variability becomes
even more pronounced.
*
It is likely that Asian summer monsoon precipitation variability will increase,
although confidence in this projection is limited by modelling difficulties.
Modelling results appear to suggest an increase in runoff in the Hong Kong and
-
37
Guangdong area, but models are coarse and the scale of any such increase is
uncertain. It is not clear whether any increases in annual rainfall would result
from larger or more intense rainstorms, or from a greater number of events, or both.
Prediction of how climate change might affect the frequency, intensity or tracks of
tropical storms is highly uncertain. Current Global Climate Models do not have
sufficiently fine spatial resolution to simulate individual storms. Some data
suggest that minimum pressures may decrease and windspeeds may increase in
tropical storms worldwide, although the projected changes are small compared to
past inter-annual variability. With increased ocean temperatures it is almost
certain that the moisture-holding capacity of the atmosphere will rise, and this
might mean precipitation increases in areas frequented by tropical storms. This
could also be relevant to Hong Kong.
A final conclusion, rather depressing if of somewhat more academic interest, is
that global mean surface temperature and sea level will continue to rise for
hundreds of years after the levels of greenhouse gas in the atmosphere are
stabilised.
3. ANALYSIS OF LOCAL CLIMATE RECORDS
3.1 Rainfall
Annual rainfall isohyets for Hong Kong covering the 49-year period from 1952 to
2000 inclusive have been obtained from the HKO. These maps were compiled using all data
available at the time and the judgement of the HKO meteorologists. The area covered
comprises the entire land area of Hong Kong, and most of the sea area, with the exception of
the westernmost part of Lantau, and the Sharp Peak/Crooked Harbour area in eastern Sai
Kung (data not available until the 1960s). The data were digitised using the Lands
Department survey grid of approximately 1.2 by 1.4 km rectangles. The data covering the
grid square in which the HKO is situated were checked against the actual annual figures at the
HKO. Agreement was good, suggesting that the digitisation was accurate.
The average rainfall across Hong Kong in each year was calculated by summing the
annual rainfall in all the cells and dividing by the total number of cells (approximately 1,150).
Overall, the average rainfall over Hong Kong appears to have increased slightly during this
49-year period (see Figure 1).
When the data are examined cell by cell a more complex picture emerges.
Regression and moving-mean analyses carried out for the 49-year period for each cell show
that annual rainfall appears to have increased significantly in the central parts of Hong Kong,
with increases concentrated on Hong Kong Island, Kowloon, Kwai Chung, Tai Mo Shan, Tai
Wai/Shatin and Tai Po. Secondary peaks are seen over SE Hong Kong Island and Po Toi,
and in the Tuen Mun area. Rainfall elsewhere appears to have either stayed about the same,
or decreased. The precise scale of the changes is a little uncertain - neither regression nor
moving-mean analyses give a clear picture. The regression analyses suggest a maximum
rise of 700 mm, while the moving mean analyses indicate that this figure might be closer to
400 to 500 mm. The changes deduced from the moving means tend to be perhaps 60-70% of
those indicated by the regressions. * However, both analyses agree on the direction of change
38 -
in a given grid square, so there is some confidence in the geographical spread (see Figure 2).
Figures 3 and 4 respectively illustrate the situation in grid squares showing a large increase
and negligible change.
It must be emphasised that the apparent trends shown in these plots are small
compared to the inter-annual variability. Conventional statistical analyses of these data
show very high scatter, and measures of significance are not high. This is not unusual when
dealing with trends over time in natural systems. The inter-annual variability of
precipitation is notoriously large. However, the geographical spread of trend values
(Figure 2) does not appear to be random, and might indicate that an underlying mechanism is
at work.
3.2 Possible Urban Heat Island Effects
Hong Kong, and the Pearl River Delta region as a whole, has seen tremendous growth
in the built environment over the last twenty to thirty years. It is widely recognised that
urbanisation and changes in land use can have a marked effect on local climate. This is not a
global warming issue, it is more a question of urban areas generating their own microclimates
through additional heating. The phenomenon is known as an "urban heat island". The
following processes are relevant.
*
Cooling from the evaporation of soil and vegetation water does not occur in
urbanised areas. Instead, buildings and roads absorb and radiate solar energy.
*
Waste heat from buildings and transport can contribute as much as one third of the
heat received from solar energy.
*
Buildings conduct heat much more efficiently than rural vegetation, and the
canyon structures created by tall buildings enhance warming by trapping solar
energy via multiple reflections and reducing infrared heat losses by absorption.
The urban heat island may increase cloudiness and precipitation in the city, as thermal
circulation sets up with the surrounding region. Strong winds will tend to reduce this effect,
so it is not a phenomenon that could be expected to have much impact during a typhoon.
However, the heavy rain events which result from intense convective activity associated with
low pressure troughs or unstable southerly airflows might perhaps be exacerbated. The
effects of triggering or increasing convection in these airflows might not be confined directly
to the built-up areas. It is possible to envisage situations in which enhanced convection
could move "downstream", or, conversely, could trigger additional convection "upstream".
If Figure 2 is viewed with this in mind, and bearing in mind the dominant southerly
airflow during much of the wet season, it is possible to see a consistent pattern in the annual
rainfall trends which might perhaps be connected to the generation of a heat island over the
major urban and industrial areas. Triggering of convection upstream and downstream from
these areas in the dominant southerly airflow could perhaps account for the apparent rainfall
increases observed over Tai Mo Shan and SE Hong Kong Island/Po Toi.
The HKO have provided average daily temperature data from the HKO site in
39 -
Kowloon and from Waglan Island (approximately 5 km south-east of Cape D'Aguilar of
Hong Kong Island, and remote from development) for the 48-year period from 1953 to 2000
inclusive. These data (average daily minima, maxima and means) also appear to reveal
some long-term trends.
Daily minima (lowest night-time temperatures) at the HKO appear to have risen
significantly during the period, when compared with the data from Waglan (see Figure 5).
This would seem to support the theory that urban Hong Kong has developed a heat island.
The average daily maxima (maximum day-time temperatures) from the HKO do not show this
trend - in fact they show a significant relative decrease when compared to the Waglan data
(which show a rise during the period - see Figure 6). This might suggest increasing relative
cloudiness over the urban area, which again is a possible symptom of the development of a
heat island. The possible implications of the measured rise in daily maxima at Waglan are
not clear - this could reflect a genuine regional trend. The average daily mean temperatures
also show an increasing trend for the HKO as compared to Waglan (see Figure 7).
The above data are fairly crude and are limited in areal coverage, but they do appear to
indicate that there have been measurable changes in relative temperature distribution over the
last 50 years, and these are consistent with the generation of an urban heat island.
In addition to temperature data, the HKO have also provided plots of total annual
sunshine duration and solar radiation at Kings Park (in the urban area of Tsimshatsui) from
the 1950s. These parameters (Figures 8 and 9) show a fairly definite drop in values over the
last 40-50 years. This is again consistent with the generation of an urban heat island
(decreasing sunshine hours and solar radiation suggesting increasing cloudiness).
4. CONCLUSIONS
It appears that there may have been localised increases in annual rainfall of up to
400-500 mm over the last 50 years. These apparent increases are concentrated in the central
parts of Hong Kong. It is not possible to demonstrate whether or not the increases are
statistically significant due to the large inter-annual variability in rainfall. This is a common
problem worldwide in the analysis of precipitation trends. These changes (if genuine) are
very unlikely to be a direct result of global climate change (although this might be having an
underlying, subsidiary, effect).
The apparent increase in annual rainfall in the central parts of Hong Kong might be
due to a "heat-island" effect. Other parameters (temperature, solar radiation and sunshine
hours) display trends over the last 40-50 years which would seem to confirm that urban Hong
Kong has indeed developed a heat-island. Whether the apparent rainfall trend will continue
is a matter of conjecture. However, it would seem reasonable to assume that areas where
development is continuing (or, possibly, where there are going to be significant changes in
existing development) might experience changes in annual rainfall. Given the possible
contribution to heat-island generation of solar reflection and radiation from bare concrete
surfaces, GEO's slope-greening initiatives should be viewed positively.
Global warming impacts in the next 50 to 100 years might be significant. Increases
in both total rainfall, number of rainstorms and rainfall intensity are possible in Hong Kong,
40
-
but there are major uncertainties related to the magnitude of any such increases.
5. RECOMMENDATIONS
Recent work (Evans & Yu, 2000) has shown significant variations in intense rainfall
parameters across Hong Kong, using records that date back to the mid-1980s when the
automatic raingauge system was first commissioned. It would be advisable to periodically
re-examine these data, as changes in rainfall intensity and/or location resulting from local
and/or global climate change will affect the calculated results. However, these statistical
evaluations are representative only of the period covered by the data, and cannot account for
possibly continuing trends. It should not be assumed that past rainfall records are
necessarily a sound basis for predicting future extreme events. It might therefore also be
prudent to regularly re-examine the rainfall return periods used in landslide risk assessments,
slope design/stability analysis and landslide emergency preparedness planning, as an
allowance for uncertainty can be introduced in this way.
6. REFERENCES
Evans, N.C. & Yu, Y.F. (2000). Regional Variation in Extreme Rainfall Values. Technical
Note TN 5/2000, Geotechnical Engineering .Office, Hong Kong SAR Government,
25 p.
IPCC (2001 a). Climate Change 2001: Impacts, Adaptation and Vulnerability. Technical
Summary. A Report of Working Group II of the Intergovernmental Panel on Climate
Change, 73 p.
IPCC (2001b). Climate Change 2001: Impacts. Adaptation and Vulnerability. Summary
for Policymakers. A Report of Working Group II of the Intergovernmental Panel on
Climate Change, 17 p.
- 41
LIST OF FIGURES
Figure
No.
Page
No.
1
Areally-averaged Annual Rainfall, 1952-2000
42
2
Possible Change in Annual Rainfall, 1952-2000
43
3
Example of Large Apparent Rise in Annual Rainfall
(HNW~6,Kwai Chung)
44
4
Example of Negligible Apparent Change in Annual
Rainfall (10SW-10, Peng Chau)
45
5
Average Daily Minimum Temperatures, HKO and
Waglan
46
6
Average Daily Maximum Temperatures, HKO and
Waglan
47
7
Average Daily Mean Temperatures, HKO and Waglan
48
8
Sunshine Duration, Kings Park, 1947-2000
49
9
Solar Radiation, Kings Park, 1957-2000
50
3000
to
500
1950
1955
1995
Legend;
Linear (annual rainfall)
15-year moving average (annual rainfall)
Figure 1 - Areally-averaged Annual Rainfall, 1952-2000
2000
Values in mm derived from regression analyses - moving means indicate values approximately 60-70% of these.
Figure 2 - Possible Change in Annual Rainfall, 1952-2000
4000
3500
500
1950
1995
1955
Legend:
Linear (annual rainfall)
15-year moving average (annual rainfall)
Figure 3 - Example of Large Apparent Rise in Annual Rainfall (11NW-6, Kwai Chung)
2000
3000
2500-
2000-
1500-
1000-
500
1950
1955
1995
Legend:
Linear (annual rainfall)
15-year moving average (annual rainfall)
Figure 4 - Example of Negligible Apparent Change in Annual Rainfall (10SW-10, Peng Chau)
2000
22.5
19
1950
1995
1955
Legend:
HKO
5-year moving average (HKO)
Waglan
5-year moving average (Waglan)
Figure 5 - Average Daily Minimum Temperatures, HKO & Waglan
2000
27
26.5 26P
25.5
<D
i-i
25
o
1 24.5 H
24-
23.523
1950
1955
1960
1965
1970
1975
Year
1980
1985
1990
Legend:
HKO
5-year moving average (HKO)
Waglan
5-year moving average (Waglan)
Figure 6 - Average Daily Maximum Temperatures, HKO & Waglan
1995
2000
24.5
00
21
1950
1995
1955
Legend:
4
HKO
5-year moving average (HKO)
Waglan
5-year moving average (Waglan)
Figure 7 - Average Daily Mean Temperatures, HKO & Waglan
2000
2500
1500
Year
Figure 8 - Sunshine Duration, Kings Park, 1947-2000
20
18-
16-
cd
rt
a
u\
14
o
3
GO
12-
__r_
10
rH
ON
*o
ON
CO
*o
ON
ON
ON
ON
rH
cn
ON
\o
ON
VO
ON
ON
^H
ON
ON
ON
ON
ON
ON
l>
ON
r-H
00
ON
co
00
ON
Year
Figure 9 - Solar Radiation, Kings Park, 1957-2000
00
ON
00
ON
ON
00
ON
ON
ON
CO
ON
ON
ON
ON
r-
ON
ON
ON
ON
ON
GEO PUBLICATIONS AND ORDERING INFORMATION
A selected list of major GEO publications is given in the next
page. An up-to-date full list of GEO publications can be found at
the CED Website http://www.info.gov.hk/ced/ on the Internet
under "Publications". Abstracts for selected documents can also
be found at the same website. Technical Guidance Notes are
published on the CED Website from time to time to provide
updates to GEO publications prior to their next revision.
http://www.iifo.gov.hk/ced/ •
•
•
Copies of GEO publications (except Sheet Reports, 1:5000
maps and other reports which are free of charge) may be
ordered either:
by writing to
Publications Sales Section,
Information Services Department,
Room 402,4th Floor, Murray Building,
Garden Road, Central, Hong Kong.
Fax: (852) 2598 7482
: (852) 2598 7482
Information Services Department's Websites
http://www.info.gov.hk/isd/puborder/order-e.htm
http:^ookstore.esdlife.com/eng/default.asp
The Information Services Department will issue an invoice upon
receipt of an order.
In Hong Kong, publications may be directly purchased from:
Government Publications Centre,
Ground Floor, Low Block,
Queensway Government Offices,
66 Queensway, Hong Kong.
Tel: (852) 2537 1910 / (852) 2537 1914
Fax:(852)25237195
http://www.info.gov.hk/isd/puborder/order-e.htm
http://bookstore.esd3ife.com/chi/default.asp
(852) 2537 1910 / (852) 2537 1914
(852) 2523 7 195
1:5 000 maps may be purchased from:
Map Publications Centre/HK,
Survey & Mapping Office, Lands Department,
23th Floor, North Point Government Offices,
333 Java Road, North Point, Hong Kong.
Tel: 22313187
Fax:(852)21160774
; 2231 3187
•(852)21160774
Requests for copies of reports which are free of charge and the
full list of GEO publications should be sent to:
For Geological Survey Sheet Reports:
Chief Geotechnical Engineer/Planning,
(Attn: Hong Kong Geological Survey Section)
Geotechnical Engineering Office,
Civil Engineering Department,
Civil Engineering Building,
101 Princess Margaret Road,
Homantin, Kowloon, Hong Kong.
Fax:(852)27140247
E-mail: sgegsjpln@ced,gov.hk
(852) 2714 0247
h [email protected]
For other free publications:
Chief Geotechnical Engineer/Special Projects,
Geotechnical Engineering Office,
Civil Engineering Department,
Civil Engineering Building,
101 Princess Margaret Road,
Homantin, Kowloon, Hong Kong.
Fax: (852) 2714 0275
E-mail: [email protected],hk
(852) 2714 0275
acospgr_spd@ced,gov.hk
'
:
,
V'i tf S ^ ^ / o
1 2 MAR 2003
MAJOR GEOTECHNICAL ENGINEERING OFFICE PUBLICATIONS
GEOTECHNICAL MANUALS
Geotechnical Manual for Slopes, 2nd Edition (1984), 300 p. (English Version), (Reprinted, 2000).
Highway Slope Manual (2000), 1 14 p.
GEOGUIDES
Geoguide 1
Guide to Retaining Wall Design, 2nd Edition (1993), 258 p. (Reprinted, 2000).
Geoguide 2
Guide to Site Investigation (1987), 359 p. (Reprinted, 2000).
Geoguide 3
Guide to Rock and Soil Descriptions (1988), 186 p. (Reprinted, 2000).
Geoguide 4
Guide to Cavern Engineering (1992), 148 p. (Reprinted, 1998).
Geoguide 5
Guide to Slope Maintenance, 2nd Edition (1998), 91 p, (English Version), (Reprinted, 1999).
• H-JEC1998) » 89
GEOSPECS
Geospec 1
Model Specification for Prestressed Ground Anchors, 2nd Edition (1989), 164 p. (Reprinted,
1997).
Geospec 2
Model Specification for Reinforced Fill Structures (1989), 135 p. (Reprinted, 1997).
Geospec 3
Model Specification for Soil Testing (2001), 340 p.
GEO PUBLICATIONS
GCO Publication
No. 1/90
Review of Design Methods for Excayations (1990), 187 p. (Reprinted, 2000).
GEO Publication
No. 1/93
Review of Granular and Geotextile Filters (1993), 141 p.
GEO Publication
No. 1/96
Pile Design and Construction (1996), 348 p. (Reprinted, 2001).
GEO Publication
No. 1/2000
Technical Guidelines on Landscape Treatment and Bio-engineering for Man-made Slopes and
Retaining Walls (2000), 146 p.
GEOLOGICAL PUBLICATIONS
The Quaternary Geology of Hong Kong, by J.A. Fyfe, R. Shaw, S.D.G. Campbell, K.W. Lai & P.A. Kirk (2000),
2 10 p. plus 6 maps.
The Pre-Quatemary Geology of Hong Kong, by RJ. Sewell, S.D.G. Campbell, CJ.N. Fletcher, K.W. Lai & P.A.
Kirk (2000), 181 p. plus 4 maps.
TECHNICAL GUIDANCE NOTES
TON 1
Technical Guidance Documents
TOG
X1SDS7317
HKP 624.151363 S95 c
Sun, H. W.
Comparative risk indicators &
possible climate change trends
in Hong Kong and implications
for the slope safety system
Date Due