Transcript WHY THE MULTI-RISK?

A MULTI RISK ASSESSMENT OF DISASTERS
RELATED TO CLIMATE CHANGES
Paolo Gasparini1
Warner Marzocchi2
Amra Scarl, Napoli
1 Dipt. Di Scienze Fisiche, Università di Napoli Federico II
2 Istituto Nazionale di Geofisica e Vulcanologia, Roma
How to focus risk mitigation
policies?
WHY THE MULTI-RISK?
FLASH FLOODS
ANTHROPOGENIC
SOURCES
What is the most dangerous
hazard for my city??
…mmm…
I don’t really know…
EARTHQUAKES
(ground shaking)
RAPID MASS MOVEMENTS
SEA LEVEL RISE
ERUPTIONS
(tephra fall, pyroclastic flows, …)
INTRODUCING MULTI-RISK…
How to focus risk mitigation policies?
What is the most dangerous hazard
for my city??
WHAT IS NEEDED TO?
1. quantitative risk assessment (probability)
needed for decision makers
1. ranking of risks
2. interaction among risks
WHAT DO WE HAVE NOW?
…mmm…
I don’t really know…
risks are considered independently, through
inhomogeneous procedures…
…they are not comparable!!!
from CLASSICAL RISK APPROACH…
starting from the ADVERSE EVENT
RISKS ARE TREATED SEPARATELY
Different approaches to Hazard:
- Geological hazard can be considered constant with time
- Hazard affected by climate change are not constant with time.
Different Time scales
Different Criteria of damage assessment
Specific vs. systemic vulnerability
Different Spatial definition
RISKS ARE NOT COMPARABLE!!!
ILLNESS
HUNGER
CLIMATIC CHANGE
TEMPERATURE
WIND
PRECIPITATION
HAZARD
PROBABILITY
SCENARIOS
RURAL VULNERABILITY
REFUGEES
MULTI RISK ASSESSMENT
URBAN VULNERABILITY
PEOPLE
COPING CAPACITY
RESILIENCE
PLACES
•INDIVIDUAL LEVEL
•COMMUNITY LEVEL
•GOVERNMENT LEVEL
THINGS
RAINFALLS
VULNERABILITY OF
URBAN AREAS
URBAN
CATCHMENTS
DISCHARGES
HYDRAULIC ROUTINE
HYDROLOGICAL ROUTINE
CLIMATIC CHANGE
SEWER NETWORK
URBAN FLASH
FLOODS
STORAGE FACILITIES
REAL TIME CONTROL
INNOVATIVE LAND USE
STRUCTURAL AND
SOCIAL DAMAGES
STRUCTURAL AND
NON STRUCTURAL
MITIGATION OPTIONS
MULTI-RISK: assessment of the potential damages caused by all the events threatening
an object (industry, city, environment, etc.).
Usually, multi-risk assessment is provided as the “sum” of independent single risk
assessment, but:
1) Single risk assessments are not always liable to be summed (i.e., different spatial
and temporal resolution, different approaches to vulnerability);
2) Risks are NOT independent: the hazard and vulnerability of one specific event may
change significantly if another event occurred. (INTERACTION AND CASCADE
EVENTS).
THIS MAY LEAD TO SEVERE UNDERESTIMATION OF THE REAL RISK.
…to MULTI-RISK APPROACH
starting from the TARGET AREA
Better consistency using DAMAGE-from-SOURCE.
RISKS TREATED COHERENTLY
Comparable Time scales
Same Type of damage
Comparable Spatial definitions
Comparable Approaches to evaluate hazard
Interaction and cascade effects easier to be accounted for
RISKS ARE COMPARABLE!!!
Why Bayesian Methods?
 The sources of Risk are aleatoric events;
 The imperfect knowledge of the processes/parameters
introduces epistemic uncertainties;
Bayesian approach allows us to take into account both aleatory
and epistemic uncertainties;
Bayesian approach allows us to merge different types of
information, such as theories, model output, data, and so on.
Why Bayesian Methods?
The Bayesian approach is particularly useful in practical problems
characterized by few data and scarce theoretical knowledge.
The Bayesian approach implies that the probability is not a single
value but it is a probability distribution.
The probability distribution has an average (the best guess of the
probability) and a standard deviation.
These two parameters estimates the aleatoric and epistemic
uncertainties.
Accounting for epistemic and aleatory
uncertainty
Each probability may be represented by a
single value or, more appropriately,
by a distribution whose central value
represent the “best guess”, and the
spreading mimics the epistemic
Bayes
theorem
uncertainty
on the
best guess.
Likelihood
(e.g.. DATA)
+
Prior
(e.g. given by models)
POSTERIOR PDF
(no epistemic uncertainty)
Risks are NOT independent: the hazard and vulnerability of one specific event
may change significantly if another event occurred.
Example: Risk for one event E1 that depends on a second one E2
R1  p(E1 | E 2 ) p(E 2 )  p(E1 | E 2 ) p(E 2 ) 


p[Ck1 | (E1, E 2 )]p(E 2 )  p[Ck1 | (E1, E 2 )] p(E 2 )Lk 
 k

 The yellow box is the hazard. The blue box is the damage.
Risks are NOT independent: the hazard and vulnerability of one specific event may
change significantly if another event occurred.
Let us consider only one hazard (due to the event E1 depending on the event E2)
H1  p(E1 | E 2 ) p(E 2 )  p(E1 | E 2 )p(E 2 )
- Usually, long-term H1 is determined by databases. If p(E2) is not changed across the
time covered by the database (i.e., the boundary conditions are the same), the
database provides directly an unbiased estimation of H1.

- If p(E2) varies with time (e.g., global warming), the database provides a biased
estimation. In this case, we need to estimate p(E2), p(E1 | E2) and p(E1| NOT E2).
- In the short-term hazard assessment, we may be interested in estimating p(E1 | E2)
instead of H1, because we know that E2 is already occurred (cascade effects).
Risks are NOT independent: the hazard and vulnerability of one specific event may change
significantly if another event occurred.
Let us consider one hazard (E1) due to the occurrence of intensive rainfall (E2; here for
simplicity E2 is dichotomic: 0 – no intensive rainfall; 1 – intensive rainfall, e.g. rainfall over a
given threshold):
H1  p(E1 | E 2 ) p(E 2 )  p(E1 | E 2 )p(E 2 )
- if no heavy rainfall occurred in the past, from the database we can estimate a biased value
of H1 that is given by p(E1 | NOT E2) (being p(NOT E2)=1). Then, p(E2) is the probability to
have a rainfall
 over the given threshold. p(E1 | E2) is the probability that we can estimate
from a scenario: the probability to have E1 given a rainfall over the given threshold
(INTERACTION).
Naples case Annual risks for human life:
•R seis = 0.0017
•R vulc = 1.37
•R flood = 4.2 10-5
•R land = 6 10-7
•R ind = 1.83 10-6 < IR < 1.83 10-8
•R env = 0.0125
Multi risk annual probabilities
Industrial accident (Toxic emission): 3.6 x 10-3
NAPLES CASE: SCENARIOS OF IN TOWN LANSLIDE TRIGGERED BY INTENSIVE PRECIPITATION
Heavy rainfall
Over threshold
No landslide
Slow landslide
Below threshold
Fast landslide
Failure of infrastructures
Loss of containment
No failure of
infrastructure
No loss of containment
GPL
Toxic release
Fire
NW
People
(residents,
workers,..)
W
Clone
SW
Clone
…
Air, soil, subsoil,
superficial water,
groundwater
Explosion
Cuenca Project (supporting agencies: BID, ETAPA)
Heavy rainfall
Over the threshold
Below the threshold
Fast landslides
Flash floods
Rio Yanuncay, Cuenca, Ecuador
Damage to tanks of
water supply network
Damage to building
and infrastructures
Fast
landslides
Damage to tanks of
water supply network
Damage to building
and infrastructures
Damage to building
and infrastructures
Failure of
sewer network
EC FP7 CLUVA
Climate Change and Urban
Vulnerability in Africa
Studied Cities:
Douala, Cameroun
Saint Louis, Senegal
Ouagadougou, Burkina Faso
Addis Ababa, Ethiopia
Dar Es Salaam, Tanzania