Transcript Document

Irish pork
GLOBAL FLOOD RISK
Upmanu Lall
Columbia Water Center, IRI
Global Flood Initiative
Hirschboeck, 1988
TRADITIONAL PERSPECTIVE
•
What is a flood? : river out of banks and inundates area for some duration
•
Design/Insurance: Estimate T-year flood using at site runoff or rainfall-runoff data
• Annual Max data or Partial duration series
• Regional Flood – use multiple locations to improve at site T-year estimates
• Loss estimates – typically direct physical loss for flood impacted area
•
Operation/Warning: Map QPF into event flood peak, volume, duration prediction using
hydro-models
•
Hydraulic analyses to map flood plain for zoning
•
Retrospective analysis of Synoptic Meteorology/Climate state associated with floods
_______________________________________________________________________
Mixtures? Climate Mechanisms? Duration, intensity, recurrence attributes?
Hirschboeck, Paleo-floods work, ENSO/interannual variability in flood incidence
Flooding affects more people worldwide than any other form of
natural disaster. And yet insurance cover against the risk of flooding is
not widespread (locally correlated risk).
- Swissre
Climate
Characteristics
Characteristics
of Catchment/
Stream/
Floodplain
System
FLOOD FLOW
ESTIMATES
River Basin Flood Risk
Analysis
Water Resource
Development + Use
Agricultural
Land Use
Urban / Industrial
Development
FLOOD LEVEL
ESTIMATES
Flood Vulnerability
Other Flood
Characteristics
Flood
Hazard
Flood
Probabilities
FLOOD
RISK
Socio-Economic
Values,
Environmental
Values,
Regulatory
Provisions,
Community
Attitudes
FPM Goals
Floodplain
Management
Strategies,
Flood Design
A GLOBAL FLOOD PERSPECTIVE
Flood: Atmospheric and terrestrial concentration of water flux into certain regions, that leads to multiple
locations with inundation over a period of time?
•
•
How do specific climate mechanisms lead to floods at different space-time scales across the world –
conditional quantification using local, regional and global factors?
•
IID: Fat tails or identifiable nonstationary, mixtures?
•
Dynamics: Persistent climate state  high frequency space-time precipitation dynamics 
with river basin topology and hydrologic dynamics: linked spatio-temporal stochastic models
A dynamic risk rather than static risk paradigm, including its spatial implications
•
•
Dynamic  time scales, lead times, space scales
Shift from purely watershed/river basin perspective to ocean-atmosphere pathways: Local
correlation structure vs global or far field correlation structure– inferred from dynamical models?
Global Impacts and Decisions:
•
Persistent and delayed socio-economic and health impacts in addition to direct physical loss
•
Global Supply Chains
•
Insurance, and infrastructure design/operation considering cumulative impacts and risk
layering
•
Disaster response
AUSTRALIAN FLOODS IMPACT GLOBAL SUPPLY CHAINS
•
The impact of the devastating floods in Queensland will be felt through global supply
chains for many months to come. Almost 70% of global steel production depends on
metallurgical or coking coal. Australia produces two-thirds of global exports of coking coal,
of which Queensland accounts for 35%.
•
Fears over coal supplies as Australia floods worsen
•
More heavy rainfall causes exports of coal, wheat and sugar to significantly decline
as country left underwater
•
Coal supply
Australia is the world's largest exporter of coking coal, supplying half the global market.
used to produce steel, and operators of around 40 mines have been affected by the
floods.
•
The supply of wheat, of which Australia is the world's fourth biggest exporter, has also
been hit.
•
Australia floods to squeeze India steel cost margins - CRISIL Reuters
PAKISTAN SUPPLY CHAIN UNDER STRAIN
•
The floods have had a significant impact on Pakistan's nascent textile industry. Local
business associations have estimated that the destruction has destroyed three million
bales of cotton. As a consequence, the cost of clothes production within the country will
rise by 20%. With apparel buyers seeking to stock inventories for the Christmas sales,
companies are concerned over the viability of the Pakistan supply chain to deliver
sufficient volume on time and on budget. Indeed, many orders have been re-directed to
suppliers in Bangladesh and Sri Lanka. Already, export orders have declined by 7-10%,
and this could fall by a further 30%.
•
The FT reports that clothing companies such as Levi Strauss and UK-based Next have
warned of inflating clothing prices.
Managing Climate Risk (Layering)
Abrupt
Anthropogenic
Climate
Change
“Natural”
Dynamic
Risk
Predictable
Long Term
Statistics
“Smooth”
Unpredictable
Near Term
Evolution
Infrastructure Design
Allocation/Operation Rules
Adaptive Operation &
Allocation
Early Warning Systems
Pizarro, Lall and Atallah, Env Finance 10(10), 2009
Residual
Risk
Financial
Instruments:
Insurance
Cat Bonds
Relief
EXPLORING THE CLIMATIC CONTEXT OF FLOODS
• Floods associated with large scale circulation patterns
• Meridional and Zonal Moisture Transport and Convergence
• Spatial Incidence of Floods  regions with high potential
• Identifiable low frequency forcing….ENSO etc
• Prediction? Hierarchical Bayesian Models of Floods
• Area Scaling
• Covariates
• Diagnosis of Large floods in a region
• Ohio River Basin
Global Flood incidence recent trends
Columbia Water Center
Global Flood Initiative
Hypothesis: Meridional water vapor transport
changes drive latitudinal shifts in flood incidence
JFM
Columbia Water Center
Global Flood Initiative
2002 JJA
L
a
t
i
t
u
d
e
Longitude
Year
No. of floods
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
60 110
100
69
60
77
91
56
47
52
2003 JJA
L
a
t
i
t
u
d
e
Longitude
Year
No. of floods
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
60 110
100
69
60
77
91
56
47
52
2004 JJA
L
a
t
i
t
u
d
e
Longitude
Year
No. of floods
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
60 110
100
69
60
77
91
56
47
52
2009 JJA
L
a
t
i
t
u
d
e
Longitude
Year
No. of floods
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
60 110
100
69
60
77
91
56
47
52
2010 JJA
L
a
t
i
t
u
d
e
Longitude
Year
No. of floods
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
60 110
100
69
60
77
91
56
47
52
JJA Flood Density by Latitude: Groups
2001
2005
2008
2002
2004
2006
2010
2009
2003
2007
Latitude
Flood Magnitude depends on Area (Scaling law)
Flood magnitude may depend on a pre-season climate covariate
Can we predict conditional flood distribution at gaged/ungaged locations?
A Hierarchical Bayesian Model -- Lima and Lall, 2010
CLIMATE INFORMED NON-STATIONARY,
REGIONAL FLOOD PREDICTION
Flood Data
Location of Basin in Brazil
• Daily naturalized series of 37 sites (Parana basin)
• Provided by ONS – Period 1931-2001
• Homogeneous sub-basins re climate (ENSO and SACZ)
Location of streamflow sites (red dots are testing sites)
Hierarchical Bayesian Model
• Simple Scaling Law: log(flow moments) ~ log(drainage area)
E [ Y  ]   E [ Y1 ]
h
hk
h
log E [Y  ]  hk log   log E [Y1 ]
h
h
• Hierarchical Bayesian Model: event based scaling
log( q ij ) ~ N ( 0 i   1 i x ( j ),  )
2
x ( j )  log( A ( j ))  log( A )
• Priors
  0 ,i 
    11 y ( i )


 ~ N  10
    y ( i ) ,   , i  1,  , N
 
21
 20

 1,i 
Climate index: NINO3 DEC(-1)
p ( )  1 .
• Hyperpriors (uniform)
p ( 10 ,  11 ,  20 ,  21 )  1
p    
 ( d 1) / 2
, d  2.
Flood Data – Drainage area pdf
Drainage areas varying from 2588 to 823555 km2
Testing sites
Results – non-stationary scaling
parameters
Results – parameters vs pre-season
NINO3 index
Slopes are statistically significant!
Results: predicting “ungaged” annual
flood series
r=0.74
r=0.71
r=0.66
Dynamic Risk: 100 year event– site 1
Q* such that P(Q(t) > Q*) = 0.01
Dynamic 100 year flood – site 2
Inverse Problem: I see a big flood….how did it get here
A very few selected examples out of many diagnostic ventures
FLOODS AND LARGE SCALE MOISTURE
TRANSPORT
Atmospheric Moisture Transport associated with one of
the top 10 floods at different locations
Source: Hyun-Han Kwon
Columbia Water Center
Global Flood Initiative
Nakamura et al, Dec 2010 AGU
Columbia Water Center
Global Flood Initiative
Columbia Water Center
Global Flood Initiative
Columbia Water Center
Global Flood Project
Columbia Water Center
Global Flood Project
Columbia Water Center
Global Flood Initiative
Columbia Water Center
Global Flood Project
Columbia Water Center
Global Flood Initiative
Columbia Water Center
Global Flood Initiative
Columbia Water Center
Global Flood Initiative
Columbia Water Center
Global Flood Project
Columbia Water Center
Global Flood Initiative
Columbia Water Center
Global Flood Initiative
DIRECTIONS……..
• Invitation to develop global flood risk initiative
• An Open Source Risk Modeling & Mitigation Effort – Climate to
Impacts to Response
• The design and exploration of a statistical-dynamical approach for the
short (-5 to 10 days) and long lead (> 1 month) prediction, and for the
conditional simulation of such events using climate (model) states.
• Inverse/Forward Modeling and Prediction at various lead times
appears possible enabling dynamic risk management
• Spatio-temporal causal structure at large and fine scales needs to
be identified and modeled (joint flood/drought incidence/extent)
• Integrating storm track dynamics and drainage network response
including infrastructure
• Loss dynamics – composite events, delayed and far field losses
• Mitigation: Risk Layering, Response and Recovery Design
Columbia Water Center
Global Flood Initiative