Flood damage modelling
using the Flood Impact Assessment Tool
Delft3D User Days 2019
Why model flood impacts?
• Cost Benefit Analyses
• Benefits of detailed measures
• Optimal design
• Spatial planning
• Impact forecasting
• Forecasting what the weather will do rather
than what the weather will be
• Settings premiums
Application of impact modelling
CBA infrastructure investments
Risk screening studies
Adaptive delta management
Setting insurance premiums
Climate change impact
Impact Based Forecasting:
Warning information and
Where to send aid first
How much money to free
up for recovery.
Initial distribution of
Category Tangible Intangible
Direct • Capital (houses, crops,
cars, factory buildings)
• Production losses, income
• Casualties, injuries,
• Social disruption, emotional
Indirect • Production losses / loss of
utility services outside
• Unemployment, migration.
• Cutting of infrastructure
• Loss of potential for attracting
• Reputation damage
Use of multiplication factor for everything that is difficult to
• Damage and loss
• Modelling needs
• Scope Delft-FIAT
Direct tangible - Business Interruption
• A flood can last between hours up to sometimes 1 year (e.g. Zeeland
• If water can flow away naturally it is short.
• If water needs to be pumped and dikes need to be repaired this can be
• Recovery time can also be long (easily 1 year)
• Shortage of contractors, waiting for permits.
• Experts need to check for mold.
• Larger total disasters need more recovery time.
• Recaptured value
• Often a lot of interruption damage can be recaptured elsewhere (e.g.
competitor does more).
• Damage depends on definition.
• Production losses outside flooded area
• (e.g. production process halted because crucial component cannot be
• Famous case of hard drives in Thailand
• Part of the losses recaptured by competition
• Modeled with several types
of economic models, highly
• Cutting of infrastructure lines
• E.g. Traffic problems,
power outages, etc.
New York Times – Nov 6, 2011
• Deadly casualties differ very strongly among floods.
(often 0 sometimes 1000s).
• Deadly casualties when: large water depths, rapid
rise rate, unexpected and unprepared people.
• Casualties are more often sick and elderly.
• Poor people in developing countries might die from
hunger or disease.
• Poor people in developing countries may become
homeless and get into major trouble.
• After a flood there are often shortages in construction labour and
• Shortages drive up prices as people compete for limited resources.
• Especially important when a flood is focused on one densely populated
area (e.g. dike breach near city).
Including demand surge
• Demand surge is a loss for some but an equal profit for others. Therefore,
often not used in an economic analysis.
• Insurance companies do take it into account.
Correcting for inequality in Cost Benefit Analyses
0 20 40 60 80 100
Equal decrease in
Unequal decrease in
Delft-FIAT – Damage only
Delft3D FM Suite:
• D-Flow FM
• D-Hydrology (wflow)
Probabilistic Toolkit (PTK) Delft-FIAT
- Mostly expert judgment
- Only few techniques available
Available approaches damage functions
• Expert or group of experts come
together and estimate a damage
• Elements of object of interest
can be assessed individually.
• Weakness is that experts
typically have one setting in
Data-driven approach (empirical)
• Regression analysis on available
data points of past flood damage.
• Weakness: Data availability and
From damage to flood risk (EAD)
• Flood damage can be calculated for an event. Yet many possible events
• The flood damage of one event alone is therefore too little to get a
complete picture and hence too little for rational decision making.
Flood risk: Expected Annual Damage (EAD)/Annual Average Loss (ALL)
• Summary statistic that combines all possible flood events, their
probabilities and their damages into one figure.
• The unit is: Euro/year
• Very useful for decision making!
Calculating flood risk (EAD)
• Combine many different flood
events into maps (or aggregate
damages) for different
exceedance probabilities .
• Take the integral to get the
expected annual damage.
• In practice calculate the area
under the graph.
𝑅𝑖𝑠𝑘 = න 𝐷𝑎𝑚𝑎𝑔𝑒 𝑝 𝑑𝑝
0 1/20 1/10 3/20 1/5 1/4
Exceedance probability (1/y)
• A risk reduction measure needs to
function for a long time
• A cost-benefit requires future risks as
input and not just current
• Hazard, Exposure and Vulnerability
changes over time
• Change needs to be predicted
Change in hazard
• Climate change
• Sea level rise
• More extremes (rain, droughts,
• Changes to the system:
• Land subsidence
• Erosion, sedimentation
• Wetland encroachment
• Change in impervious area
Change in exposure
• Extra buildings
• Population growth
• Fewer people per building
• More value per building
• GDP per capita growth
Change in vulnerability
• Often neglected, little
• Bangladesh example of
reduction in vulnerability of
loss of life
Mechler & Bouwer (2015) Climatic Change
Beyond Delft-FIAT: Machine Learning for better impact predictions
From: Damage fraction = f(water depth)
To: Damage fraction = f(water depth, warning time, wave height, …..)
DF = f(water depth) DF = f(water depth, warning time, waves height, …..)
Multi-variable damage models can be build from data with Machine
• My PhD and project to prioritize
humanitarian aid in the Philippines
• Use of historical data on damages
Machine Learning for macro level impact forecasting
RedCross data: 12 typhoons, 2012 - 2016
1600 damage data
% Total damaged houses in a municipality
Hazard : Average wind speed, rainfall
Exposure : building, people (2010)
Vulnerability : roof & wall type, GDP, slope
Situation Flood risk Colombo
• Recent floods
• Combination river
discharge, local rainfall
and sea level
• Wetland encroachment
• Proposed interventions
• WorldBank loan
80 runs (30m) different
Return period maps
projections and CBA.
Training 1 Training 2 Training 3
Ferdinand Diermanse Laurens Bouwer
Marc van Dijk
Simplified method outer areas
carried out completely by local partners
Exposure and damage functions
• Detailed data on building level
• Collected for this project
• Building type, number of floors, shanty.
• 57 damage categories
• Also vehicles, electricity and telecom.
• Created by experts
• Bills of quantities
0 5 10
Inundation depth (m)
Future damage projections
- Damage assumed to
increase with GDP per capita
- Population growth not
included because expected
move to high rise buildings
- Population growth in wetland
areas considered separately
- 5 growth scenarios
Cost Benefit Analysis
• Sum of future risk reductions should be smaller than the investment costs
of the intervention.
• Risk reductions in the future count less (discount rate)
• Discount rate Colombo difficult to estimate.
• Internal Rate of Return is the discount rate for which the sum of future risk
reductions is equal to the investment costs.
• Indirect damage discussion
Details about this project and
all additional assessments
are available in article:
measures for reducing flood