1 of 25

1

Compound Flood Exposure: A Primer and Approaches for Incorporating Climate Change

Jayantha Obeysekera (‘Obey’), Ph.D.,P.E.

Director and Research Professor

Sea Level Solutions Center

Institute of Environment

2024 SIO Flood Workshop

January 30-31, 2024

Web: https://environment.fiu.edu | http://slsc.fiu.edu Facebook: @FIUWater | Twitter: @FIUWater

2 of 25

Outline

Introduction
Definition and typology
Modeling Approaches
Nonstationarity
Case Study

2

Future changes in the return periods of concurrent meteorological drivers of compound flooding

Bevacqua et al. (2022).

https://doi.org/10.1038/s43247-020-00044-z

3 of 25

COMPOUND FLOODING IN A NON-STATIONARY WORLD: A PRIMER FOR PRACTICE (ASCE ~2024) CACC/HYDEA COMMITTEE~2024)

Editors

Rolf Olsen, Julie Pietrzak, Jayantha Obeysekera

2. Background

Poulomi Ganguli, pganguli@agfe.iitkgp.ac.in (lead author)

Shaleen Jain, Carlo De Michele, Gianfausto Salvadori

3. Hydrodynamic Process-Based Models of Compound Flooding

Liv Herdman, (lead author)

Matthew Bilskie, Antonia Sebastian, Ning Lin, Miguel Medina, Teng Wu

4. Statistical Models of Compound Flooding

Carlo De Michele, (co-lead author)

Gianfausto Salvadori, (co-lead author), Thomas Wahl, Amir AghaKouchak, Robert Jane, Emanuele Bevacqua, Ivan Haigh, Ferdinand Diermanse

5. Joint Probability Method

Norberto Nadal, (lead author)

Michelle Bensi, Madison Yawn, Victor Gonzalez

6. Linking Statistical and Hydrodynamic Process-Based Models

Hamed Moftakhari, Amir AghaKouchak, David Muñoz, Ning Lin, Ferdinand Diermanse,

7. Analysis of Changing Conditions

Rolf Olsen, (co-lead author)

Julie Pietrzak,

Jayantha Obeysekera

Gerarda Shields

Poulomi Ganguli

Avi Gori Ning Lin

Mohammad Reza Najafi

Miguel Medina,

Teng Wu

8. Risk and Uncertainty Analysis

Gerarda Shields, (co-lead author)

Jayantha Obeysekera

Rolf Olsen, Julie Pietrzak,

Poulomi Ganguli

Michelle Bensi

Antonia Sebastian

Miguel Medina

Carlo De Michele

Gianfausto Salvadori, Fabrizio Durante,

Ferdinand Diermanse, Teng Wu,

Aikaterini Kyprioti

9. Case Studies

Shubra Misra

Norberto Nadal

Credit: MOP (in_Prep)

4 of 25

Compounding Events: Drivers of Change

Sea Level Rise/Storm Surge Impacts

Hurricanes

Saltwater Intrusion

Contamination from on-site

Septic Systems

Aging Infrastructure

Stressors:

Rising Temperatures
Sea Level Rise & Storm Surge
Saltwater intrusion
Rising groundwater levels
Changes in rainfall patterns

Fronts
Thunderstorms

Frequency and magnitude Hurricanes

Urban Flooding

Natural System Impacts

5 of 25

Concepts of Flood Frequency and Increasing Risk: Threat Multipliers

A measure of flood frequency: 1% chance flood (100 Year-flood)

Road

1%

2%

10%

Risk: 26% in 30-years

Risk: 45% in 30-years

Risk: 79% in 15-years

Changing

Heavy Rainfall

Rising groundwater

Rising Seas

High Tide

Flooding

6 of 25

6

Extended Risk Framework

The combination of multiple drivers and/or hazards that contributes to societal or environmental impacts (Zscheischler et al., 2018).

https://doi.org/10.1038/s41558-018-0156-3

(Zscheischler et al., 2020).

Key Elements of Compound Events

7 of 25

Compounding Events: Typology

Sequential or preconditioned:

Sequential or preconditioned hazards indicate a background climate or weather condition that leads to an amplified impact

Multivariate

The coincidence of two or multiple hazards at the same location

Temporally compounding/Consecutive Hazard:

Results from a sequence of hazards involving similar or dissimilar types of weather or climate events that have occurred in succession affecting a geographical region

Spatially compounding:

Spatially compounding events results from accumulated impacts of the same or different hazards occurring in multiple connected locations within a short time window.

7

(Zscheischler et al., 2020).

8 of 25

Compound Flood Generating Mechanisms�

Driving Mechanisms

High coastal water levels impacting river flow due to backwater effect 🡪 prominent < 10 m

High storm surge height block/slow precipitation drainage into the sea, triggering floods

High coastal water levels and high river discharge in deltas 🡪 driven by a storm event

Precipitation on already saturated soil preceded by river floods

Geospatial setting, wind-facing direction, shape of coastlines, dictates frequency of CFs

Causal links

Physical Processes triggering CF Events

ASCE MoP (in Prep.)

9 of 25

Modeling Approaches

Hydrodynamic Process-Based Models
Statistical Models
Hybrid: Linking Statistical and Process-Based Models
AI/ML Methods (emerging)

9

10 of 25

Hydrodynamic Process-Based Models of Compound Flooding

Rainfall and Runoff
Riverine Flows
Coastal dynamics
Coupling between these processes
Important inputs for accurately capturing inundation

10

Rainfall

Coastal Storms

Coastal Nuisance Flooding

Riverine Flows

11 of 25

Different Modelling Compound Flood Requires Coupling Different Processes-based Models�

11

Coupling Technique	Methods
One-way	Computations that are transferred from one model and used as an input in another (i.e. linking technique)
Loosely	Separately-running models are coupled using information exchange in an iterative manner (i.e. two-way coupling)
Tightly	Independent models are integrated into a single modeling framework by combining their source code
Fully	Governing equations of all the physical processes considered are solved simultaneously within the same modeling framework

Source: Santiago-Collazo et al. (2019)

https://doi.org/10.1016/j.envsoft.2019.06.002

There are four ways to couple these models as classified by Santiago-Collazo, Bilskie and Hagen. You can do a one-way coupling where Computations that are transferred from one model and used as an input in another (i.e. linking technique). The models can be loosely-coupled, which means Separately-running models are coupled using information exchange in an iterative manner (i.e. two-way coupling). Or models can be tightly coupled where Independent models are integrated into a single modeling framework by combining their source code. Currently in most compound flooding applications a loosely coupled modeling approach has been utilized. The joining technique should depend on the type of physical processes that are being modeled.

One must pay attention to the conservation of mass when joining models to ensure that any exchange of water from one process model to the other does not increase or decrease the overall amount of water in the system. Often tightly coupling models that have different numerical frameworks can lead to violations of mass conservation and/or other instabilities that make a fully or tightly coupled approach difficult. This is an ongoing area of active research.

*From Félix L. Santiago-Collazo, Matthew V. Bilskie, Scott C. Hagen,

A comprehensive review of compound inundation models in low-gradient coastal watersheds,

Environmental Modelling & Software,

Volume 119,

2019,

Pages 166-181,

ISSN 1364-8152,

https://doi.org/10.1016/j.envsoft.2019.06.002.

12 of 25

Challenges of Process-Based Models

Significant underlying assumptions in formulation and parameterization,
Lack of a comprehensive record for validation of numerical models that affects the reliability,
Necessity of coupled modeling to enable the exchange of flux at the interface of a hydrologic and hydrodynamic model,
Usually the link to bio-physical properties are omitted. The choice of the most suitable variable to represent a flood driver depends on the system under consideration,
The flood drivers typically can vary in time and space, while single values are preferred for statistical analysis,
To cover the wide range of possibilities for compounding effects of flood drivers it is desirable to carry out a relatively large number of model simulations, yet in order to keep the approach practically manageable with respect to computation cost, a limited number of model simulations is preferred.

12

13 of 25

13

Statistical Methods

Archimedean Copula family

Clayton

Frank

Gumbel

Credit: Thomas Wahl

14 of 25

Copulas to the rescue!

15 of 25

Hazard Scenarios

15

See Salvadori et al. (2015) for examples10.1002/2015WR017225

e.g. Floods at a confluence

e.g. Flood Peak and Volume

e.g. Coastal erosion causing many different outcomes

16 of 25

Shortcoming of Statistical Methods

Lack of spatiotemporal coverage of overlapping records to detect and characterize the interdependencies of compound flood drivers
Given their significant underlying assumptions (i.e. stationarity), they offer limited opportunities to capture anthropogenic effects (i.e. climate change) and their impacts on the probability that coincidence/concurrence of flooding mechanisms yield in a level of impacts greater than each in isolation.

16

17 of 25

Linking Statistical and Process Based Models (Hybrid Models)

Integrating and linking statistical and process-based models can result in a more reliable estimation of compound flood risk while keeping the computational cost reasonable and providing valuable insights for resilience assessment and planning.
‘Hybrid’ approaches, under which multiple statistical and process-based models are coupled to evaluate the desired impact in a larger complicated system, have recently been proposed that lays out opportunities to more efficiently overcome these challenges with the help of reduced physics surrogate modeling and high-performance computing systems.

17

18 of 25

18

Non-stationarity – different ‘types’

Non-stationarity in the multivariate case can come from either of the relevant variables (e.g. trends, cycles etc.) or changes in their dependency.

Green: hypothetical distribution of two climatic drivers in the present climate.

Blue: a future climate with shift in mean, variability and correlation between the drivers.

Purple: future climate with an increase in dependence in the upper tail of both drivers.

19 of 25

Change-Point Approach

Non-stationary Compound Flooding. Natural and anthropogenic forcings, including climate and/or land use changes, can put at risk the assumption of stationarity of floods.

The problem can be decomposed in assessing separately the presence of non-stationarity either in the marginals, or in the Copula, or both.

The non-stationary can affect the estimate of future Failure Probabilities and Return Periods of occurrences (e.g., before and after a distributional Change-Point, as in the top/bottom figure).

Salvadori et al. 2015

Water 2018, 10, 751; doi:10.3390/w10060751

19

20 of 25

Nonstationarity in the Univariate Case

20

Risk-Based Design

Recurrent Flood Frequency

Expected Waiting Time

21 of 25

Nonstationarity Framework for Precipitation and Sea Level

Trends in and dependence between sea level and precipitation increase the probability of compound flooding along US coastlines
A Bayesian copula-based framework captures the uncertainty in the quantification
Precipitation trends play a major role in increasing the uncertainty in compound flood frequency.
To account for non-stationarity, a long historical record is required.

21

Naseri and Hummel 2022

https://doi.org/10.1016/j.jhydrol.2022.128005

Stationary

Nonstationary

22 of 25

Nonstationarity Framework for Riverine Discharge and Sea Level

22

23 of 25

Case Study: Miami, Florida

The number of years before the O-sWL in the 50-year design event derived using the bivariate approach reaches the corresponding value obtained using the original design approach according to the three SLR scenarios

https://doi.org/10.5194/nhess-20-2681-2020

10-yr

20-yr

50-yr

100-yr

24 of 25

Case Study(cont.)

24

25 of 25

Questions?