Flood and Drought Displacement Risk in Ethiopia Somalia and Sudan

Ref.
Ares(2024)5119930 - 15/07/2024
July 2024
Flood and Drought

Displacement risk
in Ethiopia, Somalia
and Sudan
D3.2 Research paper

The HABITABLE project – Linking Climate Change, Habitability and Social Tipping Points: Scenarios
for Climate Migration – aims at significantly advance our understanding of the current interlinkages
between climate change impacts and migration and displacement patterns, and to better anticipate
their future evolution.
Running for 4 years (2020-2024), HABITABLE brings together 21 partners: University of Liège, University
of Vienna, Potsdam Institute for Climate Impact Research, University of Exeter, the IDMC, Lund
University, Sapienza Università di Roma, adelphi, Université de Neuchâtel, Institut de Recherche pour
le Développement, Council of Scientific and Industrial Research, UNESCO, University of Ghana, CARE
France, University of Twente, Université Cheikh Anta Diop, Stockholm Environment Institute Asia, Raks
Thai Foundation, Addis Ababa University, Institut National de la Statistique du Mali and Samuel Hall.
HABITABLE is the largest research project on climate change and migration to have ever been funded
by the European Commission's Horizon 2020 programme.
Please visit www.habitableproject.org for more information about the project.
Lead authors:
Sylvain Ponserre – IDMC
Lauro Rossi - CIMA
Co-authors from CIMA Co-authors from CIMA Internal reviewers:
for floods: and Vrije Universiteit Jacob Schewe - PIK
Lorenzo Campo Amsterdam for droughts: Christelle Cazabat - IDMC
Andrea Libertino Daria Ottonelli Florian Debève - University of Liege
Daria Ottonelli Marthe Wens
Roberto Rudari Hans de Moel
Eva Trasforini
Suggested citation:
CIMA Research Foundation, Internal Displacement Monitoring Centre (IDMC) (2024). Floods and
Drought Displacement risk in Ethiopia, Somalia and Sudan. HABITABLE research paper. Geneva: IDMC.
Publication available on:

To be completed by the ULiège team before uploading to the HABITABLE website
Version N° Date Information

1 15 July 2024 Initial version submitted to the EC as
deliverable
Graphic design, cartography and layout: Stéphane Kluser (Komplo).
This project has received funding from the European Union's Horizon 2020 research and innovation
programme under Grant Agreement No 869395. The content reflects only the authors' views. The European
Union is not responsible for any use that may be made of the information it contains therein.
Table of content 3
Flood Displacement risk in Ethiopia, Somalia and Sudan.......... 5

Abstract..................................................................................................................6
Introduction.......................................................................................................... 7
Methods................................................................................................................10
Displacement risk concepts.........................................................................................10
Disaster displacement risk model ...............................................................................12
Exposure and vulnerability model ..............................................................................16
Flood hazard assessment . ...........................................................................................22
Flood impact assessment .................................................................................24

Ethiopia ..........................................................................................................................25
Sudan .............................................................................................................................28
Somalia ..........................................................................................................................31
Flood and Drought Displacement risk in Ethiopia, Somalia and Sudan

Discussion and conclusion ..............................................................................36
Drought Displacement risk in Ethiopia, Somalia and Sudan.... 39

Abstract............................................................................................................... 40
Introduction........................................................................................................ 41
Methods................................................................................................................42
Drought-induced displacement risk concepts..........................................................42
Drought displacement risk assessment ......................................................................45
Drought-induced displacement......................................................................56
Somalia...........................................................................................................................58
Ethiopia...........................................................................................................................60
Sudan..............................................................................................................................61
Discussion and conclusion.............................................................................. 64

References.......................................................................................................... 66
risk in Ethiopia,
Somalia and Sudan
Flood Displacement
5
Flood Displacement risk in Ethiopia, Somalia and Sudan

Abstract
This report highlights that floods have led to more Using the latest technologies and a novel
than 185 million internal displacements, or forced vulnerability assessment method, the study
movements within one’s country, globally since expresses displacement risk through average
2008. Africa experienced around 29 million flood- annual displacement (AAD) and probable
induced displacements between 2008 and 2022, maximum displacement. The results, evaluated
ranking as the second most affected region. under current and future climate conditions with
Focusing on the Horn of Africa, particularly Sudan, optimistic and pessimistic scenarios, indicate a
Ethiopia and Somalia, this study introduces a potential increase in AAD of two to four times
novel flood displacement risk model. This model, compared with current conditions, and even
developed under Work Package 3 of the higher risks for pessimistic scenarios, such as a
HABITABLE project, aims to provide calibrated ninefold increase for Sudan.
estimations of future movements, supporting The outputs can inform national and subnational
the formulation of effective policies. The disaster risk reduction measures, helping identify
methodology incorporates a unique vulnerability areas prone to large-scale displacements.
assessment, considering factors often omitted in Decision-makers can use this information for
standard risk models, such as direct impacts on risk-informed efforts to prevent and mitigate the
houses and livelihoods, and indirect impacts on impacts of displacement. The report concludes
critical facilities and services. The assessment uses with recommendations for comprehensive
a probabilistic approach, integrating climatic, policies and strategies to address flood-induced
hydrological and hydraulic modelling to estimate displacement risks and protect affected
the impacts triggering displacement. populations.
Keywords
Displacement Risk, Riverine Flood, Probable Maximum Displacement, Average Annual Displacement,
Displacement Vulnerabilities
Introduction 7
Every year, floods, droughts, storms, earthquakes IDMC began a unique probabilistic modelling
and other natural hazards force millions of people exercise for global disaster displacement in
to leave their homes, a level of displacement 2017, assessing the likelihood of such population
greater than that associated with conflict and movements in the future. It built on a risk analysis
other forms of violence. Leaving home can be the developed by the UN Office for Disaster Risk
first of many further disruptions to people’s lives: Reduction (UNDRR), based on the consideration of
they may be forced to move several times once a wide range of hazard scenarios, their likelihood
they become displaced, and it can be weeks and their potential to cause housing damage
or even months before they are able to return (the latter serving as a proxy for displacement).2
home. Those who do return often face unsafe The model used a state-of-the-art probabilistic
conditions and the prospect of being displaced approach, similar to that applied by catastrophe
again by the next disaster. modellers and the insurance industry over the
past few decades. At the time, it covered only
In 2023, the Internal Displacement Monitoring
the physical aspect by looking at the extent of
Center (IDMC) detected 26.4 million of
damage and destruction that hazards of different
movements due to consecutive disasters. And
intensities were likely to cause.
1 out of 3 disaster displacements were triggered
by floods, surpassing storms for the second time Given that people’s level of vulnerability and
since 2016.1 exposure to hazards does much to determine
the severity of a hazard’s impacts, however, it
Under the European Commission-funded project
is important to assess how these aspects may
HABITABLE, IDMC brings its unique expertise
change over space and time, and to unpack
as one of the world’s definitive sources of
the economic, social and environmental factors
data and analysis on internal displacement.
that affect disaster risk.
Since its establishment in 1998 as part of the
Norwegian Refugee Council, IDMC has offered To do so, IDMC has worked closely with partners to
a rigorous, independent and trusted service to obtain improved data on risk exposure and rethink
the international community. Its work informs how to assess vulnerability in the displacement
policy and operational decisions that improve risk equation. Given that “riskscapes” evolve
the lives of the millions of people living in internal constantly, we need to understand population

displacement or at risk of becoming displaced and socioeconomic patterns, and fluctuations
in the future. The objective of this report is in the frequency and intensity of hazards linked
to advance understandings of the current to climate change.
interlinkages between climate impacts and This report presents the future riverine flood
displacement patterns, and better anticipate displacement risk assessment at the national
their future evolution. and subnational level for three countries in the
Under Work Package 3 of the HABITABLE project, Horn of Africa, through the application of a new
we support the development of models to methodology first applied in two small island
produce empirically calibrated estimations of developing states, Fiji and Vanuatu in the Pacific.3
future human mobility for flood hazards as the It provides a more comprehensive assessment of
effects of climate change deepen in Ethiopia, vulnerability in terms of disaster displacement risk,
Sudan and Somalia. introducing socioeconomic drivers’ key factors
for displacement risk assessment.
It reveals the magnitude of displacement risk, For Work Package 3 of the HABITABLE project,
and by comparing present climate conditions we collaborated with the International Center
with various future scenarios, it shows how they for Environmental Monitoring (CIMA Research
are likely to differ. Like many countries in Africa, Foundation) on assessing flood displacement risk
exposure to flood hazards in Ethiopia, Somalia under two different climate scenarios (optimistic
and Sudan is driven by the growing concentration and pessimistic) in the medium and long term in
of people and assets in flood-prone areas. Ethiopia, Somalia and Sudan.
9
UNICEF Ethiopia/2021/ Nahom Tesfaye

Nechye Aygaliye (45), a mother of 7 was
among those displaced. She and her family
were forced to leave their Chugar village
and moved to Delerele where they were
settled into a camp for the displaced, along
with 12,000 other people, like her, displaced
by the floods from villages dotted along
the river and lake. Some 2,000 under-five
children live in the camp.
Methods
Displacement risk concepts Like disaster risk, disaster displacement risk is linked
to the nature and magnitude of a given hazard,
Disaster risk and vulnerability and to people’s exposure and vulnerability. Risk
The UN Office for Disaster Risk Reduction (UNDRR) analysis reveals the likelihood of future disasters
defines a disaster risk as “the potential loss of life, triggering displacement. Evaluating vulnerability
injury, or destroyed or damaged assets which levels proves valuable in understanding why severe
could occur to a system, society or a community events do not consistently yield proportionally
in a specific period of time, determined severe consequences for populations and
probabilistically as a function of hazard, exposure, regions, whereas events of lesser severity can
vulnerability and capacity”.4 It is important to result in profound impacts. In other words, it
consider the social and economic background clarifies why two hazards of similar intensity and
in which disaster risks occur and understand duration could have different consequences in
that people do not necessarily share the same terms of damage, people affected and people
perceptions of risk or underlying risk factors. displaced.
Vulnerability is defined as the conditions Understanding the parameters that influence

determined by physical, social, economic and vulnerability is essential to inform effective policies
environmental factors or processes that increase and strategies that prevent and manage disaster
the susceptibility of individuals, communities, displacement risk, and so reduce the number of
assets or systems to the impacts of hazards.5 The people forced to flee.
vulnerability of humans is influenced by a variety Until now, approaches to disaster displacement
of economic, social, cultural, environmental, risk assessment have typically calculated
institutional, political and psychological aspects vulnerability levels by using likely housing rendered
that mould the lives of individuals and the uninhabitable as a proxy for displacement, as
world around them. Factors including income, demonstrated in our 2017 report.7 This involved
education and healthcare access play a role assuming that if a disaster renders a residence
in determining individuals’ vulnerability levels. uninhabitable, its occupants will encounter
Depending on these elements, vulnerability levels displacement, even if only temporarily. This
tend to vary across a community or a population. approach allows for an estimation of the potential
The higher the level of vulnerability, the greater magnitude of future displacement.
the probability of being harmed by a disaster.
This vulnerability is not static; it changes over The need for a broader approach
time. It comprises various dimensions, all of
People’s vulnerability is influenced by a
which must be considered when conducting a
combination of physical and societal elements.
comprehensive evaluation.
People who live in concrete buildings are likely
to be less vulnerable to some types of hazards
Disaster displacement risk than those living in adobe huts, but their
The term disaster displacement refers to a situation vulnerability also depends on a range of other
in which people are forced to flee their homes factors that need to be assessed. They may have
or places of habitual residence because of a different sources of livelihood, income and social
disaster or to avoid the impact of an impending conditions, including access to basic services
natural hazard. Displacements generally occur such as health and education.
when people are exposed to a natural hazard Such elements, however, are not yet included in
in situations where they are too vulnerable and standard risk models even though they help to
lack the capacity to face the hazard’s impacts.6 determine whether people flee or not.
11
Consider, for example, a flood that affects 2 probability of occurrence, all elements of the risk
hectares of land. Its impact will not be the same equation (risk = hazard X exposure X vulnerability),
on all farmers. A subsistence farmer who only and their variability and uncertainty ranges are
owns 5 hectares will suffer a significant loss of all included.
livelihood, while a large-scale farmer who owns Events that have rarely been recorded but might
a greater amount of land will be much more able occur more often under climate projections are
to withstand the loss. People who depend on thus also considered. This feature is particularly
the primary sector (agriculture and livestock), useful because climate change is increasing
particularly if it is for subsistence, are at higher uncertainty about future hazard patterns. To be
risk of being displaced when a disaster strikes. prepared, societies need to calculate the worst
Disaster risk models need to analyse such possible impact. Viewed through this lens, there
socioeconomic factors if they are to inform is no valid alternative to a probabilistic analysis to
more effective policies and strategies to reduce address such uncertainty in a usable, quantitative
the number of people vulnerable to disaster way.
impacts and the number at risk of displacement. Displacement risk information, expressed in
Such policies and strategies would consider the average annual displacement (AAD) and
geography and specific population groups and probable maximum displacement (PMD), is
would include not only structural interventions calculated at the subnational regions and
– those intended to mitigate disaster impacts aggregated at country level, allowing for a
through physical construction – but also social geographic and quantitative comparison within
and economic measures. and between countries. These analyses and
Being able to move implies having sufficient comparison exercises are an important step in
economic, logistical and social resources to do risk awareness processes and key to pushing for
so. People without these resources are unlikely risk reduction, adaptation and management
to be unable to flee and may become trapped mechanisms to be put in place.
in the affected area. Such groups may include The PMD curve illustrates the probability of
extremely poor people, those with disabilities or a specific scenario leading to an estimated
illnesses and those living in isolated parts of a

number of displacements. This likelihood is usually
country. measured in terms of return period, which is often
misunderstood. A return period is the average time
Using a probabilistic risk assessment interval in years that separates two consecutive
events equal to or exceeding the given
The true benefits of a probabilistic risk assessment
magnitude. The most common misconception
are frequently misconstrued because it is
is that an event with a 100-year return period
regarded as a complex and challenging method
will only occur once a century, when instead it
to implement and follow, with a communication
means that it has an exceedance probability of
hurdle when presenting outcomes. A probabilistic
1 in 100, so events of the same or greater intensity
disaster displacement risk profile must be seen
happen once every 100 years on average. This
as a diagnostic tool, because it offers insights
does not preclude the possibility of several events
into potential hazard occurrences and their
with a 100-year return period happening within a
consequences.
century, or even the rare chance of consecutive
Such profiles cover all possible risk scenarios in a events transpiring in consecutive years. Neither
certain geographical area. Both low-frequency, does it eliminate the possibility of an entire century
high-impact events and high-frequency, passing without such an event occurring.
low-impact events are considered. Their
Applying different climate scenarios • optimistic: the scenario closest to the 20th
percentile, corresponding to an average
To explore the potential impact of climate change
temperature rise of about 1°C by 2100
on the forthcoming frequency and intensity of
extreme events, along with their associated • pessimistic: the scenario closest to the 80th
percentile, corresponding to an average
exceedance probabilities and return periods, our
temperature rise of more than 5°C by 2100
evaluation considers varying timescales:
We did not consider changes in exposure and
• under current climate conditions, with disaster
vulnerability between current and future climate
risk assessed using observed conditions from
conditions, although factors such as population
1979 to 2016
growth and distribution (for example, the rapid
• under long-term projected conditions from
urban sprawl that shrinks natural areas available
2061 to 2100
to absorb floodwater) may greatly alter the future
To capture the spread of possible climate “riskscape”.
scenarios, we compared 15 models from the
Inter-Sectoral Impact Model Intercomparison
Project (ISIMIP3b), which provides bias-corrected Disaster displacement risk model
Coupled Model Intercomparison Project Phase Our novel approach aims to address the
6 (CMIP6) climate scenarios for pre-industrial, deficiencies in existing displacement risk models
historical, SSP1-RCP2.6, SSP3-RCP7.0 and SSP5- and provide a more comprehensive assessment
RCP8.5 conditions, in terms of temperature of vulnerability to the risk associated with sudden-
and precipitation rise in 2016.8 Temperature onset hazards. Assessing displacement risk
and precipitation trends proved to be highly for slow-onset events, such as drought, would
correlated in the models considered, so we require a different approach. Under the same
referred only to temperature trends to define Work Package of the HABITABLE project, we
representative scenarios, from which we selected also collaborated with the International Center
two: for Environmental Monitoring (CIMA Research
Foundation) to assess drought displacement risk
in Ethiopia, Somalia and Sudan.
Average global annual temperature Temperature increase with respect to baseline
[°C] [°C]
22
Historical 7
Pessimistic scenario
21 Optimistic scenario
SSP126 models 6
SSP370 models
20 SSP585 models
5
19
4
18
3
17
2
16
1
15
0
14
-1
13
1980 2000 2020 2040 2060 2080 2100
Years
Figure 1: Model projections of future climate scenarios from ISIMIP3b.
This riverine flood displacement risk analysis Direct impact on houses 13
aids understandings of exposure and some This component is included in conventional
vulnerability dynamics. models for disaster displacement risk. It takes into
The methodology considers different but consideration the number of homes rendered
intrinsically linked components, as follows, in uninhabitable because of a disaster. Many
assessing the impact of disasters that have models, however, treat it in a simplified way,
already occurred. often using a threshold on the hazard intensity
parameter, which means they do not differentiate
between the different vulnerabilities to hazards
No essential
No housing No income
services
EVALUATION OF INDIVIDUAL CONTRIBUTIONS
Population is assigned to Employed population (per Residential population is

residential buildings sector) is assigned to assigned to different services
different land use classes (e.g. education facilities,

health facilities,
transportation network, water
Impacts on houses are
supply, etc.) through the
evaluated considering a Impacts on the different land
definition of a catchment
damage threshold for classes are evaluated; a
area.
identifying the displacement damage threshold - which
may vary according to the
social fabric - is defined for
identifying need Impacts on the single
displacement elements of the “network of
service” are evaluated, as
well as the loss of
functionality of the overall
service; a double threshold is
defined for identifying the
displacement.
CLEAR / DIRECT UNCLEAR / INDIRECT

Figure 2: Identification of the main components with the potential to trigger disaster displacement.
associated with different house types. The new needs of those compelled to move out of harm’s
methodology employs a complete physical way.
vulnerability model to compute the impact on To evaluate flood risk specifically, it is necessary
housing. to adapt the classical approach to evaluate
Despite efforts to improve security and promote the three main components, namely hazard,
peacebuilding, conflict displacement continues exposure and “physical” vulnerability.
to increase in many countries in Africa. The Different procedures and methodologies to
countries analysed in this project hosted more determine risk are used worldwide through a
than 11 million internally displaced people (IDPs) variety of models and approaches. Their common
as of 31 December 2022.9 This creates a significant aim is to understand the probability of different
spread of sites and camps across the countries, magnitudes of flood characteristics, such as
which were considered as an additional element depth, horizontal extent, velocity and duration,
of housing in the analyses. occurring over an extended period.
Direct impact on livelihoods Estimates can be computed for present and

projected climate conditions, considering various
This element, which goes beyond traditional
emission scenarios. This involves conducting
models, measures the direct impact of a disaster
analysis of meteorological, geological,
on people’s livelihoods, namely damage to crops,
hydrological, hydraulic and topographical
herds, shops, industries and services. Accounting
attributes of watersheds, channels and floodplains
for these additional factors provides a more
to generate comprehensive hazard maps.
comprehensive understanding of disaster-related
Consequently, this methodology facilitates the
impacts that may influence people’s decisions
prediction of water depths for specific locations
to move.
or individual infrastructure elements, across a
Indirect impact on critical facilities, services range of return periods.
and livelihoods The damage assessment is converted into “loss”
Indirect consequences, like the extended absence metrics through the computation of the expected
of vital services, should also be considered. AAD, averaged over many years, and the PMD,
People’s choices to remain in place or relocate a curve showing possible displacement with a
are also dependent on their access to necessary certain probability range.
resources like food, clean water, education and
health facilities. Therefore, the potential indirect From vulnerability to vulnerabilities
damage to critical facilities, public infrastructure While our new methodology follows a relatively
and services must be considered. This enhances standard approach to impact assessment,
the thoroughness of the estimation of disaster it introduces innovative and unconventional
displacement risk because, with these losses, the components designed to tackle the vulnerability
vulnerability of the person and consequently the aspect. The interpretation of the term vulnerability
likelihood of displacement may increase. strongly relies on its context. For instance, in terms
of economic losses within a probabilistic flood
Taken together, these elements provide a more
risk assessment, vulnerability usually pertains to
accurate estimate of the number of people who
the potential physical impact of the flood on
may be displaced by the direct and indirect
structures and infrastructure. In contrast, our novel
impacts of future sudden-onset disasters, such
approach views vulnerability as a multifaceted
as floods.
factor encompassing people’s susceptibility to
Incorporating analysis of these elements into displacement.
disaster displacement risk assessments generates
This perspective necessitates the consideration
improved risk information. This, in turn, informs
of numerous additional factors, and specific
policymaking and the development of effective,
physical vulnerability plays an important role in
localised and tailored strategies to reduce
each of them. Integrating these factors results
people’s and communities’ vulnerability to
in a more comprehensive understanding of
disaster impacts and to address the protection
vulnerability to displacement, where potential
15
Figure 3: Representation of the overall process adopted for developing the assessment of flood
displacement risk.
interactions among the different elements need the fact that they have been displaced because
to be factored in. of livelihood loss, which transforms the previous
equation as follows:
The scale of displacement in an area depends
Probability of having lost
on the probability of people being displaced both house and livelihood
because their homes have been rendered

uninhabitable, plus the probability of them being dtot = dh + dl (1 − P[dh dl])
displaced because of livelihood loss, minus the
probability of them having lost both their home
and livelihood.
The probability of people being displaced
This concept can be expressed in the following
because of livelihood loss is subdivided between
equation:

those who depend on agriculture, pastoralism,
Total number Number of displaced
of displaced people due to loss of livelihoods services or industry.
The conditional probabilities are estimated

dtot = dh + dl − dh∩ dl separately for each sector, based on the number
of people who depend on that specific livelihood
Number of displaced
due to loss of their house and the fraction of people being simultaneously
exposed in the workplace and at home.
The number of people who may lose access to

basic services, such as health and education,
is a factor that heightens vulnerability and is
In this equation, dtot is the total number of people
also quantified. As a first approximation, this
displaced, dh is the number displaced because
additional information is provided separately
of housing loss and dl is the number displaced
and not integrated into the displacement risk
because of livelihood loss.
computation. Integration has been deferred
The probability of people losing both their home because there is insufficient data available to
and livelihood can be expressed as: the probability calibrate this element.
of people being displaced because their home
has been made uninhabitable conditioned to
Exposure and vulnerability model databases; available exposure models; global
data sets on population distribution, settlement
Exposure layers are a fundamental component
identification and land use/land cover; and
in risk modelling. These layers provide spatial
sector-specific data sets.
information about the distribution and
characteristics of assets, populations and National statistical database
infrastructure in a specific area. Vulnerability is
The primary statistical data source considered
also a pivotal concept in risk modelling, playing
was the Census, which provides an official
a central role in understanding the potential
count of population, dwellings and households.
consequences of hazardous events. It refers to
For the countries analysed, official Census data
the susceptibility of individuals, communities,
from the past few years was not available, but
infrastructure and ecosystems to damage or
the following documents, found on the portal
adverse impacts resulting from various hazards.
Humanitarian Data Exchange, provided some
To evaluate the different components of the statistical information on population and gender:10
model, a spatial representation of the population,
subdivided by the various sectors, is required. • Population Estimation Survey for Somalia in
October 2013–March 2014, promoted by the
More specifically:
United Nations Population Fund (UNFPA), which
• To calculate the number of people displaced only had the population estimate at Admin
because of house damage, a spatial raster level 1.11 UNFPA investigated replacement or
data set depicting the distribution of the projection of this data set in 2021
residential population, expressed as the • Subnational Population Statistics for Ethiopia
number of people per cell, is needed. administrative level 0–3 sex and age
• To calculate the number of people displaced disaggregated projected 2022 population
because of livelihood loss, a spatial raster statistics, promoted by UNFPA
distribution of employees in each sector, such • Subnational Population Statistics for Sudan
as agriculture, service and industry, is needed. administrative level 0–1 sex and age
• To calculate the number of people more likely disaggregated projected 2022 population
to be displaced because of a lack of basic statistics, promoted by UNFPA
services, a spatial raster distribution of each
The International Labour Organization (ILO)
service and the people related to them is
database ILOSTAT was also consulted and used
needed.
for specific figures.12 This database encompasses
These population distributions should be a comprehensive range of information related to
characterised by physical vulnerability, against the labour market, including data on employment
which we assess the flood impact, as explained by economic sector, hourly wages, weekly
in the following. working hours, unemployment rates and strikes.
The first two distributions, assets and livelihoods, are The ILO data collection includes data derived
used directly to derive dh and dl, while the third is from National Labour Force Surveys, which can
used as an aggravating element that increases differ from the ILO-modelled estimates. The
the likelihood of people being displaced. ILOSTAT entry page provides access to country-
specific or subject-specific statistics, along with
With all three, whenever we determine that
information on concepts and definitions.
an asset (a home, workplace or service point)
is affected by an event, we can derive the Exposure models
number of people affected and identify them
An existing exposure database has been
as susceptible to displacement.
adopted and analysed to extract essential
To identify the residential distributions and the information for defining the specific requirements
distribution of people dependent on different in the modelling exercise within the study. Derived
livelihoods in Ethiopia, Sudan and Somalia, we from the Global Infrastructure Risk Model and
integrated and processed data from a number Resilience Index (GIRI) project led by Coalition
of different sources, such as: national statistical
for Disaster Resilient Infrastructure, it has been Population 17
refined and developed by the United Nations For the global data sets on population distribution,
Environment Programme’s (UNEP’s) Global several products are available. In this project, two
Resource Information Database (GRID-Geneva), products were considered:
funded by the Internal Displacement Monitoring
Centre (IDMC).13 • the Global Human Settlement (GHS) population
grid (R2023), epoch: 2020, resolution: 100
The Global Exposure Socio-Economic and metres, for Ethiopia16
Building Layer (GESEBL) is an update of the of • the WorldPop product (100-metre resolution)
the existing Global Risk Model developed for the for Somalia and Sudan, called in the following
UN in 2013 for the Global Assessment Report.14 This WorldPop17
is a global data set on exposure that combines
The choice of these layers has followed statistical
population data with country-specific building
and visual comparisons, to define the best
typology, usage and value in urban and rural
dataset that could represent reality. In particular,
settings. The data set uses the number of
statistical representation is adopted for describing
individuals, categorised by socioeconomic
and displaying the data, highlighting anomalies
class, residing in a specific construction style in
and obtainable differences between the two
a defined geographic area as a foundation for
layers.
distributing the exposed economic value of the
building stock.15 The choice of these layers followed statistical and
visual comparisons to define the best data set that
Recent advances in data acquisition methods
could represent reality. Statistical representation
related to building footprints, Earth observation
was adopted for describing and displaying the
using satellite imagery and the appearance
data, highlighting anomalies and obtainable
of alternative data sources have boosted
differences between the two layers.
the potential for and development of novel
techniques to model and quantify exposed assets Anomalies were located through a specific
and people at a more granular level. All the data comparison conducted between the two layers
(socioeconomic, building type and capital stock) and optical satellite images. In the geographic
is provided through the uniform geographical information system (GIS) environment, from the
unit of a reference grid at 1x1 kilometres at the data sets, statistics at different Admin levels (1
equator. The resolution of 1x1 kilometres needs and 2) were performed, as seen in figure 4. These
to be improved to properly match the increased graphs compare the values derived from the
resolution for the hazard proposed in this study data sets (Sum of WorldPop 2020 and Sum of
(90 metres). GHS-Pop2020) with those found in other sources,

such as United Nations projection and Census
Additional global datasets data.
To improve the spatial representation of the In the figure, “counts” represent the sum of
GESEBL in coherence with the hazard, a population at Admin level 1, while “errors”
downscaling procedure was performed using illustrate the relative errors of the two layers
ancillary data, such as global data sets on with the UN projection. In this example, there is
population distribution, which have a good level a substantial difference between the two data
of accuracy. In the procedure this map became sets only for Addis Ababa area. In this region, the
the built-up mask of reference. These global two layers and the optical satellite image have
layers are also essential for obtaining missed been compared in terms of spatial distribution, as
information on the characterisation of exposure in figure 5, and values of each pixel. In this case,
and vulnerability through identifying urban and GHS (red) seems to be more up to date, mapping
rural land. a part of the urbanised area that WorldPop does
not capture. Since checks like this have been
carried out in several places in the region and
GHS has always seemed more consistent with
Admin 1 Population
million
45
Census
40
UN Projected 2022
35
SUM WorldPop 2020
30 SUM GHSL-Pop 2020
25
20
15
10
0
Addis Ababa Afar Amhara Benishangul Dire Dawa Gambela Harari Oromia Sidama SNNP Somali South West Tigray
Gumz Ethiopia
Admin 1 Population - Relative Error to UN projected 2022

%
40
SUM WorldPop 2020
35
SUM GHSL-Pop 2020
30
25
20
15
10
0
Addis Ababa Afar Amhara Benishangul Dire Dawa Gambela Harari Oromia Sidama SNNP Somali South West Tigray
Gumz Ethiopia
Figure 4: Statistical distribution on population in Ethiopia, according to different sources.
Figure 5: Comparison of GHS (red), WorldPop (green) and satellite images in the Addis Ababa Region.
urban development, it has been chosen as a extracted from these primary sources and is
reference layer. presented in Table 1. The table shows, for each
country, the employment-to-population ratio
Land cover and employment by economic and the employment rate by economic activity
activitiy (agricultural, industrial and service sectors).
In terms of global products, as well as the land use/ The values of employment-to-population ratio and
land cover map, we employed the Joint Research the employment rate by industrial and service
Centre’s (JRC’s): Anomaly hot Spots of Agricultural sectors derive from our GESEBL at a resolution of
Production (ASAP), to determine the cropland and 1 kilometre, and they are in line with source as
grassland area; and GHS-SMOD R2022A settlement ILO. The agriculture data is not available in the
layers, application of the Degree of Urbanisation GESEBL data set, so the employment rate by
methodology (stage I) to GHS-POP R2022A and agricultural sector derive from ILO data. While the
GHS-BUILT-S R2022A, multitemporal (1975-2030), to employment rate by cropland and grassland was
identify the rural and urban areas.18, 19 determined by a ratio of area in the land cover
Statistical information pertinent to this study was map of reference, JRC’s ASAP.202122
SUDAN ETHIOPIA SOMALIA 19
Employment (2019) 1.04E+07 (24%) 2.81E+07 (26%) 8.21E+05 (5%)
45% 64% 27%

Agriculture Crop – 67% Crop – 71% Crop – 38%
Graz – 33% Graz – 29% Graz – 62%
Industry 2,920,000 (15%) 7,930,000 (10%) 103,000 (9%)

Services 7,430,000 (40%) 20,200,000 (26%) 718,000 (64%)
Table 1: Statistical data on employment rate.
Sector-specific national dataset Firstly, many settlements are situated in or near

flood-prone areas. Secondly, IDPs’ ability to
By the end of 2022, Sudan, Ethiopia and
move elsewhere may be restricted by ongoing
Somalia were home to 12.2 million IDPs.23 These
violence, among other reasons. More critically,
countries are exposed to a mix of conflict and
some IDPs reside in inadequate shelters or tents,
meteorological shocks every year, which not only
leaving them especially vulnerable to extreme
trigger movements but also lead to protracted
weather conditions.
displacement situations.
Sudan and Somalia were assessed to understand
While the situation of IDPs can be highly volatile,
how settlements of IDPs are exposed to floods.
this study aims to investigate how disasters impact
Conscious that the situation of IDPs is dynamic
both informal and formal settlements of IDPs. This
in these two countries, we decided to take a
exploration is motivated by two primary factors.
snapshot of the location of IDPs and estimate the
percentage of settlements exposed to severe
flooding.
Identification and location of the different

settlements of IDPs in Sudan and Somalia was
put together using two data sets:
IDP settlements at risk • Sudan: IOM-DTM Weekly Displacement

of flooding as result
of a 100-year event" (cm)
Snapshot (3) 8 September 2023, referencing
Less than 25 more than 800 points. A large majority were
25-100 georeferenced as neighbourhoods and
100-200
villages (informal settlements), with only 15
Floods Displacement risk in Ethiopia, Somalia and Sudan

200-500
More than 500
official camps (formal settlements). Before the
IDP settlements military takeover in April 2023, more than 100
as of 2nd quarter 2023
camps were georeferenced.24
• Somalia: Camp Coordination and Camp
Management (CCCM) list of sites of IDPs for
the second quarter 2023, referencing more
than thousands of points.25
MOGADISHU The limitations inherent in this analysis must be

acknowledged. Given the dynamic nature of the
situation and the limitation of input data sets, the
findings of our analysis in Sudan and Somalia are
subject to significant uncertainty. For instance,
information on the location of camps was only
available as a point-based layer (instead of
area-based), which may lead to a potential
Source: Camp Coordination and Camp Management (CCCM)
Figure 6: Somalia, locations of camps of IDP overlaid with a underestimation of displacements.

flooding of an 100 year return period event.
Vulnerability component rainy season. This methodology prevents the
disaggregation of displacement occurrences
The study is based on a physical vulnerability
based on specific locations, hindering a more
analysis. Physical damage (as a percentage) is
granular analysis.
directly derived from vulnerability curves, which
relate the hazard intensity to the expected This comparison of simulated values, computed
damage. In the displacement risk evaluation by the application of our methodology and the
procedure, concerning the home and workplace recorded values, exhibits a good agreement,
for services and industries, we adopted a specific allowing us to validate the proposed procedure
damage threshold, or loss, that could cause and calibrate the threshold on the damage
displacement (see figure 8). This threshold is a curves. A mean value among the countries is
percentage value of building damage that represented in figure 8.
means the place can no longer be used. It has The matching between the curve and the element
been calibrated using empirical data from the is done according to different characteristics,
IDMC Global Internal Displacement Database, depending on the vulnerability library. In this study,
as illustrated in figure 7. It shows a histogram with the Comprehensive Approach to Probabilistic
the values of displacement in terms of AAD for Risk Assessment (CAPRA) physical vulnerability
the different countries considered in the analyses, functions for the built-up area were adopted.26
representing the empirical data from the IDMC They are region-specific vulnerability functions,
Global Internal Displacement Database on the defined for different construction materials (mud,
events: floods, riverine floods and with or without
concrete, masonry and wood, as in figure 8)
flash floods. The period for the selection of the and for different numbers of stories. The building
events was from 2008 to 2016, consistent with typology in our GESEBL are associated to the
the assessments made on flooding, where the CAPRA vulnerability library.
historical discharge series used to create the
Furthermore, in the methodology proposed, the
scenarios ranged from 1979 to 2016.
camps of IDPs have been associated to a high-
Overestimation may be caused by the resolution vulnerability construction type, because of their
of the simulation in Sudan. The area along the Nile provisional nature.
is the most critical area in Sudan. It is where most
In agriculture, the vulnerability function is a
of the population are concentrated and where
binary parameter: “flooded” or “not flooded”. It
people are used to living with and managing
categorises a field as flooded if it cannot support
flood conditions. Buildings built on a higher terrain
agriculture and pastoralism for farmers, leading
and local flood protections are not considered
to displacement. This classification is based on a
at a 90-metre resolution, and this may lead to
water depth of 50 centimetres.
overestimation. Furthermore, the empirical
data from IDMC was recorded for Sudan in a Residential population
single event per year, encompassing the entire
The residential population from the GESEBL
corresponds to the sum of people in each of the
AAD
200,000 four income classes established by the World
Simulated
180,000
IDMC Data Bank (World Bank’s annual World Development
Report [WDR]).27 They are distributed in the
160,000 IDMC Average
without flash
140,000
floods recorded
different building typologies present in the
120,000
100,000
country. The ratio of the population associated
80,000 to a single building typology is evaluated, and
60,000 this value is downloaded to the built-up mask of
40,000
reference (GHS and WorldPop). Figure 9 shows
20,000
the population spatial distribution (at raster level)
0
Ethiopia Sudan Somalia for each country and the values summed at
Figure 7: Simulated AAD for the countries compared Admin 1.28
with empirical data in the IDMC database.
Normalized damage factor
1 21
0.9
triggers displacement
Direct damage that
0.8
0.7
0.6
0.5
0.4
0.3
0.2 Mud
Wood
0.1 Reinforced concrete
Masonry
0
0 0.5 1 1.5 2 2.5 3
Water depth [m]
Figure 8: CAPRA vulnerability curves for some building typologies and a mean
threshold on the damage ratio that causes displacement.
Residential population exposure
Less than 500,000
500,000-1,000,000
1,000,000-5,000,000
5,000,000-7,500,000
More than 7,500,000
Population
KHARTOUM
ADDIS-ABABA
MOGADISHU
Figure 9: Residential population distribution at the reference administrative level for the three countries.
Population from different

SUDAN ETHIOPIA SOMALIA
sourced in MLN
Estimated 41.76 171.31 16.55
UN 2021 45.66 120.28 17.07
Table 2: Population data from different sources.

Population by livelihood Flood hazard assessment
We used different layers to determine the Flood hazard assessment aims to systematically
distribution of people depending on different analyse the likelihood and magnitude of flood
livelihoods. occurrences, considering factors such as
topography, hydrology, climate patterns and land
The non-residents population in the GESEBL
use. This multidisciplinary process incorporates
include workers in industry and service sectors,
geospatial data, hydrological modelling and
students, represented by the sum of people
historical flood records to generate accurate
included in the labour and education sectors.
and comprehensive flood hazard maps, which
From this data, the workers in the service and
serve as essential tools for promoting resilience
industrial sectors have been defined, through the
and minimising the adverse effects of flooding
downscaling procedure on the raster reference
events. To best predict possible flood scenarios,
layer.
we used a chain of climate, hydrological and
For the agricultural sector, we determined the hydraulic models. We applied the “Continuum”
number of workers according to the areas of hydrological model to all basins in Ethiopia, Sudan
cropland and grassland based on the ASAP data and Somalia, at a resolution of 4 kilometres, to
set. obtain streamflow series for each pixel in current
We derived the number of workers per sector and projected climate conditions. We then
from percentage values in table 1. input this information into a simplified hydro-
morphological model, based on the Manning’s
Population exposure by livelihood,

Services
Less than 100,000
100,000-500,000
500,000-1,000,000
1,000,000-2,500,000
More than 2,500,000
KHARTOUM
ADDIS-ABABA
MOGADISHU
Figure 10: Population by livelihood, services

equation, to create hazard maps at a 90-metre pessimistic projected climate conditions, with a 23
resolution for different return periods. Return long-term time horizon.
periods of T=2, T=5, T=10, T=50, T=100, T=200 and To assess risk in large domains, a set of mutually
T=250 years in current and projected climate exclusive and collectively exhaustive possible
conditions were selected to represent the hazard scenarios that may occur in the country
different frequencies and intensities of possible (including the rarest and most catastrophic
events in each basin.29 ones) was synthetically generated, keeping the
On this basis, we developed hazard maps for historical spatial correlation of events.
each country under current, optimistic and
100years return period hazard

(cm)
Less than 50
50-100
100-200
200-400
More than 400
KHARTOUM
ADDIS-ABABA
MOGADISHU
Figure 11: Hazard map of 100-year return period for the three countries.
Flood impact assessment
Presenting results from flood impact assessments number of employees for the considered working
conducted in Ethiopia, Somalia and Sudan, this sectors and the corresponding percentage of
section offers insights into the comprehensive employees located in hazard-prone areas
evaluation of flood vulnerabilities. By analysing (coming from the intersection between the
factors like topography, hydrological patterns, spatial raster distribution of employees in each
and assets and livelihoods exposed, these sector and the flood hazard map for T=1,000
assessments provide a deeper understanding of years).
the potential risks posed by flooding events. The
outcomes contribute valuable information for Box 1: The “risk” of double counting.
informed decision-making, enabling effective Industry
disaster preparedness and mitigation strategies

tailored to each country’s unique situation. op
s
Cr
How did we do it?
e
ic
nt
rv
Se
Re
Emp l oym
Each scenario served as input for evaluating
si
the impact on potentially displaced individuals.
den t i a
g
az in
A value of water depth was assigned to each
l
feature in the exposure model (residential Gr
buildings for residential population, industrial or
service buildings for population working in these Double count
sectors, crop or grazing areas for the employees We present our results, incorporating our double
in agriculture, schools for the students, hospitals, counting procedure. The grey segment, labelled
and camps of internally displaced people [IDPs]). as “double count”, signifies individuals who have
experienced both housing and livelihood loss in the
Damage assessment was conducted by
same scenario. Our intention is to visually depict
integrating the raster exposure model and
the proportion of individuals affected by both
Comprehensive Approach to Probabilistic Risk
displacement risks. When presenting numerical
Assessment (CAPRA) curves, through the Rapid
results, however, we only accounted once for
Analysis and Spatialisation Of Risk (RASOR)
people deemed to have lost their home and
Platform libraries.30 Elements in the built-up raster
livelihood in the same scenario.
layer that exhibited damage beyond a specified
threshold were considered unable to fulfil their
Residential capacity was considered at the
functions (e.g. providing shelter, jobs or supporting
same resolution as the exposure model, and at
crop fields), resulting in displacement.
the beginning of the process, it was assumed
The population associated with these damaged to be equal to the residential population. An
elements were deemed susceptible to iterative procedure of assignment to residential
displacement. Individuals who experienced both areas was implemented, starting from the
housing and livelihood loss in the same scenario residential capacity within the cell itself. The
were counted only once, to avoid duplication. residential capacity was updated within the
The procedure to avoid potential double procedure (reduced) whenever employees of
counting is based on the idea of associating a specific working sector were assigned. If not
workers to both their house and their workplace, all the employees could be assigned within such
to evaluate if both were in hazard-prone areas. a cell because the residential capacity was not
As the assignment could not be performed for sufficient for all the employees, the procedure
each single worker for computational issues, we searched for residual capacity in neighbouring
adopted an aggregation based on a 5-kilometre areas, increasing the radius for the search step
regular grid. In each cell, we evaluated the by step. The assignment stored the percentage
25
of employees with a workplace in a hazard- Ethiopia

prone area. The procedure was repeated until Ethiopia is a country known for its diverse and
all workers were assigned to residential areas. At extensive river systems. Rivers are essential
the end, a set of new products was available at for various purposes, including agriculture,
the same resolution as the exposure model, each hydropower generation, transportation, and
representing the percentage for a specific sector sustaining ecosystems and communities.
of employees living in the area and working in a Ethiopia’s varied topography, ranging from
hazard-prone area. highlands to lowlands, contributes to the diversity
The number of potentially displaced people of its river systems and their significance to the
across various sectors was aggregated at the country’s development and environment. The
relevant administrative level. These impacts were main river basin is the Blue Nile, one of the two
then used to determine the average annual main tributaries of the Nile river, and it originates
displacement (AAD) and probable maximum from Lake Tana in the Ethiopian Highlands. The
displaced (PMD) values. Blue Nile contributes significantly to the overall
flow of the Nile and is a critical water source for
We produced results for Ethiopia, Somalia and
downstream countries.
Sudan from Admin level 1 and aggregated these
at national level.
0
Less than 1,000
1,000 - 5,000
5,000 - 25,000
More than 25,000
ADDIS ABABA ADDIS ABABA ADDIS ABABA
Current AAD Optimistic AAD Pessimistic AAD

38,400 100,600 150,100

Figure 12: AAD in current climate conditions and in long-term projections under optimistic and pessimistic
scenarios in Ethiopia. Results are aggregated at Admin level 1.
Region AAD Current % Male % Female

Tigray
Addis Ababa 50 47% 53%
Afar 1,500 55% 46%
Afar
Amhara
Amhara 4,000 50% 50%
Benishangul Benishangul Gumz 900 51% 49%
Gumz
Dire Dawa 0 51% 50%
Gambela 2,000 52% 48%
Gambela
Harari 10 50% 50%
Oromia 11,300 50% 50%
SNNP
SNNP 4,800 49% 51%
Somali 17,900 53% 47%
Tigray 1,800 49% 51%
Legend
Low Medium low Medium high High
Figure 13: AAD disaggregated by sex.

The analysis of flood displacement risk at a causes displacement ranging from 0.08 per cent
country level shows an AAD value of around to 15 per cent in the different Admin 1, according
38,500 people (considering the double counting to the total population of each region.
effect), corresponding to about 0.02 per cent of The disaggregation of PMD at country level
the overall population. The results under different according to the source of displacement shows
climate scenarios highlight a non-negligible that in current climate conditions, almost 42 per
influence of climate change. Both the optimistic cent of displacements are connected to loss
(the scenario closest to the 20th percentile, of housing, while the majority, 58 per cent, are
corresponding to an average temperature connected to loss of livelihoods. People working
rise of about 1°C by 2100) and the pessimistic in the agricultural sector are the most affected.
(the scenario closest to the 80th percentile,
In long-term climate projections, the amount
corresponding to an average temperature rise
of displacements linked to loss of housing
of more than 5°C by 2100) scenarios increase
rises to more than 53 per cent. Regarding the
AAD values, tripling and quadrupling the current
disaggregation of PMD by origin of displacement,
conditions respectively. While climate change
almost 80 per cent of displacements originated
strongly influences the absolute numbers of
in rural areas. This decreases to 75 per cent in
potentially displaced people (a relevant number
projected climate conditions.
of Admin 1 considerably increases the expected
level of displacement), their spatial pattern is Factors such as future population growth and
comparable when considering results in current changes in distribution (e.g. urbanisation), which
climate conditions alongside projected ones. might significantly change these estimates, are
not considered in the modelled future scenarios.
Knowledge of gender inequalities in internal
displacement is limited, yet indispensable to Lack of services is likely not the main driver
propose tailored solutions for internally displaced for displacement after a rapid-onset event.
men, women, boys and girls. Internal displacement Nevertheless, the combined criticality of loss of
generally amplifies pre-existing vulnerabilities and housing, work or both alongside a general lack of
inequalities. As women across the world are, on services in the area can be a worsening factor.
average, economically, legally, politically and In Ethiopia, there are about 23 million students
socially less empowered than men, internally (around 20 per cent of the total population).
displaced women are twice as disadvantaged. The areas characterised by high values of
Figure 13 presents the disaggregation of AAD in displacement linked to loss of housing service or
current climate con.ditions at Admin level 1 by loss of the possibility to work are also characterised
sex (male and female) by non-negligible disruption of the education
While the AAD expresses an average number Displacements
of expected flood-induced displacements, it 450,000
Current
tends to hide potential outliers. For this reason, it 400,000
Optimistic
Pessimistic
is useful to compare the PMD curves for present 350,000

Possible scenario
and projected conditions (see figure 14), which

300,000
show the number of potentially displaced people
in connection with frequent (low return periods) 250,000
or rare (high return periods) events. 200,000
In Ethiopia, displacements would increase 150,000
significantly for higher return periods. For instance, 100,000
observing the current climate conditions, the

50,000
displacements triple from frequent events (e.g.
T=5 years) to rare events (e.g. T=250 years). 0
0 50 100 150 200 250
Years
This trend is also confirmed when analysing the
PMD associated to long-term projected climate Figure 14: PMD curves for current climate conditions
and long-term projections under both optimistic
conditions, for both optimistic and pessimistic
and pessimistic scenarios.
scenarios. In this latter case, a 250-year event
Industry
27
e
r vic
Se
Industry
s
op
nt
Cr
p s
Emp l o y me
ro
Re
C
Industry
s i den t i a l
e
ic
nt
rv
Se
e
e
vic
Re
Emp l oym
g
r
Se
si
zin
t
s
den t i a
en
p
Graz i ng Cro
a
y
Re
az in
Emp l o
Gr
sidential
l
Gr Double count
Double count Double count
Total Current Optimistic Pessimistic
Displacements 69,400 169,500 334,400
Figure 15 Displacements for a 100-year return period flood event (PMD at 100yrp) disaggregated by
cause of displacement.
Urban displacements
Rural displacements
26%
31%
38% 74%
69%
62%

Displacements 69,400 169,500 334,400

Figure 16: Origin of displacements for a 100-year flood event (PMD at 100yrp) at country level.
sector connected to flood events. This can be a Displacement of students (100-year event)
0
worsening element that increases vulnerability, Less than 100
leading to a higher displacement propensity. 100 - 500

500 - 1,000
About 16,750 students, accounting for around 1,000 - 12,700
0.008 per cent of the total student population,

would need to relocate because of their inability
to access severely damaged schools in the event ADDIS ABABA
of a 100-year return period event. The Somali
region emerges as the most vulnerable to the
risk of disrupted education services.
Figure 17: Number of students who would lose

education services as a result of a 100-year flood
event.
Sudan Climate change influences the absolute numbers
of potentially displaced people (i.e. a relevant
Sudan is also home to several significant rivers
number of Admin 1 considerably increases the
that play crucial roles in its landscape, economy
expected level of displacement), and their
and the livelihoods of its population. The Nile is the
spatial pattern is coherent among the present
lifeline of Sudan, with the Blue Nile and the White
and future conditions, but in the climate change
Nile converging in the capital city of Khartoum.
projections the number of Admin 1 with relevant
Sudan is often referred to as the “breadbasket
number of displaced grows.
of the Arab world” because of its agricultural
activities, supported by the Nile’s waters. The river Knowledge of gender inequalities in internal
systems in Sudan have shaped its history, culture displacement is limited, yet indispensable to
and development, making them integral to the propose tailored solutions for internally displaced
nation’s identity and progress. men, women, boys and girls. Internal displacement
generally amplifies pre-existing vulnerabilities and
The analysis of flood displacement risk at a country
inequalities. As women across the world are, on
level shows an AAD value of nearly 170,000
average, economically, legally, politically and
people (considering the double counting effect),
socially less empowered than men, internally
corresponding to about 0.3 per cent of the overall
displaced women are twice as disadvantaged.
population. The results highlight climate change’s
Figure 19 presents the disaggregation of AAD in
strong influence. The optimistic and the pessimistic
current climate conditions at Admin level 1 by
scenarios show AAD values 5 to 10 times higher
sex (male and female).
than those in current conditions, respectively.
0
Less than 1,500
1,500 - 5,000
5,000 - 50,000
More than 50,000
KHARTOUM KHARTOUM KHARTOUM

170,000 603,600 1,644,150
scenarios in Sudan. Results are aggregated at Admin level 1.
Al Jazirah 19,800 52% 48%
Al Qadarif 630 50% 50%
Red Sea Blue Nile 2,900 49% 51%
Central Darfur 60 51% 49%
East Darfur 20 48% 52%
Kassala 700 45% 55%
Khartoum 42,000 47% 53%
North Darfur 10 49% 51%
North Darfur Kassala
North Kurdufan 0 52% 48%
North Kurdufan Northern 48,700 49% 51%
Al Red Sea 0 43% 57%
West
River Nile 51%
Qadarif
Darfur 33,200 49%
Central Sennar 19,600 51% 49%
Darfur
South Darfur 30 48% 52%
South East West Sout South Kurdufan 10 51% 49%
Darfur Darfur Kurdufan Kurdufan
West Darfur 90 51% 49%
West Kurdufan 10 51% 49%
White Nile 1,900 51% 49%
Legend

While the AAD expresses an average number projected climate conditions, for both optimistic 29
of expected flood-induced displacements, it and pessimistic scenarios. In this latter case, a
tends to hide potential outliers. For this reason, it 250-year event causes displacement ranging
is useful to compare the PMD curves for present from 0.01 per cent to 25 per cent in the different
and projected conditions (see figure 20), which Admin 1, according to the total population of
show the number of potentially displaced people each region.
in connection with frequent (low return periods) The disaggregation of PMD at country level
or rare (high return periods) events. according to the source of displacement shows
Displacements in Sudan would increase that in current climate conditions, almost 60
significantly for higher return periods. For instance, per cent of displacements are connected to
observing the current climate conditions, the loss of housing. Among the remaining 40 per
displacements may triple when comparing cent connected to loss of livelihoods, people
frequent (e.g. T=5 years) with rare (e.g. T=250 working in the agricultural sector are the most
years) events. This trend is also confirmed when affected. The same distribution is observed in
analysing the PMD associated to long-term long-term climate projections although, in this
case, the amount of displacements linked to loss
Displacements
8,000,000
of livelihood decreases to around 15 per cent.
Regarding the disaggregation of PMD by

7,000,000
origin of the displacement, almost 60 per cent
6,000,000 of displacements originated in rural areas. This
decreases significantly to 39 per cent in projected
5,000,000
climate conditions.
4,000,000
Factors such as future population growth and
3,000,000 changes in distribution (e.g. urbanisation), which
might significantly change these estimates, are
Current
Optimistic
2,000,000 Pessimistic
Possible scenario not considered in the modelled future scenarios.
1,000,000
Lack of services is likely not the main driver
0 for displacement after a rapid-onset event.
0 50 100 150 200 250
Years Nevertheless, the combined criticality of loss
Figure 20: PMD curves for current climate conditions of housing, work or both alongside a general
and long-term projections under both optimistic lack of services in the area could exacerbate
and pessimistic scenarios.
the situation. The areas characterised by high

Industry Grazing
op
Cr
Industry Grazing
en t
e
ym
vic
op
Se r
Cr
o
pl
Re
Em
en t
s i den t i a l
Industry Grazing
ym
e
Se r v i c
Re
o
Emp l
ce
rv i
si
Se
den t i a
t R
en
Emp l oym
es
Crop
ident i a
Double count
l
l
Double count
Double count
Displacements 975,700 3,110,700 6,054,000
Figure 21: Displacements for a 100-year return period flood event (PMD at 100yrp) disaggregated by
cause of diplacement.
Urban displacements
Rural displacements
63%
54%
43% 37%
46%
57%
Displacements 975,700 3,110,700 6,054,000
Figure 22: Origin of people’s displacement for a 100-year flood event (PMD at 100yrp) at country level.
values of displacement linked to homes rendered White Nile, are the most vulnerable to the risk of
uninhabitable or loss of the possibility to work disrupted education services, which can lead to
are also characterised by non-negligible lack of displacement.
access to the educational sector connected to
Conflict and violence left around 3.6 million
flood events. This can contribute to an increase
people living in internal displacement across the
in vulnerability, ultimately resulting in a higher country as of the end of 2022. As political instability
likelihood of displacement. and unresolved intercommunal tensions persist,
Around 50,000 (1.3 per cent) preschool and disaster risk reduction and resilience strategies
secondary students would need to relocate have stalled, leaving IDPs particularly vulnerable
elsewhere because of their inability to access to disasters, especially flooding.
severely damaged schools, if the country suffered Shifting displacement dynamics erupted in
a 100-year event. All the regions crossed by the Sudan in 2023, from intercommunal violence
Nile and its tributaries, the Blue Nile and the to a nationwide conflict. The situation has had
Displacement of students (100-year event) a significant impact on the millions of civilians
0 who are fleeing their homes. All the IDPs in River
Less than 10
10 - 500
Nile state, Northern state and White Nile state
500 - 5,000 have arrived from Khartoum state and most of
5,000 - 13,700
them are seeking shelter with relatives in the host
community. A large majority of IDPs, since the
military takeover in April 2023, are being hosted
in informal settlements in neighbourhoods and
villages. There are only 15 official camps (formal
settlements).31
KHARTOUM The overview of the situation in the summer of
2023 indicates that around 4 per cent of the
sites accommodating more than 250 IDPs are
susceptible to flooding in Sudan. While Gezira
state, near Khartoum, has the highest number
of sites exposed to flooding, the situation in the
states Sennar, Kharthoum, River Nile and Northern,
Figure 23: Number of students who would lose along the Nile river, is particularly serious.
education services as result of a 100-year flood
event. If the country were to experience a 100-year
event under the current climate scenario, around Somalia 31
18 per cent of the existing sites of IDPs in the Somalia, located in the Horn of Africa, has a
Northern state would face severe damage. This relatively limited number of permanent rivers
percentage would increase to almost 60 per cent because of its arid and semi-arid climate. While
under optimistic scenarios and 76 per cent under Somalia’s rivers might not be as extensive as those
pessimistic scenarios. in some other regions, they still play a crucial role
The situation in the River Nile region is even in providing water for irrigation, livestock and
more concerning. Under the optimistic climate local communities. Because of the country’s
scenario, about 43 per cent of the existing sites challenging environmental conditions, these
would be severely damaged, and this number water sources are essential for sustaining life and
could escalate to 80 per cent under pessimistic livelihoods, particularly during times of drought.
scenarios. In the Khartoum region, near the The Juba and Shabelle rivers are some of the
capital, it is anticipated that about 15 per cent most significant rivers in Somalia. They originate
of the existing informal sites would experience in the Ethiopian Highlands and flow southward
severe damage under current climate conditions. through Somalia before emptying into the
This percentage could significantly increase to Indian ocean. Both rivers are prone to seasonal
more than 60 per cent under both optimistic flooding, which can be aggravated by the El
and pessimistic scenarios. In Sennar state, along Niño Southern Oscillation (ENSO). ENSO exerts
the White Nile River, around 5 per cent of the various impacts on Somalia’s climate variability,
existing sites are susceptible to severe damage, leading to increased rainfall and flooding during
a figure that could rise to around 30 per cent El Niño events, while causing droughts during La
under optimistic scenarios and exceed 55 per Niña years.
cent under pessimistic scenarios. The analysis of flood displacement risk at country
We may be potentially underestimating the level shows an AAD value of around 43,400
actual impact because of the constraints of using people (considering the double counting
points (coordinates) instead of camp areas in effect), corresponding to about 0.26 per cent
this analysis. Considering adequacy of shelter of the overall population. The results highlight
for refugees and displaced people, a minimum a non-negligible influence of climate change.
surface area of 45 square metres per person is Both the long-term optimistic and pessimistic
recommended.32 scenarios show AAD values that triple those in
current conditions.
IDPs as of 8 September 2023 IDPs at risk of floods The present spatial risk pattern increases in
Low
Official camps
projected conditions. The most affected

(formal settlements) Medium low
administrative units will increase in projected
Neighbourhood Medium high
and village (informal settlements)
High
conditions because of climate change, along
with the absolute numbers of potentially displaced
people in administrative units already affected in
current conditions.
Knowledge of gender inequalities in internal

displacement is limited, yet indispensable to
KHARTOUM
propose tailored solutions for internally displaced
men, women, boys and girls. Internal displacement
generally amplifies pre-existing vulnerabilities and
inequalities. As women across the world are, on
average, economically, legally, politically and
socially less empowered than men, internally
displaced women are twice as disadvantaged.
Figure 25 presents the disaggregation of AAD in
Figure 24: IDP settlements at risk of flooding as result
of a 100-year event. current climate conditions at Admin level 1 by
sex (male and female).
0
Less than 1,000
1,000 - 5,000
5,000 - 10,000
More than 10,000
MOGADISHU MOGADISHU MOGADISHU

43,400 114,800 125,900
scenarios. Results are aggregated at admin level 1.

Awdal Sanaag Bari Awdal 20 51% 49%
Bakool 20 48% 52%
Woqooyi
Galbeed Sool
Banaadir 1,300 49% 51%
Togdheer Bari 30 49% 51%
Nugaal
Bay 10 51% 49%
Galguduud 80 50% 50%
Mudug
Gedo 2,500 51% 49%
Galguduud
Hiiraan 5,060 51% 49%
Jubbada Dhexe 2,980 51% 49%
Bakool
Jubbada Hoose 9,640 51% 49%
Mudug 120 52% 48%
Bay Nugaal 10 51% 49%
Sanaag 10 51% 49%
Shabeellaha Dhexe 1,030 50% 50%
Shabeellaha Hoose 20,500 50% 50%
Sool 10 49% 51%
Togdheer 40 50% 51%
Woqooyi Galbeed 50 52% 48%
Legend
While the AAD expresses an average number

Displacements of expected flood-induced displacements, it
450,000
tends to hide potential outliers. For this reason, it
400,000
is useful to compare the PMD curves for present
350,000 and projected conditions (see figure 26), which
300,000
show the number of potentially displaced people
in connection with frequent (low return periods)
or rare (high return periods) events.
250,000
200,000
Displacements would increase significantly for
150,000 higher return periods. For instance, observing the
100,000 current climate conditions, the displacements
Current
Optimistic
triple when comparing frequent (e.g. T=5 years)
50,000
Pessimistic
Possible scenario
with rare (e.g. T=250 years) events. This is also
0
0 50 100 150 200 250 confirmed when analysing the PMD associated to
Years
long-term projected climate conditions, for both
Figure 27: PMD curves for current climate conditions
optimistic and pessimistic scenarios. In the latter
and long-term projections under both optimistic
and pessimistic scenarios. case, a 250-year event causes displacement
Grazing
33
Industry
Service
Grazing Industry
Service
p
ro
C
r op
C
me n t
oy
me n t pl
y Em
Grazing Industry p lo
m Double count
Re
E
Re
Service
Double count
s i den t i a l
s i den t i a l
op
nt
me
Cr
oy
Re
pl
Em
s
i den t i a
Double count
l

Displacements 103,700 360,300 405,600
Figure 28: Displacements for a 100-year return period flood event (PMD at 100yrp) disaggregated by
cause of diplacement.
ranging from 0.01 per cent to 7 per cent in displacements originated in rural areas. This
the different Admin 1, according to the total decreases significantly to 59 per cent in projected
population of each region. climate conditions.
The disaggregation of PMD at country level by Factors such as future population growth and
origin of displacement shows that in current changes in distribution (e.g. urbanisation), which
climate conditions, almost 77 per cent of might significantly change these estimates, are
displacements are connected to housing not considered in the modelled future scenarios.
rendered uninhabitable. Among the remaining Lack of services is likely not the main driver
23 per cent connected to loss of livelihood, for displacement after a rapid-onset event.
people working in the agricultural sector are the Nevertheless, the combined criticality of loss
most affected. The same distribution is observed of housing, work or both alongside a general
in long-term climate projections, although in this lack of services in the area can be a worsening
case, displacement linked to loss of livelihood factor. The areas characterised by high values
decreases to around 14 per cent. of displacement linked to loss of housing or loss
Regarding the disaggregation of PMD by of the possibility to work are also characterised

origin of displacement, almost 69 per cent of by a non-negligible lack of access to education,
Urban displacements
Rural displacements
49% 47%
37%
51% 53%
63%

Displacements 103,700 360,300 405,600
Figure 29: Origin of people's displacement for a 100-year flood event (PMD at 100yrp) at country level.
connected to flood events. This can be a Danish Refugee Council, and 14 other internal
worsening element which increases vulnerability, and national organisations.33 The assessment
leading to a higher displacement propensity. referenced more than 1,100 sites, which varied
Around 1,650 (0.05 per cent) students would need from hosting 60,000 people in the biggest site to
to relocate because of their inability to access 180 people in the smallest.
severely damaged schools if the country suffered Although the Bay region contains the most sites
a 100-year event. exposed to flooding, Hiiraan and Mudug are the
As of the end of 2022, nearly 3.9 million people regions where sites could suffer the most. If the
were enduring displacement in Somalia. Over country suffered a 100-year event, under the
the past three decades, conflict and violence optimistic climate scenario, about 5 per cent of
have remained the primary drivers of internal the sites in Mudug would be severely damaged.
displacement in the region. Ongoing clashes This percentage will double under pessimistic
between the military and non-state armed scenarios. In Hiiraan, the situation is even more
groups, notably al-Shabaab, persistently displace severe, with around 16 per cent of sites severely
people from their homes. Many of these displaced damaged under the optimistic climate scenario,
individuals reside in temporary settlements, which and around 38 per cent under pessimistic
can expose them to the added vulnerability of scenarios.
risk of flooding and the subsequent potential for We may potentially be underestimating the
secondary displacement. actual impact because of the constraints of using
The Camp Coordination and Camp Management points (coordinates) instead of camp areas in
(CCCM) – Somalia released an assessment of sites this analysis. Considering adequacy of shelter
of IDPs in July 2023. At this time, about 4.3 million for refugees and displaced people, a minimum
people were living in displacement situations and surface area of 45 square metres per person is
2.2 million were hosted in camps managed by recommended.34
Somali Community Concern, the International
Organisation for Migration (IOM), Acted,
IDPs at risk of floods

Displacement of students Low
(100-year event) Medium low
0 Medium high
Less than 10 High
10 - 500
IDP settlements
500 - 5,000 as of 2nd quarter 2023
5,000 - 8,000
MOGADISHU MOGADISHU
Figure 30: Number of students who would lose Figure 31: IDP settlements at risk of flooding as result
education services as result of a 100-year flood of a 100-year event.
event.
35
UN Photo/Tobin Jones
12 November 2013
Thousands displaced by floods and
conflict near Jowhar, Somalia
Discussion and conclusion
Disaster displacement related to flood events a factor of three for Somalia, four for Ethiopia and
is one of the most important humanitarian and even nine for Sudan.
development challenges we face in the 21st We also conducted an initial analysis of flood
century. The Horn of Africa has experienced risks and their impact on formal and informal
severe drought year after year, but now the sites of internally displaced people (IDPs).
region is also facing the threat of floods. Across Acknowledging that the situation of IDPs can be
Ethiopia, Sudan and Somalia, the rainy season highly volatile and evolve from day to day, we
and sporadic rains in certain areas have caused aimed to emphasise that in certain areas, both
devastating floods, compounding the impacts of formal and informal settlements are situated in
the drought and resulting in significant losses of flood-prone areas. This could further affect IDPs,
livestock, livelihoods and homes. leaving them with no choice but to move again
This report creates and applies a flood as soon as a flood occurs.
displacement risk model that uses a novel method The model’s results can be used to inform national
for vulnerability assessment. The methodology and subnational disaster risk reduction measures,
gauges the potential displacement of individuals identify areas where large numbers of people
as a result of riverine floods, taking into account the could be made homeless by floods, and calculate
potential impact on homes and livelihoods. This evacuation centre capacities and the amount
includes a focus on potential loss of employment of investment needed to support displaced
and, consequently, income. people. In short, the results will allow decision-
We performed a fully probabilistic risk assessment makers to make risk-informed efforts to prevent
based on a modelling chain that integrates displacement happening in the first place and
climatic, hydrological and hydraulic modelling, reduce its impact when it does occur. Through
to estimate impacts on physical assets and this report, we present some recommendations
estimate displacement. The quantification of that can serve as a foundation for developing
risk is expressed in terms of average annual comprehensive policies and strategies to mitigate
displacement (AAD) and probable maximum displacement risks associated with floods and
displacement, computed under current climate protect the rights and wellbeing of affected
conditions and long-term projections according populations.
to optimistic and pessimistic scenarios. The We need to witness tangible steps in terms of:
outputs of this methodology, which was first
• Vulnerable Populations
applied in two small islands of the Pacific, Fiji and
Identify and target vulnerable populations who
Vanuatu, provide insightful information on the
are at higher risk of displacement. Develop
proportions of housing rendered uninhabitable
tailored strategies to protect their rights and
and loss of livelihood, both of which can trigger
wellbeing.
displacements. We present outputs at Admin
level 1 for Ethiopia, Sudan and Somalia, to • Conflict and disaster linkages
highlight where the spatial risk pattern increases Recognise and address the potential linkages
under projected conditions influenced by climate between conflict- and disaster-induced
change, considering optimistic and pessimistic displacement, especially in areas where both
scenarios. risks are prevalent.
The results show that flood displacement risk • Community-based adaptation

may increase under these optimistic scenarios, Support community-based adaptation initiatives
with AAD two to four times higher than current that empower local communities to better
conditions, depending on the country. Under prepare for and respond to floods.
pessimistic scenarios, the risk might increase by
37
• Education and awareness • Climate change mitigation

Educate the public about flood risk, preparedness Address the underlying causes of increased flood
and evacuation plans. Promote awareness and risk, including climate change. Advocate for and
community engagement. implement climate mitigation measures.
• Capacity building Continued investment in data collection and

Build the capacity of local authorities, research will also improve the understanding of
non-governmental organisations and displacement (roots and causes), vulnerabilities
communities to respond effectively to flood- and impacts. Regularly update displacement
related displacement and support recovery flood risk assessments.
efforts. Data availability, technological advances and
• Risk-informed land use planning growing international recognition of the scale
Implement land use planning that considers and increasing risk of disaster displacement mean
flood risk at the most granular resolution possible. the time is right for more and better-coordinated
Avoid construction in high-risk areas and promote action to build on good practices and address the
resilient infrastructure development. Please note challenge of designing effective displacement
that the current resolution of the model presented risk models.
in this report is not suitable for informing land use Further efforts are required to effectively quantify
and urban planning decisions. future hazard risks, especially with the dynamic
• Resilient infrastructure background of evolving “riskscapes”. The
Invest in flood-resistant infrastructure and building effectiveness of our efforts will naturally improve
standards to reduce displacement and damage with the availability of additional data to refine
to property. and calibrate our models. This necessitates a
comprehensive understanding of population and
• Insurance and financial instruments
socioeconomic patterns, along with fluctuations
Promote the use of flood insurance and other
in the frequency and intensity of hazards
financial mechanisms to help individuals and
associated with climate change. To achieve
communities recover from flood-related losses.
this, we must leverage open-access data, refine
• Resettlement

terrain models for greater accuracy and acquire
Develop clear guidelines and policies for the more hydrometeorological data.
planned and dignified resettlement of displaced
With support from the EU, this cutting-edge
populations when necessary.
methodology lays a robust foundation for the
• Early warning systems creation of innovative approaches to assess
Invest in and strengthen early warning systems to displacement risk related to sudden hazards,
provide timely and accurate information to at-risk spanning from the subnational to the global
communities.35 Ensure that these systems are scale. The risk of disaster-induced displacement
accessible and understandable for vulnerable is a global reality, present in every country. Now is
populations. the opportune time to demonstrate our collective
• Social safety nets commitment to the principle of leaving no one
Establish or strengthen social safety nets to behind, particularly those already enduring
provide support to displaced individuals and the challenges of protracted and repeated
families, including access to shelter, food and displacement.
medical care.
39
Drought
Displacement risk
in Ethiopia, Somalia Drougt Displacement risk in Ethiopia, Somalia and Sudan
and Sudan
Abstract
The report highlights that drought has led to annual displacement and probable maximum
nearly five million global internal displacements displacement. The analysis reveals that while
since 2017, with Africa accounting for about 90 the risk of flood-induced displacement may
per cent of these over six years. Focused on the significantly increase in certain areas, the impact
Horn of Africa (Sudan, Ethiopia and Somalia) of droughts may remain stable or even decrease
and developed under Work Package 3 of the in others, a phenomenon known as the East Africa
HABITABLE project, the model aims to estimate Climate Paradox. For example, in Sudan, the risk
future internal displacements and support policy of drought-induced displacement may slightly
formulation. The methodology involves quantifying decrease under both optimistic and pessimistic
the link between drought hazard drivers and scenarios. In Ethiopia, however, this risk may
observed internal displacement, projecting triple under optimistic scenarios and decrease
changes under climate-induced variations. by a factor of six under pessimistic scenarios.
Modelling drought-driven displacement employs Somalia, meanwhile, demonstrates an increase
an impact-based approach with machine in displacement risk under both scenarios, with a
learning models, particularly decision trees. It 50 per cent increase under the pessimistic one.
establishes a quantitative relationship between The outputs can inform national and subnational
observed impacts and drought hazards to disaster risk reduction measures, helping identify
understand system and sector vulnerability. geographical and socioeconomic landscapes
Using advanced technologies, the study assesses prone to drought displacements.
displacement risk using metrics such as average
Keywords
Displacement Risk, Drought, Probable Maximum Displacement, Average Annual Displacement, East
Africa Climate Paradox
Introduction 41
Every year, floods, droughts, storms, earthquakes empirically calibrated estimations of future human
and other natural hazards force millions of people mobility caused by drought hazards as the effects
to leave their homes, a level of displacement of climate change deepen in Ethiopia, Sudan
greater than that associated with conflict and and Somalia. This report accompanies the one
other forms of violence. Leaving home can be the focusing on flood displacement risk in the same
first of many further disruptions to people’s lives: area. To assess drought displacement risk, we also
they may be forced to move several times once collaborated with knowledgeable partners and
they become displaced, and it can be weeks or explored new technologies, such as machine
even months before they are able to return home. learning, to make progress in this emerging field
Those who do return often face unsafe conditions of research.
and the prospect of being displaced again by Severe and prolonged droughts result in reduced
the next disaster. soil moisture and drinking water reserves, causing
The Internal Displacement Monitoring Centre crop and livestock loss. Such drought disasters are
(IDMC) detected 36.6 million movements as a detrimental to regional economies because they
result of consecutive disasters in 2022. About 6 per produce income losses, threaten food security,
cent, totalling 2.2 million disaster displacements, and worsen conflicts and violence among
were triggered by drought. This represents the communities. Severe droughts affect millions of
highest number ever recorded by IDMC since data people in the Horn of Africa drylands.
was first published on this specific phenomenon Internal displacement must be addressed in the
in 2017. context of drought risk, especially in regions like
More intense or frequent severe weather hazards, Somalia, Ethiopia and Sudan, which are prone to
along with gradual changes in the climate and slow- and fast-onset hazards, with an increasing
environment, already affect many people and impact on human mobility in the region.37
their livelihoods. This trend is expected to worsen This report presents a future drought displacement
under future climate conditions, particularly in risk assessment at the national and subnational
vulnerable regions of the world where livelihoods level for three countries in the Horn of Africa,
are already strained as a result of various climatic, through the application of a new methodology
Drougt Displacement risk in Ethiopia, Somalia and Sudan

political and social factors.36 developed jointly by the International Center
Under the European Commission-funded project for Environmental Monitoring (CIMA Research
HABITABLE, IDMC brings its unique expertise as one Foundation, Italy) and the Vrije Universiteit
of the world’s definitive sources of data and analysis Amsterdam. Tackling drought-related mobility,
on internal displacement. Since its establishment and particularly drought displacement, from a
in 1998 as part of the Norwegian Refugee Council, data-driven perspective is still at an early stage.
IDMC has offered a rigorous, independent and More advanced methodologies have only been
trusted service to the international community. Its applied in this area in the past decade.38
work informs policy and operational decisions that For Work Package 3 of the HABITABLE project,
improve the lives of the millions of people living we reveal the magnitude of displacement risk
in internal displacement or at risk of becoming by comparing present climate conditions with
displaced in the future. The objective of this report various future scenarios, highlighting the potential
is to advance understandings of the current disparities. Given the diverse geographical and
interlinkages between climate impacts and socioeconomic landscapes of Ethiopia, Somalia
migration and displacement patterns, and better and Sudan, it is crucial to recognize how exposure
anticipate their future evolution. to drought hazards may vary and affect different
Under Work Package 3 of the HABITABLE project, regions in these countries.
we support the development of models to produce
Methods
Drought-induced displacement risk social, economic and environmental factors

concepts or processes that increase the susceptibility of
individuals, communities, assets or systems to the
Disaster displacement risk and vulnerability impacts of hazards.41 The vulnerability of humans
The term disaster displacement refers to a situation is influenced by a variety of economic, social,
in which people are forced to flee their homes cultural, environmental, institutional, political
or places of habitual residence because of a and psychological aspects that mould the lives
disaster or to avoid the impact of an impending of individuals and the world around them. Factors
natural hazard. Forced displacements generally including income, education and healthcare
occur when people are exposed to a natural access play a role in determining individuals’
hazard in situations where they are too vulnerable vulnerability levels. Depending on these
and lack the capacity to face the hazard’s elements, vulnerability levels tend to vary across
impacts.39 a community or a population. Vulnerability is not
static. Rather, it changes over time, because it
The UN Office for Disaster Risk Reduction (UNDRR)
comprises several dimensions. The higher the
defines a disaster risk as “the potential loss of life,
level of vulnerability, the greater the probability
injury, or destroyed or damaged assets which
of being harmed by a disaster.
could occur to a system, society or a community
in a specific period of time, determined
probabilistically as a function of hazard, exposure, Drought-induced displacement
vulnerability and capacity”.40 It is important to Drought combines with socioeconomic
consider the social and economic background and governance factors to set the stage for
in which disaster risks occur and understand specific triggers of displacement. These triggers
that people do not necessarily share the same encompass loss of land, loss of livelihood and
perceptions of risk or underlying risk factors. loss of access to essential resources such as food
Like disaster risk, disaster displacement risk is linked and water.
to the nature and magnitude of a given hazard, Soil degradation and rising temperatures
and to people’s exposure and vulnerability. Risk may contribute to nutrient loss and erosion,
analysis reveals the likelihood of future disasters devastating agricultural practices and severely
triggering displacement. Evaluating vulnerability affecting livestock and pastoralist activities. When
levels proves valuable in understanding why severe such factors combine with drought, people’s
events do not consistently yield proportionally livelihoods can become irretrievable. Nomadic
severe consequences for populations and pastoralist and agro-pastoralist communities, for
regions, whereas events of lesser severity can example, find themselves increasingly displaced
result in profound impacts. In other words, it from their traditional territories, relocating to
clarifies why two hazards of similar intensity and nearby small towns and peri-urban areas.
duration could have different consequences in Although there is emerging evidence of these
terms of damage, people affected and people dynamics, it remains largely anecdotal and
displaced. inadequately documented. The absence of
comprehensive data makes drought modelling
Drought disaster risk and vulnerability complex.
Drought disaster risk and vulnerability stem from Sudan, Ethiopia and Somalia were home to 16.6
the interplay of various factors. Vulnerability is million internally displaced people (IDPs) at the
defined as the conditions determined by physical, end of 2023. These countries are exposed to a
43
mix of conflicts and meteorological shocks every and erode their traditional coping strategies (such
year, which not only trigger movements but also as mobility), making livelihoods unsustainable or
lead to protracted displacement situations. unviable. Drought and its impacts are among
the most difficult hazards to monitor, more so
The need for a broader risk assessment than sudden-onset disaster events. Droughts
approach are episodic (time-limited events), complex
Methods for assessing the risk of disaster-induced and multifaceted phenomena that result from
displacement have traditionally determined intricate interactions between natural processes
vulnerability levels by employing hazard and human activities.
thresholds as a substitute for displacement, as Drought-affected mobility is a multicausal and
was done in our 2017 report.42 This implies that if multidimensional phenomenon where no driver
a disaster renders a residence uninhabitable, its can be determined as the single reason for
occupants will encounter displacement, even displacement. Its dynamics emerge, instead, from
if only temporarily. This approach allows for an a combination of socioeconomic, political and
estimation of the potential magnitude of future environmental drivers that interact at different
displacement. Evaluating drought-induced spatiotemporal scales in non-linear ways.44 When
displacement risk, however, presents greater does population displacement become forced
complexity than assessing sudden-onset hazards and unusual? The distinction between voluntary
such as storms, floods or earthquakes. Despite and forced displacement is often difficult to
this complexity, these detailed assessments must discern.45
consider various human factors.
Drought crisis events also involve intrinsic
The direct and indirect impacts of drought pose complexities, such as the duration of the hazard,
significant challenges for populations that depend uncertain onset times, the influence of compound
primarily on natural resources. The impacts effects (e.g. land degradation over the years)
can generate severe disruptions in people’s and climate patterns (e.g. El Niño/Southern
livelihoods, food security, the economy and Oscillation). There is no single definition of drought.
ecosystems43. Drought-induced displacement Researchers have defined several types, such as
occurs when the direct or indirect impacts of

agricultural, hydrological, meteorological and
drought push communities to critical thresholds socioeconomic droughts (see figure 1).
Figure 1: Different types of droughts [NOAA].

As the UNDRR states, the damage and costs data is a primary obstacle to the calibration of
caused by drought are usually underestimated parameters in displacement drought modelling
because of extended and cascading impacts approaches. For the moment, however, only
and are often not attributed to the drought in situ measurements, such as the International
because most impacts are indirect (UNDRR, Organization for Migration’s Displacement
2021). Drought, for example, may trigger food Tracking Matrix (IOM-DTM) in Ethiopia, the
and water security crises in regions where there UN Refugee Agency’s (UNHCR’s) Population
is poor water and land management. Conflict Movement Tracker, and the Protection and
over scarce resources may then arise, generating Return Monitoring Network (PRMN) data in
demographic pressures and exacerbating Somalia, are available as reliable data sources
a drought’s widespread impacts.46 These of drought internal displacement for modelling
processes are intertwined, affecting the social purposes.
and ecological dimensions that may trigger
displacement. Probabilistic risk assessment
In the past few decades, academics and
With increased data availability on the topic,
operational actors have tried to understand and
IDMC compiled data for drought-associated
model drought as a trigger of displacement. Few
displacement for the first time in 2017. An
projects or academic publications, however,
estimated 1.3 million movements were recorded,
have focused on the modelling of drought-
primarily in the Horn of Africa.
induced internal displacement from a data-driven
IDMC has identified drought displacements in 17 perspective.47
countries since 2017, resulting in 4.9 million internal
Recent advances in data acquisition methods
displacements. Notably, almost 80 per cent of
related to displacement monitoring, Earth
these displacements were recorded in Somalia
observation using satellite imagery and the
and Ethiopia (although in the latter case with only
appearance of alternative data sources have
a few data points). Unfortunately, no data was
boosted the potential for and development of
available for Sudan.
novel techniques to model and quantify the
This discrepancy in reporting on drought-induced impacts of events and estimate levels of drought
displacements can be attributed to several displacement.48
factors:
For this study, we combine probabilistic risk
1. Not all drought events result in disasters or assessment from a data-driven perspective using
trigger displacements. machine learning (ML). It is important to regard
2. Because of the slow development and creeping a probabilistic disaster displacement risk profile
nature of a drought’s impacts, determining as a diagnostic tool, because it offers insights
that it has been a trigger of displacement is into potential hazard occurrences and their likely
challenging, as drought can be a direct or consequences. Such profiles cover all possible risk
indirect trigger of internal displacement. scenarios in a certain geographical area, ranging
3. There is often a reactive approach to from low-frequency, high-impact events to high-
monitoring displacement events resulting from frequency, low-impact events.
drought. For this reason, several countries do
Events that have rarely been recorded but might
not engage in regular monitoring of it.
occur more often under climate projections are
4. The magnitude of drought-induced
thus also considered. This feature is particularly
displacement can often be small, involving
useful because climate change is increasing
only a few families, which makes it difficult to
uncertainty about future hazard patterns. To be
track and monitor accurately.
prepared, societies need to calculate the worst
The existing data gap poses a challenge in possible impact. Viewed through this lens, there
assessing long-term impacts and developing is no valid alternative to a probabilistic analysis to
robust models to evaluate short, medium and address such uncertainty in a usable, quantitative
long-term displacement risks. The scarcity of way.
Displacement risk information, expressed in 100-year return period within a century, or the rare 45
average annual displacement (AAD) and possibility of back-to-back events happening in
probable maximum displacement (PMD), is consecutive years. Neither does it eliminate the
calculated at the subnational regions and possibility of an entire century passing without
aggregated at country level, allowing for a such an event occurring.
geographic and quantitative comparison within To assess drought displacement risk for this study
and between countries. These analyses and quantitatively to estimate AAD and PMD, we have
comparison exercises are an important step in used a data-driven approach that correlates the
risk awareness processes and key to pushing for number of displacements with hazard drivers,
risk reduction, adaptation and management along with the probability risk assessment.
mechanisms to be put in place.
The AAD estimate shows the average amount Drought displacement risk
of drought-induced displaced people on yearly
assessment
basis. Most years, displacement numbers are low,
Our novel approach aims to address the
but some years, because of drought conditions,
deficiencies in existing displacement risk models
displacement will be elevated. Thus, the AAD gives
(system dynamics, gravity model, logistic
an indication of the overall size of the problem,
regression and timeseries forecasting models)
averaged over time. The PMD curve illustrates
and provide a more comprehensive assessment,
the probability of a specific scenario leading
by starting from impacts felt on the ground
to an estimated number of displacements. This
(displacement of people) and considering a wide
likelihood is usually measured in terms of return
range of hydrometeorological drought indicators
period.
to capture a variety of ways that drought
A return period is the average time interval in impacts can develop in a region (see figure XX).
years that separates two consecutive events This section describes, step by step, the model
equal to or exceeding the given magnitude. This chain (see figure XX). It starts an assessment of
concept, however, is often misunderstood. The drought hazard followed by vulnerabilities linked
most common misconception is that an event to ecological zones, evaluation of the people
with a 100-year return period will only occur exposed to droughts and observed impacts
once a century, when instead it means that it on displacement. This is all brought together in
has an exceedance probability of 1 in 100, so an ML-based risk-impact assessment model, to
events of the same or greater intensity happen evaluate the annual average and probable
once every 100 years on average. This does not maximum drought-induced displacement.
rule out the occurrence of several events with a
Figure 2: Model chain to evaluate the drought-induced displacement risk.

Drought hazard assessment As drought is a deviation from normal conditions,
it is not straightforward to visualise drought
A drought hazard assessment aims to
indicators. An average of standardised drought
systematically analyse the likelihood and
hazard indices will not show any spatial variability.
magnitude of drought occurrences, considering
Nevertheless, mapping how often (monthly)
different climate variables.
indicators are below a certain threshold value
Drought indices are frequently used in drought illustrates how variable the hydrometeorological
assessments to characterise hazard conditions. situation is, and thus gives a first indication of how
Indices are proxies that indicate the abnormality often hydrometeorological drought conditions
of the water availability in the hydrological are expected to occur.
cycle: the atmosphere (meteorological drought
To investigate how climate change may alter the
indices), hydrological system (hydrological
future frequency and intensity of extreme severe
drought indices) and the soil (agricultural drought
weather events, within exceedance probabilities
indices).
and return periods, our assessment also considers
In this study, drought indices are computed different timescales:
based on:
• under current climate conditions, using
• root-zone and surface soil moisture, potential observations from 1981 to 2022
evaporation from the Global Land Evaporation • under long-term projected conditions from
Amsterdam Model (GLEAM, 0.25-degree 2051 to 2092
resolution, Martens et al., 2017; Miralles et al.,
To capture the spread of possible future climate
2011)
scenarios, we compared 15 realisations of five
• precipitation from the Multi-Source Weighted-
models from the Inter-Sectoral Impact Model
Ensemble Precipitation (MSWEP, 0.1-degree
Intercomparison Project (ISIMIP3b), which provides
resolution, Beck et al., 2019)
bias-corrected Coupled Model Intercomparison
• temperature from the Multi-Source Weather
Project Phase 6 (CMIP6) future climate scenarios
(MSWX 0.1-degree resolution, Beck et al., 2022)
for SSP1-RCP2.6, SSP3-RCP7.0 and SSP5-RCP8.5
These physical variables are used to calculate the conditions (besides pre-industrial and historical
following drought indices at various accumulation runs), in terms of temperature and precipitation
times (1, 2, 3, 6, 9, 12 and 24 months): rise in 2016. Temperature and precipitation trends
• SPI: standardised precipitation index (identifies proved to be highly correlated in the models
rainfall deficits) considered, so we referred only to temperature
• SEI: standardised evaporation index (potential, trends to define representative scenarios. From
atmospheric water demand) the 15 realisations (three scenarios and five global
• SPEI: standardised precipitation-evaporation climate models), we selected two:
index (effective rainfall) • optimistic: the scenario closest to the 20th
• SETI: standardised precipitation/evaporation percentile, corresponding to an average
ratio index (aridity, water balance) temperature rise of about 1°C by 2100
• SSMI: standardised soil moisture index (related • pessimistic: the scenario closest to the 80th
to plant root moisture) percentile, corresponding to an average
• SSFI: standardised surface moisture index temperature rise of more than 5°C by 2100
(related to runoff, streamflow)
The climate scenarios were used to force the
• STI: standardised temperature index (identifies
fully distributed, process-based hydrological
heat waves)
model called Continuum.49, 50, 51. The hydrological
All indices are calculated on a monthly temporal processes for all basins in Ethiopia, Sudan
scale from 1980 to 2022 at Admin 2 level. The and Somalia were simulated at a 4-kilometre
inclusion of 2022 is intended to maximise the resolution, obtaining projected daily series of flow,
overlap with the drought-induced internal evapotranspiration and soil moisture, for each
displacement data. pixel in current and projected climate conditions.
Average global annual temperature
[°C]
Temperature increase with respect to baseline
[°C] 47
22
Historical 7
Pessimistic scenario
21 Optimitistic scenario
SSP126 models 6
SSP370 models
20 SSP585 models
5
19
4
18
3
17
2
16
1
15
0
14
-1
13
1980 2000 2020 2040 2060 2080 2100
Years
Figure 3: Model projections of future climate scenarios from ISIMIP3b.
We then employed the widely used “delta- On each of the figures of the three countries,
change” approach,52 which involves using the it is evident that climate change will affect
simulated projected change (anomalies) as a precipitation in a “positive” way (more rain) and
perturbation of observed data.53 temperature in a “negative” way (more heat).

Hence, SPI decreases in projected scenarios,
In figures 4, severe drought hazard conditions
while SPEI, with potential evaporation strongly
is illustrated as the annual likelihood of having
linked to temperature, increases generally in the
the considered drought index (either the
projected climate scenarios.
standardised precipitation index, SPI, or the
standardised precipitation-evaporation index, As a result of the high increases in temperature
SPEI) 1.5 standard deviations below normal for and thus atmospheric water demand, the
three consecutive months. This does not say likelihood of experiencing a “effective
anything about the impacts felt but shows how precipitation” deficit (i.e. atmospheric water
much area is hit by severe droughts and at what supply [rain] minus atmospheric water demand
frequency. [potential evaporation]) becomes almost 100
per cent in Somalia, larger part of Ethiopia and
In figure 4, the drought hazard conditions are
Sudan. This means that the effective precipitation
shown for Somalia, Ethiopia and Sudan under
conditions currently experienced as a severe
current, optimistic and pessimistic climate
drought will become “normal” under unmitigated
scenarios.
anthropogenic climate change.
0
Less than 0.025
0.025 - 0.05
0.05 - 0.075
0.075 - 0.1
More than 0.1
Current SPI Optimistic SPI Pessimistic SPI
0
Less than 0.025
0.025 - 0.05
0.05 - 0.1
0.1 - 1
More than 1
Current SPEI Optimistic SPEI Pessimistic SPEI
0
Less than 0.01
0.01 - 0.025
0.025 - 0.05
0.05 - 0.075
0.075 - 0.1
More than 0.1
Current SPI Optimistic SPI Pessimistic SPI
0
Less than 0.1
0.1 - 0.25
0.25 - 0.5
0.5 - 0.75
0.75 - 1
More than 1
Current SPEI Optimistic SPEI Pessimistic SPEI
Figure 4: Annual likelihood of having the standardised precipitation index, SPI (left) or the standardised
precipitation-evaporation index, SPEI (right) 1.5 standard deviations below normal, for three consecutive
months.
0
Less than 0.025 49
0.025 - 0.05
0.05 - 0.075
0.075 - 0.1
0.1 - 1.125
More than 1.125 KHARTOUM KHARTOUM KHARTOUM
SPI current SSP 126 SSP 585
0
Less than 0.4
04 - 0.5
0.5 - 0.6
0.6 - 0.7
0.7 - 0.8 KHARTOUM KHARTOUM KHARTOUM
More than 0.8
SPEI current SPEI 126 SPEI 585

Thus, even though current trends show the Africa Climate Paradox (Rowell et al 2015) and is
opposite, the climate model used in this study visible as a decreasing likelihood of severe rainfall
projects an increase in precipitation under global deficits in our drought hazard results.
warming. This is a phenomenon known as the East
Box 1: East Africa Climate Paradox.

In climate science, significant research efforts are underway to develop climate models that deepen our
understanding of past climate conditions and anticipate future climate scenarios. These models play a crucial
role in climate research, allowing scientists to simulate various climate scenarios and assess their potential
impacts. The recent report by the Intergovernmental Panel on Climate Change is a notable example of research
that integrates climate modelling to better support the “state of knowledge about possible climate futures”.
In East Africa, existing climate models suggest an increase in precipitation during the annual long rainy season,
which typically spans from March to May. Projections from these models indicate that this trend will persist
throughout the region until the end of the century. The actual experience of people living in this region, however,
contradicts these projections. Since the 1980s, there has been a noticeable decline in the long rainy season.
These alterations in rainfall patterns pose significant challenges to the local population. This disparity between
projected outcomes and observed trends is commonly referred to as the East African Climate Paradox54, 55, 56.
Under Work Package 3 of the HABITABLE project, IDMC has also developed a report on flood displacement
risk in Ethiopia, Somalia and Sudan. The methodology used incorporates a unique vulnerability assessment,
considering factors often omitted in standard risk models, such as direct impacts on houses, livelihoods, and
critical facilities and services. The assessment uses a probabilistic approach, integrating climatic, hydrological
and hydraulic modelling, to estimate the impacts triggering displacement. The results, evaluated under current
and future climate conditions with optimistic and pessimistic scenarios, indicate a potential increase in AAD
of two to four times compared with current conditions, with even higher risks for pessimistic scenarios, such as
a ninefold increase for Sudan.
These results underscore the significance of understanding the diverse dynamics associated with climate change
impacts across different regions. Both analyses, focusing on droughts and floods respectively, demonstrate
that while the risk of displacement linked to floods may increase drastically in certain regions, the impact of
droughts may remain stable or even decrease in certain areas.
can hinder accurate vulnerability assessments.

Vulnerabilities for drought and how to assess it
Moreover, drought vulnerability is often
In risk modelling, it is crucial to recognise that connected to other risks, such as food security,
risk deviates from vulnerability, highlighting the water scarcity and health issues. Integrating
dynamic interplay of various factors influencing these interdependencies into a comprehensive
the potential impact and likelihood of adverse model adds complexity to the risk equation.
events. Incorporating vulnerability into drought risk Additionally, vulnerability varies across different
equations presents several challenges, however. regions and over time. Developing a model
Vulnerability is a complex concept influenced by that accommodates this spatial and temporal
various physical, social and economic factors. variability requires sophisticated methodologies
Capturing this complexity and integrating it and dynamic data inputs.
into a quantifiable model poses significant In our modelling exercises, vulnerability is
challenges, including that of data availability. statistically derived as the link between
Reliable data on vulnerability factors, such specific drought drivers (combinations of
as socioeconomic conditions, infrastructure hydrometeorological indicators and observed
resilience and community preparedness, may be impact) and displacement, in Admin 2-level
lacking or incomplete. Inconsistent or sparse data
regions for Somalia, where comprehensive data For absolute population numbers for the years 51
was collected over time. Thus, region-specific 2021 and 2022, the population was estimated by
socioeconomic and environmental challenges extrapolating the population increase from 2016
are included as the sensitivity of regions to to 2020. The specific population data of 2022 was
drought triggers. As such, no a priori assumptions used for Somalia to convert absolute numbers of
need to be made about vulnerability; it is instead displacements to relative numbers. This allowed
represented by the derived link from the data us to pool and compare different regions. For
analysis. Ethiopia and Sudan, no model training was done
because of the lack of observed displacement
Exposure layers to assess drought data, and the WorldPop population for 2020 was
displacement risk used as a baseline to derive absolute numbers
Population data constitutes the exposure layer, from relative displacement percentages.
a fundamental component in risk modelling. In our climate change scenarios, we did not
Exposure layers provide spatial information about consider changes in exposure and vulnerability
the distribution and characteristics of populations between current and future climate conditions.
within a specific area. Nonetheless, it is worth noting that factors such as
Population data was retrieved from WorldPop population growth and distribution may greatly
202057, which rescales to match the corresponding alter the future “riskscape”.
UN Population Division estimates58.
Residential population exposure

Less than 500,000
500,000-1,000,000
1,000,000-5,000,000
5,000,000-7,500,000
More than 7,500,000
Population
KHARTOUM
ADDIS-ABABA
MOGADISHU
Figure 5: WorldPop and Center for International Earth Science Information Network (CIESIN) (2018).
Global High Resolution Population Denominators Project.
Observed data on internal displacement about people displaced by drought has been
from drought gathered through people from affected
households seeking assistance with shelter and
In this study, we investigated drought displacement
non-food items. Families seeking refuge with
in the Northern Horn of Africa region, more
relatives were not tracked by these assessments.
specifically in Sudan, Ethiopia and Somalia. For
Ethiopia and Somalia, national-level figures on Compared with Ethiopia, Somalia offers
drought-induced internal displacements were additional, highly detailed data, provided by
accessible for five years, from 2017 to 2022, the PRMN of UNHCR.62 This data set records flows
through the IDMC database (GIDD).59 of movements and contains monthly numbers
of displaced people, by departure location, on
While displacements were recorded in Somalia
Admin 2 level in Somalia. The identification and
and Ethiopia (although in the latter case with
location of the different settlements of IDPs in
only a few data points), unfortunately, no data
Somalia has been considered using the Camp
was available for Sudan, particularly regarding
Coordination and Camp Management (CCCM)
drought displacement.
list of sites of IDPs for the second quarter of 2023,
Sudan is known to be exposed to recurrent drought referencing more than thousands of points.63 A
episodes. According to the EM-DAT database, large advantage of this data is that the reason
four major drought events have occurred there for displacement is classified by hazard, which
since 2000, affecting millions of Sudanese6061. made it possible to select only drought-related
There is no evidence of displacement from displacements. This data set also covers a longer
these events, however, apart from anecdotal period, from 2016 to 2022, with solid information
information. For example, the most recent drought on drought-induced internal displacements.
in 2022 affected 11 million people, but there is no
To evaluate the effect of droughts on population
data confirming the number of displacements.
movements, figures were normalized by dividing
In Ethiopia, data points are available at Admin 1 the monthly number of drought-related IDPs by
level through rigorous data collection efforts by the population count per Admin 2 level from
IOM-DTM. In Ethiopia and Somalia, information WorldPop (2018). This makes the number of
Monthly IDPs, relative to population

(%)
3 30
Somalia Total
Taleex
2.5 Qansax Dheere 25
2 20
1.5 15
1 10
0.5 5
0 0
May-16 Sep-17 Feb-19 Jun-20 Oct-21 Mar-23
Figure 6: Numbers for 2016 to 2022. The line for the whole of Somalia shows that there are two large
peaks, one in early 2017 and one in early 2022. The peak in 2017 was mainly because of IDPs in the north
of Somalia (including Taleex). The peak in early 2022 was more concentrated in the centre of Somalia
(including Qansax Dheere).
drought-induced IDPs more comparable over 15 categories of displacement level, between 0 53
regions and time. Then, 15 different categories, and 5 per cent. The impact categories were then
ranging from 0 per cent to 5 per cent drought- converted to a binary classification: exceeding
induced IDPs, were defined. the displacement level or not. This resulted in an
impact record per severity category (i.e. IDPs
Figure 6 illustrates a timeline of the average
exceeding 1 per cent, 2 per cent, etc. of the
relative IDPs for two provinces across Somalia.
population) per month per region.
This indicates that the year 2017 registered a high
amount of drought-related IDPs, especially in the The third step involved calibration of the ML
provinces of Bay and Sool. No clear seasonality in models. This was done for each displacement
population movements can be seen in the data level (the 15 categories) for each region:
set. This gives confidence that the results are not • We extracted hazard data covering the
related to seasonal migration of, for instance, duration of displacement observations. This
pastoralists. data was then bootstrapped 100 times,
ensuring an even distribution of impactful and
Drought-driven displacement modelling
non-impactful events between training and
method
testing data sets. We shuffled the data each
The method we used for modelling drought- time before splitting it, using a function called
driven displacement involved employing an StratifiedShuffleSplit from the sklearn Python
impact-based approach coupled with ML package. Following this, we proceeded with
models, specifically decision trees. This approach the analysis for each of these sampled data
establishes a quantitative relationship between sets:
observed impacts and drought hazards,
facilitating insights into the vulnerability of . For the train data, the hyperparameters
“selection criterion” (“gini”, “entropy”,
affected systems and sectors. By discerning
“log_loss”), “splitter” (“best”, “random”)
which combinations of drought drivers lead to
and “depth” (2, 3) were tuned. This was
increased displacement in which regions, this
achieved by doing a grid search to identify
method enabled the estimation of associated
the specific ML model settings resulting in the
risk under various climatic conditions, including
most balanced accuracy.
present and future scenarios.
. A decision tree algorithm (from sklearn
The methodology consisted of four steps as Python package) was then used with these
follows. hyperparameters to identify hazardous
The first step entailed the computation of drought conditions likely to lead to increased

hazard indices. displacement.
Various physical drivers for drought were . Accuracy, balanced accuracy and
precision were calculated based on the
evaluated: precipitation, potential evaporation
test data set.
and soil moisture, as presented in the drought
hazard assessment section. For each of these . The fitted decision tree was applied on
the full historic data set (not only on the
physical drivers, a multitude of indices (absolute,
bootstrapped data over the observation
relative, standardised) was determined. The
period) and projected climate conditions.65
indices were based on GLEAM data.64
. Finally, by combining estimates from 100
The second step involved the identification of bootstrapped trees, we calculated the
impact data. average of the likelihood of impact per year.
Consistent, continuous data on drought-related The fourth step consisted of estimating the two
impacts is key to calibrating the approaches used main metrics: AAD and PMD. The exceedance
to quantify drought displacement risk. Records probability curve was computed by fitting a
on displacement data, as described in the generalised extreme value (GEV) distribution
“Data on internally displaced people” section, over the yearly values of displacements. The
were normalised with population and split into
integral behind that curve gave the AAD. The administrative regions were clustered based
exceedance probability curve was converted on their environmental vulnerability (water
into the PMD by expressing the rarity in return availability profile as defined by their Köppen-
period (RP) instead of probability (p=1/RP). This Geiger climate zones) to drought risks. Then, the
allowed focus on the tail of the curve, so the link between impact and hazard (decision trees
rarest events with highest return periods. fitted in Somalia) was applied to estimate risk in
regions with similar vulnerability in Ethiopia and
For Somalia, where ample observations about
Sudan.
internal displacement were available, analysis
could be performed at Admin level 2, or district The Somalian climate zone that best reflected
level. This gave more precise results, as it is Ethiopia’s was selected through a recalibration
reasonable to assume that vulnerability (sensitivity and bias correction using the few data points
and coping capacity) to droughts within one available on Admin 1 level, from IOM-DTM. For
district is similar, although it may vary between Sudan, it was assumed that all its regions show
different districts. the same environmental vulnerability as the
steppe regions in Somalia. This was done under
For Ethiopia and Sudan, observed data was not
the assumption that the hazard–impact cause–
sufficient for a direct calibration, as described
consequence relationship is similar within one
in the previous section on observed data,
cluster of regions, and that this relationship is
and extrapolation from the more data-rich
applicable in other countries too.
Somalia was necessary. For this, some Somalian
Photo: FAO/IFAD/WFP/Michael Tewelde
31 August, 2017, Somali Region, Ethiopia
- Food distribution in Gode wereda Dolo
Baad distribution centre.
Drought-induced displacement
In this section, we present the results of our Box 2 shows a spatial comparison between
analysis on drought-driven displacement. Our observed and simulated AADs, respectively in
methodology for modelling drought-induced absolute and relative terms. A similar spatial
displacement involved using an impact-based pattern between simulations and observations
approach based on machine learning models, can be observed, with some overestimations in
particularly decision trees. Through this approach, place. The main deviation can be found in the
we aim to provide insights into the patterns and southern coastal regions, west of Mogadishu
dynamics of displacement triggered by drought around Buur Hakaba district (which has almost
events. The outcomes contribute valuable 10 per cent point deviation). There are various
information for informed decision-making, reasons for this deviation (e.g. under-reporting
enabling effective disaster preparedness and of drought-induced displacements, ongoing
mitigation strategies tailored to each country’s conflicts and violent situations, local anchoring
unique situation. factors, occupation, gaps in data collection).
Nevertheless, because of the overall good
Validation process agreement in terms of spatial pattern and,
Validating the performance of a model is crucial on average, low errors, these results justify the
to ensure its reliability and accuracy in predicting statistical model’s ability to estimate drought-
internal displacement. In this section, we induced displacement.
undertake a validation process by comparing the The scatter plot in figure 8 illustrates a comparison
average annual number of movements observed between the predicted and observed drought
with the corresponding figures simulated by our AAD from the PRMN data set. A good fit is shown
model. This validation was performed in Somalia by the two series with R2 > 0.6, with a median
for the period of 2016 to 2022, when observations difference (over-estimation) of 0.4 (mean
are available at district level. deviation 1.4+-0.2).
The number of potentially displaced people A few outliers exist, with maximum differences
was aggregated at the district level in Somalia. observed in Doolow (predicted AAD of 12 per
These impacts were then used to determine cent while observed only 5 per cent) and Ceel
the average annual displacement (AAD) and Barde (predicted AAD of 15 per cent while
probable maximum displaced (PMD) values, observed 22 per cent), but only 11 of the 74
under two future climate scenarios: optimistic districts had a deviation of more than 2 per cent.
and pessimistic.
57
Box 2: Somalia observed and simulated average annual drought-induced displacements. The
reference period is 2016–2022
Less than 1
1-5
5 - 10
10 - 15
15 - 20
More than 20
MOGADISHU MOGADISHU
OBSERVED CURRENT
Figure 7a: Somalia average annual drought-induced displacements, relative to population (%)(Trained
per province using non-zero observations 2016–2022)
Less than 5,000

5,000 - 10,000
10,000 - 15,000
15,000 - 20,000
20,000 - 25,000

More than 25,000
MOGADISHU MOGADISHU
OBSERVED CURRENT
Figure 7b: Somalia average annual drought-induced displacements (using 2000 population)
Observed vs predicted AAD in Somalia
40
35
30
25
y = 0.8176x
R² = 0.6279
20
15
10
0
0 5 10 15 20 25 30 35 40
Figure 8: Observed versus predicted average AAD in Somalian districts. The reference period is 2016–2022.
Somalia The country is also exposed to the El Niño-Southern

Oscillation (ENSO). ENSO exerts various impacts on
Somalia is a country characterised by a
Somalia’s climate variability, leading to increased
predominantly arid and semi-arid climate.
rainfall and flooding during El Niño events, while
Drought is a recurrent and devastating causing droughts during La Niña years.
phenomenon in Somalia, often leading to
The analysis of drought displacement risk in
widespread food insecurity, water scarcity and
Somalia at a country level shows an AAD value of
displacement of populations. The country’s
around 610,000 people, corresponding to about
dependence on rain-fed agriculture and
3.5 per cent of the overall population. The results
pastoralism makes it particularly vulnerable to
highlight a non-negligible influence of climate
the impacts of drought, because livelihoods are
change. Both the optimistic and pessimistic
heavily reliant on seasonal rainfall patterns.
scenarios show AAD values that increase, by 22
Drought-induced displacement is a common per cent under optimistic and 55 per cent under
occurrence in Somalia, with communities often pessimistic scenarios (see figure 9).
forced to move in search of food, water and
In absolute terms, significant increase is estimated
pasture for their livestock.
Less than 5,000

5,000 - 10,000
10,000 - 15,000
15,000 - 20,000
20,000 - 25,000
More than 25,000
(Trained per province using
non-zero observations 2016–2022)
! MOGADISHU ! MOGADISHU MOGADISHU

610,000 745,000 941,000
Figure 9: AAD in Somalia under current (left), projected optimistic (middle) and projected pessimistic
(right) climate conditions.
Less than 1
1-5 59
5 - 10
10 - 15
15 - 20
More than 20
! MOGADISHU ! MOGADISHU ! MOGADISHU
Current relative AAD Optimistic relative AAD Pessimistic relative AAD
Figure 10: AAD as a percentage of the total population in Somalia, under current (left), projected
optimistic (middle) and projected pessimistic (right) climate conditions.
in the central area (provinces of Hobyo, become more even, with the hotspots observed
Gaalkacyo, Dhuusamarreb and Cadaado), in under the current climate still being high, but
the district of Burco in the north, and the districts not standing out as much. Some of the coastal
of Kurtunwaarey, Sablaale, Diinsoor, Ceel Barde, districts may become the future hotspots of IDPs
Baidoa, Qansax Dheere, Xudur and Jowhar. in Somalia.
In relative terms (compared with the total While the AAD expresses an average number of
population), however, the main increase is expected displacements as a result of drought
particularly large in the north-east and along events, it tends to hide potential outliers. Figure
the coast (see figure 10). The pattern seems to 11 compares the PMD curves for current and
projected conditions, showing the number of
Displacement
(%) potentially displaced people in connection to
8
frequent (low return period) or rare (high return
period) events.
Figure 11 represents the PMD results at country

6 level. Both projected scenarios show a general
increase in the number of displacements for return

periods less than 25 years. For a five-year event,
4.5 per cent of the population are estimated
4
to be displaced in the current conditions. This
number would increase to 5.1 per cent in the
projected optimistic scenario and to 6.2 per
2
cent in the pessimistic scenario. This is mainly
Current because of the increase in temperature and
Optimistic evapotranspiration (see section 2.2.1). Instead,
Pessimistic
Possible scenario in the optimistic scenario, drought impact would
0 slightly decrease for events rarer than 25-year
0 10 20 30 40 50
Return period return period. Uncertainty increases with return
(Years)
period, however.
Figure 11: PMD curves for Somalia. Displacements, as
a function of return period of the event, are expressed
as a percentage of total population. The continuous
black line represents the current climate scenario, while
the continuous red line and the red dotted line depict
optimistic and pessimistic climate scenarios respectively.
Ethiopia There are some drought displacement hotspots
in the eastern, drier regions of Ethiopia (for Afar
Ethiopia experiences a diverse range of climates
region, AAD is 7,500 people, corresponding
as a result of its varied topography, characterised
to 1.4 per cent of the total population of the
by a semi-arid to arid climate with the presence
region; for Somali region, AAD is 37,500 people,
of several climatic zones, including highland,
corresponding to 2 per cent of the total population
lowland and desert regions.
of the region).
The climate variability in Ethiopia is influenced by
As a result of rising rainfall, the risk appears to
several factors, including ENSO. These climate
diminish under climate change, especially in a
phenomena can lead to irregular rainfall patterns,
much hotter and wetter pessimistic scenario.
exacerbating the frequency and severity of
Under optimistic climate scenarios, however, the
drought events in the country.
risk of becoming displaced by drought increases,
Drought is a recurrent and significant challenge in because of drier soil and higher temperatures
Ethiopia, particularly in the lowland and semi-arid in Amhara (eight times higher), Tigray (four
regions. The country’s agricultural sector, which is times higher) and Oromia (three times higher).
largely rain-fed, is highly vulnerable to fluctuations This increase is counteracted by the higher
in rainfall patterns and therefore susceptible to precipitation levels forecasted in the pessimistic
the impacts of drought. scenario (see section 2.2.1), which would lead to
The analysis of drought displacement risk at a decrease in drought-induced displacements.
country level shows an AAD value of around The PMD results at country level show intricated
131,000 people, corresponding to about 0.1 per behaviour, resulting from a complex evolution of
cent of the overall population. The results highlight climate drivers.
a non-negligible influence of climate change.
In the current climate conditions, there is a
Despite the optimistic scenario projecting a
significant difference between rare and frequent
substantial increase in displacement of around
events. A 10-year event would generate a
135 per cent, the risk of drought-induced
number of displacements, corresponding to
displacement decreases significantly under the
around 1.5 per cent of the population, and this
pessimistic scenario by about -85 per cent, mainly
number doubles if we consider a 50-year event.
because of the East Africa Climate Paradox (see
In both projected scenarios, PMD curves are less
box 1).
steep, with reduced differences between high
Figure 12 and figure 13 show AAD in Ethiopia and low return periods, in comparison with the
under current climate conditions and future current situation.
projected optimistic and pessimistic conditions,
The pessimistic scenario shows a general reduction
in terms of both absolute amount of people and
in the number of displacements, except for very
percentage of the population.
0
Less than 500
500 - 1,000
1,000 - 5,000
5,000 - 10,000
10,000 - 20,000
More than 20,000

131,000 308,000 21,000
Figure 12: AAD in Ethiopia under current (left), projected optimistic (middle) and projected pessimistic
61
0
Less than 1
1-5
5 - 10
10 - 15
15 - 20 ADDIS ABABA ADDIS ABABA ADDIS ABABA
More than 20

Figure 13: AAD as a percentage of the total population in Ethiopia, under current (left), projected
low return periods. This is mainly because of Sudan

the projected increase in precipitation. In the Sudan experiences a diverse range of climates
optimistic scenario, meanwhile drought impact that vary across different regions of the country.
would increase for frequent events up to a It is generally characterised as arid and semi-arid,
25-year return period (this is in line with the slight however, with some areas experiencing desert
increase in absolute AAD for this scenario) and conditions. Overall, Sudan’s climate is largely
decrease for rarer events (higher return periods). influenced by its location in the arid and semi-arid
Displacement belt of Africa, with varying levels of rainfall and
(%)
4 temperature across different regions. Droughts,
water scarcity and desertification are significant
challenges faced by the country as a result of its
3 climatic conditions.
Although Sudan lacks empirical data about

drought internal displacement, millions of
2 Sudanese are regularly affected by drought
episodes.
Figure 15 and figure 16 show the results of AAD

1
for drought in Sudan under current climate
Current
Optimistic conditions and future projected optimistic and
Pessimistic
Possible scenario
pessimistic conditions.
0

0 10 20 30 40 50 The analysis of drought displacement risk at
Return period
(Years) country level shows an AAD value of around
Figure 14: PMD curves for Ethiopia. Displacements, as 900,000 people, corresponding to about 2 per
function of return period of the event, are expressed cent of the overall population. The results highlight
a non-negligible influence of climate change.
the continuous red line and the red dotted line depict Both the optimistic and pessimistic scenarios
optimistic and pessimistic climate scenarios respectively. project a slight decrease by about -5 per cent
and -9 per cent respectively, mainly because of
the East Africa Climate Paradox (see box 1).
Indeed, because of the projected increases in

precipitation, the number of drought-induced
displacements are expected to reduce in some
Less than 5,000
5,000 - 10,000
10,000 - 50,000
50,000 - 100,000
100,000 - 150,000
More than 150,000
(Trained per province KHARTOUM KHARTOUM KHARTOUM
using non-zero
observations
2016–2022)

901,000 852,000 817,000
Figure 15: AAD in Sudan under current (left), projected optimistic (middle) and projected pessimistic
regions under both optimistic and pessimistic under both optimistic and pessimistic scenarios.
climate change scenarios. In absolute terms, a This further underscores the projected increases
lower number of displacements is also expected in precipitation in different regions under varying
nationwide under such climate change. climate scenarios.
Some differences emerge at the subnational All curves in figure 17 present approximately the
level, however. Conditions in Gazira state same trend, with reduced differences between
could improve under the pessimistic scenario, high and low return periods. In alignment with
with a decrease of 13 per cent in drought the decrease in AAD projected under future
displacement risk. On the other hand, while climate conditions, both optimistic and pessimistic
drought displacement risk remains stable under scenarios indicate a slight reduction in the
both optimistic and pessimistic scenarios in all the number of displacements, primarily attributed
northern provinces of the country, provinces in to the anticipated increase in precipitation.
the south (such as South and East Darfur, South Specifically, in the current climate conditions,
Kordofan, and Blue Nile) could experience an a 10-year event would displace 2.4 per cent of
increase in drought-induced displacement. For the total population. This figure decreases to 2.2
instance, South Kordofan could have an increase per cent and 2 per cent in the optimistic and
of around 55 per cent under pessimistic scenarios, pessimistic climate scenarios, respectively.
while Blue Nile could experience a similar increase
Less than 0.01

0.01 - 0.0125
0.0125 - 0.015
0.015 - 0.0175
0.0175 - 0.02
More than 0.02
(Trained per province KHARTOUM KHARTOUM KHARTOUM
using non-zero
observations
2016–2022)

Figure 16: AAD as a percentage of the total population in Somalia, under current (left), projected
Displacement
(%) 63
4
1
Current
Optimistic
Pessimistic
Possible scenario
0
0 10 20 30 40 50
Return period
(Years)
Figure 17: PMD curves for Sudan. Displacements, as

function of return period of the event, are expressed
the continuous red line and the red dotted line depict
optimistic and pessimistic climate scenarios respectively.

Discussion and conclusion
The risks associated with drought extend far and hazard information available.
beyond measurable impacts, encompassing The data-driven, impact-based drought disaster
communities, ecosystems and economies. risk analysis, providing a probabilistic estimate
Vulnerable populations bear a disproportionate of drought-induced displacement, shows
burden of these consequences. Yet, a clear connection between droughts and
with an improved understanding of the reported numbers of internally displaced people
interconnectedness of global drought and other (IDPs). Indeed, for Somalia, where plenty of
complex risks, an opportunity arises for enacting observations on IDPs’ region of origin and cause
the necessary changes to reduce risks and of displacement were available, the model
enhance drought resilience. performance was good, proving reliable risk
The Horn of Africa has been suffering its worst assessments under current climate conditions.
drought in 40 years since October 2020, with Given the climate models and projections for
extended dry conditions punctuated by short precipitation and temperature under climate
intense rainfall that has often led to flash flooding. scenarios, the future risk assessment can also be
Across Ethiopia, Sudan and Somalia, the rainy seen as robust.
season and sporadic rains in certain areas have Here, the East African Climate Paradox needs to be
given rise to devastating floods, aggravating the considered, however, because in general, climate
impacts of the drought and resulting in significant models are challenged to make predictions in
losses of livestock, livelihoods and homes. There this region, and the paradox may influence input
have been five consecutive seasons of rainfall drought indicators for future scenarios. Moreover,
below normal levels. Millions of people have neither changes in socioeconomic and political
needed humanitarian assistance and have been conditions, affecting vulnerability, nor changes
at risk of acute food insecurity and potentially in demographics, altering exposure, were taken
famine. into consideration in the future risk estimations.
This report creates and applies a drought These two risk factors are dynamic but would not
displacement risk model that involves a have resulted in more robust estimates, given the
novel method, modelling drought-induced uncertainty in how they might change, which is
displacement using an impact-based approach influenced by climate and other risks and global
based on machine learning (ML) models, trends.
particularly decision trees. The 20 data points on yearly scale for regions
ML models are trained to learn the relationship in Ethiopia provided enough information to
between impact (drought-induced recalibrate the Somalian model for this country.
displacement) and drought hazard indicators or This helped in evaluating which type of training
indices per region, and this relationship was used (based on the Somali situation in tropical,
to estimate drought displacement risk (in terms steppe or desert Köppen climate zones) best fit
of likelihood of experiencing certain impacts). the Ethiopian case. As no validation data was
This was summarised in the average annual available for Sudan, the data-driven method
displacement risk metric, indicating the estimated could only be extrapolated to this region. It
average annual drought-induced displacement was assumed that similar hydrometeorological
and probable maximum displacement, or the conditions would lead to similar (relative)
probability (expressed in return period) to exceed responses in internal displacement, but future
a certain displacement amount. As a data- work with more data should verify this.
driven approach, the accuracy of estimated risk The results indicate that in Somalia, drought
depends strongly on the quality of the impact displacement risk may increase slightly under
both optimistic and pessimistic scenarios, by • Social protection programmes 65
22 per cent and 54 per cent respectively, in Implementing social protection programmes, such
comparison with current conditions. Conversely, as cash transfers, food assistance and livelihood
in Sudan, there might be a decrease in risk by support, can help vulnerable populations cope
5 per cent to 9 per cent under both scenarios. with the impacts of drought and reduce the
Ethiopia exemplifies the East African Climate likelihood of displacement.
Paradox even more starkly, with the risk of drought • Sustainable water management
displacement potentially doubling under the Investing in sustainable water management
optimistic scenario, while dropping to one-sixth practices, such as water conservation, rainwater
of current levels under the pessimistic scenario. harvesting and improved irrigation techniques,
The model’s results have extensive utility, can help mitigate the impact of droughts
informing both national and subnational on water availability and reduce population
disaster risk reduction measures. They enable displacement.
the identification of areas where drought • Early warning systems
could potentially push people to flee and help Efforts should be strengthened following the
determine the necessary investment to support release of the 2023 Global Status of Multi-
displaced populations. Hazard Early Warning Systems report at COP28.
Essentially, these results empower decision-makers Implementing effective early warning systems
to adopt risk-informed approaches to prevent to alert communities and authorities about
displacement and mitigate its impacts. Through impending drought conditions can help minimise
this report, we aim to present recommendations displacement by enabling timely preparedness
that can serve as a foundation for developing and response measures.
comprehensive policies and strategies geared Continued investment in data collection and
towards mitigating displacement risks associated research will also improve the understanding of
with drought, while safeguarding the rights and displacement (roots and causes), vulnerabilities
wellbeing of affected populations. and impacts. Further research could also include
We need to witness tangible steps in terms of: conflict, antecedent food security, long-term
desertification trends, access to roads, cities
• Vulnerable Populations
and markets, and other vulnerability proxies to
Identify and target vulnerable populations who
improve the ML-based method.
are at higher risk of displacement. Develop
tailored strategies to protect their rights and With support from the EU, this cutting-edge
wellbeing. methodology establishes a solid foundation for

developing innovative approaches to assess
• Conflict and disaster
displacement risk associated with drought,
Recognise and address the potential links
ranging from the subnational to the global scale.
between conflict- and disaster-induced
The risk of disaster-induced displacement is a
displacement, especially in areas where both
global reality, affecting every country. Now is the
risks are prevalent.
opportune time to demonstrate our collective
• Community-based adaptation commitment to the principle of leaving no one
Support community-based adaptation initiatives behind, particularly those who are already
that empower local communities to better enduring the challenges of protracted and
prepare for and respond to drought. repeated displacement.
• Livelihood diversification Given the significant improvements in data
Promoting livelihood diversification strategies, recording, it is expected that Admin level 2 data
such as alternative income-generating activities on a yearly time scale will be available soon, and
and resilient agricultural practices, can help then vulnerability dynamics can be included, too.
communities reduce their vulnerability to drought-
related displacement by providing alternative
sources of income and food security.
References
1
IDMC, “Global Report on Internal Displacement”, 2023
2
IDMC, “Global Disaster Displacement Risk - A baseline for future work”, 2017
3
IDMC, “Flood displacement risk: Assessment for Fiji and Vanuatu in current and future climate
scenarios”, 2022
4
UNDRR Prevention Web, "Understanding disaster risk", accessed on September 2023
5
UNISDR, "Terminology on DRR", 2009
6
The Nansen Initiative, “Agenda for the protection of cross-border displaced persons in the context
of disasters and climate change Vol. I” , 2015
7
8
ISIMIP, "The Inter-Sectoral Impact Model Intercomparison Project”, accessed on September 2023
9
IDMC, “Global Internal Displacement Database”, accessed on September 2023
10
HDX, “The Humanitarian Data Exchange”, accessed on September 2023
11
UNFPA, “population estimation survey 2014”, 2014
12
ILO, “Statistics and databases”, accessed on April 2023
13
CDRI, "Global infrastructure Resilience", accessed on April and December 2023
14
De Bono, M.G. Mora “A global exposure model for disaster risk assessment” Int. J. Disaster Risk
Reduc., 10 (2014), pp. 442-451
15
Ibid
16
Schiavina M., Melchiorri M., Pesaresi M. (2023): GHS-SMOD R2023A - GHS settlement layers,
application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-
BUILT-S R2023A, multitemporal (1975-2030) European Commission, Joint Research Centre (JRC) PID:
http://data.europa.eu/89h/a0df7a6f-49de-46ea-9bde-563437a6e2ba, doi:10.2905/A0DF7A6F-
49DE-46EA-9BDE-563437A6E2BA
17
Bondarenko M., Kerr D., Sorichetta A., and Tatem, A.J. 2020. Census/projectiondisaggregated
gridded population data sets, adjusted to match the corresponding UNPD 2020 estimates, for 51
countries across sub-Saharan Africa using building footprints. WorldPop, University of Southampton,
UK. doi:10.5258/SOTON/WP00683
18
Zanaga, D., Van De Kerchove, R., De Keersmaecker, W., Souverijns, N., Brockmann, C., Quast, R.,
Wevers, J., Grosu, A., Paccini, A., Vergnaud, S., Cartus, O., Santoro, M., Fritz, S., Georgieva, I., Lesiv,
M., Carter, S., Herold, M., Li, Linlin, Tsendbazar, N.E., Ramoino, F., Arino, O., 2021. ESA WorldCover 10
m 2020 v100. https://doi.org/10.5281/zenodo.5571936
19
Schiavina M., Melchiorri M., Pesaresi M. (2023): GHS-SMOD R2023A - GHS settlement layers,
application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-
BUILT-S R2023A, multitemporal (1975-2030)European Commission, Joint Research Centre (JRC) PID:
http://data.europa.eu/89h/a0df7a6f-49de-46ea-9bde-563437a6e2ba, doi:10.2905/A0DF7A6F-
49DE-46EA-9BDE-563437A6E2BA
20
Meroni, M.; Rembold, F.; Urbano, F.; Csak, G.; Lemoine, G.; Hervé, K.; Pérez-Hoyos, A. The Warning
Classification Scheme of ASAP—Anomaly Hot Spots of Agricultural Production; JRC Technical
Reports; Publications Office of the European Union: Luxemburg, 2016; ISBN 978-92-79-64529-7
21
Pérez-Hoyos, A.; Rembold, F.; Kerdiles, H.; Gallego, J. Comparison of Global Land Cover Datasets
for Cropland Monitoring. Remote Sens. 2017, 9, 1118. https://doi.org/10.3390/rs9111118
22
Rembold, F.; Meroni, M.; Urbano, F.; Lemoine, G.; Kerdiles, H.; Pérez-Hoyos, A.; Csak, G. ASAP-
Anomaly hot Spots of Agricultural Production, a new early warning decision support system
developed by the Joint Research Centre. In Proceedings of the Multitemp 2017 Conference,
Bruges, Belgium, 27–29 June 2017
67
23
24
IOM-DTM “Sudan_Weekly Displacement_Snapshot (3) 08 September 2023”, accessed on
September 2023
25
CCCM-Cluster Somalia “List of IDP sites in Somalia”, accessed on September 2023
26
Cardona, O.D., Ordaz, M., Reinoso, E., Yamín, L.E., Barbat, A.H. (2012). CAPRA— Comprehensive
approach to probabilistic risk assessment: International initiative for risk management effectiveness.
Proceedings of the 15th world conference on earthquake engineering, 24–28 September 2012,
Lisbon, Portugal
27
World Bank “World Bank’s annual World Development Report (WDR)”, accessed in April and
December 2023
28
United Nations “World Population Prospects” , accessed in April and December 2023
29
Manning, R. (1891). “On the flow of water in open channels and pipes”. Transactions of the
Institution of Civil Engineers of Ireland. 20: 161–207.
30
Arrighi C, Rossi L, Trasforini E, Rudari R, Ferraris L, Brugioni M, Castelli F (2018) Quantification of Flood
risk mitigation benefits: a building-scale damage assessment through the RASOR platform. J Environ
Manag 207:92–104
31
IOM-DTM “Sudan_Weekly Displacement_Snapshot (3) 08 September 2023”, accessed on
September 2023
32
UNHCR “Emergency Book -Camp site planning minimum standards – July 2019”, accessed on
September 2023
33
34
https://emergency.unhcr.org/emergency-assistance/shelter-camp-and-settlement/camps/
camp-site-planning-minimum-standards
35
United Nations, “Early Warnings for All”, accessed on September 2023
IPCC Sixth Assessment Report., "Climate Change 2021: The Physical Science Basis"

36
37
S. Ceola and al. "Drought and Human Mobility in Africa - Ceola - 2023 - Earth's Future - Wiley Online
Library", 2023
38
IDMC, "Drought displacement modelling, 2022"
39
The Nansen Initiative, “Agenda for the protection of cross-border displaced persons in the context
of disasters and climate change Vol. I” , 2015
40
UNDRR Prevention Web, "Understanding disaster risk", accessed on September 2023
41
UNISDR, "Terminology on DRR", 2009
42
43
UNDRR, "GAR Special Report on Drought” 2021. Accessed: Aug. 04, 2021.
44
“Migration and global environmental change: future challenges and opportunities", United Nations
Network on Migration, accessed Dec. 25, 2021
45
IDMC, "Monitoring methodology for displacement associated with drought", 2020
46
G. J. Abel, M. Brottrager, J. Crespo Cuaresma, and R. Muttarak, “Climate, conflict and
forced migration,” Glob. Environ. Change, vol. 54, pp. 239–249, Jan. 2019, doi: 10.1016/j.
gloenvcha.2018.12.003.
47
S. Ceola and al. "Drought and Human Mobility in Africa - Ceola - 2023 - Earth's Future - Wiley Online
Library", 2023
48
IDMC, "Monitoring methodology for displacement associated with drought", 2020
49
Silvestro F., Ercolani G., Gabellani S., Giordano P., Falzacappa M. "Improving real-time operational
streamflow simulations using discharge data to update state variables of a distributed hydrological
model". Hydrol. Res., 52 (6) (2021), pp. 1239-1260
50
Silvestro F., Gabellani S., Delogu F., Rudari R., Boni G. "Exploiting remote sensing land surface
temperature in distributed hydrological modelling: the example of the Continuum model", Hydrol.
Earth Syst. Sci., 17 (1) (2013), pp. 39-62
51
Silvestro F., Gabellani S., Rudari R., Delogu F., Laiolo P., Boni G. "Uncertainty reduction and
parameter estimation of a distributed hydrological model with ground and remote-sensing data",
Hydrol. Earth Syst. Sci., 19 (4) (2015), pp. 1727-1751
52
Andréasson et al., 2004; Bosshard et al., 2011; Gellens and Roulin, 1998; Graham et al., 2007a,
2007b; Lettenmaier et al., 1999; Middelkoop et al., 2001; Moore et al., 2008; Shabalova et al., 2003
53
Ibid
54
Commonwealth Scholarship Commission, "Investigating the East African Climate Paradox:
Research and youth engagement to tackle the climate challenge", accessed on February 2024
55
Chris Funk and al., "Frequent but Predictable Droughts in East Africa Driven by a Walker Circulation
Intensification", 2023
56
Baxter, A.J., Verschuren, D., Peterse, F. et al. Reversed Holocene temperature–moisture relationship
in the Horn of Africa. Nature 620, 336–343 (2023). https://doi.org/10.1038/s41586-023-06272-5
57
Bondarenko M., Kerr D., Sorichetta A., and Tatem, A.J. 2020. Census/projectiondisaggregated
gridded population datasets, adjusted to match the corresponding UNPD 2020 estimates, for 51
countries across sub-Saharan Africa using building footprints. WorldPop, University of Southampton,
UK. doi:10.5258/SOTON/WP00683
58
WorlPop, Populalion counts, "Constrained individual countries 2020 UN Adjusted", accessed on
February 2024
59
60
CRED-EMDAT, "The international Disaster Database", accessed on February 2024
61
"Hunger and malnutrition being driven by climate crisis and conflict in South Sudan", accessed on
February 2024
62
UNHCR, "Protection & Return Monitoring Network", Somalia, accessed February 2024
63
64
"Global Land Evaporation Amsterdam Model", GLEAM, , accessed on February 2024
65
They were also applied in other regions (in Sudan and Ethiopia) where no observations on IDPs are
present but hazard indices are available, and on the projected future hazard indices influenced
by climate change, so as to estimate future impacts.
69

Flood and Drought Displacement Risk in Ethiopia Somalia and Sudan

Uploaded by

Copyright:

Available Formats

Flood and Drought Displacement Risk in Ethiopia Somalia and Sudan

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Flood and Drought Displacement Risk in Ethiopia Somalia and Sudan

Uploaded by

Copyright:

Available Formats

Ref.

Flood and Drought

D3.2 Research paper

Please visit www.habitableproject.org for more information about the project.

Publication available on:

Version N° Date Information

Graphic design, cartography and layout: Stéphane Kluser (Komplo).

Flood Displacement risk in Ethiopia, Somalia and Sudan.......... 5

Disaster displacement risk model ...............................................................................12

Exposure and vulnerability model ..............................................................................16

Flood hazard assessment . ...........................................................................................22

Flood impact assessment .................................................................................24

Flood and Drought Displacement risk in Ethiopia, Somalia and Sudan

Drought Displacement risk in Ethiopia, Somalia and Sudan.... 39

Drought displacement risk assessment ......................................................................45

Discussion and conclusion.............................................................................. 64

Flood Displacement risk in Ethiopia, Somalia and Sudan

Flood Displacement risk in Ethiopia, Somalia and Sudan

UNICEF Ethiopia/2021/ Nahom Tesfaye

Vulnerability is defined as the conditions Understanding the parameters that influence

Flood Displacement risk in Ethiopia, Somalia and Sudan

Population is assigned to Employed population (per Residential population is

Flood Displacement risk in Ethiopia, Somalia and Sudan

CLEAR / DIRECT UNCLEAR / INDIRECT

Direct impact on livelihoods Estimates can be computed for present and

because their homes have been rendered

Flood Displacement risk in Ethiopia, Somalia and Sudan

The conditional probabilities are estimated

The number of people who may lose access to

Flood Displacement risk in Ethiopia, Somalia and Sudan

Admin 1 Population - Relative Error to UN projected 2022

Figure 4: Statistical distribution on population in Ethiopia, according to different sources.

45% 64% 27%

Industry 2,920,000 (15%) 7,930,000 (10%) 103,000 (9%)

Table 1: Statistical data on employment rate.

Sector-specific national dataset Firstly, many settlements are situated in or near

Identification and location of the different

IDP settlements at risk • Sudan: IOM-DTM Weekly Displacement

Floods Displacement risk in Ethiopia, Somalia and Sudan

MOGADISHU The limitations inherent in this analysis must be

Figure 6: Somalia, locations of camps of IDP overlaid with a underestimation of displacements.

Flood Displacement risk in Ethiopia, Somalia and Sudan

Population from different

Table 2: Population data from different sources.

Population exposure by livelihood,

Figure 10: Population by livelihood, services

100years return period hazard

Flood Displacement risk in Ethiopia, Somalia and Sudan

disaster preparedness and mitigation strategies

How did we do it?

of employees with a workplace in a hazard- Ethiopia

Current AAD Optimistic AAD Pessimistic AAD

Floods Displacement risk in Ethiopia, Somalia and Sudan

Region AAD Current % Male % Female

Figure 13: AAD disaggregated by sex.

is useful to compare the PMD curves for present 350,000

and projected conditions (see figure 14), which

or rare (high return periods) events. 200,000

In Ethiopia, displacements would increase 150,000

significantly for higher return periods. For instance, 100,000