Methodology for Flood Risk Assessment
Author: Hervé Bousquet Date: 24/02/2022
1 Purpose
The purpose of this document is to give an overview of the methodology selected by the UNDP Team
for the AF Drin River Basin Flood Risk Management Project.
The selection has been made after reviewing various existing methodologies and ready-made tools
(and even off-the-shelf software) in the world, and assessing what datasets could be made available
for the project either from existing databases and GIS layers, or from possible data to be collected by
surveyors on the field.
With this respect, six main questions proved to be key issues when selecting a methodology:
What data can be reasonably be made available considering the available timeframe and
budget?
What granularity and quality of data can be accepted, looking for a balance between
accuracy and efforts?
What is really the purpose of risk assessment: only rough and aggregate estimation of
damage and losses depending on statistical frequencies for comparison between areas at risk
(for example to rank communities), or detailed assessments in order to justify from a Cost
benefit Analysis dedicated investments to reduce flood risk?
Is the assessment focusing only on economic aspects (damage and losses) or also on social
vulnerability?
Will hydraulic parameters (water depths, velocities, flood duration and flood frequency)
be available for many return periods (at least 10yrs, 25 yrs, 50 yrs, 100 yrs and rare events:
250yrs or 300yrs or 500yrs), or only rough flood extend maps, or hazard maps for only one to
three return periods?
Will indirect losses be included in the assessment?
Only based on such criteria could be selected a relevant methodology for each project.
2 Possible approaches
2.1 Bespoke model Vs ready-made tool
A bespoke model is generally preferable for risk assessment as it will be designed and develop based
on the objectives and the available datasets for the project. It can rather simple, using GIS and EXCEL
for example. Nonetheless, it would require a GIS expert for the development, and additional time for
testing and documentation.
A ready-made tool could be used in particular if only a rough approach is sufficient for risk
assessment.
Nonetheless, most available tools such as CAPRA and INASAFE are probably too rough to be used
herein, as they ignore all particularities in constructions and account only for huge urban areas
altogether as can be extracted for web layers and images (such Google map or OpenStreet map), as
well as national gross figures.
Among available ready-made tools, HEC-FIA could be recommended for risk assessment at a local
(community) level, yet after customization of metrics.
2.2 Economic assessment only Vs Social and economic risk assessment
If vulnerability is a selected criterion in risk assessment, it must be integrated as needing a dedicated
dataset, disaggregated at least at community level and with geographical attributes for automatic GIS
treatment.
In addition, a scoring system combining both economic and social components is required.
Regarding economic (loss and damage) assessment, indirect losses can be assed applying a multiplier
factor based on experience and recent flood event, yet with very poor scientific justification…
2.3 Available more or less aggregated data Vs detailed datasets
When designing a bespoke model or selecting a ready-made one, the question of availability of data
(of good quality) as required by the tool is crucial. Generally speaking, data needs to be
disaggregated at community level (at least); detailed datasets required surveying and can be focused
in some pilot areas, or on a limited number of important receptors, such as infrastructures, facilities,
businesses… considering that they generally have specificities than should be ignored.
2.4 Unit costs Vs damage curves
Regarding potential damages and losses in building (dwelling) and in agriculture (after delineating
floodable fields and types of crops) two approaches are possible:
Use of unit costs, based on statistics from recent events (mean damage cost and mean value
of furniture and equipment in a house for example), regardless potential water depths and
velocities;
Application of damage curves, to be calculated from proportional available damage curves
based on national economic factors, which would also require hazard maps for many return
periods (not only three) and a rather detailed description and location of all receptors.
2.5 High granularity in receptors and attributes Vs single attributes
Risk assessment can be carried out using various possible levels in attributes describing all receptors.
For example:
Receptor: buildings. Attribute: Residential / Commercial. Sub-attributes: type (apartment…),
fabric, existing threshold…
Receptor: road. Attributes: international , national, local. Sun-attributes: bridges, tunnels,
type of vehicles, traffic…
On simple and very rough approach is to delineate so-called homogeneous urban blocks and
agricultural lands with global/general attributes.
2.6 Possible further use of database for CBA Vs one-shot assessment
If the database to be prepared for risk assessment will further be used to carry out Cost Benefit
Analysis for optioning measures to reduce flood risk, it needs details as only a small area will benefit
from the measures. Rough data will not allow to proper asses potential benefits from the project.
3 Methodology selected for the AF-UNDP Drin River Basin Flood Risk
Management Project
3.1 Preliminary remarks
We have faced to use HEC-FIA, as directly linked with HEC-RAS which is being used for hazard
assessment. It is rather simple to use, with some possibilities to customize parameters for
the country or the region.
We also faced the possibility to develop a bespoke model for ranking communities at risk and
for risk assessment in APSFRs being modelled. The selected solution has been to implement a
dedicated similar model develop for Georgia (another UNDP Project). The implementation is
on-going.
The methodology proposed by GIZ (according to their dedicated guidebook) is rather similar,
whereas less accurate in some points. The remaining question is about the availability and
reliability of data and parameter sets for Montenegro…
3.2 Main concepts and stages in the methodology
3.2.1 Stage 1: Data collection and database structure
As flood risk in assessed only over a limited number of rather small areas, data collection has been
organized according to two levels:
Available data regarding building, infrastructures, agriculture, population… have been
collected by the project GIS experts using available web sites.
Surveyors are being engaged to locally collect necessary data regarding types of crops, type
of dwellings, roads…
The level of granularity will be adapted depending on the types of receptors within floodable areas,
with the possibility to consider homogenous potential losses (uniform levels of equipment in
dwellings…), after a first assessment of vulnerable receptors in the APSFRs being studied.
3.2.2 Stage 2: Vulnerability assessment and scoring
When available, two types of information will be put into the mode: hazard maps from hydraulic
models on the one hand, GIS layers and tables detailing social and economic components.
Automatic calculation will provide scores according to various selected criteria.
Social vulnerability will thus be assessed at community level.
3.2.3 Stage3: Loss and damage assessment
For each type of receptors in the flood prone areas, loss and damages will both be assessed using
damage curves adapted for the country (or the region), with possible aggregation per community.
Indirect damages / losses will be assessed using sub-attributes, as for example potential road
disruption (after assessing flood duration) where traffic data will be obtained or appraised.
3.2.4 Risk mapping
Risk mapping is performed using specific and global parameters at community scale (even if potential
damage is assessed either at urban “block” level in some areas and at the individual level in other
areas):
Specific: social vulnerability based on gender, age, poverty…
Specific: loss and damage per type of receptor (building, agriculture, infrastructures…)
Global: loss and damage / vulnerability at community level
…
Colours to be used might be adapted, after scoring, from GIZ guidebook on risk mapping.
3.3 Needed matrices and metrics
Several matrices are required for risk assessment based on this approach, in particular:
Unit costs and damage curves. It normally takes years to elaborate damage curves in one
region or country; it has been recognised that for most types of attributes (or assets),
proportional damage curve can be successfully applies, yet introducing conversion factors
derived from simple gross economic parameters.
Business metrics to account both for direct damages and business disruption, activities being
depicted per category (agriculture, construction, finance, manufacture…).
Infrastructures and facilities: roads, sewage, water treatment plants…
Education
Health
Environment
…
Simple mathematical formulae are applied based on the inventory of assets at risk and on gross
national or regional figures.
In addition, matrices are required, for example to score risk for people depending on their
vulnerability.
4 Recommendations
One decision must be made about what the product should and could be, considering both
timeframe and budget.
All available tools will require data truthing and quality verification, as well as GIS calculation, in
addition to national economic expertise. Considering that not all data could be made available from
the internet, it will be needed to have data collection from administrations and eventually surveyors
for the characterisation of buildings and businesses (Google earth might be useful using on-ground
images).
In case of a very rough assessment, one simple solution is to delineate geographical units using
CORINE LandCover and available databases from the cadastre for buildings, and even work on the
basis of uniform urban blocks of building with similar characteristics. Social and economic data
(population, age, gender, mean gross revenue…à might be available at community level .
