Participatory approach for flood risk assessment: the case of Yeumbeul Nord (YN), Dakar, Senegal

Abstract

Flooding has emerged lately as a major threat for poor people in the suburban area of Dakar (Senegal), a densely populated area. In this region, flood events are mainly controlled by rainfall intensity and groundwater level fluctuations. To assess flood risk at a neighbourhood level, accurate data on flood extent, exposure and vulnerability is required. The objective of the present study is to obtain these data combining remote sensing data and local knowledge. Field work data were collected through interviews with inhabitants. About 500 respondents were surveyed with the support of a local association and very detailed mapping was carried out to clearly identify elements at risk. Information collected at the household level concerns: 1) socio-economic data, 2) information on the property, 3) flooded houses and 4) strategies of risk reduction. This research demonstrates that local knowledge is an important tool to obtain accurate data useful for understanding flood hazard and vulnerability patterns. It provides quantitative data at the household level that can be used to complement conventional GIS and remote sensing data.

Full text

Article
Reference
Participatory approach for flood risk assessment: the case of
Yeumbeul Nord (YN), Dakar, Senegal
SY, Bocar, et al.
Abstract
Flooding has emerged lately as a major threat for poor people in the suburban area of Dakar
(Senegal), a densely populated area. In this region, flood events are mainly controlled by
rainfall intensity and groundwater level fluctuations. To assess flood risk at a neighbourhood
level, accurate data on flood extent, exposure and vulnerability is required. The objective of
the present study is to obtain these data combining remote sensing data and local knowledge.
Field work data were collected through interviews with inhabitants. About 500 respondents
were surveyed with the support of a local association and very detailed mapping was carried
out to clearly identify elements at risk. Information collected at the household level concerns:
1) socio-economic data, 2) information on the property, 3) flooded houses and 4) strategies of
risk reduction. This research demonstrates that local knowledge is an important tool to obtain
accurate data useful for understanding flood hazard and vulnerability patterns. It provides
quantitative data at the household level that can be used to complement conventional GIS
and remote sensing [...]
SY, Bocar, et al. Participatory approach for flood risk assessment: the case of Yeumbeul Nord
(YN), Dakar, Senegal. WIT Transactions on The Built Environment, 2016, vol. 165, p.
331-342
DOI : 10.2495/UW160291
Available at:
http://archive-ouverte.unige.ch/unige:86254
Disclaimer: layout of this document may differ from the published version.
[ Downloaded 22/08/2016 at 10:54:32 ]
1 / 1

Participatory approach for flood risk
assessment: the case of Yeumbeul Nord (YN),
Dakar, Senegal
B. Sy1, C. Frischknecht1, H. Dao2, G. Giuliani4, D. Consuegra1,3,
S. Wade5 & C. Kêdowidé6
1Department of Earth Sciences, University of Geneva, Switzerland
2Department of Geography and Environment,
University of Geneva, Switzerland
3School of Business and Engineering Vaud, Switzerland
4Institute for Environmental Sciences, University of Geneva, Switzerland
5Institute of Earth Sciences, University Cheikh Anta Diop, Senegal
6African Urban Management Institute, Senegal
Abstract
Flooding has emerged lately as a major threat for poor people in the suburban area
of Dakar (Senegal), a densely populated area. In this region, flood events are
mainly controlled by rainfall intensity and groundwater level fluctuations.
To assess flood risk at a neighbourhood level, accurate data on flood extent,
exposure and vulnerability is required. The objective of the present study is to
obtain these data combining remote sensing data and local knowledge. Field work
data were collected through interviews with inhabitants. About 500 respondents
were surveyed with the support of a local association and very detailed mapping
was carried out to clearly identify elements at risk.
Information collected at the household level concerns: 1) socio-economic data,
2) information on the property, 3) flooded houses and 4) strategies of risk
reduction.
This research demonstrates that local knowledge is an important tool to obtain
accurate data useful for understanding flood hazard and vulnerability patterns. It
provides quantitative data at the household level that can be used to complement
conventional GIS and remote sensing data.
Keywords: flooding risk, P-mapping, P-GIS, Dakar, Yeumbeul Nord (YN).
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press
This paper is part of the Proceedings of the 5 International Conference
th
on Flood Risk Management and Response (FRIAR 2016)
www.witconferences.com
doi:10.2495/UW160291

1 Introduction
Natural disasters caused by natural phenomena, whether geological, hydrological
and climate affect more and more people in the world with floods constituting the
primary cause across the globe (Wang et al. [1]). In the period between 1900 and
2016, there were 910 flood disasters in Africa alone, killing 27,051 people,
affecting 71 million people, and causing 7.9 million USD of economic losses (EM-
DAT [2]). For Senegal itself, in the past 30 years (1980 to 2009 inclusive) over
900,000 people were affected, 45 were killed, and damage costs were estimated to
be more than 142 million USD (Gouvernement du Sénégal [3]). The Dakar region
is the most affected region, where several severe floods occurred in 1989, 1996,
2001, 2005, 2009, and 2012, affecting particularly the suburbs, the 2009 event
being the worst.
Many efforts have been tempted by the Government and local authorities to
deal with and minimize the negative impacts of frequent floods, consisting of
structural (construction of canalization and retention basins) and non-structural
measures (relocation). But flooding within sub-urban of Dakar still represents a
major problem for the local government. And the lack of information on hazard
and vulnerability related to floods does not allow to produce quantitative
assessment on flooding risk, necessary for the implementation of a flood risk
reduction policy and a sustainable development.
Flood risk assessments are usually carried out by three approaches: 1) field
work, 2) remote sensing analysis and/or 3) hydraulic modelling. As such, these
approaches present some specific limitations, related either to the time required to
carry out field work, weather conditions (Brivio et al. [4]) influencing the quality
of satellite images or the lack of images covering the area and the uncertainties on
results obtained by hydraulic modelling (Pappenberger et al. [5]), which is very
sensitive to the accuracy of basic data. Therefore, the main purpose of this study
is to obtain accurate data on flood risk related parameters combining remote
sensing data and local knowledge. As mentioned by Tran et al. [6], local people
have a better understanding of their surroundings and can help identify areas prone
to floods. Moreover, it is also known that local knowledge can improve the
accuracy of data (Dunn [7]). Therefore, a participatory approach has been used
implying participatory mapping (P-mapping) and participatory GIS (P-GIS) in
order to consider and integrate local knowledge on floods.
2 Study area
The area under investigation is the municipal district of Yeumbeul Nord (YN), one
of the 16 municipal districts of the city of Pikine. It is located in the suburbs part of
the capital city Dakar in Senegal (figure 1). The total area of this municipal district
covers about 9 km2 divided into 82 major neighbourhoods (Kêdowidé and Cissé
[8]). According to the 2013 census of the National Senegalese Agency for Statistic
and Demography, the total population is about 168,379 inhabitants (ANDS [9]). It
is one of the most populated districts of Senegal, with a population density of
332 Urban Water Systems and Floods
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

approximately 18,700 habitants/km2. The study area is characterized by lowlands
with elevations less than 20 m.
Floods occurring in YN are due to the combination of natural (rainfall
variability, groundwater position, topography, hydrology and lithology), and
anthropogenic factors related to uncontrolled urbanization and soil impermeability
(Mbow et al. [10]).
Figure 1: Map of Yeumbeul Nord and the 82 neighbourhoods.
3 Methodology
Acquisition of data related to floods in YN was achieved using remote sensing and
ground based data (figure 2). Satellite data were used to obtain a preliminary land-
use map including the spatial distribution of houses and the spatial distribution of
the 2009 flood extent. Then, information was completed using local spatial
knowledge. This has been achieved through active household survey and
interviews, participatory-mapping and participatory-GIS. The interviews yielded
socio-economics characteristics, physical exposure, flood depth and event
duration, as well as strategies of flood risk reduction applied at the ho usehold level .
In this study, the participatory approach refers to mapping and collecting
information on flooding from the local population. P-mapping helped getting
detailed location of elements at risk and areas of water. Finally, P-GIS was carried
out in order to cross-check results obtained from the remote sensing data,
interviews and P-mapping.
Urban Water Systems and Floods 333
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

Figure 2: Chart flow of the data acquisition process.
3.1 Remote sensing data
In order to obtain our first land use map over the study area, we used high
resolution satellite images (January–July, 2015) available on Google Earth®. The
location of different land use categories (infrastructures, agricultural area, water
bodies, etc.) and each house were photo-interpreted and digitized in Google Earth.
We then used Global mapper 15® for the rapid conversion of the KML files into
shapefiles with the reference system UTM (Zone 28N). Finally, we have imported
our preliminary database in ArcGIS 10® where a unique identification number was
attributed for each house.
The flood extent of the 2009 event has been obtained by comparing a reference
high resolution satellite image before the flood (March 11) and after (October 14)
obtained in Google Earth using its historical satellite dataset.
The Google Earth high-resolution imagery archive remains a largely
unexploited resource for the analysis and description of the Earth’s land surface
(Potere [11]). The high-resolution images (2.5 m resolution) used in this analysis
come from Digital Globe’s (e.g. Quick Bird—Ikonos) satellites.
It can be seen in figure 3 that some areas are covered with clouds causing a
partial loss of information.
334 Urban Water Systems and Floods
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

Figure 3: Free Google Earth® images used for flood extent mapping.
3.2 Household survey
Interviews were carried out on the basis of a questionnaire developed specifically
for this study. It consisted of five main sections: 1) respondent’s characteristics,
2) information on the property, 3) flooded houses, 4) socio-economic data and
5) strategies for flood risk reduction. The target of the interview was the household
head and when the household head was not available, the spouse or a person
having memories of the floods was surveyed. The questionnaires were conducted
by teams composed of young adults from the local association “Réseau
d’Information d’Education de Communication maternité de Yeumbeul” and the
students from the Institute of Earth Sciences at the University of Dakar and each
team was attributed to a specific neighbourhood for a total of 54. They were trained
before going to the field.
Moreover, the questionnaire was first administrated in a test study lasting one
day and implying 50 households. Feedback obtained from these participants and
from the investigators allowed to ensure that wording was comprehensible
and added other aspects not taken into account. In view of the low levels of
education of respondents and the high illiteracy rate, the interviewers were called
to translate the French version to the most widely spoken language in Senegal the
Wolof. Each survey lasted 30 to 45mn during which the respondents actively
interact with the investigators. Finally, 502 households were surveyed (figure 4).
In complement to the interviews, the respondents also provided information on
the water depth affecting their house during the past flood events, when possible.
With a handled GPS and the support of a paper map showing the details of each
investigated neighbourhood, investigators were instructed to identify the house on
the map and take GPS coordinates for cross checking.
Urban Water Systems and Floods 335
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

Figure 4: Distribution of the surveyed households.
3.3 Participatory mapping and participatory GIS
The P-mapping was carried out in the 82 neighbourhoods with the support of the
local association and the householders interviewed. These local representatives
brought their knowledge at three levels of the dataset. The first level was to verify
the accuracy of the preliminary land-use map obtained with remote sensing data.
The second level concerns the precise mapping of areas of water (lake, outlet,
retention basin) and the third level was to achieve a detailed survey of important
infrastructures of the area (e.g. health centres, schools, recreation areas, public
service etc.). Handled GPSs, and mobile GIS were used to carry out the detailed
mapping of the key infrastructures and water bodies. In complement to the update
of the land use map, the data obtained on each household were combined with the
GIS database existing on houses. To do this, the information gathered in the
household survey was included into ArcGis 10® by joining the excel sheet
containing the survey data and the initial GIS layer of houses based on a common
identifier which was the name of the houses (each house had the same name in the
Excel sheet and the initial database). This operation was carried out with the help
of the interviewers in order to make sure that the merging of the two datasets was
correctly done.
4 Results
4.1 Land use
The resulting land-use map of the study area was obtained from remote sensing
data and then verified with the local teams and improved by integrating local
knowledge. The result shows 7 major land cover types: forest, agriculture,
buildings, infrastructures, water bodies, livestock farming, and recreation (figure
5). The area used for buildings and infrastructure is the largest with about 47%.
The other including roads, canalisations and the nature were estimated about 33%.
Water bodies were estimated about 10%. The proportion of agriculture land is
5.8%. Forest accounts for 3.6%. The percentage of recreation area is 0.56 and
336 Urban Water Systems and Floods
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

livestock farming is 0.48 (table 1). This classification shows that the study area is
highly urbanized.
Figure 5: Land-use map.
Table 1: Land-use classes.
4.2 Flood extent
Figure 6 presents the flood extent of the 2009 event based on remote sensing
analysis and the flood extent that might have really occurred if we integrate the
places highlighted by the household interviews for this event. This result
demonstrates that household surveys can contribute to improve data provided by
remote sensing, when the images is covered by clouds and also yield more accurate
information.
Urban Water Systems and Floods 337
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

Figure 6: Flood extent based on remote sensing and P-mapping.
4.3 Flood depth and flood duration
During the survey, respondents were asked to show the depth reached by flood
events and to describe the duration. Based on this information, we can see
(figure 7) that the flood depth, independently of the event, varies from 0.2m to 1m,
Figure 7: Spatial distribution of flood depths.
338 Urban Water Systems and Floods
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

Figure 8: Spatial distribution of flood duration.
with an average of 0.5m, whereas the flood duration (figure 8) ranges from
< 1 day to 365 days. Such a long duration is probably due to the lack of means for
water evacuation and the existence of topographic depressions (Mbow et al. [10]).
But the usual water duration is between 0 and 7 days.
4.4 Socio-economic characteristics of the household survey
Table 2 summarizes the general demographic and social characteristics of the 502
households included in our sample. The majority of the respondents were women
(67%). Among the respondents 41% were the owner of the house and 59% were
not. Respondents were distributed among the following categories: 25% were the
head of household, 36% were the spouse of head of household and 39% other were
in general the most knowledgeable person in terms of flooding. The average
number of persons per household is 13. The average of number of women per
household was 6. The average number of disabled was 1. The average number of
children per household was 5. The age of the respondents varies from 15 to 87 and
the average is 44 years old.
The analysis of household head education shows that 27% of the household
head were illiterate, 18% had primary school level, and 14% had secondary school
level, 6% had university level and 32% had received another type of education.
The average monthly income per household in YN is 435 USD.
4.5 Information on the physical exposure
The survey revealed key information on the 502 households related to date of
construction, the ownership, the type of construction materials and the structure as
well as the price of the land. Results show that 12% of the houses were built
Urban Water Systems and Floods 339
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

Table 2: Respondent’s characteristics.
between 1970 and 1980, 40% between 1981 and 2002 and 14% between 2003 and
2015. 65% of the surveyed households were living in their own houses, 20% were
occupied by lodger and 15% by both owner and lodger. A majority of 88% of the
surveyed houses were residential and 12% were mixed residential and workplace.
Almost all houses were built in cement and 34% of the interior courtyards
consist of cement, 29% of sand and finally 27% are in tile. More than 91% of the
surveyed houses have a foundation and 2% do not know if there is or not. 67% of
surveyed houses had cracks. 17% of surveyed houses were having the ground floor
lower than the road.
The land surface occupied by houses varies from about 100 m2 to more than
300 m2. In terms of land price one third (36%) costs less than 2000 USD (approx.
1FCFA= 500 USD), 13% varies between 2001USD and 6000 USD, 6% between
60001 USD and 12000 USD and 7% above 12000 USD. 34% of respondents do
not know the land price of their house and 4% didn’t want to answer.
Among the 502 houses included in the surveyed, 209 were at least once flooded
(figure 9). Most of surveyed houses were flooded in 2005, 2009, and 2012.
Figure 9: Number of events per house.
4.6 Local strategies for flood risk reduction
The household survey reveals the variety of local strategies put in place for flood
risk mitigation. They can be classified in structural and non-structural measures as
summarized in table 3.
340 Urban Water Systems and Floods
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

Table 3: Flood risk mitigation.
5 Discussion and conclusion
This study illustrated how the combined use of remote sensing data and local
knowledge manages to capture key elements on flood hazard (the magnitude of
flood: flood depth, flood duration), elements at risk (houses, critical
infrastructures), vulnerability (socio-economic and physical vulnerability at the
household level) as well as local strategies for flood risk reduction.
The innovative contribution of our investigation in comparison with previous
works, such as Canevari-Luzardo et al. [12] and Musungu et al. [13] is the size of
the survey, 502 households, and the level of details obtained on each flood risk
parameters, in particular, detailed information on flood hazard. Moreover, we
demonstrated how local knowledge can complement the use of remote sensing
analysis, aspect that wasn’t integrated in Canevari-Luzardo et al. [12] and
Musungu et al. [13].
Although the participatory approach allowed us to improve the analysis of
satellite images, it has some limitations. The local population can give inaccurate
information, especially in terms of hazard mapping and spatial perception (Haynes
et al. [14]). However in our study, we used neighbourhood scale paper maps,
handled GPS and mobile SIG, therefore increasing the accuracy of mapping as
already noticed by Canevari-Luzardo et al. [12]. In order to acquire high quality
data, interviewers were trained, we organized a test study to ensure wording was
comprehensible and we used the local language to avoid misunderstanding.
However, the stability of the responses was not tested and it could be interesting
to do so, in particular regarding flood events and their intensity.
In conclusion, our integration of local knowledge together with remote sensing
shows that this can improve data, when satellite images are covered by clouds and
also yield new or more accurate information in terms of hazard intensity, exposure
and location of key infrastructures. Our approach is an alternative to the use of
expensive high-resolution satellite images, when financial resources are scarce or
when images are not available on the study area. While this study focuses on flood
risk assessment, this approach could be replicated for different risks in other
contexts.
Urban Water Systems and Floods 341
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press

References
[1] Wang, Z., Lai, C., Chen, X., Yang, B., Zhao, S., and Bai, X., Flood hazard
risk assessment model based on random forest, Journal of Hydrology, 527,
pp. 1130–1141, 2015.
[2] EM-DAT, The international disaster database, http://www.emdat.be/
disaster_profiles/index.html, accessed 3 March 2016
[3] Gouvernement du Sénégal, Rapport d’évaluation des besoins post
catastrophes: Inondations urbaines à Dakar en 2009. Gouvernement de la
République du Sénégal avec l’appui de la Banque Mondiale, du système des
Nation Unies et de la Commission Européenne, 2010.
[4] Brivio, P., Colombo, R., Maggi, M., and Tomasoni, R., Integration of
remote sensing data and GIS for accurate mapping of flooded areas,
International Journal of Remote Sensing, 23(3), pp. 429–441, 2002.
[5] Pappenberger, F., Matgen, P., Beven, K. J., Henry, J.-B., Pfister, L., and
Fraipont de, P., Influence of uncertain boundary conditions and model
structure on flood inundation predictions, Advances in Water Resources,
29(10), pp. 1430–1449, 2006.
[6] Tran, P., Shaw, R., Chantry, G., and Norton, J., GIS and local knowledge in
disaster management: a case study of flood risk mapping in Viet Nam,
Disasters, 33(1), pp. 152–169, 2009.
[7] Dunn, C. E., Participatory GIS—a people’s GIS?, Progress in human
geography, 31(5), pp. 616–637, 2007.
[8] Kêdowidé, C., and Cissé, O., Adapting to Flooding Impacts in Yeumbeul
Nord, a Suburb of Dakar, Senegal. Institut Africain de Gestion Urbaine,
2013.
[9] ANDS, Recensement Général de la population et de l’Habitat, de
l’Agriculture et de L’Elevage. République du Sénégal, Ministère
de l’Economie, des Finances et du Plan, Agence Nationale de la Statistique
et de la Démographie, 2014.
[10] Mbow, C., Diop, A., Diaw, A. T., and Niang, C. I., Urban sprawl
development and flooding at Yeumbeul suburb (Dakar-Senegal), African
Journal of Environmental Science and Technology, 2(4), pp. 75–88, 2008.
[11] Potere, D., Horizontal positional accuracy of Google Earth’s high-
resolution imagery archive, Sensors, 8(12), pp. 7973–7981, 2008.
[12] Canevari-Luzardo, L., Bastide, J., Choutet, I., and Liverman, D., Using
partial participatory GIS in vulnerability and disaster risk reduction in
Grenada, Climate and Development, pp. 1–15, 2015.
[13] Musungu, K, Motala, S., and Smit, J., Using multi-criteria evaluation and
GIS for flood risk analysis in informal settlements of Cape Town: the case
of Graveyard Pond, South African Journal of Geomatics, 1(1), pp. 92–108,
2014.
[14] Haynes, K., Barclay, J., and Pidgeon, N., Volcanic hazard communication
using maps: an evaluation of their effectiveness, Bulletin of Volcanology,
70(2), pp. 123–138, 2007.
342 Urban Water Systems and Floods
www.witpress.com, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 165, ©2016 WIT Press