Search

Deprived urban areas, commonly referred to as ‘slums,’ are the consequence of unprecedented urbanisation. Previous studies have highlighted the potential of Artificial Intelligence (AI) and Earth Observation (EO) in capturing physical aspects of urban deprivation. However, little research has explored AI’s ability to predict how locals perceive deprivation. This research aims to develop a method to predict citizens’ perception of deprivation using satellite imagery, citizen science, and AI .A deprivation perception score was computed from slum-citizens’ votes. Then, AI was used to model this score, and results indicate that it can effectively predict perception, with deep learning outperforming conventional machine learning. By leveraging AI and EO, policymakers can comprehend the underlying patterns of urban deprivation, enabling targeted interventions based on citizens’ needs. As over a quarter of the global urban population resides in slums, this tool can help prioritise citizens’ requirements, providing evidence for implementing urban upgrading policies aligned with SDG-11., SCOPUS: ar.j, info:eu-repo/semantics/published

This research aims at assessing the potential of an semi-unsupervised approach to create labeled samples for predicting deprivation probability at 100x100m. We compare Machine Learning (ML) and Deep Learning approaches (DL) with a combination of Sentinel-1 (S1) and Sentinel-2 (S2) data. Our results confirm that the propose approach for creating labelled samples using semi-supervised method works. The best performance is achieve by the ML approach combining S2 and S1 data and reach an overall accuracy of 95.34%., info:eu-repo/semantics/published

La disponibilité d’évidences scientifiques sur les Accidents de la Route (AR) est essentielle à la mise en place des politiques de sécurité routière pertinentes, efficaces et adaptées au contexte des pays en développement. Ce travail a visé trois objectifs essentiels :celui de décrire les caractéristiques des accidents de la route au Bénin, ensuite, identifier leurs facteurs de risque et de gravité et enfin, identifier les facteurs associés aux conséquences chez les victimes d’accident de la route au Bénin. Il permettra de mettre à la disposition des décideurs des données probantes, scientifiquement objectives et susceptibles d’orienter les prises de décisions en matière de sécurité routière. Afin d’atteindre les objectifs fixés, nous avons mis en place et suivi entre juillet 2019 et janvier 2021, une cohorte de 1871 victimes d’accidents de la route qui a servi de base à six (06) études présentées dans ce travail. De plus une base nationale de données des accidents de la route, provenant des constats policiers réalisés entre 2008 et 2017, a été exploitée pour la réalisation de deux (02) autres études relatives aux facteurs environnementaux et géographiques. Les résultats de ces études montrent que :o Les victimes d’accidents de la route au Bénin étaient en majorité des jeunes âgés de 15 à 44 ans (67,1%), de sexe masculin (72,7%) et avaient un emploi au moment de l’accident (71,1%) ;o La prévalence du port de protection était de 57,0% et environ 11,3% des victimes de la cohorte utilisaient leur téléphone au moment de l'accident ;o Les accidents étaient plus fréquents dans la journée entre 07 heures et 19 heures (67,0%) ;o Les principaux types d’usagers de la route impliqués dans les accidents étaient des motocyclistes (75,3%) suivis des piétons (20,1%) ;o La proportion de blessés graves parmi les victimes de la cohorte était de 19,9% ;o Les proportions de victimes suivies à 6 et 12 mois étaient respectivement de 73,9% et 62,3% o L’analyse réalisée grâce aux Systèmes, The availability of scientific evidence on road crashes is essential for the implementation of relevant and effective road safety policies adapted to the context of developing countries. This study has three main objectives: to describe the characteristics of road crashes in Benin, to identify their risk and severity factors and, finally, to identify the factors associated with the consequences of road crashes in Benin. It will provide decision-makers with scientifically objective evidence to guide decision-making on road safety. In order to achieve the set objectives, we have set up and followed up between July 2019 and January 2021, a cohort of 1871 road crash victims which served as a basis for six (06) studies presented in this work. In addition, a national data base of road crashes recorded by the police from 2008 to 2017 was used for two (02) other studies on environmental and geographical factors. The results of these studies show that: o Road crash victims in Benin were predominantly young people aged 15-44 years (67.1%), male (72.7%) and were in employment at the time of the crash (71.1%);o The prevalence of wearing protection was 57.0% and approximately 11.3% of the victims in the cohort were using their phones at the time of the crash;o Crashes were more common during the day between 7 a.m. and 7 p.m. (67.0%) ;o The main types of road users involved in the crashes were motorcyclists (75.3%) followed by pedestrians (20.1%);o The proportion of serious injuries among the victims of the cohort was 19.9% ;o The proportions of victims followed up at 6 and 12 months after the crash were 73.9% and 62.3% respectively;o Analysis using Geographic Information Systems shows that vulnerable road users (pedestrians, cyclists and motorcyclists) were at greater risk of injury at intersections on the main roads in Cotonou;o Factors associated with injury accidents were good road condition (OR = 7.0; CI95%: 2.9-17.1) and road intersections (OR = 3.3; CI95%: 1.8-6.1);o Facto, Doctorat en Santé Publique, info:eu-repo/semantics/nonPublished

This study investigates the potential of freely available Sentinel-1 imagery coupled with Google Earth Engine (GEE) for mapping and monitoring flooding in Dar es Salaam. Sentinel-1 images (n = 55) available during the rainy season (March–May) since 2016 were used and processed in GEE. For separating water and land surfaces, we used a histogram-based automatic thresholding method. The binarization accuracy was assessed using confusion matrix based on 1064 randomly generated points in GEE. Overall accuracy of 95% (Kappa = 0.90) were achieved. Dar es Salaam has experienced flood inundation per flood event on average over an area of 50 km2 in March 2019 and 2021. Territories located along the Ocean and inland water shores, built and bare ground were subject to flooding compared to other land cover types. Flooding inundations have been difficult to detect in the city center. With the current temporal and spatial resolution of Sentinel-1, flood detection in city centers remains a challenge yet. However, Sentinel-1 images, coupled with GEE cloud computing simplified flood mapping and monitoring in a large urban region and this approach can be applied in other large cities and their surroundings for countries where data gap and lack of processing tools are critical challenges., SCOPUS: ar.j, info:eu-repo/semantics/published

Military conflicts strongly affect agricultural activities. This has strong implications for people’s livelihoods when agriculture is the backbone of the economy. We assessed the effect of the Tigray conflict on farming activities using freely available remote sensing data. For detecting greenness, a normalized difference vegetation Index (NDVI) was analyzed in Google Earth Engine (GEE) using Sentinel 2 satellite images acquired in the pre-war (2020) and during war (2021) spring seasons. CHIRPS data were analyzed in GEE to understand the rainfall conditions. The NDVI of 2020 showed that farmlands were poorly covered with vegetation. However, in 2021, vegetation cover existed in the same season. The NDVI changes stretched from − 0.72 to 0.83. The changes in greenness were categorized as increase (2167 km2), some increase (18,386 km2), no change (1.6 km2), some decrease (8269 km2), and decrease (362 km2). Overall, 72% of the farmlands have seen increases in green vegetation before crops started to grow in 2021. Scattered patches with decreases in vegetation cover correspond to irrigation farms and spring-cropping rain-fed farms uncultivated in 2021. There was no clear pattern of changes in vegetation cover as a function of agro-climatic conditions. The precipitation analysis shows less rainfall in 2021 as compared to 2020, indicating that precipitation has not been an important factor. The conflict is most responsible for fallowing farmlands covered with weeds in the spring season of 2021. The use of freely accessible remote sensing data helps recognizing absence of ploughing in crisis times., SCOPUS: ar.j, info:eu-repo/semantics/published

Remote sensing has been used for decades to produce vector-borne disease risk maps aiming at better targeting control interventions. However, the coarse and climatic-driven nature of these maps largely hampered their use in the fight against malaria in highly heterogeneous African cities. Remote sensing now offers a large panel of data with the potential to greatly improve and refine malaria risk maps at the intra-urban scale. This research aims at testing the ability of different geospatial datasets exclusively derived from satellite sensors to predict malaria risk in two sub-Saharan African cities: Kampala (Uganda) and Dar es Salaam (Tanzania). Using random forest models, we predicted intra-urban malaria risk based on environmental and socioeconomic predictors using climatic, land cover and land use variables among others. The combination of these factors derived from different remote sensors showed the highest predictive power, particularly models including climatic, land cover and land use predictors. However, the predictive power remained quite low, which is suspected to be due to urban malaria complexity and malaria data limitations. While huge improvements have been made over the last decades in terms of remote sensing data acquisition and processing, the quantity and quality of epidemiological data are not yet sufficient to take full advantage of these improvements., SCOPUS: no.j, info:eu-repo/semantics/published

info:eu-repo/semantics/nonPublished

With rapid urbanization leading to the proliferation of deprived urban areas (often referred to as “slums") in sub-Saharan Africa, there is a growing number of city dwellers living in inadequate housing conditions and being exposed to multiple hazards. In this context, Earth Observation has the potential for filling gaps in spatial data availability and thereby support evidence-based policy making. We assess the potential of free open-source software, open dual-pol SAR and optical imagery (Sentinel-l and Sentinel-2), and open global datasets for producing accurate city-scale maps of areas having morphological characteristics of deprivation. Implementing a grid-based machine learning approach, we evaluate different combinations of spectral and spatial Sentinel features, and features from global data. The results show that a high accuracy can be reached with the best combinations. Since publishing maps with hard labels (e.g. deprived vs. non-deprived areas) could raise ethical concerns or even lead to misuses, the output is provided as gridded morphological deprivation probability maps., info:eu-repo/semantics/published

Machine learning algorithms such as Random Forest (RF) are being increasingly applied on traditionally geographical topics such as population estimation. Even though RF is a well performing and gen- eralizable algorithm, the vast majority of its implementations is still ‘aspatial’ and may not address spatial heterogenous processes. At the same time, remote sensing (RS) data which are commonly used to model population can be highly spatially heterogeneous. From this scope, we present a novel geographical implementation of RF, named Geographical Random Forest (GRF) as both a predictive and exploratory tool to model population as a function of RS covariates. GRF is a disaggregation of RF into geographical space in the form of local sub-models. From the first empirical results, we conclude that GRF can be more predictive when an appropriate spatial scale is selected to model the data, with reduced residual autocorrelation and lower Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) values. Finally, and of equal importance, GRF can be used as an effective exploratory tool to visualize the relationship between dependent and independent variables, highlighting interesting local variations and allowing for a better understanding of the processes that may be causing the observed spatial heterogeneity., info:eu-repo/semantics/published

The population at risk from sudden hazards, such as volcanic eruptions, is often estimated based on the location of the residents. Crisis management tools, such as evacuation and emergency plans, are usually based on this information. Therefore, they often do not reflect over- and underpopulation in areas where the present population is different from the resident population at the time of the crisis. In the context of data scarcity that characterises Goma city (North Kivu, DRC), the results of the spatial analysis presented in this paper allowed the population to be located hour by hour on each day of the week. The dual movement of those leaving and those arriving in Goma made it possible to establish the temporal variations of the total population located in Goma. The variations are similar on all weekdays but differ for Sundays. Moreover, the spatio-temporal location of the population over time appeared to be strongly influenced by the activity rhythm of the city's vital infrastructures. In other words, areas where commercial activities, employment, and schools are concentrated, are crowded during working hours, while areas with a predominantly residential character become depopulated at these times. Therefore, the various areas of Goma could be impacted differently by sudden natural hazards. If their occupancy is poorly known, the safety of the population and the satisfaction of their expectations are difficult to predict accurately. Modelling these spatio-temporal variations of the population is crucial information for quantifying the resources needed to manage a future sudden crisis in Goma., info:eu-repo/semantics/published

In the past two decades, Earth observation (EO) data have been utilized for studying the spatial patterns of urban deprivation. Given the scope of many existing studies, it is still unclear how very-high-resolution EO data can help to improve our understanding of the multidimensionality of deprivation within settlements on a city-wide scale. In this work, we assumed that multiple facets of deprivation are reflected by varying morphological structures within deprived urban areas and can be captured by EO information. We set out by staying on the scale of an entire city, while zooming into each of the deprived areas to investigate deprivation through land cover (LC) variations. To test the generalizability of our workflow, we assembled multiple WorldView-3 datasets (multispectral and shortwave infrared) with varying numbers of bands and image features, allowing us to explore computational efficiency, complexity, and scalability while keeping the model architecture consistent. Our workflow was implemented in the city of Nairobi, Kenya, where more than sixty percent of the city population lives in deprived areas. Our results indicate that detailed LC information that characterizes deprivation can be mapped with an accuracy of over seventy percent by only using RGB-based image features. Including the near-infrared (NIR) band appears to bring significant improvements in the accuracy of all classes. Equally important, we were able to categorize deprived areas into varying profiles manifested through LC variability using a gridded mapping approach. The types of deprivation profiles varied significantly both within and between deprived areas. The results could be informative for practical interventions such as land-use planning policies for urban upgrading programs., SCOPUS: ar.j, info:eu-repo/semantics/published

In the version of this Article initially published, an error appeared in the caption to Fig. 4, where the descriptions for panels b and c were reversed. The caption has been corrected to read: “b, Landslide hazard trends when we assume there is no landslide response to deforestation in new forests. c, Landslide hazard trends in the Kivu Rift when we assume the landslide response to deforestation is the same for new and primary forests.” The error has been corrected in the online version of the Article., SCOPUS: er.j, info:eu-repo/semantics/published

Urbanisation has had a profound impact in sub-Saharan Africa and can be attributed to the complex human-environment interaction. Knowledge on long-term urbanisation processes in sub-Saharan Africa is lacking. Besides, understanding the urbanisation process in sub-Saharan Africa necessitates to go beyond the global scale to the local scale to unravel the idiosyncrasies in the growth of each city. The perennial lack of data (or data meeting required specifications) is a bottleneck to such studies. Very often, the nature of the available data might present challenges to existing classification algorithms. Moreover, availability of adequate and well-curated reference datasets presents a bottleneck to the generation of accurate land-cover maps. The main aim of this research was to conduct a long-term analysis of urbanisation patterns in Central Africa by developing methodologies based on deep learning, a class of Artificial Intelligence. To this end, we address the main research question: “How can we understand long-term urbanisation patterns by applying deep learning on digital aerial images in Central Africa?” We used case-studies from three cities in Central Africa namely Goma and Bukavu in The Democratic Republic of Congo and Bujumbura in Burundi, to aid in understanding the urbanisation process in sub-Saharan Africa. We generate baseline data from an archive of historical panchromatic orthomosaics that allow for the capturing of the urbanisation before the onset of rapid urbanisation that was characteristic of Countries in sub-Saharan Africa after gaining independence from colonialism.The main contribution of this thesis is the 60-year long-term analysis of urbanisation patterns for three cities in Central Africa using a unique dataset of historical orthomosaics. The growth patterns and driving forces are analysed using spatial and qualitative data. The results show that social triggers such as wars drive urban expansion. On the contrary, biophysical drivers such as, L’urbanisation a eu un impact profond sur l’Afrique subsaharienne et peut être attribuée à l’interaction complexe entre l’homme et l’environnement. Les connaissances sur les processus d’urbanisation sur des temps longs font défaut. En outre, pour comprendre le processus d'urbanisation en Afrique subsaharienne, il faut aller au-delà de l'échelle mondiale et s'intéresser à l'échelle locale pour comprendre les particularités de la croissance de chaque ville. L'éternel manque de données (ou de données répondant aux spécifications requises) constitue des limitations de telles études. Très souvent, la nature des données disponibles peut présenter des défis pour les algorithmes de classification existants. De plus, la disponibilité d'ensembles de données de référence adéquats et fiables constitue un goulot d'étranglement pour la création de cartes précises de l'occupation des sols.L'objectif principal de cette recherche était de mener une analyse sur des temps longs des différentes modes d'urbanisation en Afrique centrale en développant des méthodologies basées sur le deep learning, une forme d'intelligence artificielle. A cette fin, nous répondons à la question de recherche suivante :"Comment pouvons-nous comprendre les modes d'urbanisation sur des temps longs en appliquant des approches de deep learning sur des images aériennes numériques en Afrique centrale ?". Nous avons étudié trois villes d'Afrique centrale, à savoir Goma et Bukavu en République démocratique du Congo et Bujumbura au Burundi, de manière à comprendre le processus d'urbanisation en Afrique subsaharienne. Nous pouvons générer des données de base à partir de mosaïques d'orthophotographies aériennes panchromatiques historiques qui permettent de saisir l'urbanisation avant le début de l'urbanisation rapide ayant caractérisé les pays d'Afrique subsaharienne depuis leur indépendance.La principale contribution de cette thèse est l'analyse, sur 60 ans, des modes d'urbanisation pour trois villes d'Afrique centra, Doctorat en Sciences, info:eu-repo/semantics/nonPublished

info:eu-repo/semantics/nonPublished

info:eu-repo/semantics/nonPublished

info:eu-repo/semantics/nonPublished

info:eu-repo/semantics/nonPublished

Sub-Saharan African (SSA) cities are some of the most rapid-growing regions inthe world. As estimated by the United Nations, more than 50% of the Africanpopulation will reside in cities by 2030 while in absolute numbers, its population willhave doubled - from 1.2 billion in 2015 to 2.5 billion by 2050. This ongoing shift hasdramatically affected the capacity of several SSA cities to provide essential servicesfor its residents, such as durable housing, employment and healthcare accessibility.On the contrary, the proliferation of slums, further marginalizing the urban poor hasbeen an undisputed observation of the last years. It is therefore profound, that effortsto improve the quality of life of urban dwellers are needed. A starting point to do so,it to provide relevant authorities, stakeholders and organizations with useful socioeconomic,demographic and health indicators of the urban dwellers. Nonetheless, Inlarge parts of SSA, this type of critical information is at best scarce and at worstnon-existent and certainly not suited for sophisticated intra-urban analyses.This thesis harnesses the strength of very-high-resolution (VHR) satellite remotesensing to provide multi-level (physical, socio-economic, demographic and epidemiological)representations of the urban surface in several major SSA cities. First, cuttingedge advances relevant to the classification of urban land cover are presented.Afterwards, an assessment and evaluation of the potential of VHR satellite imageryas input to population estimation, household wealth and malaria risk models atunprecedented resolutions is investigated. The conclusions of this work encouragethe systematic exploitation of VHR data to support evidence-based decision makingshaping the sustainable future of SSA cities., Doctorat en Sciences, In reference to IEEE copyrighted material which is used with permission in this thesis, the IEEE does not endorse any of ULB’s products or services. Internal or personal use of this material is permitted. If interested in reprinting/republishing IEEE copyrighted material for advertising or promotional purposes or for creating new collective works for resale or redistribution, please go to http://www.ieee.org/publications_standards/publications/rights/rights_link.html to learn how to obtain a License from RightsLink. If applicable, University Microfilms and/or ProQuest Library, or the Archives of Canada may supply single copies of the dissertation, info:eu-repo/semantics/nonPublished

Multitemporal environmental and urban studies are essential to guide policy making to ultimately improve human wellbeing in the Global South. Land-cover products derived from historical aerial orthomosaics acquired decades ago can provide important evidence to inform long-term studies. To reduce the manual labelling effort by human experts and to scale to large, meaningful regions, we investigate in this study how domain adaptation techniques and deep learning can help to efficiently map land cover in Central Africa. We propose and evaluate a methodology that is based on unsupervised adaptation to reduce the cost of generating reference data for several cities and across different dates. We present the first application of domain adaptation based on fully convolutional networks for semantic segmentation of a dataset of historical panchromatic orthomosaics for land-cover generation for two focus cities Goma-Gisenyi and Bukavu. Our experimental evaluation shows that the domain adaptation methods can reach an overall accuracy between 60% and 70% for different regions. If we add a small amount of labelled data from the target domain, too, further performance gains can be achieved., SCOPUS: ar.j, info:eu-repo/semantics/published

Land cover maps contribute to a large diversity of geospatial applications, including but not limited to land management, hydrology, land use planning, climate modeling and biodiversity monitoring. In densely populated and highly fragmented landscapes as observed in the Walloon region (Belgium), very high spatial resolution is required to depict all the infrastructures, buildings and most of the structural elements of the semi-natural landscapes (like hedges and small water bodies). Because of the resolution, the vertical dimension needs explicit handling to avoid discontinuities incompatible with many applications. For example, how to map a river flowing under a bridge? The particularity of our data is to provide a two-digit land cover code to label all the overlapping items. The identification of all the overlaps resulted from the combination of remote sensing image analysis and decision rules involving ancillary data. The final product is therefore semantically precise and accurate in terms of land cover description thanks to the addition of 24 classes on top of the 11 pure land cover classes. The quality of the map has been assessed using a state-of-the-art validation scheme. Its overall accuracy is as high as 91.5%, with an average producer’s accuracy of 86% and an average user’s accuracy of 91%.

Land cover maps contribute to a large diversity of geospatial applications, including but not limited to land management, hydrology, land use planning, climate modeling and biodiversity monitoring. In densely populated and highly fragmented landscapes as observed in the Walloon region (Belgium), very high spatial resolution is required to depict all the infrastructures, buildings and most of the structural elements of the semi-natural landscapes (like hedges and small water bodies). Because of the resolution, the vertical dimension needs explicit handling to avoid discontinuities incompatible with many applications. For example, how to map a river flowing under a bridge? The particularity of our data is to provide a two-digit land cover code to label all the overlapping items. The identification of all the overlaps resulted from the combination of remote sensing image analysis and decision rules involving ancillary data. The final product is therefore semantically precise and accurate in terms of land cover description thanks to the addition of 24 classes on top of the 11 pure land cover classes. The quality of the map has been assessed using a state-of-the-art validation scheme. Its overall accuracy is as high as 91.5%, with an average producer’s accuracy of 86% and an average user’s accuracy of 91%., SCOPUS: ar.j, info:eu-repo/semantics/published

Historical aerial photographs provide salient information on the historical state of the landscape. The exploitation of these archives is often limited by accessibility and the time-consuming process of digitizing the analogue copies at a high resolution and processing them with a proper photogrammetric workflow. Furthermore, these data are characterised by limited spectral information since it occurs very often in a single band. Our work presents a first application of deep learning for the extraction of land cover from historical aerial panchromatic photographs of the African cities of Goma, Bukavu and Bujumbura. We evaluate the suitability of deep learning for land cover generation from a challenging dataset of photographs from the 1940s and 1950s that covers large geographical extents and is characterised by radiometric variations between dates and locations. A fully convolutional approach is investigated by considering two network architectures with different strategies of exploiting contextual information: one used atrous convolutional layers without downsampling, whereas the second network has both downsampling and learned upsampling convolutional layers (U-NET). The networks are trained to detect three main classes namely, buildings, high vegetation and a mixed class of bare land and low vegetation class. High overall accuracies of >90% in Goma-Gisenyi and Bukavu, and >85% in Bujumbura are obtained. This work provides a novel methodology that outperforms a baseline standard machine learning classifier for the exploitation of the vast archives of historical aerial photographs that can aid long-term environmental baseline studies. Future work will entail developing domain adaptation strategies in order to make the trained network robust for different image mosaics., SCOPUS: ar.j, info:eu-repo/semantics/published