Introduction
Deprived urban areas present an immense humanitarian risk to urban dwellers throughout the world. They have low access to sanitation, housing security, safety and economic mobility, as urban governance ignores the existence of the over 1.1 billion slum dwellers throughout the world (United Nations, 2023). Slum dwellers are largely excluded from social safety nets, which often lack the capacity and interest to reintegrate these communities into the formal economy. While deprived urban areas are often represented as homogenous figures in the statistical realm, accurate and precise mapping efforts need to be undergone to reveal disparities within informal settlements. This study will serve as a starting point for prioritizing upgrades in Kibera, an informal settlement within Nairobi, Kenya.
A deprived informal settlement (DIS), otherwise known as a slum, is an area that lacks access to one or more of the key amenities defined by the United Nations Human Settlements Program (UN-Habitat, n.d.). These amenities include adequate housing, basic urban services, security of tenure and sufficient living area. While the general terminology utilized is “slum,” interviews have noted the term as having stigmatizing connotations. For this reason, literature within the field of urban sustainability should begin referring to these areas as deprived informal settlements (Kuffer et al., 2021). Given the lack of amenities in DISs, disease, poverty, crime and housing insecurity have a high prevalence and result in significant risk towards long-term urban sustainability. DISs are particularly prevalent in Africa, a region that will account for 95% of global population growth between 2025 and 2100 (United Nations, 2024).
Nairobi, Kenya's capital city, was selected because of the country's severe informality within its urban areas. Kenya is located in Eastern Africa and has over 70% of its urban population living in a DIS (UN-Habitat, 2020). Around 47% of Nairobi's population lives in a DIS, being partially attributable to the long-lasting impacts of colonialism on urban governance (Solymári and Czirják, 2024). According to a UN-Habitat report on KENSUP, the Kenya Slum Upgrading Initiative, only 22% of households in a DIS have water connections, with over 75% of them accessing water through vendors (UN-Habitat, 2007). These communities have inadequate access to sanitation and often rely upon flying toilets for defecation needs, therefore increasing disease risk. Flying toilets are defined as plastic bags that are used as a substitute for a toilet. Kibera was chosen as the community of study, as it is not only the largest DIS in Nairobi, but one of the largest in all of Africa (Desgroppes and Taupin, 2011).
As a solution to DISs, participatory upgrade projects should be implemented by municipal governments and humanitarian organizations to produce tangible benefits for the residents of these communities. This requires straying from the traditional government approach that has focused on the demolition and redevelopment of deprived areas in a top-down manner. Following this approach further exacerbates the issue at hand, as dwellers are pushed to the periphery of cities, where new informal settlements form and the economic hardship of residents persists (Dupont et al., 2013).
To set the stage for upgrade projects, site analysis must be done to prioritize high-need areas. A weighted overlay model can be used to integrate a data-driven approach into the decision-making process. To facilitate this analysis, participatory data collection can help decrease data poverty, as overlay models require in-depth data for comprehensive analysis. Preliminary studies should be done with crowd-sourced data and will need to be supplemented by government studies that utilize participatory, but formal data collection. The remainder of this paper will outline a weighted overlay approach and associated criteria to prioritize upgrade sites in Kibera.
Methods
This study combines a literature review and weighted overlay analysis to propose a preliminary study for Kibera upgrade sites. By combining these two methods, a comprehensive understanding of the limitations and requirements of site analysis can be framed to support future projects that decrease data poverty and improve informed decision-making processes. A literature review was done to understand the criteria needed for a site analysis, while a weighted overlay model was used to apply this knowledge to Kibera and suggest future improvements upon it.
For the literature review, a combination of 17 articles on DISs and mapping projects in Nairobi were used. Literature served 3 specific purposes. First, understanding the situation in Kibera. Second, understanding the shortcomings of current and past upgrading projects. Third, understanding the criteria needed for an effective site analysis. Twelve independent studies and five United Nation reports were combined to create a diverse collection of opinions that reduced bias.
A significant number of articles came from the United Nations because of the organization's activity within Nairobi, the home of the UN-Habitat headquarters. The UN-Habitat also sponsored KENSUP and produced the most comprehensive, publicly available mapping study on Kibera thus far. It is worth noting that one report dates back to 2007 and was only used when newly produced data could not be found. Independent studies were used to compare findings from UN reports and research specific topics that pertain to site analysis and crowdsourced data.
The analysis was done in ESRI’s ArcGIS Pro, version 3.5.3, and used their weighted overlay tool. This tool combines multiple rasters into a single layer that represents which cells meet the most criteria. Six different criteria were identified based on existing issues within DIS. The criteria defined were as follows: distance to toilets, distance to water, distance to healthcare facilities, population distribution, building density and tree coverage density. Prior to mapping, each criteria was assigned a rank that was used in rank-sum calculations. Distance to a toilet was assigned rank 1, population distribution was assigned rank 2, distance to a water point was assigned rank 3, distance to healthcare was assigned rank 4, building density was assigned rank 5 and tree coverage was assigned rank 6. A justification for each rank is included in the results section of this paper.
A weight-sum calculation was used to provide a manageable and effective way to assign percentage weights. Each rank was subtracted by 1, then 1 was added to the result. Afterwards, all values were summed together to a total of 21. Each rank was then divided by 21, resulting in each criteria’s percentage weight. Exact ranking results can be found in Appendix A.
After finalizing weights, rasters for each of the 6 criteria were created within ArcGIS Pro. For the weighted overlay tool, all rasters had to be in the same extent, cell size, mask, origin, and scale. The extent was defined by a rectangular polygon that included Kibera and some additional area surrounding the settlement, this allowed for faster processing and all relevant data points to be included within each processing step. The mask was defined by a buffered polygon around the settlement's 15 informal villages. The borders were buffered by 15 meters to include all cells that fell partially within the study area, but not entirely. Raster cell size was defined as 31 square meters, the resolution of META’s Kenya population distribution raster. Exact values for the cell and buffer size are included in Appendix B. The population raster was also used as the origin layer, therefore ensuring all rasters would have identical cell placement. WGS 1984 UTM Zone 37S was used as the universal projection, as it is a preferred projection for Kenya mapping analysis. These decisions were input into the project's environment settings and ensured all rasters were compatible for analysis. Following this, each raster was made with 3 different sources.
The population distribution raster was made by META, the parent company of Facebook. META produces a 30x30 high-resolution raster of population distribution for the entire world and uses a deep learning model to distribute census data throughout an area based on the built environment. This data is supported and distributed in the Humanitarian Data Exchange, a publicly accessible data library that is managed and reviewed by the United Nations Office for the Coordination of Humanitarian Affairs. Once the data was obtained, it was reprojected with bilinear interpolation resampling. The layer was automatically masked to the buffered boundary and a raster calculator was used to populate all null value cells with a value of 0. Following this, reclassify was used with a natural breaks classification scheme to place the raster on a common scale of 1-9. Natural breaks were used because the dataset had naturally occurring clusters that would have been broken using an artificial quantile classification scheme. One represents the least suitability and nine represents the highest suitability. The reclassification scale of 1-9 was used for all other layers that follow.
The tree coverage raster was created using a self-digitized polygon layer that included Kibera and its buffered boundaries. This digitization was done by Thomas Russell and was based on Maxar's high-resolution Vivid satellite imagery. Due to technical restraints, a deep learning model was not used for the creation of this dataset. Once produced, the polygon layer was projected and added to the project. A fishnet grid that aligned with the population raster was created and was intersected with the tree polygons. After this, an area field was added and was populated with the area of every intersected polygon. These polygons were dissolved and their areas were summed to produce a layer with the combined area of each collection of polygons per cell, which were then joined back to the initial grid. The summed values were used to calculate the percent of each cell that is covered by tree canopy and were added to a new field. All grid cells with a null value were populated with 0, then the grid layer was turned to a raster based on the percent of tree coverage per cell. Finally, the raster was reclassified using a geometric interval classification. Geometric interval was used because it is highly effective for showing meaningful distribution in right-skewed datasets.
The building density raster was created using Open Street Map (OSM) building data, the most accurate and comprehensive public source for DISs. Given that these areas generally do not have public data released or taken in the first place, crowd-sourced data like OSM can be incredibly useful for preliminary studies. The same workflow that produced the tree coverage layer was applied for the building density raster.
Prior to producing any distance-related rasters, a cost raster was produced for all distance calculations. This layer combined the initial building density raster and a distance from road raster using a weighted sum tool in ArcGIS. The distance from road raster was created using the distance accumulation tool, which represents each raster by how far it falls from a road line. OSM road data was used based on similar reasoning to the building data layer. Both layers were reclassified to a common scale of 1-10 and assigned weights for the weighted sum. Roads were assigned a weight of 0.25, as informal settlements have many hidden alleyways that can be used to traverse the area, while buildings were assigned a weight of 0.75, given that they represent physical barriers to movement.
The distance to the toilet raster was created using OSM point data that represented every toilet within Kibera. OSM data was used because there are no other publicly available datasets that show toilet points within Kibera. First, the point data was projected and clipped to the project extent. Next, the layer was put into the distance accumulation tool with the aforementioned cost raster. The distance raster was then reclassified and completed.
The distance to water raster was created using OSM point data that represented every water point within Kibera. A water point is defined as a place where water can be accessed by the public and can include water pumps, water tanks and other static distribution sites. The same workflow that was used for the toilet raster was used for this layer.
The distance to healthcare raster was created using OSM point data that represented each healthcare facility within Kibera. Healthcare facilities included 3 different types of service, which were clinics, pharmacies and chemists. Chemists are natural healers that use medicinal processes to cure ailments. While these services represent unorthodox approaches to healthcare, they are an important cultural facility that should be included in the healthcare sector. The same workflow that was used for the previous two distance rasters was applied to this layer.
With all layers created and weights assigned, each raster was added to the weighted overlay tool. The final weights were input, and the class range was defined as 1-9. This produced a raster that showed the highest priority cells for upgrade projects. This raster ranged from a scale of 1-8, with 1 being the lowest suitability, and 8 being the highest suitability. After producing this raster, zonal statistics were used to calculate the priority score (mean value) of each informal village within Kibera, therefore aggregating results to the village level. Other statistics were calculated using the same zonal statistics tool with different input options, such as, standard deviation, minimum, maximum, majority, majority percent and majority count. The combination of a per cell and per village raster allowed for complex analysis of the highest priority areas. Limitations to this methodology are discussed in detail further on and should be examined to understand future improvements to this process.
Results
A literature review resulted in findings regarding key criteria to be included in site analysis. Toilet access was identified as a key issue within DIS, which is supported by UN-Habitat findings. During a Kibera mapping project, the organization found that toilet distribution was not equal between the 15 informal settlements of Kibera and a significant amount of waste was disposed of in rivers and roads (UN-Habitat, 2020). Sanitation issues present unprecedented public health concerns that have implications for all three facets of sustainability, indicating that toilet access should be ranked 1st. Population distribution was ranked 2nd, as it has significant implications on all other criteria. Higher population counts represent overcrowding issues and indicate a higher number of people per facility. Overcrowded areas lack sufficient living space and prioritizing high population areas should be essential to site analysis. Drinking water access was ranked 3rd, as residents that lack water points within a close distance are more likely to overpay while relying upon water vendors (UN-Habitat, 2007). This has a significant impact on economic sustainability and is essential to the economic mobility of residents within a DIS. While water access is essential to the well-being of Kibera’s dwellers, over 60% of Kibera's settlement area has access to water points within 50 meters of walking distance, which suggests it is a lower priority issue when compared with toilet access (UN-Habitat, 2020). Distance to healthcare was ranked 4th because it represents an aftercare issue when compared with sanitation and water access. Healthcare is an invaluable resource; however, toilet and water access serve as essential tools for preventive care, a measure that can stop health issues from appearing in the first place (Center for Disease Control and Prevention, 2025). For this reason, it has been assigned a higher rank than factors pertaining to the general built environment, but a lower rank than preventive measures. Building density was ranked 5th because it is less representative of specific issues and can be just as high in formal settlements. Including this criteria is important because it shows where upgradable structures are located, but does not have the same implications as those listed above. Tree coverage was ranked 6th because of its indirect impacts on living standards. The absence of tree coverage can increase the urban heat island effect and decrease carbon sequestration; however, a lack of tree coverage is less severe than the more pressing issues of sanitation and resource access.
With criteria defined, a weighted overlay analysis was performed and the zonal statistics tool produced a ranking of villages based upon priority for upgrade projects. While Kibera has 18 villages, only 15 are classified as informal settlements. The 3 formal villages have been omitted from statistical analysis and are labelled in the village map below with the word “estate.”
Results indicated that the lowest priority village was Gatwekera, one of Kibera’s central villages. Aside from healthcare access, Gatwekera has high resource access and a relatively low population when compared with the other villages. The highest priority village was identified as Laini Saba, a village in Eastern Kibera. On average, cells in Laini Saba scored a 4.55 on the priority scale. At the village scale, average values ranged from 2.54 to 4.55, with only three scoring above a 4. In addition to Laini Saba, Lindi and Toi Market both scored above a 4. A weighted analysis map has been produced at both the village and cell scale and can be used to present findings and highlight priority areas. Results indicate that the eastern and northern regions of Kibera are the highest priority areas, while central Kibera is the lowest priority. Even though Eastern Kibera does have the highest priority, the far east village, Soweto East, has a medium priority for upgrades. Soweto East has a mean priority score of 3.55, landing it in the middle of the pack. The priority score of each village is listed in Appendix C.
The cell value map reveals more precise patterns, showing vast inner village variation. Laini Saba, the highest priority village, has a minimum value of 2 on the priority scale and a maximum value of 8 (the highest score a cell could receive). The mode of Laini Saba was 4, which accounted for a total of 80 cells, or 28% of the village. Gatwekera, the lowest priority village, had a mode of just 2, which accounted for 46% of the village. Inner village standard deviation scores showed that Laini Saba, Toi Market and Mashimoni had the most variation from the mean value (priority score) of the village. Mashimoni had a standard deviation of 1.37, Toi Market had a standard deviation of 1.30 and Laini Saba had a standard deviation of 1.29. Makongeni had the lowest standard deviation, with a value of 0.73.
North West Toi Market and North West Laini Saba had the highest priority clusters, while other regions of the villages had lower priority. Lindi, the second-highest priority village, had a relatively normal priority score distribution, though the southern region demonstrates the highest cluster of high-priority cells. Some priority clusters cross village borders, particularly between Laini Saba and Mashimoni, two villages that share a clustered region. Mashimoni’s priority score was 10th out of the 15 villages and can be explained by its high standard deviation. While the eastern region shares a cluster of high-priority cells, the western region scores very low on priority by cell.
The initial production of 6 raster layers resulted in high-resolution maps of Kibera’s resource access and population distribution. Each raster is stored at the 30-meter resolution and includes accurate and precise statistical information for the related topic. Population distribution is shown per person. Tree coverage is displayed based on the percent of area covered by tree canopy per cell. Building density is shown by percent of building cover. Toilet access, water access and healthcare access are displayed based on a numerical value that represents the physical distance and cost of travel to an access point.
Resource access is clustered mostly within central Kibera, while the fringes of the settlement have the least access. The high distance to resources within Eastern Kibera can be explained by the cost raster, which indicates a high prevalence of dense buildings and lack of road access in the area. The highest cost areas are explained by a high building density and lack of road access, which is most prevalent in Laini Saba and Lindi. Red shows the highest cost to travel per cell, with green showing the lowest.
Overall, results highlighted 3 key villages for upgrades: Laini Saba, Lindi and Toi Market. Aside from Soweto East, Eastern Kibera has the highest priority villages, with high priority cells falling mostly within this area. Central Kibera has the lowest priority, as resources are clustered in this region with a low travel cost. Significant variation within village borders was found, which is supported by standard deviation calculations. Results should be focused on the cell score map to present precise priority areas that can be targeted for initial projects. Depending on the project's goals, themed maps can be analyzed for improved decision-making and project implementation.
Discussion
Kibera and Data Collection
Kibera is located near Nairobi’s city center and has 18 villages within its borders. According to a UN-Habitat Policy Recommendation report, Kibera has an estimated population of 250,000, though population estimates vary considerably depending on the source (UN-Habitat, n.d.). The dataset used for this analysis estimates Kibera’s 15 informal settlements as having a population of 118,834. A large body of research has already examined Kibera, as Kenya's Slum Upgrading Programme (KENSUP) has targeted the settlement with two projects thus far. KENSUP is supported by the UN-Habitat and plans to improve the living standards of dwellers in all of Kenya's informal settlements. The Kibera Water and Sanitation Project (K-WATSAN) and Kibera Slum Upgrading Initiative (KSUI) are both projects that have been implemented in Soweto East to improve access to water and housing, with varied results (Solymári and Czirják, 2024). These programmes can help explain why site analysis results indicated Soweto East as a medium priority village, despite the otherwise high priority eastern villages.
Laini Saba, Lindi and Toi Market should be prioritized at the village level, while different areas within these villages should be focused on first. If basing scores on village priority, Laini Saba should be the first place of action and its northwest region should be the starting point. These results make sense, given the high population count and lack of toilet access within the area, as these two criteria held the most weight in overlay calculations.
Basing upgrading initiatives solely on village rating should be avoided, as inner village variation is high. Multiple priority cell clusters cross village boundaries and it would be a disservice to ignore this fact. Based on the findings of this paper, the benefits of K-WATSAN have not reached the villages surrounding Soweto East. If projects had a broader scope thus far, they would not only have been effective at achieving their inner village goals, but also at expanding resources outwards.
While it is important to focus on all villages, there is overwhelming resource access within Central Kibera, when compared to the surrounding regions. This is important to note, as funding is limited and programmes need to target the highest need areas. Central Kibera should be targeted in the long term, but should be put on the back burner in comparison to Eastern and Northern Kibera. Findings on Central Kibera align with those of other mapping projects, as the UN-Habitat’s “Informal Settlements’ Vulnerability Mapping in Kenya” report noted that Central neighborhoods, such as Soweto West, Getwekera, Kisumu Ndogo and Kambi Muru have the highest access (UN-Habitat, 2020).
Other UN-Habitat findings have aligned with this report, as Toi Market, a village with notably low water access in this site analysis, was found to have the least access to water points. Toi Market’s low access to water points and other resources, such as toilets, may be explained by the fact that it is a commercial hub. NGOs locate themselves in their preferred localities and are not evenly distributed throughout the settlement (UN-Habitat, 2020). With that being said, Toi Market is still a populated village that needs to be targeted with high priority, as it serves as a strong site for improving the economic mobility of workers in the village.
Economic capacity and job opportunities are essential if we seek to improve the lives of those in DISs, as extreme poverty reduces the ability of dwellers to exercise their economic productivity (UN-Habitat, n.d.). Dwellers have many skills that are often overlooked by employers in the formal economy. Many dwellers are business owners and merchants within Kibera and improving their workplace resource access improves their economic productivity and upward mobility. To further the integration of the workforce into a site analysis process, data on businesses, travel patterns and land use should be integrated into the model, which is currently restricted by severe data poverty. Since DISs are often invisible to government agencies, there is a lack of formal data collection at the household and building level, therefore decreasing the capacity of decision makers to practice sustainable urbanism.
While OSM and other crowd-sourced data are effective for mapping and analyzing underserved communities, questions of completeness are often raised. In some areas of the world, OSM data can be most complete in DISs, as users target areas that are most in need. However, it is not always sufficient for the final decision-making process. When OSM data is used, completeness tests should be run to confirm the validity of the data (Porto de Albuquerque, 2021). Given the difficulties of mapping complex networks in DISs, preliminary site analysis, such as this study, can present target areas for data collection that help scale down the requirements of government data initiatives. Once target areas are defined, or if there is a capacity to do city-wide collection, detailed census data should be taken through a participatory approach.
To integrate a participatory approach into data collection, dwellers can collect data on themselves, both through surveying and data collection training (Seckel, 2018). By training residents on data collection and compensating them for their time, governments can improve the economic capacity of residents while decreasing the formal workload of their agencies. However, for this to work, measures must be taken to ensure the secured tenure of residents, as it is unlikely anyone would participate in such measures were their housing to be in danger.
Criteria and Weighted Overlay
This study defined 6 key criteria for site analysis on DISs, however, there are a plethora of other criteria that can and should be implemented to formal analysis initiatives. Aside from the criteria included within this study, socioeconomic criteria must be integrated into the weighted overlay model. Some important socioeconomic data can include household income, price of housing and length of informal land tenure, among many other factors. This is demonstrated within “Spatial Information Gaps on Deprived Urban Areas (Slums) in Low-and-Middle-Income-Countries: A User-Centered Approach,” as experts noted there was a need for data on socioeconomic conditions to supplement their analysis (Kuffer et al., 2021).
Furthering the precision and detail of existing resources is necessary for analysis, as grouping degraded and operational assets together results in distorted analysis. In the case of toilets, while general access is essential, not all produce the same benefits. In Kibera, “most households share public toilets, which are locked and require a fee of 5 to 10 shillings to unlock,” therefore rendering them somewhat useless to the majority of dwellers (Wang, 2024). When this data is publicly available, it is important that these conditions be integrated into the overlay process, therefore improving the results and reducing data distortions. Some additional criteria about the built environment should also be integrated, particularly access to handwashing stations and vulnerability assessments (Olthuis et al., 2015). It is vital that future site analysis includes a well-rounded criterion to reduce the possibility of skewed results. Even though upgrading initiatives occur over long periods, resources need to be targeted at the right locations from the start.
Utilizing site analysis and the weighted overlay model is a must for future initiatives, as data-driven decision-making is one of the strongest tools for urban governance. Informed policy decisions require comprehensive analysis and can be done with a weighted overlay model. As noted by multiple United Nations studies on upgrading programmes, such as KENSUP, all projects start with a socioeconomic mapping phase. While expert opinion is essential for analyzing the results, using a statistically driven model will always outperform the alternatives.
To improve the weighted overlay model, integrating the Analytical Hierarchy Process into the weighted overlay model can be useful for defining weights, as community stakeholder opinions can be transformed into a participatory ranking system. This allows for the effective distribution of weights among different criteria (Waheeb et al., 2023). Doing so helps decrease the subjectivity of weight rankings, a key limitation of the weighted overlay model.
A few notable limitations were found within this study and are particularly pertinent to the aforementioned issue of data poverty. Since OSM data was utilized for a large portion of the criteria, uncertainties surrounding the completeness of the data could be presented. To reduce this limitation, initial data and results were cross-referenced with UN-Habitat findings, which indicated an overall completeness of the OSM networks. Another limitation present is the lack of stakeholder input in weight calculations. While literature review drove the initial ranking process, dwellers of DISs need to be included in ranking calculations. Community organizations should also be used as stakeholder inputs, as they have extensive experience with the area of study. A final limitation was the spatial relationship between the building density and tree coverage data. Since the weighted overlay model does not account for spatial relationships between different criteria, this could skew the results of the study. However, since both criteria had relatively low weights and were important for understanding priority locations, neither were omitted.
Conclusion
Deprived informal settlements are an immense humanitarian concern for future population growth and present risks of disease, overcrowding and poverty to billions of future urban residents. As of now, a significant number of urban dwellers throughout the world are already subjected to housing, job and health insecurity, due to the conditions they face in their communities. To reduce the prevalence of these areas, while also improving the living conditions of the people who reside within them, in-situ upgrading projects need to be undertaken for urban sustainability.
Kenya, a populous country in Eastern Africa, faces high rates of DISs in multiple cities throughout its land. Nairobi, the capital of Kenya, is one of these cities, and it is home to Kibera, one of the largest deprived areas in all of Africa. While multiple measures to improve Kibera have been undertaken, they have only been so effective. Future initiatives will have to be focused on the highest priority areas, ensuring that long-term investments result in the highest return. A weighted overlay model can be used to prioritize areas within deprived communities, therefore paving the way for effective upgrading initiatives. Some criteria that can be useful for these studies can include toilet access, water access, healthcare access, population distribution, building density and tree coverage.
After performing a weighted overlay analysis on Kibera, it was discovered that Laini Saba, Lindi and Toi Market should be priority villages. Smaller-scale priority areas were found within each village, as inner-village variation was high. By utilizing the maps included in a weighted overlay analysis, decision makers can narrow down on project goals to target the most pressing issues. This will pave the way for informed, data-driven decision-making and will be at the forefront of urban sustainability in the remainder of this century.