The national scale data are valuable to inform policy and to set strategic goals such as the Millennium Goals (MG) of the Sustainable Development Goals (SDGs). However, this data, as presented in the State of World Cities report by the UN-Habitat [11
] or the World Development Indicators [12
], are compiled with estimates and not actual measurements. Furthermore, such entities measure informal settlements and development using national indexes, which can be compared across regions since they focus on the national scale; however, this method lacks detail at the urban scale. To date, work that evaluates the global scale of urban informal settlements, or slums, as they are commonly known, [4
] is based on such data. These national indexes are necessary to evaluate the evolution of informal settlements and overall development in regions. There are great examples of scholarly work using national indexes that concentrate on the role of informal settlements in the process of urbanization in Latin America [3
]. They focus on these indexes to tell the story of the region. While these studies capture the spirit of their data, they still have little reliability at the city scale. The shortcomings of the data on which this group of research is based are the result of the complexity of mapping informal settlements. Most developing nations do not have the technical capacity, or the resources, to map informal areas in detail, and even when the capability exists, some governments regulations forbid or impede the mapping, recording, or sharing of public data of such populations [13
]. Limitations on resources, or regulatory approaches in some cases, leave empty spaces where informal settlements exist on maps [16
]. In this way, entire neighborhoods are kept invisible and outside of official documents sometimes in deliberate ways [17
]. However, many other variables also contribute to concealing or revealing informal settlements, such as their relationship to topography and urban morphology [18
]. Visibility, recognition, and classification of informal settlement are crucial for communities to access to secure tenure [19
2.1. Towards an Inclusive Definition of Informal Settlements
Among the obstacles to comparative work on informal settlements is the epistemological complexity of defining what constitutes an informal settlement. Different organizations, national and municipal government agencies, and scholars disagree on which variables are determinant of an informal settlement [21
]. Defining informal settlements is a complicated endeavor; the literature on informal settlements, both current and historic, spends a significant amount of time attempting to determine the limits between what is formal and informal [22
]. Authors call attention to the need for more specificity in the way the phenomena of informal settlements are defined and classified [27
]. International agencies find it challenging to assess indicators of informal settlements, as currently defined, and examples of such problems can be found in the Sustainable Development Goals [1
]. No consistency of evaluative variables exists between different countries or cities [29
]. At meta-level, informal settlements are the most common process of city-making. Some authors agree that “urban informality is hence conceived as a negotiation process through which spatial value is produced” [30
]. However, characterization is still an elusive subject because the definition of informal settlements can have social, economic, or political implications. Each definition of informal settlements has its limitations. For the purpose of exposing the various forms of informal settlements globally, we argue that the more specific the definition is, the less we can know about the phenomena at a global scale. We take this stance with the understanding that the more open the description is, the higher level of variability is introduced into the sample, and thus less comparability can be found. Even at the city scale, we can find variations in determining what is considered informal. A previous study pointed out that “in Rio de Janeiro, what defines a favela changes over the years and depends on the state initiative at the time” [10
]. In Medellin, areas marked as informal by city officials in the 70s are not recognized as such by the same city department in the next decade, even without any upgrading happening [10
On a global scale, the definition problems start with the multitude of names used to label informal settlements; each country in the world has a different word to define what an informal settlement is. Nor do scholars agree on a precise definition of informal settlement. Some authors define informal settlements, for example, as the predominant characteristic of the urban form of African cities [31
]. However, informal settlement as urban phenomena can exist in every continent. A non-contested defining feature is that a “slum” is a group of buildings and not a single one [32
], and a mainly urban phenomenon even if rural areas have deficient housing conditions that could be labeled as informal.
The most common description of informal settlements portrays them as areas where there is a lack of basic infrastructure, poor housing, illegal dwelling, non-secure tenure, high urban density, lack of sanitation, poverty, and exclusion [33
]. The United Nations defines “slums” with qualitative measures, such as the lack of durable housing [34
]. Informal settlements in literature are traditionally presented as the other side of the coin from the formal. Therefore, they are defined by the lack of features present in the formal city. In this tension between the formal and the informal, the settlements can be established as places lacking access to those faculties of the formal city such as access to safe water, access to improved sanitation, durable structures, sufficient living area (overcrowding), and access to secure tenure [35
]. We recognize these remain vague characterizations of the complexity that informal settlements embody. Yet, in studying the phenomena, we start from the current state of knowledge, in its many forms, and contribute new insights to refine our understanding of informal settlements.
presents several variables applied as defining characteristics of informal settlements across multiple scholars and institutions [10
]. The table presents intersections and contradictions within the multiple definitions of informal settlements. It reveals the challenges of choosing one definition over another. Depending on the agency, government, or scholar that is looking at the phenomenon, an informal settlement can be classified in many ways. Adding or subtracting variables would inevitably delete or add places. Scholars argue about the difficulties of creating generalizations of the informal settlements phenomenon because of the lack of data across continents. Taubenböck and Kraff argue that “a systematic approach to measuring morphological characteristics of slums in different cities across the world beyond test cases is still absent” [41
]. For this study, which engages a global morphological approach to map informal settlements, the ultimate goal is to create a database that permits evaluation across multiple cases to build theory through cross case comparison.
The goal of the Atlas of Informality (AoI) is to create a globally representative database for informal settlements and a methodology from the sample categories that can be used across the globe. This research then seeks to use an inclusive definition of what an informal settlement is. The AoI database incorporates any case designated by a scholar, practitioner, publication, or community as an informal settlement in any of their multiple definitions and names. We selected this overarching way to determine what an informal settlement is and to provide a tool to incorporate the broad spectrum of informal settlements as an urban form globally in the effort to create a consistent and inclusive mapping.
2.2. Mapping Informal Neighborhoods Globally
Despite initiatives to diminish the prominence of informal settlements as an urban reality [45
], on a global scale, it is estimated that the number of informal settlements continues to grow. According to the United Nations Economic and Social Council’s evaluation of the progress of the Sustainable Development Goals (SDGs), “despite some gains, the absolute number of urban residents who live in slums continued to grow, owing in part to accelerating urbanization, population growth, and lack of appropriate land and housing policies” [1
]. While percentages have diminished over the last 14 years by approximately 9% [45
], the total number of world slum dwellers has increased by 11.11% in the same period, some estimating the increase as high as 28% [46
]. The World Bank assessment shows a dramatic reduction in slums per region over recent decades, from 35% in 1990 to 20% in 2014 in Latin America and the Caribbean; from 67 to 55% in South Saharan Africa; from 57 to 31% in South Asia; and 47 to 26% in East Asia and the Pacific in the same period; and from 39 to 28% in the Middle East and North Africa in 1990–2005 [47
]. However, this macro-level analysis seems to hide within the definition of “slum” the real percentage of change of such spaces.
Existing data about informal settlements are not available, accurate, or complete and, in most cases, obsolete [48
]. The mapping of cadastral areas in informal settlements is crucial for the transformation of these areas [49
]. Imprecise data about the scale of informal settlements hinder agencies, city officials, and scholars’ efforts to inform appropriate policies about the phenomena of informality. The UN-Habitat highlights a lack of data at the sub-city level [50
] as one of the challenges in dealing with urban poverty globally. The question then is: what is the best way to solve this vacuum of information? The AoI response to this problem lies in the intersection of these methods. The main goal is to systematically collect data at the same resolution level and over the same period, to apply when possible remote sensing tools but using manual input so it can bypass today’s technological bottleneck.
The mapping of informal settlements can be divided into five methodologies: community mapping, single case selection, national indexes, remote sensing, and urban morphologies. Each one of these methods has its virtues and challenges. Community mapping is born out of the vacuum of information and the need for communities to create data to make the needs of their neighborhoods visible. Since state agencies fail to map or abstain from mapping informal settlements’ living conditions, among the most effective systems to gather reliable data is self-mapping by communities [51
]. This self-mapping produces a precise level of detail of living conditions. Along with the mapping data, the community benefits from the empowerment that emanates out of community self-identification. However, this happens haphazardly in a neighborhood or, in a few cases, a citywide scale, without consistency across communities. The generalization of community mapping is limited since most maps reflect local needs. The isolation of communities also contributes to the accessibility of the information for groups or individuals not related to these mapping exercises. Another methodology of mapping is single case selection; these are the mappings by scholars, NGOs, and municipalities that focus on the neighborhood or city as a case [10
]. Here, the output is more detailed, as new aerial photography and remote sensing data make the creation of highly detailed maps of neighborhoods possible. Unique case mapping allows bypassing state regulations since non-state agencies can perform them. A disadvantage that this project exposes is the challenge of generalizability of the single cases. National indexes lend themselves to generalization; these mappings draw on the capacity of nations to collect demographic data based on variables of census tracts that then get compared at a global scale. The data collected at this scale are the data of the international agencies such as the World Bank, United Nations/UN-Habitat, and the International Development Bank. However, the broad nature of national indexes is characterized by low-resolution data at the urban scale. National indexes data such as census undergo a filtering process to protect individuals and populations, and these data separate the information from the geographic location of communities to levels that make it difficult to relate this information to informal settlements at the neighborhood scale. However, most of the scholarly understanding of the informal world at a global scale comes out of this type of data [4
]. Remote sensing (RS) technologies bridge the gap between comparability and the level of detail necessary to create global analyses [41
]. The use of high-resolution satellite data permits researchers to visit multiple sites and apply variables to analyze the qualities of urban form [56
]. The use of algorithms applied to this data, such as object-oriented, radial casting, and contour model (snakes), permits extraction and recognition of unique features of the landscape [48
]. These mapping projects use these algorithms alongside remote sensing imagery to collect, identify, and map informal settlements. GIS literature in remote sensing focuses on how to train algorithms to determine the location of informal settlements [62
]. The RS method helps identify any unknown settlements, particularly when there is not sufficient information about the locations of these settlements. However, old consolidated settlements or settlements with regular urban patterns are difficult to pinpoint with such an approach. RS mapping is perhaps the most promising upcoming technology. Today, there is an emergence of RS-based studies focused on exploring the morphological features of informal settlements to identify informal settlements, much development has occurred in the automatic classification of locations from optical resources [60
], and other methods such as the use of radar have also shown promise and success [66
]. RS automated methods have also demonstrated validity in the feature detection in landscape archaeology [68
] and detection of anthropogenic geomorphology [70
]. And the use to identify environmental risk in traditional unmapped areas such as in archeology [71
]. Spatial-contextual information can also be incorporated through Object Based Image Analysis (OBIA), which is also currently the most common strategy for the classification of informal settlement areas [63
]. The use of Very High Resolution (VHR) with the help of the Level Co-occurrence Matrix (GLCM) has permitted the extraction of informal settlement areas [74
]. However, this method still has limitations. So, new capabilities have helped automatize the classification from machine-learning (ML) [75
] to computer vision using convolutional neural networks (CNNs); this approach has demonstrated higher accuracy for detecting informal settlements from VHR images [77
]. Local directional pattern (LDP) has also provided advances over the traditional gray level co-occurrence matrix (GLCM) [78
]; a challenge in VHR arises in the limitation of up-to-date thematic maps of informal areas for which the use of Unmanned Aerial Vehicles (UAVs) has served as a tool to breach the time gap but increase the work at the ground level [79
]. After all advances in RS automatization, manual delineation still offers advantages over automatic classification achieving an overall accuracy of 80.5% when compared to manual delineation [80
]. At the rate of growth of the quality of imagery available and the perfecting of RS methodologies in the future, mapping of any urban form would be a wholly automatic endeavor. However, even as this promise exists, at present, there is still no global mapping of informal settlements or a systematic inventory of its morphologic types across the globe [81
]. Finally, Urban Morphology (UM)—the morphology of the built environment, in particular that of the informal settlements, shares unique characteristic patterns across settlements: scale, size, shape, and distribution [82
]. Texture measures can be potentially used to represent the contrast between planned and unplanned settlements [83
]. Analyzing the urban environment’s physical features such as green space, structure of layout, density of built-up areas, and size of buildings can help to identify heterogeneity in urban areas that helps in the determination of sub-standard residential areas (informal settlements) [84
]. The self-built process in which informal settlements are created tends to follow particular morphological processes [85
], which can be analyzed using methods such as space syntax [86
]. Interest in the urban form of informal settlements developed in an emerging body of knowledge on the morphologies and morphogenesis of informal settlements [8
]. An emerging body of work focuses on exploring the change in the morphology of informal settlements [39
]. Variation over time in the extent of informal areas presents challenges for planning and city management [91
]. A second effort of the AoI database is to investigate the variable of time—in particular, how the land surface occupied by informal settlements varies with time. If one of the defining features of informality is their ever-changing nature, what does that change represent on a global scale? Focusing on areas instead of population percentage estimations can provide a more empirical approach to understanding the changing rates of informal settlements globally. It can also direct the establishment of guidelines for urban policy and adaptive capacity.
2.3. A Methodological Approach to Map and Measuring Informal Settlements
Mapping informal settlements is a complicated endeavor, new tools have provided ways to do so. However, there is still some level of imprecision that stems from the non-uniform nature of informality globally [81
]. How can the nature of growth be captured in these places, and how can the data be collected from such different environments? Currently, digital mapping and new sensing technology have created new opportunities to map informal settlements using remote tools while reducing the gap of traditional studies, allowing mapping of cities and particular urban forms of informal settlements using the same methodology that opens the opportunity for comparability [41
]. However, new sensing technology is limited in its capacity to differentiate informal settlements from formal city spaces. For this project, we took a different approach: we developed an inclusive definition of informal settlements then selected a few significant cases per city/country as examples of informality in that place. By inclusive definition of informality, we mean that we chose a process that permits capturing an informal neighborhood in any of its definitions. In this research, we first included neighborhoods already designated as an informal settlement by a state, city, community, or scholar; we looked for geographic locations and also for mapped areas with defined settlement boundaries; in addition, we used morphological features of those places such as homogeneous scale and texture of constructions to further determine settlement boundaries. Although imperfect, we chose this method because it permitted us to capture the various ways in which the phenomenon of informality is presented around the world.
For the mapping process, we followed a five-step approach: (1) crowdsourcing of cases, followed by (2) the creation of a mapping protocol, distributed across a similar number of undergraduate environmental design students—one case per student for (3) the application of the mapping protocol with direct mapping. (4) A curation of data followed that included a peer-review process examining completeness and validating the correctness of measurement for each case. This ensured each case complied with the standard of the Atlas. Finally, (5) data compiled with the mapping using Google Earth Pro (GEP) was transferred to Arc-GIS for analysis.
Crowdsourcing of cases: Informal settlements are neighborhoods; we chose this as the unit of analysis. The project endeavored to capture the phenomenon in its various global forms. We aspired to capture a balanced number of cases per city and country, aiming to reach most countries where informal settlements are present. For the initial selection of cases to the AoI, a crowdsourcing process was used to obtain the first set of examples. The idea was to start with the most known cases, those included in the informal settlement literature, those known as the exemplary cases in each city, and those that scholars have already identified through each one of the multiple definitions of informal settlements. We invited scholars of informal settlements to collaborate on the creation of the first list. Additionally, a literature review and web search produced a total of 405 cases. The main objective of the informal settlement selection was to cover most of the world’s countries and cities using one or a few representative cases per city.
To find and corroborate the classification of settlements as informal, a Google Scholar search was performed using keywords such as the name of country or city and the words slum and informal settlement, and any local term used to identify an informal settlement in that geographic area, such as Tugurios, Favelas, Bidonvilles, Chabolas, Pueblos Nuevos, Coreas, Barracas, Kampung, Morros, Ashwa’iyya, Squatters, or Shanty Towns. As the sample grew, priority was given to places (nation/cities) not already covered in the selection. For vacuums in the literature, we used a Google search engine entering the same keywords in popular media (local and national news organizations), state reports (local and national), and organization reports (such UN-Habitat, World Bank, IDB). We intentionally avoided over-representation of places with robust informal settlement mapping already in place. For example, there is vast information about informal settlements in Latin America; Rio de Janeiro in Brazil has ample GIS shapefiles that map the boundaries of each one of its 1040 favelas [20
Complete and accurate mapping of informal settlements is difficult since these special urban forms are constantly changing. As a result, finding comprehensively mapped places is uncommon. However, there are several cases that demonstrate that it is possible. Samper’s historical research of informality in Medellin presents a compilation of maps identifying informal settlements created by the Departamento Administrativo de Planeación (planning department) of the city of Medellin from 1957 to 2014 [10
]. The Instituto Municipal de Urbanismo Pereira Passos (IPP) In Rio de Janeiro, Brazil, maintains a database with mappings from 1999 to the present that includes favela limits, updated regularly by the Sistema de Informações Urbanas (SIURB) of the mayor office (Prefeitura da Cidade do Rio de Janeiro) free to access online [92
] The city of Bogota maintains a record of asentamientos humanos
(Informal Settlements) accessible at the open access data page of the Colombian Government [93
] and updated as part of their city master plan (Plan de Ordenamiento Territorial POT). The organization Techo has a database of open access of Asentamientos (Informal Settlements) for countries such as Chile, Argentina, Colombia, and Paraguay. However, for this research, the goal was to create an equal distribution across the world, to the extent possible with the existing tools and resources.
The Mapping Protocol
: The protocol contains a detailed set of instructions on how to find the assigned settlement, determine boundaries, map the current settlement boundary, review historical aerial images, and redraw the settlement boundary, recording dates of change observed. Beginning with the location identified through search and corroboration processes previously described, locations were loaded into GEP for close examination of urban form using the satellite imagery. Specifically, we examined the density of structures as compared to other parts of the city, the regularity, or irregularity of the division of space (grid vs. curvilinear, buildings seemingly overlaid vs. those with distinct edges), and materiality (appearance of concrete, steel, glass, and wood vs. piecemeal materials with less sturdy appearance). The protocol also includes guidance on how to use the GEP software and InDesign layout to record sources and additional information. In addition, protocol included modeling of sample areas in SketchUp software of each settlement for area density comparison; however, that part of the data is not included in this paper. Students were instructed on the use of the software and advised on ways to find more resources to determine the extension and location of the settlement (for a sample of mapping protocol see Supplemental Material
). Since data sources varied, this entailed one of the most time-consuming parts of the research. Several scholarly papers refer to informal areas in cities; however, they give little guidance on the location of those areas of research in the cities.
For the actual mapping, a combination of remote sensing in the form of satellite historical images and direct mapping was selected as the most viable mapping method. Using readily available satellite photography provided a standard base for each entry. Google Earth Pro (GEP) provided a widely accessible data set that could be used across cases and did not require the purchase of mapping software. An ongoing goal of the project is to continue additional mapping to reach other stakeholders to map additional settlements. Using direct mapping allows us to define clear limits based on the nuanced variations in urban form and triangulate information from different sources that cannot automatically be captured by a learning algorithm. For example, this method allows identification of temporary structures that do not represent the expansion of the settlement, differentiation between city boundaries and settlement encroachments, and inclusion of areas that have unique morphological urban development patterns that are not similar to those in informal settlements in other regions, such as the case of informal areas in Mongolia, whose units of circular shapes differ substantially from settlements in places such as India. Computer vision can be effective at identifying different urban forms, and it is advantageous to distinguish formal from non-formal settlements. However, it is more difficult for that software to determine differences between similar urban forms, such as the limits between two adjacent non-mapped informal settlements. To make determinations on these ambiguities, humans tend to be more proficient; they can interpolate non-geographical resources (such as reference to an academic paper or a media publication) to make such determinations. The limits of a neighborhood are not only based on the unique morphological features but also on their history and its members’ demarcations. For example, in Colombia, the informal settlement of Independencia in on the city map as a single place; however, it is comprised of three different neighborhoods for its residents, each one with a unique history and particular local governance structure. This social difference has determined a different morphological evolutionary process for each area [10
]. This project used the two sources of the morphological identification applying RS and the secondary academic and media sources to shed light on those features not readily identifiable by following only the morphology. The academic and media sources permit the identification of geolocation of settlements, and in some cases the estimated boundary of the settlements. The morphological exploration looks at homogenous urban typologies of settlement build texture to determine a more accurate measure of limits [83
] This morphological exploration serves to create a baseline measurement to explore changes over time. We acknowledge that with more technological tools, staff, and budgetary resources, all these issues could be automated; however, at this stage of the research, we do not have those resources. For this project, we used manual digitization from high-resolution satellite images, a data methodology used by other mapping projects [94
] to compensate for the lack of precision found in other remote sensing tools [63
The first step of the protocol was to find the settlement and determine its boundaries (Figure 1
). Determining the limits of an informal settlement is not a precise endeavor [96
]. However, informal settlements present significant morphological differences, contrasting formal areas that surround them [34
]. For this project, we determined that in addition to knowing locations, to distinguish the boundaries of an informal settlement we would use those physical characteristics of continued urban patterns, unregulated housing distribution, high density, and small (substandard) building sizes. This analysis of spatial patterns using space syntax tools has been used to identify Favelas in Rio de Janeiro [98
]. The direct measure is a more labor-intensive process; however, it provides a more accurate and standardized way to collect the required information. Other forms to determine the limits are municipal maps, scholarly articles, and web maps. The key was to find the boundary of the settlement at the time of mapping. At this stage, the selected researcher would find the exact location and perimeter of the settlement using GEP and would geolocate the site by drawing its current boundary, collecting a 4k resolution aerial image of the site as a reference. After identifying and mapping change for each settlement, we coded surrounding physical features for each entry on the sample to understand how the physical context influences how settlements expand. We codified physical characteristics that could affect or impede settlement expansion based on the three physical variables: surrounding development, topography, and water features. For surrounding development, we evaluated if the urban form permitted the settlement to expand; for this, we classified each settlement as open or closed depending on how the settlement perimeter abutted open land or developed areas. We used the same procedure for water features, visually verifying using the GEO databases for features that represented water limits such as oceans, lakes, or rivers. For topography, we used the GEP database and the World Topographic Map which compiles data from U.S. Geological Survey (USGS), U.S. Environmental Protection Agency (EPA), U.S. National Park Service (NPS), Food and Agriculture Organization of the United Nations (FAO), Department of Natural Resources Canada (NRCAN), HERE, and Esri. Data for select areas were sourced from OpenStreetMap contributors. The map provides coverage for the world down to a scale of ~1:72k. Coverage is provided down to ~1:4k for the following areas: Africa, Australia, and New Zealand; Europe and Russia; India; most of the Middle East; Pacific Island nations; Alaska; Canada; Mexico; South America and Central America. Coverage is available down to ~1:2 and ~1:1k in select urban areas. Settlements located in, or adjacent to, areas with more than 30% slope were classified as having topographic limitations to development.
The second step of the protocol focused on what we considered the most salient and unexplored feature of informal settlements: their changing nature [8
]. Studies of multitemporal change to map growth in informal settlements in Kenya used aerial images over multiple years to help determine the variation of the urban form [39
]. A standard tool to measure urban change is multispectral imagery, used to differentiate built areas from the rest. Vast databases of this information exist for the U.S and Europe, making them a great tool for examining cities in these places with spatial resolutions from 15 to 30-m. However, for the rest of the world, resolution varies to a scale of a 60-m resolution, which made this method inaccurate to measure urban change at the neighborhood scale globally. The Landsat database offers global coverage, but resolutions are different depending on place, varying from 30 to 60-m resolutions. Other sources such as Sentinel-2 offer better resolution—10 m. However, this is still considered low for the study of informal settlements when built units can be as small as 2 m [76
]. In this case, the mixing method of VHR and LR can provide appropriate results if used in the context of Machine Learning. A DigitalGlobe 30 cm VHR image offers an adequate and accessible form. We chose the GEP database because it had an acceptable resolution, is free to use, and permitted the easy training of personnel. It allowed us to set up the protocol as a tool that can be deployed to local communities worldwide in an accessible manner.
For this research, we selected manual digitization. While this method is labor-intensive and has its limiting factors, the inclusion of a human in the decision making process about the limits of a settlement permitted inclusion of a case by case evaluation of other variables that were not easily accessed through RS, such as political delimitation, infrastructure projects impacts, history of the place, and cultural approaches to urbanization. Particularly since the majority of the sample is located in the global south, the employment of available multispectral imagery created a low-resolution base map from which to create a detailed measure of urban change at the neighborhood scale. Figure 2
shows an example of mapping using this method in the settlement Independencias in Medellin, Colombia, using LandsatGLS-multispectral from 1990 to 2010 using the setting false-color. According to ArcGIS, “multispectral Landsat GLS image service [has been] created from the Global Land Survey (GLS) data from epochs 1975, 1990, 2000, 2005 & 2010. GLS datasets are created by the United States Geological Survey (USGS) and the National Aeronautics and Space Administration (NASA), using Landsat images. This service includes imagery from Landsat 4, Landsat 5 TM and Landsat 7 ETM, at the 30-m resolution and Landsat MSS at 60-m resolution, bands 5,4,1. It can be used for mapping and change detection of agriculture, soils, vegetation health, water-land features, and boundary studies” [102
]. This method presents a low scale of detail. For this reason, the researchers selected historical aerial images as a way to consistently determine and measure change. The GEP historical aerial images database compiles files from multiple sources. The older the aerial images provide less consistency between the places and times of recording of the imagery. This inconsistency of base imagery creates differences when the delimiting of the perimeter occurs. Since it is not possible to have maps for every year, in this project, we standardized mapped samples by dividing the area of change between the first measure and the final measure of the perimeters by the number of years between measurements. This standardization method erases the uniqueness of how deceleration or acceleration of the process of urban area change happens. However, it permits a comparison of mapped areas at the same level. GEP imagery spatial resolution varies from 6” (15 cm) to 12” (30 cm); the positional accuracy (CE 90) is of less than 1 m (accuracy will achieve 1-m CE90 in most areas. Accuracy may not meet 1-m CE-90 in areas of significant relief due to Digital Elevation Model-related errors) [103
The goal of this task was to find evidence of growth in the chosen area. Using the time tool in GEP, the protocol directed student researchers to record perimeter change in each settlement for each site at four time points across the 20 years of historic imagery available, starting with the first available aerial image and ending with the most updated one. Since urban change is not consistent and marking perimeters for each year not possible, four inflection points were selected per case—four moments in which the settlement area changed (see Figure 3
). For settlements with no variation, a new perimeter was also drawn as evidence of no change. GEP data vary for each city but high-resolution imagery is available between 1990 and 2020. Using GEP, available imagery provided a restricted period to examine the process of growth across the settlements identified. However, the foundational date in 94% of settlements was prior to the years available in aerial images of GEP databases. This restriction presented the opportunity to see how mature informal settlements evolve. The final step was the recording and filing of all data gathered in the previous steps alongside measures collected by other sources. These include year of settlement foundation, area, and any additional images readily available on the web.
After mapping all settlements, the research team performed a review of mapping correctness and standardization of the sample. Each entry location was manually reviewed and approved, corrected, or deleted from the AoI sample. Mapping correctness, in this study, was established through a peer review process to determine that the place mapped was considered an informal settlement and that its limits were accurately drawn. For determination of area as an informal settlement in any of the accepted definitions, a reference to a document affirming such determination must be provided. For limits, the student researcher provided geographical or bibliographical evidence that helped to determine an accepted delimitation of project boundary in the current year. For historical imagery, four aerial photographs were used to assess the evolution of area change. Then, the oldest image was selected as a starting point, and a new limit was drawn. The peer-review process forces determination of settlement boundaries to be reviewed by, at a minimum, three people before it is included in the AoI. From the original 405 entries, 145 were removed during the review stage as they did not meet all the requisite criteria, such as evidence of informal settlement determination or agreement of perimeter limits between researchers. Selected settlements locations and areas first and last mapped extension can be found and explored in the webapp for the AoI: www.atlasofinformality.com
(see Figure 4