Assessing the impact of building footprint dataset choice for health programme planning: a case study of indoor residual spraying (IRS) in Zambia

Assessing the impact of building footprint dataset choice for health programme planning: a case study of indoor residual spraying (IRS) in Zambia | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Assessing the impact of building footprint dataset choice for health programme planning: a case study of indoor residual spraying (IRS) in Zambia Heather R. Chamberlain, Derek Pollard, Anna Winters, Silvia Renn, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5898324/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 24 May, 2025 Read the published version in International Journal of Health Geographics → Version 1 posted 9 You are reading this latest preprint version Abstract Background The increasing availability globally of building footprint datasets has brought new opportunities to support a geographic approach to health programme planning. This is particularly acute in settings with high disease burdens but limited geospatial data available to support targeted planning. The comparability of building footprint datasets has recently started to be explored, but the impact of utilising a particular dataset in analyses to support decision making for health programme planning has not been studied. Here, we quantify the impact of utilising four different building footprint datasets in analyses to support health programme planning, with an example of malaria vector control initiatives in Zambia. Methods Using the example of planning indoor residual spraying (IRS) campaigns in Zambia, we identify priority locations for deployment of this intervention based on criteria related to the area, proximity and counts of building footprints per settlement. We apply the same criteria to four different building footprint datasets and quantify the count and geographic variability in the priority settlements that are identified. Results We show that nationally the count of potential priority settlements for IRS varies by over 230% with different building footprint datasets, considering a minimum threshold of 25 sprayable buildings per settlement. Differences are most pronounced for rural settlements, indicating that the choice of dataset may bias the selection to include or exclude settlements, and consequently population groups, in some areas. Conclusions The results of this study show that the choice of building footprint dataset can have a considerable impact on the potential settlements identified for IRS, in terms of (i) their location and count, and (ii) the count of building footprints within priority settlements. The choice of dataset potentially has substantial implications for campaign planning, implementation and coverage assessment. Given the magnitude of the differences observed, further work should more broadly assess the sensitivity of health programme planning metrics to different building footprint datasets, and across a range of geographic contexts and health campaign types. Building footprints Malaria Microplanning Geospatial Satellite imagery Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Background Health programme planning is an inherently geographic problem, requiring an understanding of populations and infrastructure to estimate where, when and what resources are required. From provision of specific services to planning national vaccination programmes, information on target populations and where they are located is key (Chaney et al. (2021), Tatem (2022), Ali et al. (2020)). The COVID-19 pandemic brought health programme planning into the spotlight, with efforts needed to rapidly plan and implement programmes for patient care, testing and subsequently vaccination (Hengel et al. (2020), WHO (2021)). Effective health programme planning for a range of communicable diseases, informed by an understanding of populations and infrastructure, far predates the COVID-19 pandemic though. Early documented examples include the 1854 London cholera outbreak, where the Broad Street pump was identified as a source of infection following collection of data on cases in relation to the resident population in the vicinity (Snow, 1856). More recently, comprehensive mapping of settlements and tracking of vaccination teams to ensure high coverage rates of target populations, contributed to the elimination of wild polio virus in northern Nigeria (Barau et al. (2014), Touray et al. (2016), Borowitz et al. (2023)). To support a geographic approach to health programme planning, detailed data are needed to understand the spatial distribution of populations, infrastructure and the built and natural environments (Dummer (2008), Chaney et al. (2021)). In the early stages of COVID-19 mass vaccination campaigns, such data enabled assessment of population access to vaccination facilities and identification of where long travel times may be an impediment to a population being vaccinated (Muchiri et al. (2022), Whitehead et al. (2021), Leithäuser et al. (2021), Rader et al. (2022)). Similar analyses have highlighted gaps in provision of health services (Macharia et al. (2017), Dotse-Ggorgbortsi et al., (2022), Hierink et al. (2023)) and vaccination programmes for many communicable diseases (Joseph et al. (2020), Utazi et al. (2019), Utazi et al. (2024)). The availability, completeness and quality of high-resolution geospatial data on populations, settlements and infrastructure however varies between countries. In settings with well-developed geospatial data systems and comprehensive national address databases, authoritative data on the locations of residential populations may be readily available, whilst in other settings such data may be very limited or outdated, if it exists at all. Low-income countries tend to have less well-developed geospatial data systems and high burdens of communicable diseases, resulting in a high demand for planning health programmes and campaigns, but limited geospatial data to support this planning. Developments in computing power, satellite imagery and machine-learning algorithms over the past decade have enabled the growth of new sources of geospatial data relevant for health programme planning, including increasingly detailed land cover (Brown et al. (2022), Karra et al. (2021), Zanaga et al. (2022)), settlement (Pesaresi and Politis (2023), Esch et al. (2022)) and population (Doxsey-Whitfield et al. (2015), Sorichetta et al. (2015), Freire et al. (2016), Tiecke et al. (2017)) datasets at global and continental scales. In the past five years, advances in feature extraction algorithms, combined with availability of high-resolution satellite imagery, have enabled detailed mapping to the level of individual buildings at scale (Chamberlain et al., 2024). Several national and multi-country building footprint datasets have been published, predominantly by commercial companies such as Google ( https://sites.research.google/open-buildings ), Microsoft ( https://www.microsoft.com/e n-us/maps/building-footprints) and Ecopia ( https://www.ecopiatech.com/global-feature-extraction ). For many countries, these new datasets have provided building footprint data for the first time without extensive manual digitisation efforts. Prior to these developments, no national building footprint datasets from authoritative sources were openly available for many countries with high burdens of communicable diseases, including all countries in Africa (Biljecki et al., 2021). In their most basic form, building footprints consist of mapped digital outlines of buildings (polygons) and provide information on the location, size and shape of buildings. Some building footprint datasets may also include additional attributes, such as a building’s type or height. They enable building and settlement locations to be mapped, and their proximity and spatial arrangement with neighbouring buildings understood (Jochem et al., 2021). The potential of new sources of high-resolution geospatial data in planning and delivery of health campaigns is increasingly being harnessed, supported by published guidance such as the Geo-Enabled Microplanning Handbook (WHO-UNICEF COVAX GIS Working Group, 2023). Geospatial datasets are also being integrated into user-friendly tools such as the Reveal platform ( https://revealprecision.com ), which supports health campaign planning and delivery by integrating geospatial data on building footprints, points of interest, and health and administrative boundaries into an interactive map interface (Pinchoff et al. (2016), Kamanga et al., (2015), Martin et al. (in review)). This is accompanied by an interface to plan out campaigns including calculating commodity requirements and developing microplans (detailed plans for delivery of health interventions to target populations (WHO (2008), WHO (2019)). During campaign delivery, teams can use the map interface to navigate to and record data against individual buildings, acting as tasks for field service delivery teams, enhancing coverage and enabling campaign progress to be monitored and visualised. The Reveal platform was initially developed to support malaria control programmes but has been used for a range of health interventions, including geo-enabled microplanning for vaccination campaigns and mass drug administration (Winters et al., (2024), Arambepola et al. (2021)). Vector control measures are a mainstay of malaria control programmes, with both long-lasting insecticide-treated nets (LLINs) and indoor residual spraying (IRS) being widely used (WHO, 2023) to reduce vector abundance and minimise transmission. LLINs are intended to provide protection for household-members sleeping under a net, by physically preventing individuals from being bitten by infected mosquitoes whilst asleep. Insecticide-treated nets kill mosquitoes which land on the net, thus helping to reduce the vector population and providing some degree of community-level protection, in addition to the immediate household-level protection. In contrast, IRS does not prevent individuals from being bitten by an infected mosquito but instead is intended to kill endophilic mosquitoes that rest on walls of residential buildings after feeding, and in this way reduce the vector population and disrupt the malaria transmission cycle. For IRS to be effective in reducing malaria incidence in a community, high- and uniform-coverage is needed; a minimum threshold of 80% coverage of eligible structures is commonly used as an operational target (WHO, 2023). For IRS and LLIN campaigns to reach the high levels of coverage needed, data on the count and location of households, residential buildings and population are required. The availability of such data can be limited and quickly become outdated. Surveys or household listing exercises to collect new data are expensive, time-consuming, and resource-intensive, but until recently were the sole option for updating such data. The time and resources required for enumeration from satellite imagery in comparison to field-based enumeration, in the context of planning and monitoring IRS campaigns in Zambia, was found to be 22 times faster, and 10 times less costly than field-based enumeration (Kamanga et al. (2015)). High-resolution geospatial data on population, settlements and infrastructure, have been used extensively in studies on malaria risk, control and elimination (Hay et al. (2010), Reid et al. (2010), Gething et al. (2011), Clements et al. (2013)), and are increasingly being used to support programme planning for malaria (Pinchoff et al. (2016), Bridges et al. (2018), Borkovska et al. (2022), Kyomuhangi et al. (2023)). The need to utilise good quality data to maximise malaria control impact through sub-national targeting, is emphasised in the WHO High Burden to High Impact (HBHI) initiative (WHO, 2018). New building footprint data products can help in maximising coverage of IRS campaigns by estimating the volume of insecticide and number of spray teams required, supporting spray teams in locating buildings, enabling identification of missed buildings and subsequent calculation of coverage rates.It can also help in identifying vulnerable communities that are remote and/or hard to reach, and therefore often missed during health service delivery. The growth in building footprint data products has brought new opportunities for detailed geospatial data to be utilised in supporting health programmes, but the degree to which these building footprint data products are interchangeable has only recently started to be explored (Chamberlain et al. (2024); Gevaert et al. (2024)). Knowledge of the impact of utilising different building footprint data products in subsequent analyses and decision making remains limited. In this study, we explore the impact of utilising different building footprint data products in analyses to support health intervention planning. We focus on Zambia, with a case study on planning implementation of malaria vector control measures and in particular the identification of priority settlements for IRS campaigns. Our results show that when the same criteria are applied with different building footprint datasets, there is considerable variation in the number and location of priority settlements that are identified. Methods To quantify the impact of utilising different building footprint datasets in health programme planning, we use the example of IRS campaign planning in Zambia. To identify priority locations for IRS campaigns, we employ target criteria used in recent campaigns, conducted by implementing partners of the National Malaria Elimination Programme (NMEP). The target criteria relate to the area of, distance between and count of building footprints, and we apply the same criteria with four different building footprint data products. We separately apply two methods to identify suitable settlements or clusters of building footprints that meet the target criteria for potential inclusion in IRS campaigns. The first approach utilises existing mapped data on settlement extents and the second derives clusters of structures, solely based on the building footprint data with a proximity threshold applied. Study setting There were 7,050,968 malaria cases in Zambia in 2021, with a case incidence of 340/1,000 population (NMEC, 2022). Malaria risk in Zambia is spatially heterogeneous, with high incidence particularly in the north, north-west and east of the country, as well as in some areas near national borders. The NMEP has used both IRS and LLINs as vector-control strategies in recent years. IRS was prioritised as a primary vector control approach alongside LLINs in the National Malaria Elimination Strategic Plan (NMESP) for 2017–2021 (NMEC, 2017). The current NMESP for 2022–2026 prioritises LLINs as the primary intervention for vector control, with IRS targeted in high burden areas (NMEC, 2022). Geospatial data on buildings has been used to plan vector control activities under both strategic plans. District IRS planning and implementation maps, which included counts of manually digitised building footprints and malaria risk data per settlement, were first used in 2014 in 15 districts in Luapula and Central provinces to prioritise available IRS resources (Kamanga et al. (2015)). In 2015 and 2019, these building footprint enumeration maps were expanded to 42 districts, across five provinces, to support IRS planning. Sources of building footprints used during this time included manual delineation and enumeration of buildings from satellite imagery, field-verified footprint data and OpenStreetMap. The development by Ecopia of the first multi-country building footprint dataset for sub-Saharan Africa in 2019, provided building footprints nationally for Zambia (under a licence permitting use for humanitarian purposes), without requiring bespoke exercises to manually digitise buildings. The Ecopia building footprints enabled counts of buildings and derived settlement extents (CIESIN et al., 2021) to be integrated into malaria microplanning maps for the first time in the 2020-21 vector control campaign. Subsequently, the development of open datasets of building footprints, has widened the availability of data further, with Google Open Buildings data then used in two districts to support IRS plannings in 2022 and 2023. One of the challenges with using satellite-derived building footprint datasets is that footprints are not labelled as residential or non-residential structures. Without attribute data on building function or use, vector control interventions in Zambia have previously applied minimum and maximum area thresholds to exclude probable non-residential buildings (Kamanga et al., 2015). Alternatively, the proportion of structures that were residential has been estimated, based on samples of field-verified data. These approaches have enabled estimation of counts of residential buildings per settlement and identification of priority settlements for IRS (Kamanga et al. (2015), Pinchoff et al. (2016), Borkovska et al. (2022), Martin et al. (in review)). Table 1 Details of the building footprint data products available as of October 2023, and included in the comparative analysis for Zambia. Dataset Version; release date Method; source Notes Ecopia Digitize Africa building footprints “Year 1”; 2019 - Automated feature extraction from satellite imagery (Maxar 30-50cm imagery) - https://www.ecopiatech.com/ - Custom, commercial license which restricts usage predominantly to humanitarian applications Digitize Africa building footprints “Year 2”; 2021 Google Open Buildings v2; 2022 - Automated feature extraction from satellite imagery (Google Maps imagery with 50cm spatial resolution) - https://sites.research.google/open-buildings/ - CC BY 4.0 License or ODbL (Open Database License) v1.0 - Includes data at building level on precision confidence score. Open Buildings v3; 2023 Microsoft “Global” building footprints Downloaded January 2023 - Automated feature extraction from satellite imagery (Bing Maps imagery, including Maxar and Airbus) - https://github.com/microsoft/ GlobalMLBuildingFootprints - ODbL v1.0 License - No clear versioning system “Global” building footprints Downloaded October 2023 OpenStreetMap (OSM) OSM features tagged building=* Downloaded January 2023 - Predominantly manual digitization from satellite imagery, aided by some automated feature detection (imagery from various sources) - https://www.openstreetmap.org/ - ODbL v1.0 License - No versioning system as constantly being updated given VGI-nature of OSM OSM features tagged building=* Downloaded October 2023 Data Four different building footprint data products were included in the comparative analysis (Table 1 ): Google Open Buildings (Sirko et al., 2021), Microsoft “Global” building footprints (Microsoft, 2022), Ecopia DigitizeAfrica building footprints (Price and Hallas, 2019) and buildings extracted from OpenStreetMap (OSM). Building footprints from Ecopia, Google and Microsoft are all examples of data products created by commercial companies, using automated feature extraction methods and high-resolution satellite imagery (Li et al., 2022). In contrast, OSM buildings are an example of volunteered geographic information (VGI) product, with features digitised by a community of mappers and added to OpenStreetMap. Over the past decade, billions of features globally have been added to OSM, but given it's VGI-nature, inevitably is spatially heterogenous in terms of geographic coverage. Prior to the recent developments of multi-country building footprint datasets through automated feature extraction, OSM was the only source of mapped building data available for many countries. A third potential source of building footprints is authoritative datasets from government agencies; however no authoritative national dataset of building footprints is available for Zambia (Biljecki et al., 2021). Given the rapidly evolving data landscape, two versions of each building footprint data product were considered (8 datasets in total: Table 1 ). All building footprint data products consisted of vector polygons with limited, if any, additional attribute information. Some information related to building use or construction materials is available for a limited subset of OSM buildings, but not for other data products so it was not considered in these analyses. In the first method of quantifying counts of building footprints per settlement, predefined settlement extents were used – specifically the GRID3 v2.0 Settlement Extent dataset (CIESIN and Novel-T, 2023). This dataset consists of vector polygons, delineating outlines of settlements or groups of buildings which have been created through geospatial processing of Ecopia Digitize Africa “year 2” building footprints. These settlement extents represent geographic groupings of building footprints, and therefore do not necessarily correspond to individual settlements recognised by local or national administrations or the communities residing there. Settlement extents are also not named. Consequently, a single named settlement may be represented in the GRID3 settlement extent dataset as multiple polygons, and conversely a single settlement extent polygon may represent multiple named settlements. The GRID3 v2.0 settlement extents include a settlement type attribute based on the Degree of Urbanisation approach (Dijkstra et al., 2021). We used this Level 1 Degree of Urbanisation attribute (rural, urban cluster or urban centre) to stratify settlement extents by rural/urban types. Data processing For each building footprint dataset, polygons were converted to a common geographic coordinate system (WGS84, EPSG: 4326) and the area of each building footprint polygon was calculated in square metres. Any building footprint polygon with a geometric centroid outside the national boundary of Zambia was excluded. As IRS campaigns are specifically targeted at buildings in which residents sleep, only the subset of building footprints that were most likely to be residential, were included (henceforth referred to as “potentially residential”). As has been used previously in IRS campaigns in Zambia, minimum and maximum area thresholds were applied to exclude probable non-residential buildings (Kamanga et al., 2015). Thus, for each dataset, building footprint polygons that were 330 m 2 (likely non-residential structures such as churches, warehouses or commercial premises) in area, were excluded from the analyses (Fig. 1 ). The first method of calculating counts of building footprints per settlement, utilised the GRID3 v2.0 settlement extents dataset as the basis for defining settlements. The GRID3 settlement extents are derived from “year 2” Ecopia building footprints, which have been buffered by 50 m and then overlapping buffers dissolved. Utilising these settlement extents with other sources of building footprints will likely result in some building footprints being located outside settlement extents, particularly in locations experiencing settlement growth. For this reason, we expanded the spatial coverage of each GRID3 settlement extent by up to 1 km, with any building footprints that were within 1 km of a settlement extent, included in the count of building footprints of the settlement extent that they were nearest to. A second method was also tested, which did not utilise pre-defined settlement extents and instead, for each building footprint dataset in turn, groupings of neighbouring building footprints were created based on a proximity threshold. Recent IRS campaigns in Zambia have used this approach with a range of proximity thresholds, from 50 m (Pinchoff et al., 2016) up to 250 m (Bridges et al., 2018). For these campaigns, proximity thresholds were decided based on both expected travel distances of mosquitoes and practical considerations around minimising travel time for IRS field teams, so as to maximise spraying of neighbouring structure and community-level protection. In this analysis, we utilised a proximity threshold of 100 m (in line with both previous IRS campaigns and the buffer distance used in deriving GRID3 v2.0 settlement extents), such that building footprints within 100 m of each other, were grouped into a cluster. For each building footprint dataset in turn, clusters were delineated by buffering all potentially residential building footprints by 50 m and dissolving all overlapping buffered areas. The resulting delineated clusters of building footprints are henceforth referred to as “derived settlement clusters”. For both settlement definitions, the counts of potentially residential building footprints per settlement extent/derived settlement cluster were then calculated for each of the four building footprint data products (Figs. 2 and 3 ). Stratification of settlements In identifying potential settlements suitable for IRS, settlements were classified based on threshold counts of potentially residential building footprints. Recent IRS campaigns have utilised a minimum threshold in terms of the count of potentially residential structures/buildings per settlement, with a threshold of 25 residential buildings commonly employed to ensure operational efficiency (Pinchoff et al., 2016). For IRS campaigns in Zambia, this threshold was initially applied based on groupings of buildings that were all located within a specified distance of each other (i.e. following the second definition of settlement described above). More recently, the same threshold has been applied to counts of potentially residential buildings within pre-defined GRID3 settlement extents (Borkovska et al., 2022). In our analysis we have employed both definitions of settlement, with the same count threshold of 25 potentially residential building footprints per (i) expanded GRID3 settlement extent and (ii) derived settlement cluster, applied. For each building footprint dataset in turn and considering only the subset identified as potentially residential (≥ 9 m 2 and ≤ 330 m 2 ), settlements were stratified into four classes: no building footprints (class A), 1–5 building footprints (class B), 6–24 building footprints (class C) and 25 or more building footprints (class D; Fig. 3 ). From the stratified settlements, class D settlements were considered as the priority locations for IRS implementation, in line with criteria used in recent IRS campaigns in Zambia. Comparative analysis To assess the impact of the choice of building footprint dataset on the identification of priority settlements for IRS campaigns, we calculated summary statistics related to the count, location and spatial similarity of class D settlements (those with 25 or more potentially residential building footprints). Our comparative analysis focussed on the most recent version of each of the four building footprint data products. The agreement in class D settlement extents across building footprint datasets was assessed in terms of the count (n = 1–4) and combination of datasets (Figs. 5 and 6 and Supplementary Figures A.1 and A.2). For each building footprint dataset, the count of class D settlement extents per province (administrative unit level 1) with rural/urban stratification was calculated, along with the count of building footprints within these settlement extents (Fig. 7 and Supplementary Table A.5). To further assess geographic similarity in class D settlement extents, Jaccard coefficient values were calculated for each province, on a pairwise basis between datasets (Supplementary Figure A.3). The Jaccard coefficient (𝐽(X,Y) = | X ∩ Y | / | X ∪ Y|) was calculated as the count of settlement extents that were identified as class D with two building footprint datasets (X and Y), divided by the count of class D settlement extents identified with building footprint dataset X and/or Y. This calculation was not weighted by the area of settlement, such that settlement extents with 25 buildings and settlement extents with, for example, 5000 buildings were both considered equally. For the subset of settlement extents classified as class D with all four building footprint datasets, statistics related to the variability in counts of building footprints per settlement extent were also calculated (Supplementary Tables A.3, A.4 and Supplementary Figure A.4). Results Comparing counts of building footprints and potentially residential building footprints per province across Zambia (Fig. 1 ), shows considerable variation between data products (Ecopia, Google, Microsoft and OSM). For the majority of provinces, counts of building footprints were highest for Ecopia (eight provinces) or Google (two provinces – Lusaka and Copperbelt), and lowest for Microsoft (six provinces) or OSM (four provinces). For Central and Muchinga province however, the count of OSM building footprints is greater than the count of Google footprints. For each building footprint data product, two datasets were included in this comparison of building footprint counts, shown as pairs of bars in Fig. 1 . The two datasets are either the two most recent versions, or in the case of OSM and Microsoft, data extracted at two time points (January and October 2023). The comparison of building footprint counts per province, between pairs of datasets, show variations in counts, with patterns differing between provinces. Counts of Ecopia building footprints per province increased between the “year 1” and “year 2” datasets. Counts of Google building footprints increased between v2 and v3 for a couple of provinces (Copperbelt and Lusaka) but decreased for all other provinces. Without details of the satellite imagery used or changes to the feature extraction or post-processing algorithms, it is not possible to know the cause of the differences observed, but these may be due to temporal differences in satellite imagery, reduced cloud cover in imagery, and/or revised definitions of what constitutes a single building. Counts of OSM and Microsoft building footprints per province showed little change between dataset versions, except for Northern province, where counts of OSM buildings increased considerably from 243,627 in January 2023 to 350,379 in October 2023. In using building footprints for health programme planning and implementation, it may be that only a subset of buildings is of interest. For example, in Zambia, IRS campaigns are specifically targeted at residential (sleeping) structures and therefore area thresholds have been applied to exclude the largest and smallest buildings that are unlikely to be residential. When minimum (≥ 9 m 2 ) and maximum (≤ 330 m 2 ) area thresholds are applied, the proportion of the total building footprints that are excluded varies between provinces and data products (Fig. 1 ). For most provinces, the greatest reduction in count of building footprints is observed for Ecopia, with minimal change in the count of Microsoft building footprints, except for Copperbelt and Lusaka provinces (Fig. 1 ). Note that for the remaining analyses, only the subset of building footprints considered to be potentially residential (≥ 9 m 2 and ≤ 330 m 2 ) are included. Calculating counts of potentially residential building footprints per settlement extent also shows considerable variation between building footprint data products. Figure 2 shows the mean count of potentially residential building footprints per settlement extent, stratified by province and Level 1 degree of urbanisation (Dijkstra et al., 2021). For rural settlement extents, the mean count of potentially residential building footprints per settlement is generally highest for Ecopia and lowest for Microsoft or OSM. In rural areas of Muchinga and Central provinces, mean counts for Ecopia, Google and OSM are very similar, with mean counts of Microsoft building footprints being lower. In both Urban Clusters and Urban Centres, Google has the highest mean count of potentially residential building footprints per settlement extent for most provinces, and OSM the lowest. For settlement extents classified as Urban Centres, the magnitude of difference in mean counts between building footprint datasets in some provinces is massive, for example in Eastern province, the mean count of potentially residential building footprints per settlement extent is 23,961 for OSM (Jan. 2023) and 54,940 for Google (v3). Similarly, for Luapula province, the mean count of potentially residential building footprints per Urban Centre extent ranges from 19,123 (OSM Jan. 2023) to 46,481 (Google v3). There is only a single settlement extent classified as an Urban centre in Eastern, Luapula, North-Western, Northern and Southern provinces, and thus the mean counts per Urban centre for these provinces, reflect values only from one settlement extent. Counts of building footprints per settlement have been used in planning past IRS campaigns in Zambia to identify priority settlements. Figure 3 shows, for each building footprint dataset, the count of settlements in classes A-D, when settlements are classified based on the count of potentially residential building footprints per settlement (considering only the most recent version of each data product). Nationally, the count of settlements with 25 or more potentially residential building footprints (class D) is greatest when Ecopia year 2 building footprints are used (30,749 settlement extents and 29,032 derived settlement clusters). The second highest count of class D settlement is for Google v3 (24,977 settlement extents and 22,806 derived settlement clusters), followed by OSM (20,330 settlement extents and 19,202 derived clusters). The count of class D settlement extents based on Microsoft building footprints is the lowest of any data product (12,982 settlement extents and 12,043 derived clusters). Figure 3 also shows the breakdown of the remaining settlements when threshold counts of 0 (class A), 1–5 (class B) and 6–24 potentially residential building footprints (class C) are applied for each data product. Counts of class B and C settlement extents and derived clusters were highest for Ecopia v2 building footprints. Conversely, the count of settlement extents with no potentially residential building footprints (class A) was highest for Microsoft. Counts of potentially residential building footprints within each settlement class are summarised in Supplementary Tables A.1 and A.2. The count of settlements in each class nationally with different building footprint data products (Fig. 3 ) however hides any geographic variation in the classification of settlements, when different building footprint data products are used. The count of settlement extents in each class (the upper part of Fig. 3 ), is shown in Supplementary Figure A.1, with the addition of nested bars. These nested bars show the subset of identical settlement extents that are classified as the same class, with each building footprint dataset. For example, of the 12,982 class D settlement extents selected when a threshold of 25 is applied with Microsoft building footprints, 12,754 (Ecopia), 12,326 (Google) and 10,135 (OSM) of the same settlement extents are also classified as class D when the same threshold is applied with the other building footprint datasets. Of the 30,749 class D settlement extents with Ecopia building footprints, a subset of 21,922, 12,754 and 17,904 are class D when the same threshold is applied to Google, Microsoft and OSM building footprints respectively. For some locations, class D settlements are quite similar between building footprint datasets, but for other locations the choice of building footprint dataset and/or the settlement definition has a considerable impact on which settlements meet the threshold for class D. In Fig. 4 , the class D settlements are shown for both GRID3 settlement extents (left side of both panels) and derived settlement clusters (right side), for two example locations. The location marked X shows class D settlement extents (and derived settlement clusters) with Ecopia, but no class D settlement extents for the other three datasets. Location Y has class D derived clusters with Microsoft and Ecopia only, but class D settlement extents are identified in the same location with Microsoft, Ecopia and Google. Location Z has class D settlement extents with OSM, Google and Ecopia, but no class D derived clusters for OSM, and much smaller clusters for Google and Ecopia. In addition, the settlement definition can have considerable impact on which settlements are identified as class D (see for example the differences between class D settlement extents and derived clusters with Google v3 in panel (a) and Microsoft in panel (b) of Fig. 4 ). Figure 5 further explores the dissimilarity in settlement extents classified as class D with each building footprint dataset. In total across the four datasets, 35,861 settlement extents were identified as class D (IRS priority settlements) as they had 25 or more potentially residential building footprints in one or more building footprint datasets. Of these, there were 9,738 settlement extents identified with all four building footprint datasets (EGMO). Conversely there were 10,123 settlement extents identified in only one (E, G, M or O) of the four building footprint datasets, with over half of those settlement extents only being identified in terms of counts of Ecopia building footprints (n = 5,571). A smaller number of settlement extents were only selected with Google (n = 2,506) and OSM (n = 1,957), with a minimal number selected only with Microsoft (n = 89). For settlement extents classified as class D with two or three datasets, the majority included combinations of Ecopia with Google and/or OSM (e.g. EG, EO, EGM and EGO). Supplementary Figure A.2 shows the locations of the selected settlement extents for each building footprint dataset combination. Considering all settlement extents classified as class D with one or more building footprint datasets, the percentage classified as class D with each building footprint dataset is mapped at district level (Fig. 6 a). The percentage of class D settlement extents found with each building footprint varies considerably, with Ecopia having the most districts (n = 102/116) with over 75% of all class D settlement extents selected. In contrast, Microsoft had the fewest districts with over 75% of all class D settlement extents selected (n = 2). With Google and OSM, more spatial variability in the percentage of class D settlements with each dataset, is observed (Fig. 6 a). In Fig. 6 b, the percentage of settlement extents classified as class D in only one dataset, and in two, three or four datasets is mapped per district. The proportion of settlement extents classified as class D with just one dataset is greatest in selected districts in Southern and Western provinces, with seven districts having over 45% of class D settlement extents classified as such with just one building footprint dataset. Conversely, four districts in Eastern province, two districts in North-Western and one district in Luapula had less than 15% of class D settlement extents as class D with just one building footprint dataset. These same districts all had more than 45% of class D settlement extents classified as such with four buildings footprint datasets, along with six districts in Luapula and one in Northern. Sinda district in Eastern province had the highest proportion of class D settlement extents classified as such with all four building footprint datasets (82%). The spatial dissimilarity of class D settlement extents is further explored in Supplementary Figure A.3, with Jaccard Coefficient values for each pairwise combination of building footprint datasets, stratified by province. In terms of class D settlement extents, the highest Jaccard Coefficient values (greatest spatial similarity in settlement extents) are found with Google and Ecopia building footprints for all provinces apart from Muchinga and Central (Supplementary Figure A.3). For Muchinga and Central provinces, the spatial similarity between class D settlement extents is greatest between Ecopia and OSM. There is considerable variation in Jaccard Coefficient values between provinces, with Luapula and Eastern provinces tending to have higher values, and Western province lower values, for most pairwise combinations. The same pairwise combination of datasets can vary a lot between provinces, for example OSM and Microsoft has a Jaccard coefficient of 0.66 for Luapula, but only 0.24 for Western province. Given the geographic differences observed in class D settlements with different building footprint datasets, the choice of dataset may bias the selection of settlements, such that certain locations (and consequently population groups) could be inadvertently included or excluded from IRS campaigns. In Fig. 7 , the count of class D settlement extents for each building footprint dataset is stratified by province and settlement type. In rural areas, the count of class D settlement extents is highest based on Ecopia building footprints in 8 provinces, with the count of class D settlement extents based on Google building footprints being higher in two provinces (Lusaka and Copperbelt). Across all provinces, the number of rural class D settlement extents with Microsoft building footprints is considerably less than the number of class D settlement extents with Ecopia building footprints; the count of Microsoft rural class D settlement extents being less than 65% of the count of Ecopia rural class D settlement extents, and as low as 21% in Western province. In some rural settings, the number of OSM class D settlement extents is comparable to Ecopia or Google (e.g. Central and Southern), but in other provinces (e.g. Lusaka) there are far fewer OSM class D settlement extents, likely reflecting geographic differences in OSM coverage. In urban clusters and centres, the total number of class D settlement extents is expectedly fewer (although the extents themselves are larger) and there is also less variation in the total number between building footprint datasets. For Eastern, Luapula, North-Western, Northern and Southern provinces, there is only a single class D settlement extent classified as an urban centre (Fig. 7 ). For urban centres in most other provinces, the count of class D settlement extents is highest for Google building footprints. Even with very similar counts of class D settlement extents with different building footprint datasets, there is considerable variation in the total count of building footprints within class D settlement extents (Supplementary Table A.5). In the most extreme example, within class D urban centre settlement extents in North-Western province, there are 92,739 Ecopia, 106,216 Google, 64,280 Microsoft and only 6,123 OSM building footprints. Considering just the subset of settlement extents that are classified as class D with all four building footprint datasets (n = 9,738, labelled as EGMO in Fig. 5 ), there is still major variability in counts of building footprints per settlement extent (Supplementary Tables A.3/A4 and Supplementary Figure A.4). Discussion High-resolution geospatial data on population, settlements and infrastructure are increasingly being utilised for planning and implementing public health interventions. The rapid growth of multi-country building footprint data products over the last five years has provided new opportunities for population and settlements to be mapped (Chamberlain et al. (2024); Darin et al. (2024)), facilitating geo-enabled microplanning to be more widely integrated into public health campaigns and programmes (Borkovska et al. (2022), Barau et al. (2014)). In the context of malaria control programmes, data on the locations of buildings is key for planning, implementing and evaluating household campaigns (Bridges et al. (2018), Kamanga et al. (2015)). For vector control measures to be effective in reducing malaria transmission, high- and uniform-coverage of communities is needed (WHO (2023), Rehman et al. (2011)). Here we focus on a case study of planning for IRS campaigns in Zambia, where new geospatial datasets and technology have been integrated over the past decade. We show that if settlements are classified by count and proximity of buildings, the settlements identified for potential inclusion in IRS campaigns differ considerably depending on the choice of building footprint dataset. The results of our analysis show that both the overall count and the geographic location of class D settlements (those settlements with at least 25 potentially residential building footprints), varies notably between building footprint datasets. Focussing on the most recent version of each of the four building footprint datasets, the count of class D settlement extents nationally varied between 12,982 and 30,749, when Microsoft (October 2023 version) and Ecopia year 2 building footprint datasets were used respectively (Fig. 3 ). Considering all class D settlement extents identified with any of the four building footprint datasets, the percentage of settlements identified as class D with each dataset in turn also shows considerable geographic variation between datasets (Fig. 6 ). When stratified by province (administrative unit level 1), the count of class D settlements in rural clusters was typically highest with Ecopia building footprints, and lowest with Microsoft building footprints (Fig. 7 ). Whilst in urban clusters, the count of class D settlements was highest with Google building footprints in most provinces, and lowest with either OSM or Microsoft buildings footprints. These results highlight how the choice of building footprint dataset may bias the selection of settlements, potentially prioritising locations with certain characteristics (e.g. predominantly urban areas) which could create inequalities in campaign provision. Although not the primary focus of this study, the settlement definition used as the basis for calculating counts of building footprints per settlement (GRID3 settlement extents or derived settlement clusters), also affected the number and geographic location of settlements selected (Figs. 3 and 4 ). Whilst settlement and building footprint data are not the sole data used in decision making concerning malaria vector control initiatives, the use of a particular dataset has the potential to influence decisions and strategies. The WHO guidance for IRS does not specify an absolute coverage target, acknowledging that 100% coverage is rarely possible, but a minimum spray target of 80% of eligible structures is commonly used (WHO, 2023). In this context, an accurate count of geo-located residential buildings is necessary for (i) planning to ensure that available resources are targeted to appropriate locations to maximise reductions in malaria transmission, (ii) ensuring that sufficient spray teams, equipment and insecticide are allocated and delivered to the right locations, and (iii) to assess spray coverage after campaign completion. The results of our analysis quantify the relative differences in counts of building footprints per settlement (Figs. 2 , 7 and Supplementary Figure A.4) and highlight the sensitivity of these metrics to the choice of building footprint dataset (Supplementary Tables A.3 and A.4). For example, if the subset of settlement extents that are classified as class D with all four building footprint datasets were all identified as appropriate locations for IRS, for Western province this would mean planning and implementing IRS in 532 settlements (Supplementary Table A.4). For these selected settlements, the choice of building footprint dataset still would have a big impact on the estimation of required resources and assessment of spray coverage. Considering a hypothetical IRS campaign, in which 50,000 buildings were sprayed in Western province, coverage rates across all target settlements in Western province would be calculated as 60.4% (out of 83,803 Ecopia building footprints), 74.4% (out of 67,225 Google building footprints), 129.8% (out of 38,531 Microsoft building footprints) or 82.9% (out of 60,326 OSM building footprints). The detailed nature of building footprint polygon data can give an illusion of accuracy, more so than, for example, raster land cover classification maps. Building footprint datasets can commonly have some buildings omitted (false negatives) and other features misclassified as buildings and thus included (false positives) (Heris et al., 2020). These can occur due to gaps in satellite imagery coverage, satellite imagery being outdated, limitations of the feature extraction algorithm, or buildings being obscured in imagery due to tree canopy cover, for example. A lack of metadata can further compound these problems, as the definition of what is considered as an individual building, details of the feature extraction algorithms and post-processing of extracted features, and the temporal imagery coverage, can all vary between data products (Chamberlain et al., 2024). As building footprint datasets evolve and are used in different fields, there is a need for data producers to take steps to address these issues and support data users. Those working with building footprint data to support decision making, in the context of health programme planning and more widely, also need to be aware of differences between data products, to consider potential impacts of choosing a particular dataset and to be cognisant of their limitations. As of October 2023, in addition to buildings from OpenStreetMap, there were at least three building footprint data products in existence for Zambia, with two or more versions of each dataset (Fig. 1 ). Apart from Ecopia building footprints, no other datasets include attribute information on the date of source satellite imagery, making it impossible to know the timepoints that building footprints are representative of. From the building footprint data products included in our analysis, we would suggest that generally the dataset coverage is most complete for Ecopia and Google building footprints. However, this recommendation has several caveats: our comparative analysis does not include an authoritative reference dataset, the Ecopia building footprints are based on satellite imagery from 2020 or earlier (over half of building footprints were extracted from satellite imagery collected in 2019) and no granular information is provided on source imagery dates for Google. In locations where there have been recent, comprehensive campaigns to digitise buildings in OSM, these data may be more comprehensive in their coverage and better reflect the structure of settlements, but this is difficult to assess at scale. Recent work indicates a likely move towards more dynamic updates of building footprints and settlement datasets in the future (Sirko et al. (2021), Metzger et al. (2024)), as well as potential improvements in building height and typology estimation (Frantz et al. (2021), Esch et al. (2022)). In the context of health programme planning, these developments would likely be beneficial in improving estimation of population numbers (Palacios-Lopez et al. (2022)) and hence programme denominators, as well as enabling settlement change detection with temporally-explicit building/settlement datasets. This study has considered the use of building footprint data products in isolation, but with a growing number and availability of data products, the potential to use data products in combination increases. Further work is needed to develop systematic and robust methods and guidelines to do this. Opportunities to efficiently conduct field validation of satellite-derived datasets, as part of field-deployed health campaigns, should also be pursued. The results of field validation can be used to further improve building datasets, ultimately accelerating planning and delivery of health interventions. Although building footprint datasets extracted from satellite imagery provide new opportunities to support geo-enabled microplanning at scale (Mendes et al. (2021), Borkovska et al. (2022), Alexander et al. (2024)), they are not a replacement for local knowledge or comprehensive field data collection. Enumeration of populations, households and/or buildings may be collected by health facilities through community-based service delivery campaigns and for headcount registers, but these are often not up to date, complete or accurate due to resources and capacity limitations (Kyomuhangi et al. (2023)). Geospatial data on counts and locations of buildings that have been field-verified (for example during resourced campaigns), likely provide the most accurate data on residential buildings for resource planning and delivery (Keating et al. (2021)). This is especially true if field service delivery teams use maps to navigate to all mapped building footprints, then verify that a building exists, add any unmapped buildings and label whether a building is residential or not. Field-verification of this sort has been conducted in Zambia using the Reveal platform mobile app which enables in situ updating of building data and denominators for each implementation cycle (Martin et al. (in review)). However, this data is usually limited to selected campaign prioritised areas, and therefore not available to all health facilities and districts – making it challenging, but not impossible, to use at scale. In Zambia, building footprints extracted from satellite imagery and field-verified data have been used in combination to inform resource needs during campaign planning and to monitor delivery (Keating et al. (2021), Arambepola et al. (2021)). It is crucial that decision makers still critically assess any such data against existing sources, such as those available at health facility level, in order to ensure that service delivery targets and denominators are as accurate as possible. Aside from IRS planning in Zambia, building footprints or similar datasets have been used for geo-enabled microplanning in a range of other public health interventions (e.g. Mendes et al. (2021), Ouédraogo et al. (2020)), often utilising a single source of data. Our findings provide insights into the potential variability in estimates when different building footprint datasets are used in other applications, but findings concerning particular data products cannot be assumed to necessarily be transferrable to other countries or contexts. In particular, the coverage of some building footprint data products is highly heterogeneous between countries (Chamberlain et al., 2024), with no data available for some datasets in some areas (for example, conflict-affected locations have been excluded from Google Open Buildings datasets) and a rapidly evolving data landscape. Our analysis has several caveats. In identifying potential priority locations for IRS, we applied a set of criteria concerning area of, distance between and counts of building footprints. As the national strategy in Zambia has changed over time, the criteria and guidelines used for planning vector control campaigns have similarly evolved (NMEC (2017), NMEC (2022)). The criteria used in our comparative analysis are based on campaigns implemented with the NMEP over the past decade, and provide a framework within which to assess the impact of utilising different building footprint datasets for IRS planning. Our analysis has considered solely building footprint data to identify the subset of settlements that might be considered as suitable for IRS, and has not incorporated the important contextual knowledge of district teams to select the final settlements for inclusion in an IRS campaign. In addition, none of the feature-extracted building footprint datasets (Google, Ecopia and Microsoft) included attributes on building type or use, necessitating an approximation of potentially residential buildings based on simple area thresholds (as has been used in previous work e.g. by Kamanga et al. (2015)). The subset of building footprints considered to be potentially residential based on simple area thresholds, will inevitably include some non-residential structures such as churches, shops or health facilities. Further work should explore more sophisticated classification approaches to identify non-residential/non-sprayable structures (e.g. Sturrock et al. (2018), Lloyd et al. (2020)), and assess the sensitivity of such classification approaches to different building footprint products. Given the lack of an authoritative national reference dataset, our analysis has been limited to comparisons between available building footprint datasets. As new sources of data become available, for example geolocated household data from the 2020 round of national population and housing censuses, these should be included in similar comparative analyses. Conclusions The recent growth in the global availability of detailed building footprint data has provided new opportunities for integrating geo-enabled microplanning into public health interventions, particularly in previously data-scarce locations. This proliferation of datasets however also brings challenges for data users and decision-makers. The results of this study show that the choice of building footprint dataset has a substantial impact on locations selected when criteria are based on counts and proximity of buildings. This has been explored in the context of planning IRS campaigns in Zambia, but similarly large differences would be expected if count-based criteria are used for planning and evaluating other resource provision campaigns. Further work is needed to quantify the sensitivity of the choice of building footprint dataset in such analyses across a range of geographic contexts. Data producers could also better support data users through improved metadata, by routinely disclosing information on spatiotemporal coverage of datasets, definitions of features considered to be buildings and rates of omission and commission. As the building footprint and settlement data landscape continues to grow and diversify, understanding the characteristics and comparability of datasets will remain fundamental to robust data use. Caution will continue to be needed in selecting building footprint data products to integrate into analyses, and where possible, any conclusions drawn should be supported by local, contextual knowledge and field-verified data. Abbreviations IRS – Indoor Residual Spraying LLIN – Long-lasting Insecticide-treated Net NMEC – National Malaria Elimination Centre NMESP – National Malaria Elimination Strategic Plan OSM - OpenStreetMap UNICEF – United Nations Children’s Fund WHO – World Health Organisation COVID-19 – Coronavirus Disease 2019, caused by the SARS-CoV-2 virus Declarations Ethics approval Not applicable Consent for publication Not applicable Availability of data and materials Google v3 building footprints are available to download in tiled format (level 4 S2 cells) from: https://sites.research.google/open-buildings/#open-buildings-download, or accessed via Google Earth Engine: https://developers.google.com/earth-engine/datasets/catalog/GOOGLE_Research_open-buildings_v3_polygons. OpenStreetMap building footprints can be extracted using various plugins, or pre-prepared national files updated daily are available from GeoFabrik: https://download.geofabrik.de/africa/zambia.html. Microsoft “global” building footprints are available to download from: https://github.com/microsoft/GlobalMLBuildingFootprints. Ecopia building footprints are not publicly available, but access can be requested for humanitarian purposes from Ecopia (https://www.ecopiatech.com/). The GRID3 v2.0 settlement extent dataset is available to download via the GRID3 Data Hub: https://data.grid3.org/datasets/e51042367feb445bb0126cb6ee39893f_0/. Competing interests The authors declare that they have no competing interests. Funding This work was initiated with funding from the Bill and Melinda Gates Foundation as part of the GRID3 project [INV-045694]. The work was further developed and completed with funding from the United Kingdom's Foreign, Commonwealth & Development Office (FCDO), as part of the Data for Development - Unleashing the Data Dividend project [GB-GOV-1-400025-405]. Authors' contributions HRC conceptualised the study, and undertook the analysis with input from AJT and DP. HRC wrote the original manuscript draft. DP, AW, OB, CAM, GM, SR, ANL and AJT reviewed, edited and approved the manuscript for submission. ANL and AJT acquired funding. Acknowledgements This study is an output of the WorldPop Research Group at the University of Southampton, United Kingdom. Preparation of this manuscript was supported by HRC attending the University of Southampton Faculty of Medicine/Faculty of Environmental and Life Sciences Writing Retreat (January 2024). References Alexander C, McKay A, Bhatt K, da Costa Lourenço ALR, Kaplan B, Krishnan RSSG. Pre-trained regional models for extracting buildings from high resolution satellite imagery to support public health initiatives. Remote Sens Applications: Soc Environ. 2024;101270. https://doi.org/10.1016/j.rsase.2024.101270 . Ali D, Levin A, Abdulkarim M, Tijjani U, Ahmed B, Namalam F, Dougherty L. A cost-effectiveness analysis of traditional and geographic information system-supported microplanning approaches for routine immunization program management in northern Nigeria. Vaccine. 2020;38(6):1408–15. https://doi.org/10.1016/j.vaccine.2019.12.002 . Arambepola R, Yang Y, Hutchinson K, et al. Using geospatial models to map zero-dose children: factors associated with zero-dose vaccination status before and after a mass measles and rubella vaccination campaign in Southern province, Zambia. BMJ Global Health. 2021;6:e007479. https://doi.org/10.1136/bmjgh-2021-007479 . Barau I, Zubairu M, Mwanza MN, Seaman VY, Improving Polio Vaccination Coverage in Nigeria Through the Use of Geographic Information System Technology, The Journal of Infectious Diseases, Volume 210, Issue, suppl_1 N. 2014, Pages S102–S110. https://doi.org/10.1093/infdis/jiu010 Borkovska O, et al. Developing High-Resolution Population and Settlement Data for Impactful Malaria Interventions in Zambia. J Environ Public Health. 2022. https://doi.org/10.1155/2022/2941013 . Borowitz M, Zhou J, Azelton K, Nassar I-Y. Examining the value of satellite data in halting transmission of polio in Nigeria: A socioeconomic analysis. Data Policy. 2023;5:e16. https://doi.org/10.1017/dap.2023.12 . Bridges DJ, et al. Accuracy and impact of spatial aids based upon satellite enumeration to improve indoor residual spraying spatial coverage. Malar J. 2018;17. https://doi.org/10.1186/s12936-018-2236-2 . Brown CF, Brumby SP, Guzder-Williams B, Birch T, Hyde SB, Mazzariello J, Tait AM. Dynamic World, Near real-time global 10 m land use land cover mapping. Sci Data. 2022;9(1):251. https://doi.org/10.1038/s41597-022-01307-4 . Center for International Earth Science Information Network (CIESIN). Columbia University; Flowminder Foundation; United Nations Population Fund (UNFPA); WorldPop, University of Southampton. Mapping and Classifying Settlement Locations. Palisades, NY: Georeferenced Infrastructure and Demographic Data for Development (GRID3). https://doi.org/10.7916/d8-gzxf-s834 (2021). Chamberlain HR, Darin E, Adewole WA, Jochem WC, Lazar AN, Tatem AJ. Building footprint data for countries in Africa: To what extent are existing data products comparable? Comput Environ Urban Syst. 2024;110C:102104. https://doi.org/10.1016/j.compenvurbsys.2024.102104 . Chaney S, Mechael P, Thu N, Diallo M, Gachen C. Every Child on the Map: A Theory of Change Framework for Improving Childhood Immunization Coverage and Equity Using Geospatial Data and Technologies. J Med Internet Res. 2021;23(8):e29759. https://doi.org/10.2196/29759 . CIESIN and Novel-T. GRID3 Zambia Settlement Extents, Version 2.0, Center for International Earth Science Information Network (CIESIN)., Columbia University and Novel-T. Palisades, NY: Geo-Referenced Infrastructure and Demographic Data for Development (GRID3). https://doi.org/10.7916/wqmn-f746 (2023). Clements AC, Reid HL, Kelly GC, Hay SI. Further shrinking the malaria map: how can geospatial science help to achieve malaria elimination? Lancet Infect Dis. 2013;13(8):709–18. https://doi.org/10.1016/S1473-3099(13)70140-3 . Darin E, Dicko AH, Galal H, Jimenez RM, Park H, Tatem AJ, Qader S. (2024). Mapping refugee populations at high resolution by unlocking humanitarian administrative data. Journal of International Humanitarian Action, 9(1), 14, https://doi.org/10.1186/s41018-024-00157-6 Dijkstra L, Florczyk AJ, Freire S, Kemper T, Melchiorri M, Pesaresi M, Schiavina M. Applying the degree of urbanisation to the globe: A new harmonised definition reveals a different picture of global urbanisation. J Urban Econ. 2021;125:103312. https://doi.org/10.1016/j.jue.2020.103312 . Dotse-Gborgbortsi W, Nilsen K, Ofosu A, et al. Distance is a big problem: a geographic analysis of reported and modelled proximity to maternal health services in Ghana. BMC Pregnancy Childbirth. 2022;22:672. https://doi.org/10.1186/s12884-022-04998-0 . Doxsey-Whitfield E, MacManus K, Adamo SB, Pistolesi L, Squires J, Borkovska O, Baptista SR. Taking advantage of the improved availability of census data: a first look at the gridded population of the world, version 4. Papers Appl Geogr. 2015;1(3):226–34. https://doi.org/10.1080/23754931.2015.1014272 . Dummer TJ. Health geography: supporting public health policy and planning. CMAJ. 2008;178(9):1177–80. https://doi.org/10.1503/cmaj.071783 . Esch T, Brzoska E, Dech S, Leutner B, Palacios-Lopez D, Metz-Marconcini A, Zeidler J. World Settlement Footprint 3D-A first three-dimensional survey of the global building stock. Remote Sens Environ. 2022;270:112877. https://doi.org/10.1016/j.rse.2021.112877 . Esch T, Brzoska E, Dech S, Leutner B, Palacios-Lopez D, Metz-Marconcini A, Zeidler J. World Settlement Footprint 3D-A first three-dimensional survey of the global building stock. Remote Sens Environ. 2022;270:112877. https://doi.org/10.1016/j.rse.2021.112877 . Frantz D, Schug F, Okujeni A, Navacchi C, Wagner W, van der Linden S, Hostert P. National-scale mapping of building height using Sentinel-1 and Sentinel-2 time series. Remote Sens Environ. 2021;252:112128. https://doi.org/10.1016/j.rse.2020.112128 . Freire S, MacManus K, Pesaresi M, Doxsey-Whitfield E, Mills J. (2016). Development of new open and free multi-temporal global population grids at 250 m resolution, Geospatial Data in a Changing World; Association of Geographic Information Laboratories in Europe (AGILE) (Organiser). AGILE; 2016. JRC100523. Gething, P. W., Patil, A. P., Smith, D. L., Guerra, C. A., Elyazar, I. R., Johnston,G. L., … Hay, S. I. (2011). A new world malaria map: Plasmodium falciparum endemicity in 2010. Malaria journal, 10, 1–16. https://doi.org/10.1186/1475-2875-10-378. Gevaert CM, Buunk T., & Van Den Homberg, M. J. (2024). Auditing geospatial datasets for biases: using global building datasets for disaster risk management. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. https://doi.org/10.1109/JSTARS.2024.3422503 Hay, S. I., Sinka, M. E., Okara, R. M., Kabaria, C. W., Mbithi, P. M., Tago, C. C.,… Godfray, H. C. J. (2010). Developing global maps of the dominant Anopheles vectors of human malaria. PLoS medicine, 7(2), e1000209. https://doi.org/10.1371/journal.pmed.1000209. Hengel, B., Causer, L., Matthews, S., Smith, K., Andrewartha, K., Badman, S., … Guy,R. (2021). A decentralised point-of-care testing model to address inequities in the COVID-19 response. The Lancet Infectious Diseases, 21(7), e183-e190, https://doi.org/10.1016/S1473-3099(20)30859-8. Hierink F, Oladeji O, Robins A, et al. A geospatial analysis of accessibility and availability to implement the primary healthcare roadmap in Ethiopia. Commun Med. 2023;3:140. https://doi.org/10.1038/s43856-023-00372-z . Jochem WC, Leasure DR, Pannell O, Chamberlain HR, Jones P, Tatem AJ. Classifying settlement types from multi-scale spatial patterns of building footprints. Environ Plann B: Urban Analytics City Sci. 2021;48(5):1161–79. https://doi.org/10.1177/2399808320921208 . Joseph NK, Macharia PM, Ouma PO, et al. Spatial access inequities and childhood immunisation uptake in Kenya. BMC Public Health. 2020;20:1407. https://doi.org/10.1186/s12889-020-09486-8 . Kamanga A, et al. Open-source satellite enumeration to map households: planning and targeting indoor residual spraying for malaria. Malar J. 2015;14:345. https://doi.org/10.1186/s12936-015-0831-z . Karra K, Kontgis C, Statman-Weil Z, Mazzariello JC, Mathis M, Brumby SP. (2021, July). Global land use/land cover with Sentinel 2 and deep learning. In 2021 IEEE international geoscience and remote sensing symposium IGARSS (pp. 4704–4707). IEEE. Keating J, et al. Retrospective evaluation of the effectiveness of indoor residual spray with pirimiphos-methyl (Actellic) on malaria transmission in Zambia. Malar J. 2021;20:173. https://doi.org/10.1186/s12936-021-03710-5 . Kebba Touray P, Mkanda SG, Tegegn P, Nsubuga TB, Erbeto R, Banda A, Etsano F, Shuaib RG, Vaz S. 67–S72. https://doi.org/10.1093/infdis/jiv493 Kyomuhangi I, et al. Assessing national vector control micro-planning in Zambia using the 2021 malaria indicator survey. Malar J. 2023;22:365. https://doi.org/10.1186/s12936-023-04807-9 . Leithäuser N, Schneider J, Johann S, et al. Quantifying Covid19-vaccine location strategies for Germany. BMC Health Serv Res. 2021;21:780. https://doi.org/10.1186/s12913-021-06587-x . Li J, Huang X, Tu L, Zhang T, Wang L. A review of building detection from very high resolution optical remote sensing images. GIScience Remote Sens. 2022;59(1):1199–225. https://doi.org/10.1080/15481603.2022.2101727 . Lloyd, et al. Using GIS and Machine Learning to Classify Residential Status of Urban Buildings in Low and Middle Income Settings. Remote Sens. 2020;12(23):3847. https://doi.org/10.3390/rs12233847 . Lubinda J, et al. Spatio-temporal monitoring of health facility-level malaria trends in Zambia and adaptive scaling for operational intervention. Commun Med. 2022;2:79. https://doi.org/10.1038/s43856-022-00144-1 . Macharia PM, Ouma PO, Gogo EG, Snow RW, Noor AM. Spatial accessibility to basic public health services in South Sudan. Geospat Health. 2017;12(1):510. https://doi.org/10.4081/gh.2017.510 . Martin A, Bwalya F, Pollard D et al. Use of a geo-enabled digital global good for microplanning and delivery of indoor residual spray (IRS) in Zambia: a case study, 2016–2020. (2024 - in review). Mendes, A., Palmer, T., Berens, A., Espey, J., Price, R., Mallya, A., … Kaplan, B.(2021). Mapathons versus automated feature extraction: a comparative analysis for strengthening immunization microplanning. International Journal of Health Geographics,20(1), 27. https://doi.org/10.1186/s12942-021-00277-x. Metzger N, Daudt RC, Tuia D, Schindler K. High-resolution population maps derived from sentinel-1 and sentinel-2. Remote Sens Environ. 2024;314:114383. https://doi.org/10.1016/j.rse.2024.114383 . Microsoft. 2022. Worldwide building footprints derived from satellite imagery (GitHub Repository). https://github.com/microsoft/GlobalMLBuildingFootprints [DATASET] (accessed 10th January 2023, and 1st October 2023). Muchiri, S. K., Muthee, R., Kiarie, H., Sitienei, J., Agweyu, A., Atkinson, P. M.,… Alegana, V. A. (2022). Unmet need for COVID-19 vaccination coverage in Kenya. Vaccine,40(13), 2011–2019. https://doi.org/10.1016/j.vaccine.2022.02.035. National Malaria Elimination Centre (NMEC). National malaria elimination strategic plan 2017–2021. Zambia: Lusaka; 2017. https://static1.squarespace.com/static/58d002f017 bffcf99fe21889/t/5b28d7f1575d1ff0942dbce1/1529403401067/National +Malaria+Elimination+Strategic+Plan+2017-Final_PRINT.pdf . National Malaria Elimination Centre (NMEC). National malaria elimination strategic plan 2022–2026. Zambia: Lusaka. 2022, https://static1.squarespace.com/static/58d002f017b ffcf99fe21889/t/632a4cb0fcd87c13d0165372/1663716530614/Z NMESP+2022+to+2026_SIGNED+120722.pdf Ouédraogo AL, Zhang J, Tinto H, et al. A microplanning model to improve door-to-door health service delivery: the case of Seasonal Malaria Chemoprevention in Sub-Saharan African villages. BMC Health Serv Res. 2020;20:1128. https://doi.org/10.1186/s12913-020-05972-2 . Palacios-Lopez, D., Esch, T., MacManus, K., Marconcini, M., Sorichetta, A., Yetman,G., … Reinartz, P. (2022). Towards an improved large-scale gridded population dataset:A Pan-European study on the integration of 3D settlement data into population modelling.Remote Sensing, 14(2), 325. https://doi.org/10.3390/rs14020325. Pesaresi M, Politis P, GHS-BUILT-S. R2023A - GHS built-up surface grid, derived from Sentinel2 composite and Landsat, multitemporal (1975–2030)European Commission, Joint Research Centre (JRC), 10.2905/9F06F36F-4B11-47EC-ABB0-4F8B7B1D72EA Pinchoff J, et al. Targeting indoor residual spraying for malaria using epidemiological data: a case study of the Zambia experience. Malar J. 2016;15(11). https://doi.org/10.1186/s12936-015-1073-9 . Price R, Hallas M. (2019, December). Mapping every building and road in sub-Saharan Africa. In AGU Fall Meeting Abstracts (Vol. 2019, pp. IN41A-02). Rader B, Astley CM, Sewalk K, et al. Spatial modeling of vaccine deserts as barriers to controlling SARS-CoV-2. Commun Med. 2022;2:141. https://doi.org/10.1038/s43856-022-00183-8 . Rehman AM, Coleman M, Schwabe C, Baltazar G, Matias A, Roncon Gomes I, et al. How Much Does Malaria Vector Control Quality Matter: The Epidemiological Impact of Holed Nets and Inadequate Indoor Residual Spraying. PLoS ONE. 2011;6(4):e19205. https://doi.org/10.1371/journal.pone.0019205 . Reid, H., Haque, U., Clements, A. C., Tatem, A. J., Vallely, A., Ahmed, S. M., … Haque,R. (2010). Mapping malaria risk in Bangladesh using Bayesian geostatistical models.The American journal of tropical medicine and hygiene, 83(4), 861. https://doi.org/10.4269/ajtmh.2010.10-0154. Sirko, W., Brempong, E. A., Marcos, J. T., Annkah, A., Korme, A., Hassen, M. A., …Quinn, J. (2023). High-Resolution Building and Road Detection from Sentinel-2. arXiv preprint arXiv:2310.11622. Sirko, W., Kashubin, S., Ritter, M., Annkah, A., Bouchareb, Y. S. E., Dauphin, Y.,… Quinn, J. (2021). Continental-scale building detection from high resolution satellite imagery. arXiv preprint, https://doi.org/10.48550/arXiv.2107.12283. Snow J. Cholera and the water supply in the south districts of London in 1854. J Public Health Sanit Rev. 1856;2(7):239. Sorichetta A, Hornby GM, Stevens FR, Gaughan AE, Linard C, Tatem AJ. High-resolution gridded population datasets for Latin America and the Caribbean in 2010, 2015, and 2020. Sci data. 2015;2(1):1–12. https://doi.org/10.1038/sdata.2015.45 . Sturrock, et al. PLoS ONE. 2018;13(9):e0204399. https://doi.org/10.1371/journal.pone.0204399 . Predicting residential structures from open source remotely enumerated data using machine learning. Tatem AJ. Small area population denominators for improved disease surveillance and response. Epidemics. 2022;41:100641. https://doi.org/10.1016/j.epidem.2022.100641 . Tiecke, T. G., Liu, X., Zhang, A., Gros, A., Li, N., Yetman, G., … Dang, H. A. (2017).Mapping the world population one building at a time. arXiv preprint. https://doi.org/10.48550/arXiv.1712.05839. Utazi CE, Olowe I, Chan THM, Dotse-Gborgortsi W, Wagai J, Umar JA, Etamesor S, Atuhaire B, Fafunmi B, Crawford J, Adeniran A, Tatem AJ. Geospatial Variation in Vaccination Coverage and Zero-Dose Prevalence at the District, Ward and Health Facility Levels Before and After a Measles Vaccination Campaign in Nigeria. Preprints 2024, 2024100445. https://doi.org/10.20944/preprints202410.0445.v1 Utazi CE, Thorley J, Alegana VA, et al. Mapping vaccination coverage to explore the effects of delivery mechanisms and inform vaccination strategies. Nat Commun. 2019;10:1633. https://doi.org/10.1038/s41467-019-09611-1 . Whitehead J, Scott N, Atatoa-Carr P, Lawrenson R. Will access to COVID-19 vaccine in Aotearoa be equitable for priority populations? N Z Med J. 2021;134(1535):25–34. WHO. (2009). Microplanning for immunization service delivery using the Reaching Every District (RED) strategy; 2009. https://www.who.int/publications/i/item/microplanning-for-immunization-service-delivery-using-the-reaching-every-district-(-red)-strategy WHO. (2018). Global Polio Eradication Initiative: Best Practices in Microplanning for Polio Eradication; 2018. http://polioeradication.org/wp-content/uploads/2018/12/Best-practices-in-mircoplanning-for-polio-eradication.pdf WHO. (2021), Guidance on developing a national deployment and vaccination plan for COVID-19 vaccines: interim guidance, 1 June 2021. Geneva: World Health Organization; 2021 (WHO/2019-nCoV/Vaccine_deployment/2021.1). WHO. Operational manual on indoor residual spraying: control of vectors of malaria, Aedes-borne diseases, Chagas disease, leishmaniases and lymphatic filariasis. Geneva: World Health Organization; 2023. https://iris.who.int/bitstream/handle/10665/375978/9789240083998-eng.pdf . WHO-UNICEF COVAX GIS Working Group. Geo-Enabled Microplanning Handbook. https://www.digitalhealthcoe.org/_files/ugd/55ae33_f30815f09639437586ef8edd4475354f.pdf (2023). Winters A, Riley C, Soobramoney L, Pollard D, Jere E, Bwalya F, Silumbe K, Hamainza B. Cost and cost effectiveness of geospatial planning and delivery tools added to standard health campaigns in Luapula Province, Zambia. Oxf Open Digit Health. 2024;2(Supplement2):ii66–74. https://doi.org/10.1093/oodh/oqae040 . World Health Organisation. High burden to high impact: a targeted malaria response. https://www.who.int/publications/i/item/WHO-CDS-GMP-2018.25 World Health Organization, Geneva (2018). World Health Organization. (2023). World malaria report 2023. World Health Organization. https://www.who.int/publications/i/item/9789240086173 Zanaga D, Van De Kerchove R, Daems D, De Keersmaecker W, Brockmann C, Kirches G, Wevers J, Cartus O, Santoro M, Fritz S, Lesiv M, Herold M, Tsendbazar N-E, Xu P, Ramoino F, Arino O. (2022). ESA WorldCover 10 m 2021 v200 (Version v200) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7254221 . Additional Declarations No competing interests reported. Supplementary Files SUPPLEMENTARYINFORMATIONAssessingtheimpactofbuildingfootprintdatasetchoiceforhealthprogrammeplanning.docx Cite Share Download PDF Status: Published Journal Publication published 24 May, 2025 Read the published version in International Journal of Health Geographics → Version 1 posted Editorial decision: Revision requested 08 Mar, 2025 Reviews received at journal 27 Feb, 2025 Reviews received at journal 24 Feb, 2025 Reviewers agreed at journal 04 Feb, 2025 Reviewers agreed at journal 03 Feb, 2025 Reviewers invited by journal 02 Feb, 2025 Editor assigned by journal 26 Jan, 2025 Submission checks completed at journal 26 Jan, 2025 First submitted to journal 24 Jan, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5898324","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":414561827,"identity":"0981dd6a-1788-4c4a-aa12-464d93d263a9","order_by":0,"name":"Heather R. Chamberlain","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABA0lEQVRIie2RsWrDMBCGzxxIi7wrJHFfQcZQUyj0VaI9Q7t1MK4h0CzO7tCSlyhoFhy4i+kTZGgWTy0UurRDoHamDFGTbh30DSc48fHfcQAezz9EAbDusQC8r7d7X/Kogn1tAMSfFAjuT1BSwPb1e7qO0hlu6GaVR1d8YeErA70sDisXBUvjhWmTETFFS0OJEC+ToKxBP7gGs4LJ0JCuEBSFxupSThWEBejVL8pga+iuQv5B4WO+U4LtEWXYpUwkii6lwJ2CfYpzMGLnw7Fp4wrFNYm626VpFI1qmbjWV8+zdvBu1meSz58+RZZHfF7Gm7fsclxZRwwealr3VTwej8dzCj+f4lXZbkZHhgAAAABJRU5ErkJggg==","orcid":"","institution":"University of Southampton","correspondingAuthor":true,"prefix":"","firstName":"Heather","middleName":"R.","lastName":"Chamberlain","suffix":""},{"id":414561828,"identity":"8b870a1f-446c-47c7-9b55-a0f01d6ebfed","order_by":1,"name":"Derek Pollard","email":"","orcid":"","institution":"Akros Research","correspondingAuthor":false,"prefix":"","firstName":"Derek","middleName":"","lastName":"Pollard","suffix":""},{"id":414561829,"identity":"e634dcf1-39f5-4187-ae8b-4442dee93165","order_by":2,"name":"Anna Winters","email":"","orcid":"","institution":"Akros Inc","correspondingAuthor":false,"prefix":"","firstName":"Anna","middleName":"","lastName":"Winters","suffix":""},{"id":414561830,"identity":"d90f68a4-37a1-4e99-bc7a-97c705f9ea72","order_by":3,"name":"Silvia Renn","email":"","orcid":"","institution":"GRID3 Inc","correspondingAuthor":false,"prefix":"","firstName":"Silvia","middleName":"","lastName":"Renn","suffix":""},{"id":414561831,"identity":"0cbf9efb-cd18-419a-ada3-3972e44a9f0e","order_by":4,"name":"Olena Borkovska","email":"","orcid":"","institution":"GRID3 Inc","correspondingAuthor":false,"prefix":"","firstName":"Olena","middleName":"","lastName":"Borkovska","suffix":""},{"id":414561832,"identity":"2622fbbd-bd88-4f1c-8a6b-8c9154f70fca","order_by":5,"name":"Chisenga Abel Musuka","email":"","orcid":"","institution":"GRID3 Inc","correspondingAuthor":false,"prefix":"","firstName":"Chisenga","middleName":"Abel","lastName":"Musuka","suffix":""},{"id":414561833,"identity":"85f02ccd-24a4-421d-95a4-c3c60f4e1e11","order_by":6,"name":"Garikai Membele","email":"","orcid":"","institution":"GRID3 Inc","correspondingAuthor":false,"prefix":"","firstName":"Garikai","middleName":"","lastName":"Membele","suffix":""},{"id":414561834,"identity":"1052e4f5-00de-4c03-93a2-10f3782a6dd2","order_by":7,"name":"Attila N. Lazar","email":"","orcid":"","institution":"University of Southampton","correspondingAuthor":false,"prefix":"","firstName":"Attila","middleName":"N.","lastName":"Lazar","suffix":""},{"id":414561835,"identity":"b02b869b-1fd2-453c-9fbd-a9ea9f7f8212","order_by":8,"name":"Andrew J. Tatem","email":"","orcid":"","institution":"University of Southampton","correspondingAuthor":false,"prefix":"","firstName":"Andrew","middleName":"J.","lastName":"Tatem","suffix":""}],"badges":[],"createdAt":"2025-01-24 21:38:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5898324/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5898324/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12942-025-00398-7","type":"published","date":"2025-05-24T15:58:10+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":76686639,"identity":"3d60b48a-e6d3-4595-bc90-376bcfebaedc","added_by":"auto","created_at":"2025-02-19 16:14:28","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":121090,"visible":true,"origin":"","legend":"\u003cp\u003eThe count of building footprints per province, for each building footprint data product. Total counts are displayed as solid bars. The subset of “potentially residential” building footprints (≥9 m\u003csup\u003e2\u003c/sup\u003e and ≤330 m\u003csup\u003e2\u003c/sup\u003e) are indicated by the striped fill.\u003c/p\u003e","description":"","filename":"groupimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/f26e8c6606caa28c424bea14.jpeg"},{"id":76685746,"identity":"b754e80c-1c14-44ab-985f-bfa76b5cddbc","added_by":"auto","created_at":"2025-02-19 16:06:29","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":14247,"visible":true,"origin":"","legend":"\u003cp\u003eMean counts of potentially residential building footprints per GRID3 v2.0 settlement extent, shown for each dataset and stratified by province and L1 degree of urbanisation.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/74eae50a960b9139a07b8842.png"},{"id":76686642,"identity":"956e4abe-9304-468f-9b3b-b3850f327c18","added_by":"auto","created_at":"2025-02-19 16:14:29","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":219567,"visible":true,"origin":"","legend":"\u003cp\u003eThe count of settlements in classes A-D, based on the count of potentially residential building footprints per settlement (upper row: GRID3 v2.0 settlement extents, lower row: derived settlement clusters), for each building footprint dataset. Derived settlement clusters are formed based on the presence of building footprints and so no Class A settlements are produced with this method. Note that the y-axis values differ between classes A-C and class D.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/92f0b6551e685747a1670bb4.png"},{"id":76685736,"identity":"dbd89080-d4f4-498f-b3c5-df609c1300c7","added_by":"auto","created_at":"2025-02-19 16:06:28","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1068313,"visible":true,"origin":"","legend":"\u003cp\u003eMapped examples of class D settlement shown for two example locations (A and B). For these locations, the class D settlement extents (upper) and class D derived settlement clusters (lower) are shown for each building footprint dataset (left to right: Ecopia, Google, Microsoft and OSM). Three example locations with clear differences between datasets are marked as X, Y and Z.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/9f539d481116397ea04ee290.png"},{"id":76685738,"identity":"c43eb993-e904-4e39-90df-5e5b8f5a8939","added_by":"auto","created_at":"2025-02-19 16:06:28","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":974436,"visible":true,"origin":"","legend":"\u003cp\u003eCumulative count of Class D settlement extents stratified by the count of building footprint datasets in agreement (n = 1-4). Within each stratum (X-axis), the total count of class D settlements is labelled by the combination of building footprint datasets (E=Ecopia year 2, G=Google v3, M=Microsoft (Oct. 2023 download) and O=OSM (Oct. 2023 download)).\u003c/p\u003e","description":"","filename":"floatimage42.png","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/ee41351beed72b50f8f5724a.png"},{"id":76685745,"identity":"bbd47864-49aa-4fc6-bebd-4985a6de6f8d","added_by":"auto","created_at":"2025-02-19 16:06:29","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":405646,"visible":true,"origin":"","legend":"\u003cp\u003eFor all settlement extents classified as class D with one or more building footprint datasets (n=35,861), (a) the percentage that are classified as class D with each dataset, and (b) the percentage classified as class D with 1-4 datasets, is mapped at district level.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/27f80adc7a23e701745f69bf.png"},{"id":76685741,"identity":"1bc77a8a-be13-423b-8879-07875a0ea325","added_by":"auto","created_at":"2025-02-19 16:06:29","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":14841,"visible":true,"origin":"","legend":"\u003cp\u003eThe count of class D settlement extents selected with each building footprint dataset, stratified by province, and the L1 degree of urbanisation class. Note that the y-axis values vary between the L1 degree of urbanisation strata.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/d851b044adca586d16e9a186.png"},{"id":83460082,"identity":"d7a475d9-4d0e-45da-b941-02cc6e7f676a","added_by":"auto","created_at":"2025-05-26 16:10:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3401444,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/eb61a35b-e2df-491c-92b4-6a9cecdea756.pdf"},{"id":76685733,"identity":"2e80c714-0b3d-4d77-98f4-bbb574fc36cf","added_by":"auto","created_at":"2025-02-19 16:06:28","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1122513,"visible":true,"origin":"","legend":"","description":"","filename":"SUPPLEMENTARYINFORMATIONAssessingtheimpactofbuildingfootprintdatasetchoiceforhealthprogrammeplanning.docx","url":"https://assets-eu.researchsquare.com/files/rs-5898324/v1/1dec1381ceaac33eb32f11ed.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Assessing the impact of building footprint dataset choice for health programme planning: a case study of indoor residual spraying (IRS) in Zambia","fulltext":[{"header":"Background","content":"\u003cp\u003eHealth programme planning is an inherently geographic problem, requiring an understanding of populations and infrastructure to estimate where, when and what resources are required. From provision of specific services to planning national vaccination programmes, information on target populations and where they are located is key (Chaney et al. (2021), Tatem (2022), Ali et al. (2020)). The COVID-19 pandemic brought health programme planning into the spotlight, with efforts needed to rapidly plan and implement programmes for patient care, testing and subsequently vaccination (Hengel et al. (2020), WHO (2021)). Effective health programme planning for a range of communicable diseases, informed by an understanding of populations and infrastructure, far predates the COVID-19 pandemic though. Early documented examples include the 1854 London cholera outbreak, where the Broad Street pump was identified as a source of infection following collection of data on cases in relation to the resident population in the vicinity (Snow, 1856). More recently, comprehensive mapping of settlements and tracking of vaccination teams to ensure high coverage rates of target populations, contributed to the elimination of wild polio virus in northern Nigeria (Barau et al. (2014), Touray et al. (2016), Borowitz et al. (2023)).\u003c/p\u003e \u003cp\u003eTo support a geographic approach to health programme planning, detailed data are needed to understand the spatial distribution of populations, infrastructure and the built and natural environments (Dummer (2008), Chaney et al. (2021)). In the early stages of COVID-19 mass vaccination campaigns, such data enabled assessment of population access to vaccination facilities and identification of where long travel times may be an impediment to a population being vaccinated (Muchiri et al. (2022), Whitehead et al. (2021), Leithäuser et al. (2021), Rader et al. (2022)). Similar analyses have highlighted gaps in provision of health services (Macharia et al. (2017), Dotse-Ggorgbortsi et al., (2022), Hierink et al. (2023)) and vaccination programmes for many communicable diseases (Joseph et al. (2020), Utazi et al. (2019), Utazi et al. (2024)). The availability, completeness and quality of high-resolution geospatial data on populations, settlements and infrastructure however varies between countries. In settings with well-developed geospatial data systems and comprehensive national address databases, authoritative data on the locations of residential populations may be readily available, whilst in other settings such data may be very limited or outdated, if it exists at all. Low-income countries tend to have less well-developed geospatial data systems and high burdens of communicable diseases, resulting in a high demand for planning health programmes and campaigns, but limited geospatial data to support this planning.\u003c/p\u003e \u003cp\u003eDevelopments in computing power, satellite imagery and machine-learning algorithms over the past decade have enabled the growth of new sources of geospatial data relevant for health programme planning, including increasingly detailed land cover (Brown et al. (2022), Karra et al. (2021), Zanaga et al. (2022)), settlement (Pesaresi and Politis (2023), Esch et al. (2022)) and population (Doxsey-Whitfield et al. (2015), Sorichetta et al. (2015), Freire et al. (2016), Tiecke et al. (2017)) datasets at global and continental scales. In the past five years, advances in feature extraction algorithms, combined with availability of high-resolution satellite imagery, have enabled detailed mapping to the level of individual buildings at scale (Chamberlain et al., 2024). Several national and multi-country building footprint datasets have been published, predominantly by commercial companies such as Google (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://sites.research.google/open-buildings\u003c/span\u003e\u003cspan address=\"https://sites.research.google/open-buildings\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), Microsoft (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.microsoft.com/e\u003c/span\u003e\u003cspan address=\"https://www.microsoft.com/e\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e n-us/maps/building-footprints) and Ecopia (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ecopiatech.com/global-feature-extraction\u003c/span\u003e\u003cspan address=\"https://www.ecopiatech.com/global-feature-extraction\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). For many countries, these new datasets have provided building footprint data for the first time without extensive manual digitisation efforts. Prior to these developments, no national building footprint datasets from authoritative sources were openly available for many countries with high burdens of communicable diseases, including all countries in Africa (Biljecki et al., 2021). In their most basic form, building footprints consist of mapped digital outlines of buildings (polygons) and provide information on the location, size and shape of buildings. Some building footprint datasets may also include additional attributes, such as a building’s type or height. They enable building and settlement locations to be mapped, and their proximity and spatial arrangement with neighbouring buildings understood (Jochem et al., 2021).\u003c/p\u003e \u003cp\u003eThe potential of new sources of high-resolution geospatial data in planning and delivery of health campaigns is increasingly being harnessed, supported by published guidance such as the Geo-Enabled Microplanning Handbook (WHO-UNICEF COVAX GIS Working Group, 2023). Geospatial datasets are also being integrated into user-friendly tools such as the Reveal platform (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://revealprecision.com\u003c/span\u003e\u003cspan address=\"https://revealprecision.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), which supports health campaign planning and delivery by integrating geospatial data on building footprints, points of interest, and health and administrative boundaries into an interactive map interface (Pinchoff et al. (2016), Kamanga et al., (2015), Martin et al. (in review)). This is accompanied by an interface to plan out campaigns including calculating commodity requirements and developing microplans (detailed plans for delivery of health interventions to target populations (WHO (2008), WHO (2019)). During campaign delivery, teams can use the map interface to navigate to and record data against individual buildings, acting as tasks for field service delivery teams, enhancing coverage and enabling campaign progress to be monitored and visualised. The Reveal platform was initially developed to support malaria control programmes but has been used for a range of health interventions, including geo-enabled microplanning for vaccination campaigns and mass drug administration (Winters et al., (2024), Arambepola et al. (2021)).\u003c/p\u003e \u003cp\u003eVector control measures are a mainstay of malaria control programmes, with both long-lasting insecticide-treated nets (LLINs) and indoor residual spraying (IRS) being widely used (WHO, 2023) to reduce vector abundance and minimise transmission. LLINs are intended to provide protection for household-members sleeping under a net, by physically preventing individuals from being bitten by infected mosquitoes whilst asleep. Insecticide-treated nets kill mosquitoes which land on the net, thus helping to reduce the vector population and providing some degree of community-level protection, in addition to the immediate household-level protection. In contrast, IRS does not prevent individuals from being bitten by an infected mosquito but instead is intended to kill endophilic mosquitoes that rest on walls of residential buildings after feeding, and in this way reduce the vector population and disrupt the malaria transmission cycle. For IRS to be effective in reducing malaria incidence in a community, high- and uniform-coverage is needed; a minimum threshold of 80% coverage of eligible structures is commonly used as an operational target (WHO, 2023).\u003c/p\u003e \u003cp\u003eFor IRS and LLIN campaigns to reach the high levels of coverage needed, data on the count and location of households, residential buildings and population are required. The availability of such data can be limited and quickly become outdated. Surveys or household listing exercises to collect new data are expensive, time-consuming, and resource-intensive, but until recently were the sole option for updating such data. The time and resources required for enumeration from satellite imagery in comparison to field-based enumeration, in the context of planning and monitoring IRS campaigns in Zambia, was found to be 22 times faster, and 10 times less costly than field-based enumeration (Kamanga et al. (2015)). High-resolution geospatial data on population, settlements and infrastructure, have been used extensively in studies on malaria risk, control and elimination (Hay et al. (2010), Reid et al. (2010), Gething et al. (2011), Clements et al. (2013)), and are increasingly being used to support programme planning for malaria (Pinchoff et al. (2016), Bridges et al. (2018), Borkovska et al. (2022), Kyomuhangi et al. (2023)). The need to utilise good quality data to maximise malaria control impact through sub-national targeting, is emphasised in the WHO High Burden to High Impact (HBHI) initiative (WHO, 2018). New building footprint data products can help in maximising coverage of IRS campaigns by estimating the volume of insecticide and number of spray teams required, supporting spray teams in locating buildings, enabling identification of missed buildings and subsequent calculation of coverage rates.It can also help in identifying vulnerable communities that are remote and/or hard to reach, and therefore often missed during health service delivery.\u003c/p\u003e \u003cp\u003eThe growth in building footprint data products has brought new opportunities for detailed geospatial data to be utilised in supporting health programmes, but the degree to which these building footprint data products are interchangeable has only recently started to be explored (Chamberlain et al. (2024); Gevaert et al. (2024)). Knowledge of the impact of utilising different building footprint data products in subsequent analyses and decision making remains limited. In this study, we explore the impact of utilising different building footprint data products in analyses to support health intervention planning. We focus on Zambia, with a case study on planning implementation of malaria vector control measures and in particular the identification of priority settlements for IRS campaigns. Our results show that when the same criteria are applied with different building footprint datasets, there is considerable variation in the number and location of priority settlements that are identified.\u003c/p\u003e "},{"header":"Methods","content":"\u003cp\u003eTo quantify the impact of utilising different building footprint datasets in health programme planning, we use the example of IRS campaign planning in Zambia. To identify priority locations for IRS campaigns, we employ target criteria used in recent campaigns, conducted by implementing partners of the National Malaria Elimination Programme (NMEP). The target criteria relate to the area of, distance between and count of building footprints, and we apply the same criteria with four different building footprint data products. We separately apply two methods to identify suitable settlements or clusters of building footprints that meet the target criteria for potential inclusion in IRS campaigns. The first approach utilises existing mapped data on settlement extents and the second derives clusters of structures, solely based on the building footprint data with a proximity threshold applied.\u003c/p\u003e\u003cp\u003eStudy setting\u003c/p\u003e\u003cp\u003eThere were 7,050,968 malaria cases in Zambia in 2021, with a case incidence of 340/1,000 population (NMEC, 2022). Malaria risk in Zambia is spatially heterogeneous, with high incidence particularly in the north, north-west and east of the country, as well as in some areas near national borders. The NMEP has used both IRS and LLINs as vector-control strategies in recent years. IRS was prioritised as a primary vector control approach alongside LLINs in the National Malaria Elimination Strategic Plan (NMESP) for 2017–2021 (NMEC, 2017). The current NMESP for 2022–2026 prioritises LLINs as the primary intervention for vector control, with IRS targeted in high burden areas (NMEC, 2022).\u003c/p\u003e\u003cp\u003eGeospatial data on buildings has been used to plan vector control activities under both strategic plans. District IRS planning and implementation maps, which included counts of manually digitised building footprints and malaria risk data per settlement, were first used in 2014 in 15 districts in Luapula and Central provinces to prioritise available IRS resources (Kamanga et al. (2015)). In 2015 and 2019, these building footprint enumeration maps were expanded to 42 districts, across five provinces, to support IRS planning. Sources of building footprints used during this time included manual delineation and enumeration of buildings from satellite imagery, field-verified footprint data and OpenStreetMap. The development by Ecopia of the first multi-country building footprint dataset for sub-Saharan Africa in 2019, provided building footprints nationally for Zambia (under a licence permitting use for humanitarian purposes), without requiring bespoke exercises to manually digitise buildings. The Ecopia building footprints enabled counts of buildings and derived settlement extents (CIESIN et al., 2021) to be integrated into malaria microplanning maps for the first time in the 2020-21 vector control campaign. Subsequently, the development of open datasets of building footprints, has widened the availability of data further, with Google Open Buildings data then used in two districts to support IRS plannings in 2022 and 2023.\u003c/p\u003e\u003cp\u003eOne of the challenges with using satellite-derived building footprint datasets is that footprints are not labelled as residential or non-residential structures. Without attribute data on building function or use, vector control interventions in Zambia have previously applied minimum and maximum area thresholds to exclude probable non-residential buildings (Kamanga et al., 2015). Alternatively, the proportion of structures that were residential has been estimated, based on samples of field-verified data. These approaches have enabled estimation of counts of residential buildings per settlement and identification of priority settlements for IRS (Kamanga et al. (2015), Pinchoff et al. (2016), Borkovska et al. (2022), Martin et al. (in review)).\u003c/p\u003e\u003cp\u003e \u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDetails of the building footprint data products available as of October 2023, and included in the comparative analysis for Zambia.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDataset\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVersion; release date\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMethod; source\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNotes\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eEcopia\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDigitize Africa building footprints\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e“Year 1”; 2019\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e- Automated feature extraction from satellite imagery (Maxar 30-50cm imagery)\u003c/p\u003e \u003cp\u003e- \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ecopiatech.com/\u003c/span\u003e\u003cspan address=\"https://www.ecopiatech.com/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e- Custom, commercial license which restricts usage predominantly to humanitarian applications\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDigitize Africa building footprints\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e“Year 2”; 2021\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eGoogle\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOpen Buildings\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ev2; 2022\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e- Automated feature extraction from satellite imagery (Google Maps imagery with 50cm spatial resolution)\u003c/p\u003e \u003cp\u003e- \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://sites.research.google/open-buildings/\u003c/span\u003e\u003cspan address=\"https://sites.research.google/open-buildings/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e- CC BY 4.0 License or ODbL (Open Database License) v1.0\u003c/p\u003e \u003cp\u003e- Includes data at building level on precision confidence score.\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOpen Buildings\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ev3; 2023\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eMicrosoft\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e“Global” building footprints\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDownloaded January 2023\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e- Automated feature extraction from satellite imagery (Bing Maps imagery, including Maxar and Airbus)\u003c/p\u003e \u003cp\u003e- \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/microsoft/\u003c/span\u003e\u003cspan address=\"https://github.com/microsoft/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eGlobalMLBuildingFootprints\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e- ODbL v1.0 License\u003c/p\u003e \u003cp\u003e- No clear versioning system\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e“Global” building footprints\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDownloaded October 2023\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eOpenStreetMap (OSM)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOSM features tagged building=*\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDownloaded January 2023\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e- Predominantly manual digitization from satellite imagery, aided by some automated feature detection (imagery from various sources)\u003c/p\u003e \u003cp\u003e- \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.openstreetmap.org/\u003c/span\u003e\u003cspan address=\"https://www.openstreetmap.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e- ODbL v1.0 License\u003c/p\u003e \u003cp\u003e- No versioning system as constantly being updated given VGI-nature of OSM\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOSM features tagged building=*\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDownloaded October 2023\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003eData\u003c/p\u003e\u003cp\u003eFour different building footprint data products were included in the comparative analysis (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e): Google Open Buildings (Sirko et al., 2021), Microsoft “Global” building footprints (Microsoft, 2022), Ecopia DigitizeAfrica building footprints (Price and Hallas, 2019) and buildings extracted from OpenStreetMap (OSM). Building footprints from Ecopia, Google and Microsoft are all examples of data products created by commercial companies, using automated feature extraction methods and high-resolution satellite imagery (Li et al., 2022). In contrast, OSM buildings are an example of volunteered geographic information (VGI) product, with features digitised by a community of mappers and added to OpenStreetMap. Over the past decade, billions of features globally have been added to OSM, but given it's VGI-nature, inevitably is spatially heterogenous in terms of geographic coverage. Prior to the recent developments of multi-country building footprint datasets through automated feature extraction, OSM was the only source of mapped building data available for many countries. A third potential source of building footprints is authoritative datasets from government agencies; however no authoritative national dataset of building footprints is available for Zambia (Biljecki et al., 2021).\u003c/p\u003e\u003cp\u003eGiven the rapidly evolving data landscape, two versions of each building footprint data product were considered (8 datasets in total: Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). All building footprint data products consisted of vector polygons with limited, if any, additional attribute information. Some information related to building use or construction materials is available for a limited subset of OSM buildings, but not for other data products so it was not considered in these analyses.\u003c/p\u003e\u003cp\u003eIn the first method of quantifying counts of building footprints per settlement, predefined settlement extents were used – specifically the GRID3 v2.0 Settlement Extent dataset (CIESIN and Novel-T, 2023). This dataset consists of vector polygons, delineating outlines of settlements or groups of buildings which have been created through geospatial processing of Ecopia Digitize Africa “year 2” building footprints. These settlement extents represent geographic groupings of building footprints, and therefore do not necessarily correspond to individual settlements recognised by local or national administrations or the communities residing there. Settlement extents are also not named. Consequently, a single named settlement may be represented in the GRID3 settlement extent dataset as multiple polygons, and conversely a single settlement extent polygon may represent multiple named settlements. The GRID3 v2.0 settlement extents include a settlement type attribute based on the Degree of Urbanisation approach (Dijkstra et al., 2021). We used this Level 1 Degree of Urbanisation attribute (rural, urban cluster or urban centre) to stratify settlement extents by rural/urban types.\u003c/p\u003e\u003cp\u003eData processing\u003c/p\u003e\u003cp\u003eFor each building footprint dataset, polygons were converted to a common geographic coordinate system (WGS84, EPSG: 4326) and the area of each building footprint polygon was calculated in square metres. Any building footprint polygon with a geometric centroid outside the national boundary of Zambia was excluded. As IRS campaigns are specifically targeted at buildings in which residents sleep, only the subset of building footprints that were most likely to be residential, were included (henceforth referred to as “potentially residential”). As has been used previously in IRS campaigns in Zambia, minimum and maximum area thresholds were applied to exclude probable non-residential buildings (Kamanga et al., 2015). Thus, for each dataset, building footprint polygons that were \u0026lt; 9 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e (probable toilets or granaries) or \u0026gt; 330 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e (likely non-residential structures such as churches, warehouses or commercial premises) in area, were excluded from the analyses (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe first method of calculating counts of building footprints per settlement, utilised the GRID3 v2.0 settlement extents dataset as the basis for defining settlements. The GRID3 settlement extents are derived from “year 2” Ecopia building footprints, which have been buffered by 50 m and then overlapping buffers dissolved. Utilising these settlement extents with other sources of building footprints will likely result in some building footprints being located outside settlement extents, particularly in locations experiencing settlement growth. For this reason, we expanded the spatial coverage of each GRID3 settlement extent by up to 1 km, with any building footprints that were within 1 km of a settlement extent, included in the count of building footprints of the settlement extent that they were nearest to.\u003c/p\u003e\u003cp\u003eA second method was also tested, which did not utilise pre-defined settlement extents and instead, for each building footprint dataset in turn, groupings of neighbouring building footprints were created based on a proximity threshold. Recent IRS campaigns in Zambia have used this approach with a range of proximity thresholds, from 50 m (Pinchoff et al., 2016) up to 250 m (Bridges et al., 2018). For these campaigns, proximity thresholds were decided based on both expected travel distances of mosquitoes and practical considerations around minimising travel time for IRS field teams, so as to maximise spraying of neighbouring structure and community-level protection. In this analysis, we utilised a proximity threshold of 100 m (in line with both previous IRS campaigns and the buffer distance used in deriving GRID3 v2.0 settlement extents), such that building footprints within 100 m of each other, were grouped into a cluster. For each building footprint dataset in turn, clusters were delineated by buffering all potentially residential building footprints by 50 m and dissolving all overlapping buffered areas. The resulting delineated clusters of building footprints are henceforth referred to as “derived settlement clusters”.\u003c/p\u003e\u003cp\u003eFor both settlement definitions, the counts of potentially residential building footprints per settlement extent/derived settlement cluster were then calculated for each of the four building footprint data products (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eStratification of settlements\u003c/p\u003e\u003cp\u003eIn identifying potential settlements suitable for IRS, settlements were classified based on threshold counts of potentially residential building footprints. Recent IRS campaigns have utilised a minimum threshold in terms of the count of potentially residential structures/buildings per settlement, with a threshold of 25 residential buildings commonly employed to ensure operational efficiency (Pinchoff et al., 2016). For IRS campaigns in Zambia, this threshold was initially applied based on groupings of buildings that were all located within a specified distance of each other (i.e. following the second definition of settlement described above). More recently, the same threshold has been applied to counts of potentially residential buildings within pre-defined GRID3 settlement extents (Borkovska et al., 2022). In our analysis we have employed both definitions of settlement, with the same count threshold of 25 potentially residential building footprints per (i) expanded GRID3 settlement extent and (ii) derived settlement cluster, applied. For each building footprint dataset in turn and considering only the subset identified as potentially residential (≥ 9 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e and ≤ 330 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e), settlements were stratified into four classes: no building footprints (class A), 1–5 building footprints (class B), 6–24 building footprints (class C) and 25 or more building footprints (class D; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). From the stratified settlements, class D settlements were considered as the priority locations for IRS implementation, in line with criteria used in recent IRS campaigns in Zambia.\u003c/p\u003e\u003cp\u003eComparative analysis\u003c/p\u003e\u003cp\u003eTo assess the impact of the choice of building footprint dataset on the identification of priority settlements for IRS campaigns, we calculated summary statistics related to the count, location and spatial similarity of class D settlements (those with 25 or more potentially residential building footprints). Our comparative analysis focussed on the most recent version of each of the four building footprint data products. The agreement in class D settlement extents across building footprint datasets was assessed in terms of the count (n = 1–4) and combination of datasets (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e and Supplementary Figures A.1 and A.2). For each building footprint dataset, the count of class D settlement extents per province (administrative unit level 1) with rural/urban stratification was calculated, along with the count of building footprints within these settlement extents (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e and Supplementary Table A.5).\u003c/p\u003e\u003cp\u003eTo further assess geographic similarity in class D settlement extents, Jaccard coefficient values were calculated for each province, on a pairwise basis between datasets (Supplementary Figure A.3). The Jaccard coefficient (𝐽(X,Y) = | X ∩ Y | / | X ∪ Y|) was calculated as the count of settlement extents that were identified as class D with two building footprint datasets (X and Y), divided by the count of class D settlement extents identified with building footprint dataset X and/or Y. This calculation was not weighted by the area of settlement, such that settlement extents with 25 buildings and settlement extents with, for example, 5000 buildings were both considered equally. For the subset of settlement extents classified as class D with all four building footprint datasets, statistics related to the variability in counts of building footprints per settlement extent were also calculated (Supplementary Tables A.3, A.4 and Supplementary Figure A.4).\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eComparing counts of building footprints and potentially residential building footprints per province across Zambia (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), shows considerable variation between data products (Ecopia, Google, Microsoft and OSM). For the majority of provinces, counts of building footprints were highest for Ecopia (eight provinces) or Google (two provinces \u0026ndash; Lusaka and Copperbelt), and lowest for Microsoft (six provinces) or OSM (four provinces). For Central and Muchinga province however, the count of OSM building footprints is greater than the count of Google footprints.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFor each building footprint data product, two datasets were included in this comparison of building footprint counts, shown as pairs of bars in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The two datasets are either the two most recent versions, or in the case of OSM and Microsoft, data extracted at two time points (January and October 2023). The comparison of building footprint counts per province, between pairs of datasets, show variations in counts, with patterns differing between provinces. Counts of Ecopia building footprints per province increased between the \u0026ldquo;year 1\u0026rdquo; and \u0026ldquo;year 2\u0026rdquo; datasets. Counts of Google building footprints increased between v2 and v3 for a couple of provinces (Copperbelt and Lusaka) but decreased for all other provinces. Without details of the satellite imagery used or changes to the feature extraction or post-processing algorithms, it is not possible to know the cause of the differences observed, but these may be due to temporal differences in satellite imagery, reduced cloud cover in imagery, and/or revised definitions of what constitutes a single building. Counts of OSM and Microsoft building footprints per province showed little change between dataset versions, except for Northern province, where counts of OSM buildings increased considerably from 243,627 in January 2023 to 350,379 in October 2023.\u003c/p\u003e \u003cp\u003eIn using building footprints for health programme planning and implementation, it may be that only a subset of buildings is of interest. For example, in Zambia, IRS campaigns are specifically targeted at residential (sleeping) structures and therefore area thresholds have been applied to exclude the largest and smallest buildings that are unlikely to be residential. When minimum (\u0026ge;\u0026thinsp;9 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e) and maximum (\u0026le;\u0026thinsp;330 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e) area thresholds are applied, the proportion of the total building footprints that are excluded varies between provinces and data products (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). For most provinces, the greatest reduction in count of building footprints is observed for Ecopia, with minimal change in the count of Microsoft building footprints, except for Copperbelt and Lusaka provinces (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Note that for the remaining analyses, only the subset of building footprints considered to be potentially residential (\u0026ge;\u0026thinsp;9 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e and \u0026le;\u0026thinsp;330 m\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e) are included.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eCalculating counts of potentially residential building footprints per settlement extent also shows considerable variation between building footprint data products. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the mean count of potentially residential building footprints per settlement extent, stratified by province and Level 1 degree of urbanisation (Dijkstra et al., 2021). For rural settlement extents, the mean count of potentially residential building footprints per settlement is generally highest for Ecopia and lowest for Microsoft or OSM. In rural areas of Muchinga and Central provinces, mean counts for Ecopia, Google and OSM are very similar, with mean counts of Microsoft building footprints being lower. In both Urban Clusters and Urban Centres, Google has the highest mean count of potentially residential building footprints per settlement extent for most provinces, and OSM the lowest. For settlement extents classified as Urban Centres, the magnitude of difference in mean counts between building footprint datasets in some provinces is massive, for example in Eastern province, the mean count of potentially residential building footprints per settlement extent is 23,961 for OSM (Jan. 2023) and 54,940 for Google (v3). Similarly, for Luapula province, the mean count of potentially residential building footprints per Urban Centre extent ranges from 19,123 (OSM Jan. 2023) to 46,481 (Google v3). There is only a single settlement extent classified as an Urban centre in Eastern, Luapula, North-Western, Northern and Southern provinces, and thus the mean counts per Urban centre for these provinces, reflect values only from one settlement extent.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eCounts of building footprints per settlement have been used in planning past IRS campaigns in Zambia to identify priority settlements. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows, for each building footprint dataset, the count of settlements in classes A-D, when settlements are classified based on the count of potentially residential building footprints per settlement (considering only the most recent version of each data product). Nationally, the count of settlements with 25 or more potentially residential building footprints (class D) is greatest when Ecopia year 2 building footprints are used (30,749 settlement extents and 29,032 derived settlement clusters). The second highest count of class D settlement is for Google v3 (24,977 settlement extents and 22,806 derived settlement clusters), followed by OSM (20,330 settlement extents and 19,202 derived clusters). The count of class D settlement extents based on Microsoft building footprints is the lowest of any data product (12,982 settlement extents and 12,043 derived clusters). Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e also shows the breakdown of the remaining settlements when threshold counts of 0 (class A), 1\u0026ndash;5 (class B) and 6\u0026ndash;24 potentially residential building footprints (class C) are applied for each data product. Counts of class B and C settlement extents and derived clusters were highest for Ecopia v2 building footprints. Conversely, the count of settlement extents with no potentially residential building footprints (class A) was highest for Microsoft. Counts of potentially residential building footprints within each settlement class are summarised in Supplementary Tables A.1 and A.2.\u003c/p\u003e \u003cp\u003eThe count of settlements in each class nationally with different building footprint data products (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) however hides any geographic variation in the classification of settlements, when different building footprint data products are used. The count of settlement extents in each class (the upper part of Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), is shown in Supplementary Figure A.1, with the addition of nested bars. These nested bars show the subset of identical settlement extents that are classified as the same class, with each building footprint dataset. For example, of the 12,982 class D settlement extents selected when a threshold of 25 is applied with Microsoft building footprints, 12,754 (Ecopia), 12,326 (Google) and 10,135 (OSM) of the same settlement extents are also classified as class D when the same threshold is applied with the other building footprint datasets. Of the 30,749 class D settlement extents with Ecopia building footprints, a subset of 21,922, 12,754 and 17,904 are class D when the same threshold is applied to Google, Microsoft and OSM building footprints respectively.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFor some locations, class D settlements are quite similar between building footprint datasets, but for other locations the choice of building footprint dataset and/or the settlement definition has a considerable impact on which settlements meet the threshold for class D. In Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, the class D settlements are shown for both GRID3 settlement extents (left side of both panels) and derived settlement clusters (right side), for two example locations. The location marked X shows class D settlement extents (and derived settlement clusters) with Ecopia, but no class D settlement extents for the other three datasets. Location Y has class D derived clusters with Microsoft and Ecopia only, but class D settlement extents are identified in the same location with Microsoft, Ecopia and Google. Location Z has class D settlement extents with OSM, Google and Ecopia, but no class D derived clusters for OSM, and much smaller clusters for Google and Ecopia. In addition, the settlement definition can have considerable impact on which settlements are identified as class D (see for example the differences between class D settlement extents and derived clusters with Google v3 in panel (a) and Microsoft in panel (b) of Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e further explores the dissimilarity in settlement extents classified as class D with each building footprint dataset. In total across the four datasets, 35,861 settlement extents were identified as class D (IRS priority settlements) as they had 25 or more potentially residential building footprints in one or more building footprint datasets. Of these, there were 9,738 settlement extents identified with all four building footprint datasets (EGMO). Conversely there were 10,123 settlement extents identified in only one (E, G, M or O) of the four building footprint datasets, with over half of those settlement extents only being identified in terms of counts of Ecopia building footprints (n\u0026thinsp;=\u0026thinsp;5,571). A smaller number of settlement extents were only selected with Google (n\u0026thinsp;=\u0026thinsp;2,506) and OSM (n\u0026thinsp;=\u0026thinsp;1,957), with a minimal number selected only with Microsoft (n\u0026thinsp;=\u0026thinsp;89). For settlement extents classified as class D with two or three datasets, the majority included combinations of Ecopia with Google and/or OSM (e.g. EG, EO, EGM and EGO). Supplementary Figure A.2 shows the locations of the selected settlement extents for each building footprint dataset combination.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eConsidering all settlement extents classified as class D with one or more building footprint datasets, the percentage classified as class D with each building footprint dataset is mapped at district level (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). The percentage of class D settlement extents found with each building footprint varies considerably, with Ecopia having the most districts (n\u0026thinsp;=\u0026thinsp;102/116) with over 75% of all class D settlement extents selected. In contrast, Microsoft had the fewest districts with over 75% of all class D settlement extents selected (n\u0026thinsp;=\u0026thinsp;2). With Google and OSM, more spatial variability in the percentage of class D settlements with each dataset, is observed (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). In Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb, the percentage of settlement extents classified as class D in only one dataset, and in two, three or four datasets is mapped per district. The proportion of settlement extents classified as class D with just one dataset is greatest in selected districts in Southern and Western provinces, with seven districts having over 45% of class D settlement extents classified as such with just one building footprint dataset. Conversely, four districts in Eastern province, two districts in North-Western and one district in Luapula had less than 15% of class D settlement extents as class D with just one building footprint dataset. These same districts all had more than 45% of class D settlement extents classified as such with four buildings footprint datasets, along with six districts in Luapula and one in Northern. Sinda district in Eastern province had the highest proportion of class D settlement extents classified as such with all four building footprint datasets (82%).\u003c/p\u003e \u003cp\u003eThe spatial dissimilarity of class D settlement extents is further explored in Supplementary Figure A.3, with Jaccard Coefficient values for each pairwise combination of building footprint datasets, stratified by province. In terms of class D settlement extents, the highest Jaccard Coefficient values (greatest spatial similarity in settlement extents) are found with Google and Ecopia building footprints for all provinces apart from Muchinga and Central (Supplementary Figure A.3). For Muchinga and Central provinces, the spatial similarity between class D settlement extents is greatest between Ecopia and OSM. There is considerable variation in Jaccard Coefficient values between provinces, with Luapula and Eastern provinces tending to have higher values, and Western province lower values, for most pairwise combinations. The same pairwise combination of datasets can vary a lot between provinces, for example OSM and Microsoft has a Jaccard coefficient of 0.66 for Luapula, but only 0.24 for Western province.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eGiven the geographic differences observed in class D settlements with different building footprint datasets, the choice of dataset may bias the selection of settlements, such that certain locations (and consequently population groups) could be inadvertently included or excluded from IRS campaigns. In Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e, the count of class D settlement extents for each building footprint dataset is stratified by province and settlement type. In rural areas, the count of class D settlement extents is highest based on Ecopia building footprints in 8 provinces, with the count of class D settlement extents based on Google building footprints being higher in two provinces (Lusaka and Copperbelt). Across all provinces, the number of rural class D settlement extents with Microsoft building footprints is considerably less than the number of class D settlement extents with Ecopia building footprints; the count of Microsoft rural class D settlement extents being less than 65% of the count of Ecopia rural class D settlement extents, and as low as 21% in Western province. In some rural settings, the number of OSM class D settlement extents is comparable to Ecopia or Google (e.g. Central and Southern), but in other provinces (e.g. Lusaka) there are far fewer OSM class D settlement extents, likely reflecting geographic differences in OSM coverage.\u003c/p\u003e \u003cp\u003eIn urban clusters and centres, the total number of class D settlement extents is expectedly fewer (although the extents themselves are larger) and there is also less variation in the total number between building footprint datasets. For Eastern, Luapula, North-Western, Northern and Southern provinces, there is only a single class D settlement extent classified as an urban centre (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). For urban centres in most other provinces, the count of class D settlement extents is highest for Google building footprints. Even with very similar counts of class D settlement extents with different building footprint datasets, there is considerable variation in the total count of building footprints within class D settlement extents (Supplementary Table A.5). In the most extreme example, within class D urban centre settlement extents in North-Western province, there are 92,739 Ecopia, 106,216 Google, 64,280 Microsoft and only 6,123 OSM building footprints. Considering just the subset of settlement extents that are classified as class D with all four building footprint datasets (n\u0026thinsp;=\u0026thinsp;9,738, labelled as EGMO in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e), there is still major variability in counts of building footprints per settlement extent (Supplementary Tables A.3/A4 and Supplementary Figure A.4).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eHigh-resolution geospatial data on population, settlements and infrastructure are increasingly being utilised for planning and implementing public health interventions. The rapid growth of multi-country building footprint data products over the last five years has provided new opportunities for population and settlements to be mapped (Chamberlain et al. (2024); Darin et al. (2024)), facilitating geo-enabled microplanning to be more widely integrated into public health campaigns and programmes (Borkovska et al. (2022), Barau et al. (2014)). In the context of malaria control programmes, data on the locations of buildings is key for planning, implementing and evaluating household campaigns (Bridges et al. (2018), Kamanga et al. (2015)). For vector control measures to be effective in reducing malaria transmission, high- and uniform-coverage of communities is needed (WHO (2023), Rehman et al. (2011)). Here we focus on a case study of planning for IRS campaigns in Zambia, where new geospatial datasets and technology have been integrated over the past decade. We show that if settlements are classified by count and proximity of buildings, the settlements identified for potential inclusion in IRS campaigns differ considerably depending on the choice of building footprint dataset.\u003c/p\u003e \u003cp\u003eThe results of our analysis show that both the overall count and the geographic location of class D settlements (those settlements with at least 25 potentially residential building footprints), varies notably between building footprint datasets. Focussing on the most recent version of each of the four building footprint datasets, the count of class D settlement extents nationally varied between 12,982 and 30,749, when Microsoft (October 2023 version) and Ecopia year 2 building footprint datasets were used respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Considering all class D settlement extents identified with any of the four building footprint datasets, the percentage of settlements identified as class D with each dataset in turn also shows considerable geographic variation between datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). When stratified by province (administrative unit level 1), the count of class D settlements in rural clusters was typically highest with Ecopia building footprints, and lowest with Microsoft building footprints (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). Whilst in urban clusters, the count of class D settlements was highest with Google building footprints in most provinces, and lowest with either OSM or Microsoft buildings footprints. These results highlight how the choice of building footprint dataset may bias the selection of settlements, potentially prioritising locations with certain characteristics (e.g. predominantly urban areas) which could create inequalities in campaign provision. Although not the primary focus of this study, the settlement definition used as the basis for calculating counts of building footprints per settlement (GRID3 settlement extents or derived settlement clusters), also affected the number and geographic location of settlements selected (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWhilst settlement and building footprint data are not the sole data used in decision making concerning malaria vector control initiatives, the use of a particular dataset has the potential to influence decisions and strategies. The WHO guidance for IRS does not specify an absolute coverage target, acknowledging that 100% coverage is rarely possible, but a minimum spray target of 80% of eligible structures is commonly used (WHO, 2023). In this context, an accurate count of geo-located residential buildings is necessary for (i) planning to ensure that available resources are targeted to appropriate locations to maximise reductions in malaria transmission, (ii) ensuring that sufficient spray teams, equipment and insecticide are allocated and delivered to the right locations, and (iii) to assess spray coverage after campaign completion. The results of our analysis quantify the relative differences in counts of building footprints per settlement (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e and Supplementary Figure A.4) and highlight the sensitivity of these metrics to the choice of building footprint dataset (Supplementary Tables A.3 and A.4). For example, if the subset of settlement extents that are classified as class D with all four building footprint datasets were all identified as appropriate locations for IRS, for Western province this would mean planning and implementing IRS in 532 settlements (Supplementary Table A.4). For these selected settlements, the choice of building footprint dataset still would have a big impact on the estimation of required resources and assessment of spray coverage. Considering a hypothetical IRS campaign, in which 50,000 buildings were sprayed in Western province, coverage rates across all target settlements in Western province would be calculated as 60.4% (out of 83,803 Ecopia building footprints), 74.4% (out of 67,225 Google building footprints), 129.8% (out of 38,531 Microsoft building footprints) or 82.9% (out of 60,326 OSM building footprints).\u003c/p\u003e \u003cp\u003eThe detailed nature of building footprint polygon data can give an illusion of accuracy, more so than, for example, raster land cover classification maps. Building footprint datasets can commonly have some buildings omitted (false negatives) and other features misclassified as buildings and thus included (false positives) (Heris et al., 2020). These can occur due to gaps in satellite imagery coverage, satellite imagery being outdated, limitations of the feature extraction algorithm, or buildings being obscured in imagery due to tree canopy cover, for example. A lack of metadata can further compound these problems, as the definition of what is considered as an individual building, details of the feature extraction algorithms and post-processing of extracted features, and the temporal imagery coverage, can all vary between data products (Chamberlain et al., 2024). As building footprint datasets evolve and are used in different fields, there is a need for data producers to take steps to address these issues and support data users. Those working with building footprint data to support decision making, in the context of health programme planning and more widely, also need to be aware of differences between data products, to consider potential impacts of choosing a particular dataset and to be cognisant of their limitations. As of October 2023, in addition to buildings from OpenStreetMap, there were at least three building footprint data products in existence for Zambia, with two or more versions of each dataset (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Apart from Ecopia building footprints, no other datasets include attribute information on the date of source satellite imagery, making it impossible to know the timepoints that building footprints are representative of.\u003c/p\u003e \u003cp\u003eFrom the building footprint data products included in our analysis, we would suggest that generally the dataset coverage is most complete for Ecopia and Google building footprints. However, this recommendation has several caveats: our comparative analysis does not include an authoritative reference dataset, the Ecopia building footprints are based on satellite imagery from 2020 or earlier (over half of building footprints were extracted from satellite imagery collected in 2019) and no granular information is provided on source imagery dates for Google. In locations where there have been recent, comprehensive campaigns to digitise buildings in OSM, these data may be more comprehensive in their coverage and better reflect the structure of settlements, but this is difficult to assess at scale. Recent work indicates a likely move towards more dynamic updates of building footprints and settlement datasets in the future (Sirko et al. (2021), Metzger et al. (2024)), as well as potential improvements in building height and typology estimation (Frantz et al. (2021), Esch et al. (2022)). In the context of health programme planning, these developments would likely be beneficial in improving estimation of population numbers (Palacios-Lopez et al. (2022)) and hence programme denominators, as well as enabling settlement change detection with temporally-explicit building/settlement datasets. This study has considered the use of building footprint data products in isolation, but with a growing number and availability of data products, the potential to use data products in combination increases. Further work is needed to develop systematic and robust methods and guidelines to do this. Opportunities to efficiently conduct field validation of satellite-derived datasets, as part of field-deployed health campaigns, should also be pursued. The results of field validation can be used to further improve building datasets, ultimately accelerating planning and delivery of health interventions.\u003c/p\u003e \u003cp\u003eAlthough building footprint datasets extracted from satellite imagery provide new opportunities to support geo-enabled microplanning at scale (Mendes et al. (2021), Borkovska et al. (2022), Alexander et al. (2024)), they are not a replacement for local knowledge or comprehensive field data collection. Enumeration of populations, households and/or buildings may be collected by health facilities through community-based service delivery campaigns and for headcount registers, but these are often not up to date, complete or accurate due to resources and capacity limitations (Kyomuhangi et al. (2023)). Geospatial data on counts and locations of buildings that have been field-verified (for example during resourced campaigns), likely provide the most accurate data on residential buildings for resource planning and delivery (Keating et al. (2021)). This is especially true if field service delivery teams use maps to navigate to all mapped building footprints, then verify that a building exists, add any unmapped buildings and label whether a building is residential or not. Field-verification of this sort has been conducted in Zambia using the Reveal platform mobile app which enables in situ updating of building data and denominators for each implementation cycle (Martin et al. (in review)). However, this data is usually limited to selected campaign prioritised areas, and therefore not available to all health facilities and districts \u0026ndash; making it challenging, but not impossible, to use at scale. In Zambia, building footprints extracted from satellite imagery and field-verified data have been used in combination to inform resource needs during campaign planning and to monitor delivery (Keating et al. (2021), Arambepola et al. (2021)). It is crucial that decision makers still critically assess any such data against existing sources, such as those available at health facility level, in order to ensure that service delivery targets and denominators are as accurate as possible.\u003c/p\u003e \u003cp\u003eAside from IRS planning in Zambia, building footprints or similar datasets have been used for geo-enabled microplanning in a range of other public health interventions (e.g. Mendes et al. (2021), Ou\u0026eacute;draogo et al. (2020)), often utilising a single source of data. Our findings provide insights into the potential variability in estimates when different building footprint datasets are used in other applications, but findings concerning particular data products cannot be assumed to necessarily be transferrable to other countries or contexts. In particular, the coverage of some building footprint data products is highly heterogeneous between countries (Chamberlain et al., 2024), with no data available for some datasets in some areas (for example, conflict-affected locations have been excluded from Google Open Buildings datasets) and a rapidly evolving data landscape.\u003c/p\u003e \u003cp\u003eOur analysis has several caveats. In identifying potential priority locations for IRS, we applied a set of criteria concerning area of, distance between and counts of building footprints. As the national strategy in Zambia has changed over time, the criteria and guidelines used for planning vector control campaigns have similarly evolved (NMEC (2017), NMEC (2022)). The criteria used in our comparative analysis are based on campaigns implemented with the NMEP over the past decade, and provide a framework within which to assess the impact of utilising different building footprint datasets for IRS planning. Our analysis has considered solely building footprint data to identify the subset of settlements that might be considered as suitable for IRS, and has not incorporated the important contextual knowledge of district teams to select the final settlements for inclusion in an IRS campaign. In addition, none of the feature-extracted building footprint datasets (Google, Ecopia and Microsoft) included attributes on building type or use, necessitating an approximation of potentially residential buildings based on simple area thresholds (as has been used in previous work e.g. by Kamanga et al. (2015)). The subset of building footprints considered to be potentially residential based on simple area thresholds, will inevitably include some non-residential structures such as churches, shops or health facilities. Further work should explore more sophisticated classification approaches to identify non-residential/non-sprayable structures (e.g. Sturrock et al. (2018), Lloyd et al. (2020)), and assess the sensitivity of such classification approaches to different building footprint products. Given the lack of an authoritative national reference dataset, our analysis has been limited to comparisons between available building footprint datasets. As new sources of data become available, for example geolocated household data from the 2020 round of national population and housing censuses, these should be included in similar comparative analyses.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThe recent growth in the global availability of detailed building footprint data has provided new opportunities for integrating geo-enabled microplanning into public health interventions, particularly in previously data-scarce locations. This proliferation of datasets however also brings challenges for data users and decision-makers. The results of this study show that the choice of building footprint dataset has a substantial impact on locations selected when criteria are based on counts and proximity of buildings. This has been explored in the context of planning IRS campaigns in Zambia, but similarly large differences would be expected if count-based criteria are used for planning and evaluating other resource provision campaigns. Further work is needed to quantify the sensitivity of the choice of building footprint dataset in such analyses across a range of geographic contexts. Data producers could also better support data users through improved metadata, by routinely disclosing information on spatiotemporal coverage of datasets, definitions of features considered to be buildings and rates of omission and commission. As the building footprint and settlement data landscape continues to grow and diversify, understanding the characteristics and comparability of datasets will remain fundamental to robust data use. Caution will continue to be needed in selecting building footprint data products to integrate into analyses, and where possible, any conclusions drawn should be supported by local, contextual knowledge and field-verified data.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cul\u003e\n \u003cli\u003eIRS \u0026ndash; Indoor Residual Spraying\u003c/li\u003e\n \u003cli\u003eLLIN \u0026ndash; Long-lasting Insecticide-treated Net\u003c/li\u003e\n \u003cli\u003eNMEC \u0026ndash; National Malaria Elimination Centre\u003c/li\u003e\n \u003cli\u003eNMESP \u0026ndash; National Malaria Elimination Strategic Plan\u003c/li\u003e\n \u003cli\u003eOSM - OpenStreetMap\u003c/li\u003e\n \u003cli\u003eUNICEF \u0026ndash; United Nations Children\u0026rsquo;s Fund\u003c/li\u003e\n \u003cli\u003eWHO \u0026ndash; World Health Organisation\u003c/li\u003e\n \u003cli\u003eCOVID-19 \u0026ndash; Coronavirus Disease 2019, caused by the SARS-CoV-2 virus\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"Declarations","content":"\u003ch2\u003eEthics approval\u003c/h2\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003ch2\u003eConsent for publication\u003c/h2\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003ch2\u003eAvailability of data and materials\u003c/h2\u003e\n\u003cp\u003eGoogle v3 building footprints are available to download in tiled format (level 4 S2 cells) from: https://sites.research.google/open-buildings/#open-buildings-download, or accessed via Google Earth Engine: https://developers.google.com/earth-engine/datasets/catalog/GOOGLE_Research_open-buildings_v3_polygons. OpenStreetMap building footprints can be extracted using various plugins, or pre-prepared national files updated daily are available from GeoFabrik: https://download.geofabrik.de/africa/zambia.html. Microsoft \u0026ldquo;global\u0026rdquo; building footprints are available to download from: https://github.com/microsoft/GlobalMLBuildingFootprints. Ecopia building footprints are not publicly available, but access can be requested for humanitarian purposes from Ecopia (https://www.ecopiatech.com/). The GRID3 v2.0 settlement extent dataset is available to download via the GRID3 Data Hub: https://data.grid3.org/datasets/e51042367feb445bb0126cb6ee39893f_0/.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eCompeting interests\u003c/h2\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThis work was initiated with funding from the Bill and Melinda Gates Foundation as part of the GRID3 project [INV-045694]. The work was further developed and completed with funding from the United Kingdom\u0026apos;s Foreign, Commonwealth \u0026amp; Development Office (FCDO), as part of the Data for Development - Unleashing the Data Dividend project [GB-GOV-1-400025-405].\u003c/p\u003e\n\u003ch2\u003eAuthors\u0026apos; contributions\u003c/h2\u003e\n\u003cp\u003eHRC conceptualised the study, and undertook the analysis with input from AJT and DP. HRC wrote the original manuscript draft. DP, AW, OB, CAM, GM, SR, ANL and AJT reviewed, edited and approved the manuscript for submission. ANL and AJT acquired funding.\u003c/p\u003e\n\u003ch2\u003eAcknowledgements\u003c/h2\u003e\n\u003cp\u003eThis study is an output of the WorldPop Research Group at the University of Southampton, United Kingdom. Preparation of this manuscript was supported by HRC attending the University of Southampton Faculty of Medicine/Faculty of Environmental and Life Sciences Writing Retreat (January 2024).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlexander C, McKay A, Bhatt K, da Costa Louren\u0026ccedil;o ALR, Kaplan B, Krishnan RSSG. Pre-trained regional models for extracting buildings from high resolution satellite imagery to support public health initiatives. Remote Sens Applications: Soc Environ. 2024;101270. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rsase.2024.101270\u003c/span\u003e\u003cspan address=\"10.1016/j.rsase.2024.101270\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAli D, Levin A, Abdulkarim M, Tijjani U, Ahmed B, Namalam F, Dougherty L. A cost-effectiveness analysis of traditional and geographic information system-supported microplanning approaches for routine immunization program management in northern Nigeria. Vaccine. 2020;38(6):1408\u0026ndash;15. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.vaccine.2019.12.002\u003c/span\u003e\u003cspan address=\"10.1016/j.vaccine.2019.12.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArambepola R, Yang Y, Hutchinson K, et al. Using geospatial models to map zero-dose children: factors associated with zero-dose vaccination status before and after a mass measles and rubella vaccination campaign in Southern province, Zambia. BMJ Global Health. 2021;6:e007479. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/bmjgh-2021-007479\u003c/span\u003e\u003cspan address=\"10.1136/bmjgh-2021-007479\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarau I, Zubairu M, Mwanza MN, Seaman VY, Improving Polio Vaccination Coverage in Nigeria Through the Use of Geographic Information System Technology, The Journal of Infectious Diseases, Volume 210, Issue, suppl_1 N. 2014, Pages S102\u0026ndash;S110. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/infdis/jiu010\u003c/span\u003e\u003cspan address=\"10.1093/infdis/jiu010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorkovska O, et al. Developing High-Resolution Population and Settlement Data for Impactful Malaria Interventions in Zambia. J Environ Public Health. 2022. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1155/2022/2941013\u003c/span\u003e\u003cspan address=\"10.1155/2022/2941013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorowitz M, Zhou J, Azelton K, Nassar I-Y. Examining the value of satellite data in halting transmission of polio in Nigeria: A socioeconomic analysis. Data Policy. 2023;5:e16. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/dap.2023.12\u003c/span\u003e\u003cspan address=\"10.1017/dap.2023.12\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBridges DJ, et al. Accuracy and impact of spatial aids based upon satellite enumeration to improve indoor residual spraying spatial coverage. Malar J. 2018;17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-018-2236-2\u003c/span\u003e\u003cspan address=\"10.1186/s12936-018-2236-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrown CF, Brumby SP, Guzder-Williams B, Birch T, Hyde SB, Mazzariello J, Tait AM. Dynamic World, Near real-time global 10 m land use land cover mapping. Sci Data. 2022;9(1):251. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41597-022-01307-4\u003c/span\u003e\u003cspan address=\"10.1038/s41597-022-01307-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCenter for International Earth Science Information Network (CIESIN). Columbia University; Flowminder Foundation; United Nations Population Fund (UNFPA); WorldPop, University of Southampton. Mapping and Classifying Settlement Locations. Palisades, NY: Georeferenced Infrastructure and Demographic Data for Development (GRID3). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7916/d8-gzxf-s834\u003c/span\u003e\u003cspan address=\"10.7916/d8-gzxf-s834\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChamberlain HR, Darin E, Adewole WA, Jochem WC, Lazar AN, Tatem AJ. Building footprint data for countries in Africa: To what extent are existing data products comparable? Comput Environ Urban Syst. 2024;110C:102104. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compenvurbsys.2024.102104\u003c/span\u003e\u003cspan address=\"10.1016/j.compenvurbsys.2024.102104\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChaney S, Mechael P, Thu N, Diallo M, Gachen C. Every Child on the Map: A Theory of Change Framework for Improving Childhood Immunization Coverage and Equity Using Geospatial Data and Technologies. J Med Internet Res. 2021;23(8):e29759. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2196/29759\u003c/span\u003e\u003cspan address=\"10.2196/29759\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCIESIN and Novel-T. GRID3 Zambia Settlement Extents, Version 2.0, Center for International Earth Science Information Network (CIESIN)., Columbia University and Novel-T. Palisades, NY: Geo-Referenced Infrastructure and Demographic Data for Development (GRID3). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7916/wqmn-f746\u003c/span\u003e\u003cspan address=\"10.7916/wqmn-f746\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClements AC, Reid HL, Kelly GC, Hay SI. Further shrinking the malaria map: how can geospatial science help to achieve malaria elimination? Lancet Infect Dis. 2013;13(8):709\u0026ndash;18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S1473-3099(13)70140-3\u003c/span\u003e\u003cspan address=\"10.1016/S1473-3099(13)70140-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDarin E, Dicko AH, Galal H, Jimenez RM, Park H, Tatem AJ, Qader S. (2024). Mapping refugee populations at high resolution by unlocking humanitarian administrative data. Journal of International Humanitarian Action, 9(1), 14, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41018-024-00157-6\u003c/span\u003e\u003cspan address=\"10.1186/s41018-024-00157-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDijkstra L, Florczyk AJ, Freire S, Kemper T, Melchiorri M, Pesaresi M, Schiavina M. Applying the degree of urbanisation to the globe: A new harmonised definition reveals a different picture of global urbanisation. J Urban Econ. 2021;125:103312. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jue.2020.103312\u003c/span\u003e\u003cspan address=\"10.1016/j.jue.2020.103312\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDotse-Gborgbortsi W, Nilsen K, Ofosu A, et al. Distance is a big problem: a geographic analysis of reported and modelled proximity to maternal health services in Ghana. BMC Pregnancy Childbirth. 2022;22:672. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12884-022-04998-0\u003c/span\u003e\u003cspan address=\"10.1186/s12884-022-04998-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDoxsey-Whitfield E, MacManus K, Adamo SB, Pistolesi L, Squires J, Borkovska O, Baptista SR. Taking advantage of the improved availability of census data: a first look at the gridded population of the world, version 4. Papers Appl Geogr. 2015;1(3):226\u0026ndash;34. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/23754931.2015.1014272\u003c/span\u003e\u003cspan address=\"10.1080/23754931.2015.1014272\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDummer TJ. Health geography: supporting public health policy and planning. CMAJ. 2008;178(9):1177\u0026ndash;80. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1503/cmaj.071783\u003c/span\u003e\u003cspan address=\"10.1503/cmaj.071783\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEsch T, Brzoska E, Dech S, Leutner B, Palacios-Lopez D, Metz-Marconcini A, Zeidler J. World Settlement Footprint 3D-A first three-dimensional survey of the global building stock. Remote Sens Environ. 2022;270:112877. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2021.112877\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2021.112877\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEsch T, Brzoska E, Dech S, Leutner B, Palacios-Lopez D, Metz-Marconcini A, Zeidler J. World Settlement Footprint 3D-A first three-dimensional survey of the global building stock. Remote Sens Environ. 2022;270:112877. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2021.112877\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2021.112877\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrantz D, Schug F, Okujeni A, Navacchi C, Wagner W, van der Linden S, Hostert P. National-scale mapping of building height using Sentinel-1 and Sentinel-2 time series. Remote Sens Environ. 2021;252:112128. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2020.112128\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2020.112128\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFreire S, MacManus K, Pesaresi M, Doxsey-Whitfield E, Mills J. (2016). Development of new open and free multi-temporal global population grids at 250 m resolution, Geospatial Data in a Changing World; Association of Geographic Information Laboratories in Europe (AGILE) (Organiser). AGILE; 2016. JRC100523.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGething, P. W., Patil, A. P., Smith, D. L., Guerra, C. A., Elyazar, I. R., Johnston,G. L., \u0026hellip; Hay, S. I. (2011). A new world malaria map: Plasmodium falciparum endemicity in 2010. Malaria journal, 10, 1\u0026ndash;16. https://doi.org/10.1186/1475-2875-10-378.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGevaert CM, Buunk T., \u0026amp; Van Den Homberg, M. J. (2024). Auditing geospatial datasets for biases: using global building datasets for disaster risk management. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/JSTARS.2024.3422503\u003c/span\u003e\u003cspan address=\"10.1109/JSTARS.2024.3422503\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHay, S. I., Sinka, M. E., Okara, R. M., Kabaria, C. W., Mbithi, P. M., Tago, C. C.,\u0026hellip; Godfray, H. C. J. (2010). Developing global maps of the dominant Anopheles vectors of human malaria. PLoS medicine, 7(2), e1000209. https://doi.org/10.1371/journal.pmed.1000209.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHengel, B., Causer, L., Matthews, S., Smith, K., Andrewartha, K., Badman, S., \u0026hellip; Guy,R. (2021). A decentralised point-of-care testing model to address inequities in the COVID-19 response. The Lancet Infectious Diseases, 21(7), e183-e190, https://doi.org/10.1016/S1473-3099(20)30859-8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHierink F, Oladeji O, Robins A, et al. A geospatial analysis of accessibility and availability to implement the primary healthcare roadmap in Ethiopia. Commun Med. 2023;3:140. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s43856-023-00372-z\u003c/span\u003e\u003cspan address=\"10.1038/s43856-023-00372-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJochem WC, Leasure DR, Pannell O, Chamberlain HR, Jones P, Tatem AJ. Classifying settlement types from multi-scale spatial patterns of building footprints. Environ Plann B: Urban Analytics City Sci. 2021;48(5):1161\u0026ndash;79. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/2399808320921208\u003c/span\u003e\u003cspan address=\"10.1177/2399808320921208\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJoseph NK, Macharia PM, Ouma PO, et al. Spatial access inequities and childhood immunisation uptake in Kenya. BMC Public Health. 2020;20:1407. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12889-020-09486-8\u003c/span\u003e\u003cspan address=\"10.1186/s12889-020-09486-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKamanga A, et al. Open-source satellite enumeration to map households: planning and targeting indoor residual spraying for malaria. Malar J. 2015;14:345. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-015-0831-z\u003c/span\u003e\u003cspan address=\"10.1186/s12936-015-0831-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarra K, Kontgis C, Statman-Weil Z, Mazzariello JC, Mathis M, Brumby SP. (2021, July). Global land use/land cover with Sentinel 2 and deep learning. In 2021 IEEE international geoscience and remote sensing symposium IGARSS (pp. 4704\u0026ndash;4707). IEEE.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeating J, et al. Retrospective evaluation of the effectiveness of indoor residual spray with pirimiphos-methyl (Actellic) on malaria transmission in Zambia. Malar J. 2021;20:173. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-021-03710-5\u003c/span\u003e\u003cspan address=\"10.1186/s12936-021-03710-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKebba Touray P, Mkanda SG, Tegegn P, Nsubuga TB, Erbeto R, Banda A, Etsano F, Shuaib RG, Vaz S. 67\u0026ndash;S72. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/infdis/jiv493\u003c/span\u003e\u003cspan address=\"10.1093/infdis/jiv493\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKyomuhangi I, et al. Assessing national vector control micro-planning in Zambia using the 2021 malaria indicator survey. Malar J. 2023;22:365. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-023-04807-9\u003c/span\u003e\u003cspan address=\"10.1186/s12936-023-04807-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeith\u0026auml;user N, Schneider J, Johann S, et al. Quantifying Covid19-vaccine location strategies for Germany. BMC Health Serv Res. 2021;21:780. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12913-021-06587-x\u003c/span\u003e\u003cspan address=\"10.1186/s12913-021-06587-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi J, Huang X, Tu L, Zhang T, Wang L. A review of building detection from very high resolution optical remote sensing images. GIScience Remote Sens. 2022;59(1):1199\u0026ndash;225. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/15481603.2022.2101727\u003c/span\u003e\u003cspan address=\"10.1080/15481603.2022.2101727\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLloyd, et al. Using GIS and Machine Learning to Classify Residential Status of Urban Buildings in Low and Middle Income Settings. Remote Sens. 2020;12(23):3847. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/rs12233847\u003c/span\u003e\u003cspan address=\"10.3390/rs12233847\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLubinda J, et al. Spatio-temporal monitoring of health facility-level malaria trends in Zambia and adaptive scaling for operational intervention. Commun Med. 2022;2:79. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s43856-022-00144-1\u003c/span\u003e\u003cspan address=\"10.1038/s43856-022-00144-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacharia PM, Ouma PO, Gogo EG, Snow RW, Noor AM. Spatial accessibility to basic public health services in South Sudan. Geospat Health. 2017;12(1):510. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4081/gh.2017.510\u003c/span\u003e\u003cspan address=\"10.4081/gh.2017.510\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMartin A, Bwalya F, Pollard D et al. Use of a geo-enabled digital global good for microplanning and delivery of indoor residual spray (IRS) in Zambia: a case study, 2016\u0026ndash;2020. (2024 - in review).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMendes, A., Palmer, T., Berens, A., Espey, J., Price, R., Mallya, A., \u0026hellip; Kaplan, B.(2021). Mapathons versus automated feature extraction: a comparative analysis for strengthening immunization microplanning. International Journal of Health Geographics,20(1), 27. https://doi.org/10.1186/s12942-021-00277-x.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMetzger N, Daudt RC, Tuia D, Schindler K. High-resolution population maps derived from sentinel-1 and sentinel-2. Remote Sens Environ. 2024;314:114383. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2024.114383\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2024.114383\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMicrosoft. 2022. Worldwide building footprints derived from satellite imagery (GitHub Repository). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/microsoft/GlobalMLBuildingFootprints [DATASET]\u003c/span\u003e\u003cspan address=\"https://github.com/microsoft/GlobalMLBuildingFootprints [DATASET]\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (accessed 10th January 2023, and 1st October 2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMuchiri, S. K., Muthee, R., Kiarie, H., Sitienei, J., Agweyu, A., Atkinson, P. M.,\u0026hellip; Alegana, V. A. (2022). Unmet need for COVID-19 vaccination coverage in Kenya. Vaccine,40(13), 2011\u0026ndash;2019. https://doi.org/10.1016/j.vaccine.2022.02.035.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNational Malaria Elimination Centre (NMEC). National malaria elimination strategic plan 2017\u0026ndash;2021. Zambia: Lusaka; 2017. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://static1.squarespace.com/static/58d002f017 bffcf99fe21889/t/5b28d7f1575d1ff0942dbce1/1529403401067/National +Malaria+Elimination+Strategic+Plan+2017-Final_PRINT.pdf\u003c/span\u003e\u003cspan address=\"https://static1.squarespace.com/static/58d002f017bffcf99fe21889/t/5b28d7f1575d1ff0942dbce1/1529403401067/National+Malaria+Elimination+Strategic+Plan+2017-Final_PRINT.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNational Malaria Elimination Centre (NMEC). National malaria elimination strategic plan 2022\u0026ndash;2026. Zambia: Lusaka. 2022, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://static1.squarespace.com/static/58d002f017b ffcf99fe21889/t/632a4cb0fcd87c13d0165372/1663716530614/Z NMESP+2022+to+2026_SIGNED+120722.pdf\u003c/span\u003e\u003cspan address=\"https://static1.squarespace.com/static/58d002f017bffcf99fe21889/t/632a4cb0fcd87c13d0165372/1663716530614/ZNMESP+2022+to+2026_SIGNED+120722.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOu\u0026eacute;draogo AL, Zhang J, Tinto H, et al. A microplanning model to improve door-to-door health service delivery: the case of Seasonal Malaria Chemoprevention in Sub-Saharan African villages. BMC Health Serv Res. 2020;20:1128. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12913-020-05972-2\u003c/span\u003e\u003cspan address=\"10.1186/s12913-020-05972-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePalacios-Lopez, D., Esch, T., MacManus, K., Marconcini, M., Sorichetta, A., Yetman,G., \u0026hellip; Reinartz, P. (2022). Towards an improved large-scale gridded population dataset:A Pan-European study on the integration of 3D settlement data into population modelling.Remote Sensing, 14(2), 325. https://doi.org/10.3390/rs14020325.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePesaresi M, Politis P, GHS-BUILT-S. R2023A - GHS built-up surface grid, derived from Sentinel2 composite and Landsat, multitemporal (1975\u0026ndash;2030)European Commission, Joint Research Centre (JRC), \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2905/9F06F36F-4B11-47EC-ABB0-4F8B7B1D72EA\u003c/span\u003e\u003cspan address=\"10.2905/9F06F36F-4B11-47EC-ABB0-4F8B7B1D72EA\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePinchoff J, et al. Targeting indoor residual spraying for malaria using epidemiological data: a case study of the Zambia experience. Malar J. 2016;15(11). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12936-015-1073-9\u003c/span\u003e\u003cspan address=\"10.1186/s12936-015-1073-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrice R, Hallas M. (2019, December). Mapping every building and road in sub-Saharan Africa. In AGU Fall Meeting Abstracts (Vol. 2019, pp. IN41A-02).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRader B, Astley CM, Sewalk K, et al. Spatial modeling of vaccine deserts as barriers to controlling SARS-CoV-2. Commun Med. 2022;2:141. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s43856-022-00183-8\u003c/span\u003e\u003cspan address=\"10.1038/s43856-022-00183-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRehman AM, Coleman M, Schwabe C, Baltazar G, Matias A, Roncon Gomes I, et al. How Much Does Malaria Vector Control Quality Matter: The Epidemiological Impact of Holed Nets and Inadequate Indoor Residual Spraying. PLoS ONE. 2011;6(4):e19205. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0019205\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0019205\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReid, H., Haque, U., Clements, A. C., Tatem, A. J., Vallely, A., Ahmed, S. M., \u0026hellip; Haque,R. (2010). Mapping malaria risk in Bangladesh using Bayesian geostatistical models.The American journal of tropical medicine and hygiene, 83(4), 861. https://doi.org/10.4269/ajtmh.2010.10-0154.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSirko, W., Brempong, E. A., Marcos, J. T., Annkah, A., Korme, A., Hassen, M. A., \u0026hellip;Quinn, J. (2023). High-Resolution Building and Road Detection from Sentinel-2. arXiv preprint arXiv:2310.11622.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSirko, W., Kashubin, S., Ritter, M., Annkah, A., Bouchareb, Y. S. E., Dauphin, Y.,\u0026hellip; Quinn, J. (2021). Continental-scale building detection from high resolution satellite imagery. arXiv preprint, https://doi.org/10.48550/arXiv.2107.12283.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSnow J. Cholera and the water supply in the south districts of London in 1854. J Public Health Sanit Rev. 1856;2(7):239.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSorichetta A, Hornby GM, Stevens FR, Gaughan AE, Linard C, Tatem AJ. High-resolution gridded population datasets for Latin America and the Caribbean in 2010, 2015, and 2020. Sci data. 2015;2(1):1\u0026ndash;12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/sdata.2015.45\u003c/span\u003e\u003cspan address=\"10.1038/sdata.2015.45\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSturrock, et al. PLoS ONE. 2018;13(9):e0204399. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0204399\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0204399\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Predicting residential structures from open source remotely enumerated data using machine learning.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTatem AJ. Small area population denominators for improved disease surveillance and response. Epidemics. 2022;41:100641. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.epidem.2022.100641\u003c/span\u003e\u003cspan address=\"10.1016/j.epidem.2022.100641\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTiecke, T. G., Liu, X., Zhang, A., Gros, A., Li, N., Yetman, G., \u0026hellip; Dang, H. A. (2017).Mapping the world population one building at a time. arXiv preprint. https://doi.org/10.48550/arXiv.1712.05839.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUtazi CE, Olowe I, Chan THM, Dotse-Gborgortsi W, Wagai J, Umar JA, Etamesor S, Atuhaire B, Fafunmi B, Crawford J, Adeniran A, Tatem AJ. Geospatial Variation in Vaccination Coverage and Zero-Dose Prevalence at the District, Ward and Health Facility Levels Before and After a Measles Vaccination Campaign in Nigeria. Preprints 2024, 2024100445. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.20944/preprints202410.0445.v1\u003c/span\u003e\u003cspan address=\"10.20944/preprints202410.0445.v1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUtazi CE, Thorley J, Alegana VA, et al. Mapping vaccination coverage to explore the effects of delivery mechanisms and inform vaccination strategies. Nat Commun. 2019;10:1633. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-019-09611-1\u003c/span\u003e\u003cspan address=\"10.1038/s41467-019-09611-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhitehead J, Scott N, Atatoa-Carr P, Lawrenson R. Will access to COVID-19 vaccine in Aotearoa be equitable for priority populations? N Z Med J. 2021;134(1535):25\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWHO. (2009). Microplanning for immunization service delivery using the Reaching Every District (RED) strategy; 2009. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.who.int/publications/i/item/microplanning-for-immunization-service-delivery-using-the-reaching-every-district-(-red)-strategy\u003c/span\u003e\u003cspan address=\"https://www.who.int/publications/i/item/microplanning-for-immunization-service-delivery-using-the-reaching-every-district-(-red)-strategy\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWHO. (2018). Global Polio Eradication Initiative: Best Practices in Microplanning for Polio Eradication; 2018. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://polioeradication.org/wp-content/uploads/2018/12/Best-practices-in-mircoplanning-for-polio-eradication.pdf\u003c/span\u003e\u003cspan address=\"http://polioeradication.org/wp-content/uploads/2018/12/Best-practices-in-mircoplanning-for-polio-eradication.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWHO. (2021), Guidance on developing a national deployment and vaccination plan for COVID-19 vaccines: interim guidance, 1 June 2021. Geneva: World Health Organization; 2021 (WHO/2019-nCoV/Vaccine_deployment/2021.1).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWHO. Operational manual on indoor residual spraying: control of vectors of malaria, Aedes-borne diseases, Chagas disease, leishmaniases and lymphatic filariasis. Geneva: World Health Organization; 2023. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://iris.who.int/bitstream/handle/10665/375978/9789240083998-eng.pdf\u003c/span\u003e\u003cspan address=\"https://iris.who.int/bitstream/handle/10665/375978/9789240083998-eng.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWHO-UNICEF COVAX GIS Working Group. Geo-Enabled Microplanning Handbook. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.digitalhealthcoe.org/_files/ugd/55ae33_f30815f09639437586ef8edd4475354f.pdf\u003c/span\u003e\u003cspan address=\"https://www.digitalhealthcoe.org/_files/ugd/55ae33_f30815f09639437586ef8edd4475354f.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWinters A, Riley C, Soobramoney L, Pollard D, Jere E, Bwalya F, Silumbe K, Hamainza B. Cost and cost effectiveness of geospatial planning and delivery tools added to standard health campaigns in Luapula Province, Zambia. Oxf Open Digit Health. 2024;2(Supplement2):ii66\u0026ndash;74. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/oodh/oqae040\u003c/span\u003e\u003cspan address=\"10.1093/oodh/oqae040\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organisation. High burden to high impact: a targeted malaria response. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.who.int/publications/i/item/WHO-CDS-GMP-2018.25\u003c/span\u003e\u003cspan address=\"https://www.who.int/publications/i/item/WHO-CDS-GMP-2018.25\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e World Health Organization, Geneva (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization. (2023). World malaria report 2023. World Health Organization. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.who.int/publications/i/item/9789240086173\u003c/span\u003e\u003cspan address=\"https://www.who.int/publications/i/item/9789240086173\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZanaga D, Van De Kerchove R, Daems D, De Keersmaecker W, Brockmann C, Kirches G, Wevers J, Cartus O, Santoro M, Fritz S, Lesiv M, Herold M, Tsendbazar N-E, Xu P, Ramoino F, Arino O. (2022). ESA WorldCover 10 m 2021 v200 (Version v200) [Data set]. Zenodo. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5281/zenodo.7254221\u003c/span\u003e\u003cspan address=\"10.5281/zenodo.7254221\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-health-geographics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijhg","sideBox":"Learn more about [International Journal of Health Geographics](http://ij-healthgeographics.biomedcentral.com/)","snPcode":"12942","submissionUrl":"https://submission.nature.com/new-submission/12942/3","title":"International Journal of Health Geographics","twitterHandle":"@IJHGeo","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Building footprints, Malaria, Microplanning, Geospatial, Satellite imagery","lastPublishedDoi":"10.21203/rs.3.rs-5898324/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5898324/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eThe increasing availability globally of building footprint datasets has brought new opportunities to support a geographic approach to health programme planning. This is particularly acute in settings with high disease burdens but limited geospatial data available to support targeted planning. The comparability of building footprint datasets has recently started to be explored, but the impact of utilising a particular dataset in analyses to support decision making for health programme planning has not been studied. Here, we quantify the impact of utilising four different building footprint datasets in analyses to support health programme planning, with an example of malaria vector control initiatives in Zambia.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eUsing the example of planning indoor residual spraying (IRS) campaigns in Zambia, we identify priority locations for deployment of this intervention based on criteria related to the area, proximity and counts of building footprints per settlement. We apply the same criteria to four different building footprint datasets and quantify the count and geographic variability in the priority settlements that are identified.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eWe show that nationally the count of potential priority settlements for IRS varies by over 230% with different building footprint datasets, considering a minimum threshold of 25 sprayable buildings per settlement. Differences are most pronounced for rural settlements, indicating that the choice of dataset may bias the selection to include or exclude settlements, and consequently population groups, in some areas.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe results of this study show that the choice of building footprint dataset can have a considerable impact on the potential settlements identified for IRS, in terms of (i) their location and count, and (ii) the count of building footprints within priority settlements. The choice of dataset potentially has substantial implications for campaign planning, implementation and coverage assessment. Given the magnitude of the differences observed, further work should more broadly assess the sensitivity of health programme planning metrics to different building footprint datasets, and across a range of geographic contexts and health campaign types.\u003c/p\u003e","manuscriptTitle":"Assessing the impact of building footprint dataset choice for health programme planning: a case study of indoor residual spraying (IRS) in Zambia","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-19 16:06:23","doi":"10.21203/rs.3.rs-5898324/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-03-08T07:28:11+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-02-27T15:28:43+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-02-24T13:27:12+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"70687134928374957408567698330565137667","date":"2025-02-04T20:13:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"218961256102512925309762616816259479577","date":"2025-02-03T09:50:18+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-02-02T05:35:24+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-01-27T00:21:38+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-01-27T00:20:08+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal of Health Geographics","date":"2025-01-24T21:29:48+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-health-geographics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijhg","sideBox":"Learn more about [International Journal of Health Geographics](http://ij-healthgeographics.biomedcentral.com/)","snPcode":"12942","submissionUrl":"https://submission.nature.com/new-submission/12942/3","title":"International Journal of Health Geographics","twitterHandle":"@IJHGeo","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"930c3137-6223-4624-aac8-305ba290cc2a","owner":[],"postedDate":"February 19th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-05-26T16:03:20+00:00","versionOfRecord":{"articleIdentity":"rs-5898324","link":"https://doi.org/10.1186/s12942-025-00398-7","journal":{"identity":"international-journal-of-health-geographics","isVorOnly":false,"title":"International Journal of Health Geographics"},"publishedOn":"2025-05-24 15:58:10","publishedOnDateReadable":"May 24th, 2025"},"versionCreatedAt":"2025-02-19 16:06:23","video":"","vorDoi":"10.1186/s12942-025-00398-7","vorDoiUrl":"https://doi.org/10.1186/s12942-025-00398-7","workflowStages":[]},"version":"v1","identity":"rs-5898324","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5898324","identity":"rs-5898324","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}