1 Background
Vector-borne diseases (VBD s), such as malaria and dengue, are well-established as highly sensitive to environmental conditions. Local land use practices and meteorological factors directly determine the spatial and temporal distributions of mosquito larval aquatic habitats (Fornace et al. 2021, Patz et al. 2004, Thomson et al. 2006, Vittor et al. 2009). Wider climatic changes and disruptions to existing ecosystems can alter the abundance and behaviour of mosquito species and lead to expansion of these vectors into new locations (Caminade et al. 2019, Iwamura et al. 2020, Jones et al. 2008, Metcalf et al. 2017). These ecological changes interact with socioeconomic systems in complex and context-specific ways to impact VBD transmission risks (Baeza et al. 2017, de Castro et al. 2006, Keiser et al. 2005).
With unprecedented levels of global environmental change, it is increasingly critical to monitor changing VBD dynamics. Increasing quality and availability of Earth observation (EO) data provides new opportunities to characterise the impacts of environmental change on VBD transmission and develop targeted surveillance methods. EO refers to the collection of physical and biological data on planetary surfaces obtained using remote sensing methods, typically sensors based on satellites or aircraft. These data can be used to monitor factors such as land cover, vegetation levels, water body distribution and extents of human settlements (Finer et al. 2018, Lloyd et al. 2017, Pekel et al. 2016, Wardrop et al. 2018, Wimberly et al. 2021). Since the launch of the satellite Landsat-1 in 1972, the quantity and quality of satellite-based EO data has grown exponentially (Finer et al. 2018). Simultaneously, aerial technologies, such as low-cost drones, are increasingly accessible and are now widely used to collect EO data at user-defined locations (Koh and Wich 2012). This expansion of EO data has been accompanied by technological developments in computing and imagery analysis methods enabling processing and interpretation of these data sources.
The potential utility of EO data for VBD epidemiological studies has been widely recognised since the 1970s (Cline 1970). Motivated by increasing mosquito resistance to insecticides and bans on the widespread use of chemicals such as DDT and outdoor disease transmission, early uses of EO data focused on characterising aquatic mosquito larval habitats to better target control measures. In 1971, the National Aeronautics and Space Administration (NASA) conducted targeted aerial surveys in New Orleans, demonstrating the capacity of multispectral imaging using the visible to near-infrared region of the electromagnetic spectrum to identify mosquito breeding sites (NASA 1973). These techniques were later deployed operationally within other areas of the United States to target the breeding sites of mosquito vectors during an outbreak of St. Louis encephalitis virus (Wagner et al. 1979). Subsequent studies illustrated how different types of aerial and space-borne EO data could be used to locate aquatic habitats of mosquito larvae and characterise wider types of land cover associated with abundance of adult mosquitoes and malaria transmission (Bergquist et al. 2021, Hay 2000, Hay et al. 1998).
Since these early uses of EO data to characterise vector habitats at local scales, there has been a rapid increase in the use of spatial and environmental data for VBD studies. Initiatives such as the Malaria Atlas Project use geo-referenced data and covariates (e.g. precipitation, temperature, vegetation) derived from planetary-level EO data to develop high resolution maps of global vector and VBD distributions (e.g. (Bhatt et al. 2013, Sinka et al. 2016, Weiss et al. 2019)). EO data is also increasingly incorporated into VBD early warning systems, with metrics derived from EO data used to inform predictions of disease outbreaks and trigger VBD control measures (e.g. (Lowe et al. 2013, Thomson et al. 2006)). These data are also used to plan and inform surveillance activities, such as mapping health facility catchment areas, defining survey populations or identifying high priority areas for vector monitoring (e.g. (Fornace et al. 2018, Franke et al. 2015, Noor et al. 2009, Stresman et al. 2014)). At much finer scales, very high resolution EO data has been used to monitor changes in land cover, mosquito breeding sites and human and wildlife disease reservoir distribution to understand the mechanisms driving infectious disease emergence (e.g. (Carrasco-Escobar et al. 2019, Fornace et al. 2014, Minakawa et al. 2005)).
However, despite these advances, only a relatively small proportion of available EO data is actively used for VBD surveillance. There remain significant barriers to integrating EO data into health surveillance systems in many high VBD burden countries; these challenges often include limited access to equipment, computing infrastructure or specialised technical remote sensing expertise within disease control departments. Additionally, with the wide range of types of EO data and analysis methods now available, it can be challenging to identify the best data for specific use cases. Within this chapter, we review available EO data sources, characteristics and analysis methods. We then examine existing applications of EO data to study VBD epidemiology, present examples of how EO data can be leveraged to improve VBD surveillance in changing environments and discuss considerations for the use of EO data, highlighting the need for knowledge of local vector ecology to inform EO data selection and applications.
2 Sources and characteristics of Earth observation data
EO involves the use of sensors to collect data used to measure characteristics of planetary surfaces from reflected or emitted radiation from a distance. Although EO can be conducted using aerial and ground-based platforms, spaceborne satellites are by far the largest source of EO data. Since the first launch of the EO satellite Sputnik 1 by the former Soviet Union in 1957, over a thousand EO satellites have been launched (Figure 1), (Tatem et al. 2008, Union of Concerned Scientists 2021). These data are available through a range of public and private sources, such as the European Space Agency, NASA and commercial companies such as Maxar Technologies (https://www.maxar.com/) and Planet (https://www.planet.com/). Costs for these data range from freely available to thousands of dollars for specifically tasking commercial satellites to cover target areas. While most free EO data is available through government or intergovernmental agencies, corporate partnerships and outreach programmes have vastly increased access to high resolution EO data. For example, a partnership between Planet and Norway’s International Climate and Forests Initiative provides free access to high resolution satellite data for tropical areas for non-profit uses (https://www.planet.com/nicfi/). These data can be further supplemented by local EO data collection using aerial platforms such as drones and other aircraft (e.g. (Carrasco-Escobar et al. 2019, Fornace et al. 2014, Hardy et al. 2017, Wagner et al. 1979)).
Earth Observation satellites launched by year (data obtained from Union of Concerned Scientists 2021)
The characteristics of the EO data collected are determined by the specific type of sensor used for data collection. Passive sensors measure light reflected by the sun while active sensors have their own source of electromagnetic energy and measure the backscatter reflected from this source. While passive sensors include instruments such as cameras, active systems use technologies such as radar and lidar. These systems are characterised by radiometric resolution based on the sensitivity of sensors to measure these differences in reflected or emitted energy. The region of the electromagnetic spectrum covered and the number of bands or channels used to sample this region of the spectrum determines the spectral resolution of a system. This spectral resolution represents the degree of precision to which reflected energy from the Earth’s surface is being sampled. For example, low spectral resolution systems (e.g. many cameras carried by low-cost drones) sample the visible part of the spectrum in three bands: blue, green and red. In contrast, more sophisticated systems sample the spectrum from the visible to the near infrared and shortwave infrared over a number of bands (e.g. Landsat TM system with six bands, MODIS with 32 bands) which enables spectral detection of more complex features, such as different habitat types or calculation of vegetation indices (Wimberly et al. 2021).
All EO data is additionally characterised by temporal resolution, the frequency at which data are collected, and spatial resolution, the pixel size of data collected. With satellite-based EO data, the temporal resolution is determined by the satellite orbit or specific tasking requests determining the frequency at which a satellite passes over a specific location and collects data. Temporal resolution may be additionally determined by the availability of usable data; critically, for optical satellite-based systems, cloud cover is a major limitation for obtaining usable data over areas within the tropics (Burke et al. 2021). These limitations have led to a rapid expansion of other types of platforms used to collect EO data underneath the clouds, such as drones (also known as unmanned aerial vehicles or UAV s). These technologies allow targeted collection of EO data at user defined time intervals. Drones additionally allow collection of data with a much higher spatial resolution than most satellites (e.g. <10 cm/pixel for standard drone data vs >10 m/pixel for most freely available satellite data). These resolutions, as well as the cost and feasibility of using different platforms, impact the suitability of different types of EO data for VBD applications (Figure 2). Due to the difficulties in obtaining cloud-free satellite imagery, increasingly, studies look towards the use of radar imagery resulting from active satellite-based systems operate within the microwave part of the spectrum that is not inhibited by cloud cover. Radar imagery can provide valuable information for vector disease applications, such as wetland mapping, land cover mapping and forest cover. Some sources of radar imagery are free and are routinely collected (e.g. Sentinel-1 with a 12-day return period for the tropics) adding to its value within operational monitoring programs. But, like optical satellite imagery, its spatial resolution is currently limited to 10 m which may be inadequate for many vector disease applications.
Example of sources, spatial resolutions and spectral capacity of Earth Observation data in April 2021 in Sabah, Malaysia (A) PlanetScope Tropical Visual Archive; (B) Landsat 8 (Band 1 = Blue, Band 2 = near infrared (NIR), Band 3 = short wave infrared (SWIR); (C) Elevation (SRTM 1 arc-second); (D) Classified forest cover (0, 1) (Hansen et al. 2013); (E) Sentinal-1 radar image (Band 1 = VV, Band 2 = VH, Band 3 = VV/VH composite); (F) Normalised Difference Vegetation Index (NVDI), derived from Landsat 8; (G) Land cover thematic map (WorldCover)
3 Analysing Earth observation data
After acquiring EO data, these data are typically analysed to identify features of interest. For example, these may include specific land classes associated with vector habitats, types of water bodies where vectors breed or locations of houses comprising the target population at risk. EO-derived metrics are often used as covariates in models to identify risk factors or predict risks of vector occurrence or disease risks (Weiss et al. 2015). On more operational levels, EO data may inform planning of VBD surveillance and control measures through identification of high-risk areas or target populations.
The intended end use of the data, the feature(s) of interest and available resources determine the analysis approach used for EO data. Prior to analysis, a number of pre-processing steps are typically conducted to improve the data quality before analysis. This can include radiometric calibration to transform the radiance measured by the sensor to reflectance, the proportion of solar radiation reflected back to the sensor, to adjust for atmospheric effects. Additional pre-processing steps may include masking clouds or unreliable values (i.e. due to a fault in the sensor) and excluding these from further analysis. Orthorectification methods may also be used to adjust for distortions from the terrain and satellite or platform motion.
After pre-processing steps have been completed, the EO data can be analysed using a range of different methods. One of the most frequently used methods is the calculation of spectral indices as a ratio between two or more bands in optical data. Commonly used indices include the normalised vegetation index (NDVI), enhanced vegetation index (EVI) and soil adjusted vegetation index (SAVI), measures of greenness used to assess vegetation (Weiss et al. 2015, Wimberly et al. 2021). Different band combinations may also be used to create natural colour and false colour composites to visualise EO imagery and aid manual identification of key features of interest.
EO data are also frequently classified into thematic maps to identify key features of interest, such as water bodies or land cover types. These classification approaches can use a wide range of methods spanning simple visual identification to advanced machine learning and artificial intelligence methods. Visualising and manually digitising EO data remains one of the simplest, and often the most accurate and quickest, methods of identifying features in complex images. This requires informed users to detect and label features or classes of interest. Although this can be labour-intensive, increasingly citizen science approaches have been used to engage volunteers in labelling large numbers of images using online platforms such as Zooniverse (Simpson et al. 2014). Additionally, digitally assisted technological approaches, such as region growing, allow users to identify features and rapidly label similar surrounding pixels more efficiently (Hardy et al. 2022a).
Alternatively, model-based methods can be used to automatically identify features in EO data. These model-based methods can be split into two categories: supervised learning, in which labelled data are used to fit a model of predetermined features of interest, and unsupervised learning, which analyses unlabelled datasets to group features into different categories (Mather and Koch 2010). Supervised learning approaches require labelled training data of the specific classes to be identified; these can include ground-truthed data (such as from field surveys of vector breeding sites) or labelled examples of features of interest. While unsupervised approaches do not require training data, results may be more challenging to interpret and require an operator to match an unsupervised class to an information class of interest.
Traditionally, pixel-based classification is one of the most commonly utilised methods of image classification. This method assigns individual pixels into different class types, typically on the basis of spectral information available for individual pixels. In contrast, object-oriented classification methods classify groups of similar pixels using both pixel-level values and the spatial relationships of pixels to each other (Liu and Fan 2010). Images are typically segmented into groups of similar pixels which are subsequently assigned to specific objects or classes. More recently, deep learning approaches, such as convolutional neural networks (CNN), have revolutionised image analysis. These methods allow efficient analysis of image textures, patterns and spectral characteristics by using artificial intelligence approaches to identify features in complex environments (Cheng et al. 2020). Although computationally intensive, more efficient deep-learning architecture and cloud-based computing are increasingly accessible and applied for ecological analysis of EO data, e.g. (Bravo et al. 2021, Gray et al. 2019, Kattenborn et al. 2019). While these approaches can capture complex patterns in the data to improve classification output, these typically require very large training datasets.
4 Existing use of Earth observation-derived data for vector-borne disease surveillance: a scoping review
These different data sources and analysis methods can be used to generate a wide range of environmental covariates to integrate into VBD epidemiological studies. To evaluate the most commonly used EO-derived metrics and data sources, a brief scoping review was conducted using Medline (through PubMed) and Scopus since 2012. We sought studies which used satellite data to describe, analyse or forecast human vector-borne diseases. We restricted the search strategy on the last 10 years but did not restrict the studies to population ages or countries.
Based on identification of relevant VBD, the following search strategy was used: (‘Chikungunya’ OR ‘Lyme Disease’ OR ‘Plague’ OR ‘Relapsing Fever’ OR ‘Rocky Mountain Spotted Fever’ OR ‘Tularemia’ OR ‘Typhus’ OR ‘West Nile Virus’ OR ‘Zika’ OR ‘trypanosomiasis’ OR ‘schistosomiasis’ OR ‘leishmaniasis’ OR ‘dengue’ OR ‘malaria’ OR ‘vector-borne disease’ OR ‘mosquito-borne disease’ OR ‘WNV’ OR ‘chikungunya’ OR ‘filariasis’ OR ‘yellow fever’ OR ‘Japanese encephalitis’ OR ‘onchocerciasis’ OR ‘tungiasis’ OR ‘crimean-congo haemorrhagic fever’ OR ‘lyme disease’ OR ‘spotted fever’ OR ‘Q fever’ OR ‘chagas’) AND (‘Earth observation’ OR ‘Remote sensing’ OR ‘Satellite data’). We used EndNote20 and Rayyan to remove duplicates from the search results. Data on the disease focus, EO data source and metrics used were extracted by four reviewers; discrepancies were solved by consensus. A qualitative synthesis was conducted whereby the characteristics of the variables which were used by the satellites.
Use of sources of Earth observation data by VBD (results from scoping review)
Within this review, there was an initial total of 963 studies identified; out of these, 714 studies were excluded due to duplicates, out of scope or not specifying the vector-borne disease they were studying. From the 249 studies which were included, we extracted the name of the EO source data, type of data used, and the vector-borne disease which were studied. Figure 3 summarizes the number of EO data sources used for each vector-borne disease´s studies. The most popular EO data sources are the Moderate-Resolution Imaging Spectroradiometer (MODIS) and the Landsat satellite, possibly due to the open data repositories and long-term existence, respectively. In terms of mosquito-borne diseases, the use of these EO data sources is predominantly reported for malaria, dengue, leishmaniasis, and West Nile virus. However, a substantial number of studies using MODIS, Landsat, and the Shuttle Radar Topography Mission (SRTM) sources assessed infectious diseases transmitted by other arthropod and non-arthropod vectors such as freshwater snails (schistosomiasis), fleas (bubonic plague), triatomine bugs, and flies (trypanosomiasis).
Earth observation-derived variables used per vector-borne disease
The distribution and epidemiology of all vectors transmitting these diseases depend on their interactions with the environment, with varying characteristics important for the ecology of different vectors. Table 1 describes the most common EO-derived variables used for each vector-borne disease. Normalized Difference Vegetation Index (NDVI), precipitation and land surface temperature are the most common types of data used to study the distribution of these diseases, especially for mosquito-borne diseases. In contrast, other variables are more frequently used for non-mosquito-borne diseases; for example, hydrological variables are most frequently reported for studies on schistosomiasis due to the habitats of the snail vector. It is important to take into consideration that the use of a specific type of data will depend on the type of vector, reservoirs, transmission mechanism, and the distribution of each disease.
5 Applications of Earth observation-derived data for VBD surveillance and control
These studies use EO-derived data for a wide range of applications, from understanding the basic epidemiology and distributions of the vectors to designing early warning systems and surveillance approaches. EO data can be used to describe habitats of vectors or hosts, identify larval breeding sites or target control measures. Within this section, we describe case studies of specific applications of EO data and integration into VBD surveillance and control measures.
5.1 Spatial distribution of vector species
One of the primary uses of EO-derived data in VBD epidemiology is the development of maps of distribution of vector species. This spatial distribution of particular vector species has often been used as a proxy for risk of disease exposure (Kraemer et al. 2019, Messina et al. 2019, Ryan et al. 2019), contributing to strategic planning and informed management decisions. Ecological niche models (ENM s, also known as Species Distribution Models), which determine environmental conditions that meet a species’ ecological requirements and predict the relative suitability of habitat have been employed in spatial epidemiology to determine environmental conditions associated with disease occurrence (Peterson 2014, Phillips et al. 2006). These models can be also used to predict the range of vectors, usually merging ground-based observations (e.g. from surveillance records) and potential environmental predictors. For example, in Ecuador, the distribution of Ae. aegypti was projected to the year 2050 under a variety of models of climate change, predicting shifts in the mosquito range and reinforcing surveillance on human populations that may be at risk of exposure to diseases vectored by Aedes (Lippi et al. 2019). Species distribution models have been also used to optimise a tsetse fly eradication campaign in Senegal, allowing more efficient control operations for the deployment of insecticide-treated targets and release of sterile males in the fight against human and animal trypanosomiasis (Dicko et al. 2014). Pareyn et al. (2021) identified that suitable habitats for Phlebotomus pedifer, vector of cutaneous leishmaniasis in Ethiopia, are more extensive than thought and that the main environmental drivers of their distribution were mean annual temperature, precipitation seasonality and Enhanced Vegetation Index, among others (Pareyn et al. 2020). In seven predominant Culex mosquito species in the Americas, temperature, humidity, urban/built-up land class, and cultivated and managed vegetation were important environmental drivers structuring the spatial distribution of these species (Gorris et al. 2021).
Example drone imagery in VBD surveillance, detecting (A) goats in Kenya; (B) human settlements in Zanzibar; (C) vector breeding sites in Peru; (D) landscape modification in Malaysia
On finer spatial scales, monitoring changes in landscape can be used to identify changes to vector habitats and design targeted entomological surveillance. A study in Sabah, Malaysia and Palawan in the Philippines demonstrated the use of UAV s to map environmental risk factors for transmission of the zoonotic malaria Plasmodium knowlesi (Fornace et al. 2014). Due to the difficulty obtained cloud-free satellite data, a UAV was used to provide data with higher spatial and temporal resolution. Consolidated and adaptive mapping of an area over a set time period (December 2013 to May 2014) also allows for repeat sampling of areas seen to be undergoing rapid land use change, which generates datasets of land cover that are accurate for the desired timeframes. This ability to adapt and map study areas in real-time and produce maps with high temporal and spatial granularity is an essential tool for understanding the dynamic interactions between environmental factors, vector ecology and local transmission patterns (Figure 4). Resulting maps can then be used to systematically characterise land use types over the study area and create accurate spatial sampling frames for entomological surveys in areas undergoing rapid deforestation (Byrne et al. 2021a).
5.2 Surveillance of invasive mosquito species
The use of EO might be particularly useful to identify pathways for vector introduction from abroad and to support strong surveillance systems and effective interventions (Caminade et al. 2012). Anopheles stephensi, an endemic malaria vector in Asia, has been detected in Africa, particularly in the Horn of Africa – Djibouti (Faulde et al. 2014), Ethiopia (Carter et al. 2018), Somalia (World Health Organisation 2019) – and now spread to several regions in the continent (Ahmed et al. 2021), raising concerns about the impact on malaria transmission. This mosquito species thrives in clean water containers, abundant in urban landscapes. Because most malaria vector control strategies in Africa target rural environments, the prospect of this malaria vector in urban areas is of great concern for the malaria control and elimination programs (World Health Organisation 2019). Due to the confirmed vector competence in An. stephensi Ethiopian and Djibutian populations (Seyfarth et al. 2019), its association to malaria outbreaks and increased reported cases in Djibuti (Faulde et al. 2014), extended mosquito presence in the dry season (Seyfarth et al. 2019) and reduced susceptibility to multiple insecticides classes (Yared et al. 2020), WHO considers the spread of An. stephensi a major potential threat to malaria control and elimination in Africa and Southern Asia (World Health Organisation 2019). As a result, several measures have been recommended to tackle its invasion and spread. One of the core interventions is to conduct active surveillance for An. stephensi aquatic stages in urban and peri-urban areas – in addition to routine surveillance in rural environments. EO data could be extremely advantageous for mapping and monitoring typical breeding sites – water storage containers outside the home, rainwater collections, roof tops, wells or large human-made cisterns – to identify high risk urban areas in which An. stephensi will have the best conditions to establish.
The potential for An. stephensi invasion across sub-Saharan Africa has been predicted combining occurrence data from 1985–2016 with satellite imagery data sets (Sinka et al. 2020). The habitat suitability modelling study, which includes several environmental variables extracted from MODIS (annual mean temperature, Index and Tasselled Cap Wetness, Enhanced Vegetation and land cover classification) ((Deblauwe et al. 2016), CHIRPS (seasonal precipitation)), GMIA (Irrigation – Global Map of irrigated areas FAO, United Nations, and International Geosphere and Biosphere Programme (GBP) (human population density and Crop mosaic), projected that the invasion of this mosquito into African urban environments could place an additional 126 million people at malaria risk (Sinka et al. 2020). EO data is essential to design effective surveillance systems to monitor areas suitable for An. stephensi.
5.3 Detection of surface-water bodies for effective mosquito larvae control
For vectors with an aquatic life stage, systematic identification of dynamic hydrological systems can be valuable in identifying breeding sites and inform larval source management. While both satellite and aerial-based EO can be used to identify breeding sites, an increasing number of studies are using drones due to the high spatial resolution, ability to detect breeding sites within a specified time frame and ownership of these systems by disease control organisations.
In Unguja, Zanzibar Archipelago, a Spatial Intelligence System (SIS) was designed to identify all surface water bodies within malaria transmission hotspots over a wide area, defined using satellite radar and optical EO imagery (Hardy et al. 2017). In this study, off-the-shelf DJI Phantom 4 drones, fitted with a standard 4k RGB (Red-Green-Blue) camera, were used to collect detailed imagery of potential breeding sites at a high spatial resolution (0.1 m/pixel). Temporal resolution was determined by the hydrological conditions of the study area with new drone surveys being commissions where rainfall events led to a significant change in surface water extent. Surface water were very varied, including large inundated rice paddies, natural swamps with aquatic vegetation, and small (<2 m in diameter) ponding of water in tyre ruts and paths where soil infiltration rates are low. This complex array of water bodies was mapped using manual digitising assisted using a freely available region growing tool (Hardy et al. 2022a). Subsequent information on water body location and extent was uploaded to a smartphone app (Zzapp Malaria: https://www.zzappmalaria.com/) which was used by field operatives to locate and systematically treat potential Anopheles sp. breeding sites.
In Cote d’Ivoire, a technical workflow was established to integrate satellite data and drone surveys with mosquito larval sampling to identify large, semi-permanent water bodies that provide potential breeding sites for malaria vector Anopheles funestus (Byrne et al. 2021b). Using Sentinel-2 high-resolution multispectral satellite imagery, the study area was stratified into four environmental strata according to spectral profiles and derived vegetation indices from satellite data and sampled proportionally to identify semi-permanent water bodies across variable rural landscapes. Red-Green-Blue (RGB) mapping of areas surrounding surface water bodies was conducted using an off-the-shelf DJI Phantom 4 drone. To achieve the required spatial resolution for fine-scale mapping, drones were flown at 150 m altitude, generating a high resolution of 0.04 m/pixel. With the aim to develop a systematic approach and inform deep-learning approaches to EO data classification, this illustrates drone surveys conducted in a spatially representative sampling frame developed in conjunction with freely available satellite datasets.
In Maynas Province, Peru, the use of drone-based imagery has also been demonstrated in identification of Nyssorhynchus darlingi aquatic breeding sites in malaria hotspots (Carrasco-Escobar et al. 2019). Here, local vector ecology, the hydrological context and research aims determine a different set of requirements. To identify the most productive water bodies in a riverine environment, the study prioritised high-resolution images over a small study area (~1 km) and multispectral imagery. Conventional RGB imagery was captured using a DJI Phantom 4 Pro quadcopter with a DJI 4K RGB flown at an altitude of 100 m to give a spatial resolution of 0.1 m/pixel. In addition, multispectral imagery was collected using a 3DR Solo quadcopter, fitted with a Parrot Sequoia sensor composed of single-band cameras (Green, Red, Red Edge and Near Infrared – NIR). The 3DR Solo was flown at an altitude of 50 m, which gives a ground sampling distance of 0.02 m/pixel. This demonstrates the flexibility and richness of data available with tailored use of UAV technology and specific sensors. The study design itself provided proof-of-concept for using high-resolution imagery from UAV to differentiate the spectral profiles of productive water bodies where Ny. darlingi is likely to breed in Amazonian Peru, an innovation which could be applied to integrating larval source management in malaria elimination strategies.
5.4 Early warning systems and automated mosquito surveillance using Earth observation
EO data is also widely used to develop predictive models enabling deployment of targeted control measures. In the last decade, the emergence and re-emergence of VBD reinforced the need for an effective early warning, surveillance and control of vectors and preparedness (Weaver 2013). Due to arthropod vectors being especially sensitive to changes in climate, forecasting systems driven by EO data have been useful for predicting disease risk and to guide epidemiological surveillance and decision-making process (Thomson et al. 2006, Thomson et al. 2008). The publication of guidelines of Malaria Early Warning Systems for Africa by the WHO, offered a valuable framework for integrated approach response strategies and epidemic preparedness (DaSilva et al. 2004, Thomson and Connor 2001, World Health Organisation 2001). Social, environmental and epidemiological factors are involved in malaria epidemics prediction and together with weather monitoring and seasonal climate forecasts, can be incorporated in malaria early warning systems (Thomson et al. 2006, World Health Organisation 2001). Dengue ecology and its strong relationship with climate variation is another example of how risk prediction studies and modelling approaches could help in the forecast of outbreaks (Colón-González et al. 2018, Colon-Gonzalez et al. 2021, Lauer et al. 2018, Lowe et al. 2016, Petrova et al. 2019) (see Finch et al. Chapter 12 for detailed environmental EO predictors and EWS in VBD).
Other innovative approaches for VBD surveillance using EO, automated sensors and algorithms/machine learning are still underway. For instance, VECTRACK (https://vectrack.avia-gis.com/) is an Earth Observation service for preventive control of insect disease vectors. This project seeks to generate an automated surveillance system combining an Earth Observation (EO) Sentinel service and satellite data with a network of ground optoelectronic sensors and traps to enable fully remote and automated counting and classification of the target mosquitoes (McEntaggart et al. 2020). This combined data will then feed into spatial models, and along with data on meteorological and environmental conditions, this can be used to generate large-scale risk maps. The use of this approach might help to significantly reduce costs and time of monitoring and surveillance and allows for more rapid intervention; however, this has yet to be fully evaluated.
5.5 Use of Earth observation data to map animal reservoirs and human populations at risk
The use of EO data in ecological or conservation studies is already well established; for example, in aerial surveys of large wild mammals over sparse areas (Koh and Wich 2012). However, in the study of vector-borne diseases with sylvatic life cycles in wildlife reservoirs, ecology intersects with public health and EO data of human and animal hosts can also provide critical information on transmission cycles and potential distribution of disease.
Yellow Fever (YF) is a zoonotic disease endemic in neotropical forests, transmitted by Haemagogus spp. mosquito vectors and with reservoirs of Yellow Fever Virus (YFV) circulating in non-human primate hosts. In Brazil, satellite-derived EO data has been used to estimate the distribution of the primate host species Alouatta caraya and A. guariba clamitans, which is then built into ecological niche models using maximum entropy modelling (Maxent) to identify geographic areas at risk of sylvatic outbreaks and human populations at risk (de Almeida et al. 2019). Explanatory variables include topography derived from the SRTM (Jarvis et al. 2022) and climatic variables derived from WorldClim (e.g. rainfall, temperature, solar radiation, water vapor pressure, wind speed), which uses meteorological station data validated by MODIS-derived satellite data (Fick and Hijmans 2017) at 1 km resolution. Here, use of satellite EO data to predict the distribution of reservoir species provides a valuable tool in building spatial risk models to inform public health interventions.
EO data has also be used to provide detailed estimations of human populations for VBD surveillance and management (Tatem et al. 2012). Sub-national metrics for population density are vital to mitigate VBD, with population denominator data required to define infection rates, allocate resources and measure the impact of interventions. However, accurate population data can be lacking in resource-poor settings. In addition, non-uniform distributions of people within administrative boundaries can lead to miscalculations. Recent computational approaches have been used to map population density as a continuous surface, using available survey/ census data and drawing on high-resolution satellite-derived remote sensing variables (e.g. topography, elevation, mean reflectance in specific bands, vegetation indices) to create a gridded population distribution (Qiu et al. 2022). Advances in high-quality and high-resolution (<10 m) satellite imagery has further improved global estimates, allowing human-built structures to be identified, counted and incorporated into models to generate spatially disaggregated estimates of human population distribution at resolutions of 90 m (Lloyd et al. 2017, Wardrop et al. 2018). In geographic areas that are at risk of VBD transmission but lack the infrastructure for updated or fine-scale demographic information, WorldPop population distribution surfaces can be vital in ensuring accurate risk mapping and targeted interventions.
At more localised spatial scales, UAV s can be used to map animal reservoirs of VBD s. Rift Valley Fever (RVF) is a viral zoonotic disease of wild ungulates in sub- Saharan Africa, that can be transmitted to humans through infective fluids or through bites from mosquito vectors. In Kenya, the commercially available senseFly eBee drone was used to identify the density and distribution of livestock and wildlife hosts of RVF (https://www.zooniverse.org/projects/rfv-drones/rift-valley-fever-drones). The UAV was fitted with the senseFly S.O.D.A sensor, a drone-specific photogrammetry camera that can capture sharp images and 3D digital surface models. From this, orthomosaic images were produced with a resolution of 0.03 m/pixel. This study required high resolution imagery across large study areas, with a view to identifying goats, camels and cattle and managing the disease reservoir. Imagery of this kind can be used to train machine learning algorithms to track and identify key mammal species in sparse landscapes that are integral to disease transmission cycles.
In Kinabatangan, Malaysia, a study aimed to identify and count the roosting sites of non-human primates (Jumail et al. 2021). Non-human primates are a reservoir for Plasmodium knowlesi, the aetiological agent of zoonotic malaria across Southeast Asia and most notably in East Malaysia. Unlike the broad swathes of savannah imaged in the previous example, this study looked at fine-scale 1 km transects of riverine forest. To obtain the required data, teams used a custom built hexacopter drone fitted with a FLiR Thermal Camera. Thermal images could then be used for visual identification of monkeys and counting of roosts across the transects. This is an example of the adaptability and user customisation possible with UAV technology, with equipment and sensors selected and modified to meet the requirements of both the ecological context and research aims.
6 Considerations for use of Earth observation data for VBD surveillance
Although these examples highlight how EO data can contribute to VBD surveillance, EO data remains underutilised in many VBD-endemic countries. While many sources of EO data are now freely available, one of the primary barriers to uptake remains technical expertise on remote sensing within disease control departments. This can be further challenged by the wide array of data sources and analysis platforms currently available. Decisions on EO data sources and analysis methods should be guided by local knowledge of vector ecology and the control strategies to be deployed. Factors to be considered include the following.
6.1 Area of interest
The geographical location and extent of the target area of interest will have determine the applicability of different data types. For example, fine-scale mapping may use targeted drone surveys while more extensive risk-mapping at larger spatial scales requires satellite-based EO data. Additionally, specific countries or regions may have targeted satellite or aerial EO data collection, such as through satellites run through national mapping services. The location also determines the feasibility of using different data sources; optical satellite data coverage may be limited in tropical forested areas with high cloud cover and drones may not be feasible in cities or areas with military activities.
6.2 Mapping objectives
The choice of EO data and analysis methods are also highly dependent on the objective of the mapping exercise and the corresponding features to be detected. For many risk mapping exercises, this may consist of a list of potential environmental covariates associated with a particular vector or VBD distribution. In other cases, the objective may be to detect specific vector breeding sites to deploy larval source control measures or to locate houses to plan a survey. In all cases, the specific features to be detected, and the characteristics of these features, should be identified. This may also consider specific types of EO data frequently used to detect these features, e.g. LIDAR data to measure canopy structure or optical data including NIR to calculate NDVI as a measure of greenness.
6.3 Spatial and temporal resolution
The spatial and temporal resolutions of these features should inform the selection of EO data with corresponding resolutions. For example, An. funestus typically breeds in large, semi-permanent water bodies which can be detected by coarser satellite-based EO data collected at irregular intervals. Other studies aiming to map the distribution of short-lived temporary water bodies, such as the breeding sites for An. arabiensis, or movements of disease hosts would be required to identify EO data with higher temporal resolutions to capture these dynamic features. Similar factors should be considered when mapping changing environments such as actively deforested areas. This should identify the temporal and spatial resolutions needed to capture these changes. Conversely, much coarser data may be required for large-scale mapping to capture broad environmental trends. Studies should consider the minimum resolutions of EO data required and trade-offs between computational requirements and higher resolution data.
6.4 Cost and logistics
Despite many sources of EO data being freely available, all mapping exercises require resources, such as personnel time, expertise, computing infrastructure or equipment purchase. These costs may increase when specialised equipment or data is required. For example, this may include the purchase of drones and sensors or tasking commercial satellites to cover specific areas. For studies requiring field activities, such as drone mapping or collection of ground-truthing data, these will require resources to support field staff, local transport and other consumables. Additionally, these studies typically require advanced permissions to collect samples, personnel data or to conduct drone mapping. This will often require permissions from Civil Aviation Authorities or other mapping agencies as well as local community acceptance and consent (Hardy et al. 2022b).
6.5 Availability of training data
One of the main determinants of analysis methods to be used is the availability of training data used to fit supervised models (such as machine learning and Artificial Intelligence approaches). The types and characteristics of training data needed are highly dependent on the target features to be identified. For example, algorithms designed to detect breeding sites of specific vector species are largely dependent on data from ground-based entomological surveys. Studies aiming to identify wider habitat types (e.g. forests or large water bodies) may be able to generate training data remotely by labeling high resolution imagery using GIS or specialised software tools. For specific applications, additional labelled training datasets may be available through online repositories (e.g. labelled agricultural data through Radiant Earth, https://www.radiant.earth/). The amount of training data required is dependent on the complexity of the feature to be identified as well as the type of algorithm used and may be a limiting factor in choosing specific algorithm types. However, there are example studies where training data has been generated automatically for machine learning models applied to satellite EO data providing an end-to-end automated mapping system, although such approaches have not yet been applied to drone-EO data (Hardy et al. 2019).
6.6 Technical expertise and computing infrastructure
Decisions on the types of EO data and analysis methods should also consider the availability of technical expertise needed to process EO data and develop algorithms. This should also assess how the data will be used, the time frames required and the primary end user. For example, if local health departments need to use information from drone surveys immediately after flights are conducted, the most feasible approaches may be manually digitising images in field settings offline. In other cases where mapping is repeatedly conducted over wide areas, it may be more feasible and cost-effective to develop deep-learning algorithms to automatically identify target features. Specific governments agencies or research programmes may additionally have access to dedicated servers or infrastructure to support repeated data analysis. In contrast, other smaller research projects or disease control agencies may need to rely on local infrastructure or cloud-based computing. The level of technical expertise available should be considered when planning any mapping activities and may impact the feasibility of different approaches.
7 Conclusions
Most VBD are governed by geographical components and decisions made in terms of controlling these diseases are often made a large (regional or national) scales. As such, EO has, and will continue to play a vital role in tackling VBD s. The rapid expansion of EO data and analysis methods offers new opportunities for VBD surveillance, from tracking habitats of invasive vector species to developing early warning systems or planning vector control activities. Rapidly changing environmental conditions increase the need for these methods and data sources to characterise changing VBD epidemiology. Within this chapter, we have given an overview of EO data sources and analysis methods as well as example applications. We have additionally outlined key considerations for the integration of EO data into VBD surveillance and priorities for determining the feasibility of different data sources and analysis strategies. While these technologies have the potential to transform VBD surveillance, EO data applications need to be designed to be appropriate for specific contexts and ecological settings.
Acknowledgements
This project was supported by a Sir Henry Dale fellowship jointly funded by the Wellcome Trust and the Royal Society (Grant no. 221963/Z/20/Z). Additional funding was provided by the project ‘International collaborative network for the integration, standardisation and assessment of the use of drones in malaria vector control strategies’ funded by the Global Challenges Research Fund (GCRF) Networking Grant (GCRFNGR4\1140). AJH received support from the Bill and Melinda Gates Foundation Innovation Fund INV-010583 through a grant managed by the Innovative Vector Control Consortium (IVCC). ML would like to acknowledge support from the National Institute of Health (USA) grant R03AI156580. The authors thank Antony Barja for processing the satellite data sources and availability and Edson J. Ascencio, Enrique A. Hernandez-Bustamante, Ximena J. Parraga, and Carmen Sofia Uribe Calama for their contribution to the scoping review in this chapter.
References
Ahmed A, Khogali R, Elnour M-AB, Nakao R and Salim B (2021) Emergence of the invasive malaria vector Anopheles stephensi in Khartoum State, Central Sudan. Parasites & vectors 14: 511. 10.1186/s13071-021-05026-4.
Baeza A, Santos-Vega M, Dobson AP and Pascual M (2017) The rise and fall of malaria under land-use change in frontier regions. Nat Ecol Evol 1: 108. 10.1038/s41559-017-0108.
Bergquist R, Luvall JC and Malone JB (2021) The changing risk of vector-borne diseases: Global satellite remote sensing and geospatial surveillance at the forefront. Geospat Health 16. 10.4081/gh.2021.1047.
Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL, Drake JM, Brownstein JS, Hoen AG, Sankoh O, Myers MF, George DB, Jaenisch T, Wint GR, Simmons CP, Scott TW, Farrar JJ and Hay SI (2013) The global distribution and burden of dengue. Nature 496: 504–507. 10.1038/nature12060.
Bravo DT, Lima GA, Alves WAL, Colombo VP, Djogbenou LS, Pamboukian SVD, Quaresma CC and Alves de Araujo S (2021) Automatic detection of potential mosquito breeding sites from aerial images acquired by unmanned aerial vehicles. Comput Environ Urban Syst 90.
Burke M, Driscoll A, Lobell DB and Ermon S (2021) Using satellite imagery to understand and promote sustainable development. Science 371. 10.1126/science.abe8628.
Byrne I, Aure W, Manin BO, Vythilingam I, Ferguson HM, Drakeley CJ, Chua TH and Fornace KM (2021a) Environmental and spatial risk factors for the larval habitats of Plasmodium knowlesi vectors in Sabah, Malaysian Borneo. Sci Rep 11: 11810. 10.1038/s41598-021-90893-1.
Byrne I, Chan K, Manrique E, Lines J, Wolie RZ, Trujillano F, Garay GJ, Del Prado Cortez MN, Alatrista-Salas H, Sternberg E, Cook J, N’Guessan R, Koffi A, Ahoua Alou LP, Apollinaire N, Messenger LA, Kristan M, Carrasco-Escobar G and Fornace K (2021b) Technical Workflow Development for Integrating Drone Surveys and Entomological Sampling to Characterise Aquatic Larval Habitats of Anopheles funestus in Agricultural Landscapes in Côte d’Ivoire. J Environ Public Health 2021: 3220244. 10.1155/2021/3220244.
Caminade C, McIntyre KM and Jones AE (2019) Impact of recent and future climate change on vector-borne diseases. Ann N Y Acad Sci 1436: 157–173. 10.1111/nyas.13950.
Caminade C, Medlock JM, Ducheyne E, McIntyre KM, Leach S, Baylis M and Morse AP (2012) Suitability of European climate for the Asian tiger mosquito Aedes albopictus: recent trends and future scenarios. J R Soc Interface 9: 2708–2717. 10.1098/rsif.2012.0138.
Carrasco-Escobar G, Manrique E, Ruiz-Cabrejos J, Saavedra M, Alava F, Bickersmith S, Prussing C, Vinetz JM, Conn JE, Moreno M and Gamboa D (2019) High-accuracy detection of malaria vector larval habitats using drone-based multispectral imagery. PLoS neglected tropical diseases 13: e0007105. 10.1371/journal.pntd.0007105.
Carter TE, Yared S, Gebresilassie A, Bonnell V, Damodaran L, Lopez K, Ibrahim M, Mohammed S and Janies D (2018) First detection of Anopheles stephensi Liston, 1901 (Diptera: culicidae) in Ethiopia using molecular and morphological approaches. Acta Trop 188: 180–186. https://doi.org/10.1016/j.actatropica.2018.09.001.
Cheng G, Xie X, Han J, Guo L and Xia G (2020) Remote sensing image scene classification meets deep learning: challenges, methods, benchmarks and opportunities. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 13: 3735–3756.
Cline BL (1970) New eyes for epidemiologists: aerial photography and other remote sensing techniques. Am J Epidemiol 92: 85–89. 10.1093/oxfordjournals.aje.a121188.
Colón-González FJ, Harris I, Osborn TJ, Bernardo CSS, Peres CA, Hunter PR, Warren R, Vuurene Dv and Lake IR (2018) Limiting global-mean temperature increase to 1.5–2 °C could reduce the incidence and spatial spread of dengue fever in Latin America. Proceedings of the National Academy of Sciences 115: 6243–6248. doi:10.1073 /pnas.1718945115.
Colon-Gonzalez FJ, Soares Bastos L, Hofmann B, Hopkin A, Harpham Q, Crocker T, Amato R, Ferrario I, Moschini F, James S, Malde S, Ainscoe E, Sinh Nam V, Quang Tan D, Duc Khoa N, Harrison M, Tsarouchi G, Lumbroso D, Brady OJ and Lowe R (2021) Probabilistic seasonal dengue forecasting in Vietnam: A modelling study using superensembles. PLoS medicine 18: e1003542. 10.1371/journal.pmed.1003542.
DaSilva J, Garanganga B, Teveredzi V, Marx SM, Mason SJ and Connor SJ (2004) Improving epidemic malaria planning, preparedness and response in Southern Africa. Report on the 1st Southern African Regional Epidemic Outlook Forum, Harare, Zimbabwe, 26–29 September, 2004. Malaria journal 3: 37. 10.1186/1475-2875-3-37.
de Almeida MAB, Dos Santos E, Cardoso JDC, da Silva LG, Rabelo RM and Bicca-Marques JC (2019) Predicting Yellow Fever Through Species Distribution Modeling of Virus, Vector, and Monkeys. EcoHealth 16: 95–108. 10.1007/s10393-018-1388-4.
de Castro MC, Monte-Mor RL, Sawyer DO and Singer BH (2006) Malaria risk on the Amazon frontier. Proceedings of the National Academy of Sciences of the United States of America 103: 2452–2457. 10.1073/pnas.0510576103.
Deblauwe V, Droissart V, Bose R, Sonké B, Blach-Overgaard A, Svenning J-C, Wieringa JJ, Ramesh BR, Stévart T and Couvreur TLP (2016) Remotely sensed temperature and precipitation data improve species distribution modelling in the tropics. Global Ecology and Biogeography 25: 443–454. https://doi.org/10.1111/geb.12426.
Dicko AH, Lancelot R, Seck MT, Guerrini L, Sall B, Lo M, Vreysen MJB, Lefrançois T, Fonta WM, Peck SL and Bouyer J (2014) Using species distribution models to optimize vector control in the framework of the tsetse eradication campaign in Senegal. Proceedings of the National Academy of Sciences 111: 10149–10154. doi:10.1073/pnas.1407773111.
Faulde MK, Rueda LM and Khaireh BA (2014) First record of the Asian malaria vector Anopheles stephensi and its possible role in the resurgence of malaria in Djibouti, Horn of Africa. Acta Trop 139: 39–43. https://doi.org/10.1016/j.actatropica.2014.06.016.
Fick SE and Hijmans RJ (2017) Worldclim 2: New 1-km spatial resolution climate surfaces for global land areas. International Journal of Climatology.
Finer M, Novoa S, Weisse MJ, Petersen R, Mascaro J, Souto T, Stearns F and Martinez RG (2018) Combating deforestation: From satellite to intervention. Science 360: 1303–1305. 10.1126/science.aat1203.
Fornace KM, Diaz AV, Lines J and Drakeley CJ (2021) Achieving global malaria eradication in changing landscapes. Malaria journal 20: 69. 10.1186/s12936-021-3599-0.
Fornace KM, Drakeley CJ, William T, Espino F and Cox J (2014) Mapping infectious disease landscapes: unmanned aerial vehicles and epidemiology. Trends in parasitology 30: 514–519. 10.1016/j.pt.2014.09.001.
Fornace KM, Surendra H, Abidin TR, Reyes R, Macalinao MLM, Stresman G, Luchavez J, Ahmad RA, Supargiyono S, Espino F, Drakeley CJ and Cook J (2018) Use of mobile technology-based participatory mapping approaches to geolocate health facility attendees for disease surveillance in low resource settings. International journal of health geographics 17: 21. 10.1186/s12942-018-0141-0.
Franke J, Gebreslasie M, Bauwens I, Deleu J and Siegert F (2015) Earth observation in support of malaria control and epidemiology: MALAREO monitoring approaches. Geospat Health 10: 335. 10.4081/gh.2015.335.
Gorris ME, Bartlow AW, Temple SD, Romero-Alvarez D, Shutt DP, Fair JM, Kaufeld KA, Del Valle SY and Manore CA (2021) Updated distribution maps of predominant Culex mosquitoes across the Americas. Parasites & vectors 14: 547. 10.1186/s13071-021-05051-3.
Gray PC, Fleishman AB, Klein DJ, McKown MW, Bezy VS, Lohmann KJ and Johnston DW (2019) A convolutional neural network for detecting sea turtles in drone imagery. Methods in Ecology and Evolution 00: 1–11.
Hansen MC, Potapov PV, Moore R, Hancher M, Turubanova SA, Tyukavina A, Thau D, Stehman SV, Goetz SJ, Loveland TR, Kommareddy A, Egorov A, Chini L, Justice CO and Townshend JR (2013) High-resolution global maps of 21st-century forest cover change. Science 342: 850–853. 10.1126/science.1244693.
Hardy A, Ettritch G, Cross DE, Bunting P, Liywalii F, Sakala J, Silumesii A, Singini D, Smith M, Willis T and Thomas CJ (2019) Automatic Detection of Open and Vegetated Water Bodies Using Sentinel 1 to Map African Malaria Vector Mosquito Breeding Habitats. Remote Sensing 11: 593.
Hardy A, Makame M, Cross D, Majambere S and Msellem M (2017) Using low-cost drones to map malaria vector habitats. Parasites & vectors 10: 29. 10.1186/s13071-017-1973-3.
Hardy A, Oakes G, Hassan J and Yussuf Y (2022a) Improved use of drone imagery for malaria vector control through technology-assisted digitising (TAD). Remote Sensing 14.
Hardy A, Proctor M, MacCallum C, Shawe J, Abdalla S, Ali R, Abdalla S, Oakes G, Rosu L and Worrall E (2022b) Conditional trust: Community perceptions of drone use in malaria control in Zanzibar. Technology in Society 68: 101895. https://doi.org/10.1016/j.techsoc.2022.101895.
Hay SI (2000) An overview of remote sensing and geodesy for epidemiology and public health application. Advances in parasitology 47: 1–35.
Hay SI, Snow RW and Rogers DJ (1998) Predicting malaria seasons in Kenya using multitemporal meteorological satellite sensor data. Transactions of the Royal Society of Tropical Medicine and Hygiene 92: 12–20.
Iwamura T, Guzman-Holst A and Murray KA (2020) Accelerating invasion potential of disease vector Aedes aegypti under climate change. Nature communications 11: 2130. 10.1038/s41467-020-16010-4.
Jarvis A, Reuter A and Nelson EG (2022) Hole-filled SRTM for the globe version 4.0. In: Database C-CSm, ed.
Jones KE, Patel NG, Levy MA, Storeygard A, Balk D, Gittleman JL and Daszak P (2008) Global trends in emerging infectious diseases. Nature 451: 990–993. 10.1038/nature06536.
Jumail A, Liew TS, Salgado-Lynn M, Fornace KM and Stark DJ (2021) A comparative evaluation of thermal camera and visual counting methods for primate census in a riparian forest at the Lower Kinabatangan Wildlife Sanctuary (LKWS), Malaysian Borneo. Primates 62: 143–151. 10.1007/s10329-020-00837-y.
Kattenborn T, Eichel J and Fassnacht FE (2019) Convolutional Neural Networks enable efficient, accurate and fine-grained segmentation of plant species and communities from high resolution UAV imagery. Sci Rep 9.
Keiser J, De Castro MC, Maltese MF, Bos R, Tanner M, Singer BH and Utzinger J (2005) Effect of irrigation and large dams on the burden of malaria on a global and regional scale. The American journal of tropical medicine and hygiene 72: 392–406.
Koh L and Wich S (2012) Dawn of drone ecology: low-cost autonomous aerial vehicles for conservation. Tropical Conservation Science 5: 121–132.
Kraemer MUG, Reiner RC, Jr., Brady OJ, Messina JP, Gilbert M, Pigott DM, Yi D, Johnson K, Earl L, Marczak LB, Shirude S, Davis Weaver N, Bisanzio D, Perkins TA, Lai S, Lu X, Jones P, Coelho GE, Carvalho RG, Van Bortel W, Marsboom C, Hendrickx G, Schaffner F, Moore CG, Nax HH, Bengtsson L, Wetter E, Tatem AJ, Brownstein JS, Smith DL, Lambrechts L, Cauchemez S, Linard C, Faria NR, Pybus OG, Scott TW, Liu Q, Yu H, Wint GRW, Hay SI and Golding N (2019) Past and future spread of the arbovirus vectors Aedes aegypti and Aedes albopictus. Nat Microbiol 4: 854–863. 10.1038/s41564-019-0376-y.
Lauer SA, Sakrejda K, Ray EL, Keegan LT, Bi Q, Suangtho P, Hinjoy S, Iamsirithaworn S, Suthachana S, Laosiritaworn Y, Cummings DAT, Lessler J and Reich NG (2018) Prospective forecasts of annual dengue hemorrhagic fever incidence in Thailand, 2010–2014. Proceedings of the National Academy of Sciences 115: E2175–E2182. doi:10.1073/pnas .1714457115.
Lippi CA, Stewart-Ibarra AM, Loor M, Zambrano JED, Lopez NAE, Blackburn JK and Ryan SJ (2019) Geographic shifts in Aedes aegypti habitat suitability in Ecuador using larval surveillance data and ecological niche modeling: Implications of climate change for public health vector control. PLoS neglected tropical diseases 13: e0007322. 10.1371/jour nal.pntd.0007322.
Liu D and Fan X (2010) Assessing object-based classification: advantages and limitations. Remote Sensing Letters 1: 187–194.
Lloyd CT, Sorichetta A and Tatem AJ (2017) High resolution global gridded data for use in population studies. Sci Data 4: 170001. 10.1038/sdata.2017.1.
Lowe R, Bailey TC, Stephenson DB, Jupp TE, Graham RJ, Barcellos C and Carvalho MS (2013) The development of an early warning system for climate-sensitive disease risk with a focus on dengue epidemics in Southeast Brazil. Stat Med 32: 864–883. 10.1002/sim .5549.
Lowe R, Coelho CA, Barcellos C, Carvalho MS, Catao Rde C, Coelho GE, Ramalho WM, Bailey TC, Stephenson DB and Rodo X (2016) Evaluating probabilistic dengue risk forecasts from a prototype early warning system for Brazil. Elife 5. 10.7554/eLife.11285.
Mather PM and Koch M (2010) Computer processing of remotely-sensed images (4th edition). Chichester, Wiley.
McEntaggart K, Fillet J, Marinoiu B, Quaresma H and de Paolis S (2020) The use of emerging technologies for regulation, Department for Business, Energy and Industrial Strategy.
Messina JP, Brady OJ, Golding N, Kraemer MUG, Wint GRW, Ray SE, Pigott DM, Shearer FM, Johnson K, Earl L, Marczak LB, Shirude S, Davis Weaver N, Gilbert M, Velayudhan R, Jones P, Jaenisch T, Scott TW, Reiner RC, Jr. and Hay SI (2019) The current and future global distribution and population at risk of dengue. Nat Microbiol 4: 1508–1515. 10.1038/s41564-019-0476-8.
Metcalf CJE, Walter KS, Wesolowski A, Buckee CO, Shevliakova E, Tatem AJ, Boos WR, Weinberger DM and Pitzer VE (2017) Identifying climate drivers of infectious disease dynamics: recent advances and challenges ahead. Proc Biol Sci 284. 10.1098/rspb.2017.0901.
Minakawa N, Munga S, Atieli F, Mushinzimana E, Zhou G, Githeko AK and Yan G (2005) Spatial distribution of anopheline larval habitats in Western Kenyan highlands: effects of land cover types and topography. The American journal of tropical medicine and hygiene 73: 157–165.
NASA (1973) The use of remote sensing in mosquito control, National Aeronautics and Space Administration, Lyndon B. Johnson Space Center, Houston, USA.
Noor AM, Alegana VA, Gething PW and Snow RW (2009) A spatial national health facility database for public health sector planning in Kenya in 2008. International journal of health geographics 8: 13. 10.1186/1476-072X-8-13.
Pareyn M, Rutten A, Merdekios B, Wedegärtner REM, Girma N, Regelbrugge L, Shibru S and Leirs H (2020) High-resolution habitat suitability model for Phlebotomus pedifer, the vector of cutaneous leishmaniasis in southwestern Ethiopia. Parasites & vectors 13: 467. 10.1186/s13071-020-04336-3.
Patz JA, Daszak P, Tabor GM, Aguirre AA, Pearl M, Epstein J, Wolfe ND, Kilpatrick AM, Foufopoulos J, Molyneux D, Bradley DJ, Working Group on Land Use C and Disease E (2004) Unhealthy landscapes: policy recommendations on land use change and infectious disease emergence. Environ Health Perspect 112: 1092–1098.
Pekel JF, Cottam A, Gorelick N and Belward AS (2016) High-resolution mapping of global surface water and its long-term changes. Nature 540: 418–422. 10.1038/nature20584.
Peterson AT (2014) Mapping Disease Transmission Risk. Baltimore, USA , Johns Hopkins University Press.
Petrova D, Lowe R, Stewart-Ibarra A, Ballester J, Koopman SJ and Rodó X (2019) Sensitivity of large dengue epidemics in Ecuador to long-lead predictions of El Niño. Climate Services 15: 100096. https://doi.org/10.1016/j.cliser.2019.02.003.
Phillips SJ, Anderson RP and Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecol Modell 190: 231–259. https://doi.org/10.1016/j.ecolmodel.2005.03.026.
Qiu Y, Zhao X, Fan D, Li S and Zhao Y (2022) Disaggregating population data for assessing progress of SDGs: methods and applications. International Journal of Digital Earth 15: 2–29. 10.1080/17538947.2021.2013553.
Ryan SJ, Carlson CJ, Mordecai EA and Johnson LR (2019) Global expansion and redistribution of Aedes-borne virus transmission risk with climate change. PLoS neglected tropical diseases 13: e0007213. 10.1371/journal.pntd.0007213.
Seyfarth M, Khaireh BA, Abdi AA, Bouh SM and Faulde MK (2019) Five years following first detection of Anopheles stephensi (Diptera: Culicidae) in Djibouti, Horn of Africa: populations established – malaria emerging. Parasitol Res 118: 725–732. 10.1007/s00436- 019-06213-0.
Simpson R, Page KR and De Roure D (2014) Zooniverse: observing the world’s largest citizen science platform. Proceedings of the 23rd International Conference on the World Wide Web: 1049–1054.
Sinka ME, Golding N, Massey NC, Wiebe A, Huang Z, Hay SI and Moyes CL (2016) Modelling the relative abundance of the primary African vectors of malaria before and after the implementation of indoor, insecticide-based vector control. Malaria journal 15: 142. 10.1186/s12936-016-1187-8.
Sinka ME, Pironon S, Massey NC, Longbottom J, Hemingway J, Moyes CL and Willis KJ (2020) A new malaria vector in Africa: Predicting the expansion range of Anopheles stephensi and identifying the urban populations at risk. Proceedings of the National Academy of Sciences 117: 24900–24908. doi:10.1073/pnas.2003976117.
Stresman GH, Stevenson JC, Owaga C, Marube E, Anyango C, Drakeley C, Bousema T and Cox J (2014) Validation of three geolocation strategies for health-facility attendees for research and public health surveillance in a rural setting in western Kenya. Epidemiol Infect 142: 1978–1989. 10.1017/S0950268814000946.
Tatem AJ, Adamo S, Bharti N, Burgert CR, Castro M, Dorelien A, Fink G, Linard C, John M, Montana L, Montgomery MR, Nelson A, Noor AM, Pindolia D, Yetman G and Balk D (2012) Mapping populations at risk: improving spatial demographic data for infectious disease modeling and metric derivation. Population Health Metrics 10: 8. 10.1186/ 1478-7954-10-8.
Tatem AJ, Goetz SJ and Hay SI (2008) Fifty Years of Earth Observation Satellites: views from above have lead to countless advances on the ground in both scientific knowledge and daily life. Am Sci 96: 390–398. 10.1511/2008.74.390.
Thomson MC and Connor SJ (2001) The development of Malaria Early Warning Systems for Africa. Trends in parasitology 17: 438–445. 10.1016/s1471-4922(01)02077-3.
Thomson MC, Doblas-Reyes FJ, Mason SJ, Hagedorn R, Connor SJ, Phindela T, Morse AP and Palmer TN (2006) Malaria early warnings based on seasonal climate forecasts from multi-model ensembles. Nature 439: 576–579. 10.1038/nature04503.
Thomson MC, Garcia-Herrera R and Beniston M (2008) Seasonal forecasts, climate change and human health, Advances in Global Change Research. Dordrecht, Springer.
Union of Concerned Scientists (2021) UCS Satellite Database.
Vittor AY, Pan W, Gilman RH, Tielsch J, Glass G, Shields T, Sanchez-Lozano W, Pinedo VV, Salas-Cobos E, Flores S and Patz JA (2009) Linking deforestation to malaria in the Amazon: characterization of the breeding habitat of the principal malaria vector, Anopheles darlingi. The American journal of tropical medicine and hygiene 81: 5–12.
Wagner VE, Hill-Rowley R, Narlock SA and Newson HD (1979) Remote sensing: a rapid and accurate method of data acquisition for a newly formed mosquito control district. Mosquito News 39: 283–287.
Wardrop NA, Jochem WC, Bird TJ, Chamberlain HR, Clarke D, Kerr D, Bengtsson L, Juran S, Seaman V and Tatem AJ (2018) Spatially disaggregated population estimates in the absence of national population and housing census data. Proceedings of the National Academy of Sciences of the United States of America 115: 3529–3537. 10.1073/pnas.1715305115.
Weaver SC (2013) Urbanization and geographic expansion of zoonotic arboviral diseases: mechanisms and potential strategies for prevention. Trends Microbiol 21: 360–363. 10.1016/j.tim.2013.03.003.
Weiss DJ, Lucas TCD, Nguyen M, Nandi AK, Bisanzio D, Battle KE, Cameron E, Twohig KA, Pfeffer DA, Rozier JA, Gibson HS, Rao PC, Casey D, Bertozzi-Villa A, Collins EL, Dalrymple U, Gray N, Harris JR, Howes RE, Kang SY, Keddie SH, May D, Rumisha S, Thorn MP, Barber R, Fullman N, Huynh CK, Kulikoff X, Kutz MJ, Lopez AD, Mokdad AH, Naghavi M, Nguyen G, Shackelford KA, Vos T, Wang H, Smith DL, Lim SS, Murray CJL, Bhatt S, Hay SI and Gething PW (2019) Mapping the global prevalence, incidence, and mortality of Plasmodium falciparum, 2000–17: a spatial and temporal modelling study. Lancet 394: 322–331. 10.1016/S0140-6736(19)31097-9.
Weiss DJ, Mappin B, Dalrymple U, Bhatt S, Cameron E, Hay SI and Gething PW (2015) Re-examining environmental correlates of Plasmodium falciparum malaria endemicity: a data-intensive variable selection approach. Malaria journal 14: 68. 10.1186/s12936-015- 0574-x.
Wimberly MC, de Beurs KM, Loboda TV and Pan WK (2021) Satellite Observations and Malaria: New Opportunities for Research and Applications. Trends in parasitology 37: 525–537. 10.1016/j.pt.2021.03.003.
World Health Organisation (2001) Malaria Early Warning Systems: Concepts, Indicators and Partners: A Framework for Field Research in Africa, World Health Organisation, Roll Back Malaria, Technical Support Network for Prevention and Control of Malaria, Geneva.
World Health Organisation (2019) Meeting report of the WHO technical consultation on the spread of Anopheles stephensi, World Health Organisation, Geneva.
Yared S, Gebressielasie A, Damodaran L, Bonnell V, Lopez K, Janies D and Carter TE (2020) Insecticide resistance in Anopheles stephensi in Somali Region, eastern Ethiopia. Malar J 19: 180. 10.1186/s12936-020-03252-2.