**Abstract**

Disease mapping utilizes disease maps as visual representations of sophisticated geographic data that provide a general overview of the disease situation in a defined geographic area. Epidemiology is concerned with investigating the causes of diseases, and often, these causes vary in frequency and in space. This variation in space gave a niche to remote sensing to find its way into the public health domain as disease researchers sought to investigate the explaining environmental and climatic factors. Studies have demonstrated the potential offered by remote sensing application to disease mapping and epidemiology and to support surveillance and control efforts. We used some examples from a case study conducted in Eswatini in Southern Africa. Remote sensing imagery when combined with GIS spatial analyses techniques could support and guide existing disease surveillance and control programs at local, regional, and even continental scales. Researchers have also studied factors influencing the patterns and distributions of vector-borne diseases at a variety of landscape scales. However, successful application of remote sensing technology depends on the ability of nonexperts' remotely sensed data and end users to access, retrieve, and analyze the data captured from satellites. The exploration of some of the opportunities presented by remote sensing to disease mapping and epidemiology is still unfolding as new opportunities are being presented.

**Keywords:** disease mapping, epidemiology, geostatistics, remote sensing, GIS

## **1. Introduction**

Remote sensing could be described as the science of scanning the earth using sensors onboard a satellite platform launched into space or high flying aircraft to obtain information and also monitor land use and land cover changes on the earth surface [1]. Often, the monitored land use and land cover changes emanate from human activities and their interaction with the environment. The observation of the earth by the orbiting satellites is done at different geographic scales and at different intervals or revisit periods, which are both widely referred to as spatial and temporal resolutions, respectively. Over the years, there had been noticeable improvements in the spatial and temporal resolutions which had been accompanied by an interesting visibility and readability of the captured images. In addition, the number of spectral bands used by the sensors to capture images had also increased, aiding in an appealing appearance to the human eye. Consequently, even nonexpert remote sensing users had been drawn into remote sensing by the *beautiful pictures* and availability of some of the end products of remote sensing. There had been, therefore, an increase in the application of remote sensing products by other

disciplines and fields of study. In public health, remotely sensed (RS) data products had been widely adopted and used in disease mapping and epidemiology. We used some examples from a case study conducted in Eswatini in Southern Africa.

Disease mapping referrers to the use of disease maps as visual representations of sophisticated geographic data that provide a general birdeye overview of the disease situation in a defined geographic area. On the other hand, epidemiology is a branch of medicine that deals with the incidence, distribution, and strategies for disease containment and control as well as other factors relating to human and animal health. Epidemiology is concerned with investigating the cause of disease, and often, these causes vary in frequency and in space. This variation in space is what gave a niche to remote sensing to find its way into the public health domain as disease researchers sought to investigate the explaining environmental and climatic factors. For this reason, remote sensing which provides a birdeye overview of the earth surface had continued to be applied in disease mapping for rapid risk assessment and monitoring efforts. As a result, remote sensing products have for quite a reasonable while been prolifically applied and used in disease mapping and epidemiology. Products of remote sensing include various vegetation indices and environmental proxies which are derived from satellite images and used to elucidate land use and land cover changes as well as to approximate environmental and climatic conditions on the earth surface.

Vegetation indices are mathematical combinations of different spectral bands that are designed to numerically separate or stretch the pixel value of different features in an image [2, 3]. RS data products had been used in a number of disease mapping and epidemiological studies such as in risk mapping of malaria [4, 5], soiltransmitted helminths [6], schistosomiases, and prediction of high risk areas for leishmaniasis in Brazil. Previous work include incorporation of RS data in human health studies and spatial targeting of trachoma control in Southern Sudan [7] by developing a national risk map and mapping tsetse fly habitat suitability among others [8]. In addition, identification of environmental risk factors for cholera using satellite-derived remotely sensed data products had been undertaken by [9]. Determination of population living in a city using remotely sensed data products was carried out in a study by Karume et al. [10], whereby a GeoEye satellite image at 50-m resolution was used, and population of the city was obtained by taking the number of houses times an average number of habitants per house. To date, an inexhaustible number of exploratory research studies in disease mapping and epidemiology had been undertaken using remote sensing vegetation indices as environmental and climatic proxies in combination with various modeling approaches.

One of the main advantages of using RS data products in disease mapping and epidemiology is its near real-time availability for rapid assessment of at risk areas and prediction of disease distribution, especially in inaccessible areas that may also lack baseline data [11]. The increase in the launch of higher resolution satellites sensors and advances in data processing techniques have enabled a wider adoption of RS data [12]. In economically disadvantaged areas with poor ground measurement meteorological station networks, RS data had often been preferred and used as environmental and climatic proxies in disease risk mapping and prediction. As new sensors with better spatial and temporal resolutions become available, new opportunities had been presented and explored in the application of remote sensing products in disease mapping and epidemiology [13].

From the early generation of ecological studies that demonstrated the capability of RS data products in disease mapping [13–16], there had been a sustained proliferation of such studies in disease mapping and epidemiology. The application of geostatistical techniques to identify spatial heterogeneities in disease distributions, patterns, and trends as well as forecasting for epidemic preparedness planning had been demonstrated in studies by Chilès and Delfiner [17]; and Tran et al. [18] *inter* 

#### *Remote Sensing Applications in Disease Mapping DOI: http://dx.doi.org/10.5772/intechopen.93652*

*alia*. Geostatistics is a branch of statistics that is used to analyze and predict the values associated with spatial and temporal phenomena [17]. It is often incorporated into modeling through the use of coordinates attached to the data that are being analyzed. An almost similar terminology also commonly used in such analysis is the term "biostatistics," which also refers to a similar approach except that the data being analyzed would involve biological data. Such models are then referred to as space–time models, especially due to the fact that they also include the observation dates of the mapped data. The theory behind incorporation of RS data in disease mapping and epidemiology was based on the field-observed association between environmental conditions and some of the disease-causing vectors [18–20], in particular how they vary in geographic space. For instance, some studies have demonstrated the association between radiation reflectance as measured by satellites and certain land cover types which have been used as environmental and climatic proxies for measurement of presence or absence of a disease and its vectors [21].

Incorporating remotely sensed data products in spatial modeling had been common in most aspects of geographic analyses. From the early usage of aerial photographs taken onboard aircraft to the launch of satellite-based sensors, remotely sensed data products have been in the forefront of scientific research on the land use and land cover changes of the earth. Many other disciplines such as public health and epidemiology have until recently overwhelmingly taken up remotely sensed data products and utilized them in disease mapping and in improvement of their understanding of environmentally driven diseases. The utilization of RS data products, especially in disease mapping, for instance, in mapping vector-borne diseases, had been prolific over the years. As mentioned earlier, products derived from remote sensing had been widely used in models as environmental proxies to analyze various behaviors of observed spatial phenomenon. Furthermore, environmental proxies had been incorporated as covariates in statistical models aimed at mapping, analyzing, and predicting spatial phenomenon relating to disease epidemiology.

In statistical models, disease analysis had been pursued by adjusting models with spatial covariates derived from remote sensing and regressed with any outcome of interest. In such cases, spatial regression models had often been parameterized using data variables derived from remotely sensed products. As a result, there had been an unprecedented upsurge in the utilization of environmental variables and proxies derived from remote sensing in statistical analysis models attempting to map diseases such as vector-borne diseases, including malaria, dengue fever, chikungunya fever, zika virus, and leishmaniasis *inter alia* [22]. Studies mapping the spatial distribution of such vector-borne diseases had often applied remotely sensed data products as environmental proxies by approximating either environmental conditions or land use and land cover features. When the spatial covariates had been used in such a manner in models, they would often be referred to as predictors or risk factors since they tend to covary with the severity of the disease. This association of diseases with their environment was first noted in some of the early ecological studies that demonstrated the capability of remote sensing products in disease mapping, including authors such as Beck et al. [15]; Hay [16], and Thomson et al. [14] among others. As an example, malaria had on numerous cases been referred to as an environmental disease, and describing environmental risk factors associated with this disease had driven the research in malaria risk mapping [23].

Whereas, the early uptake of remotely sensed products was initially limited by the slow processing power of first-generation computers as well as the limited and cumbersome storage facility required to store remotely sensed and geographic information system (GIS) data sets; the advent of fast processing computer power had made it possible to include remotely sensed data in mapping studies. In addition, storage facilities of geographic data sets had improved from the large and

difficult-to-handle cassettes to small, easy-to-carry hard drives and universal serial bus (usb) including portable external hard drives. Furthermore, the open access to remote sensing end products even for civilian usage had stimulated the interest of the scientific community and facilitated the uptake and application of remote sensing in disease mapping and epidemiology.

In areas with limited ground-based environmental data observation stations, remotely sensed data products became the only option available for mapping disease risk distribution. Such inaccessible areas included those with rugged terrain, areas in armed conflict, and those with limited resources which would not be enough to undertake field-based studies. Also, in cases where for example, ground-based weather stations had been used, they had often been limited by data discrepancies emanating from human interference, human error, instrument jamming, and in some cases, power failure. This would often result in missing data as observation could not be completed on those days when the weather station failed. During the processing and analysis of such data, the missing values in the data would often require sophisticated methods of data imputation which would not escape criticism from the scientific peer review community. Therefore, remote sensing products had been preferred because of the reliability of availability often in near real-time compared to other data sources which may require field surveys to be undertaken and thus too expensive to be repeated.
