Demographic and environmental factors associated with the distribution of Aedes albopictus in Cameroon

Abstract Aedes‐transmitted arboviruses have spread globally due to the spread of Aedes aegypti and Aedes albopictus. Its distribution is associated with human and physical geography. However, these factors have not been quantified in Cameroon. Therefore, the aim was to develop an Ae. albopictus geo‐referenced database to examine the risk factors associated with the vector distribution in Cameroon. Data on the Ae. albopictus presence and absence were collated and mapped from studies in published scientific literature between 2000 and 2020. Publicly available earth observation data were used to assess human geography, land use and climate risk factors related to the vector distribution. A logistic binomial regression was conducted to identify the significant risk factors associated with Ae. albopictus distribution. In total, 111 data points were collated (presence = 87; absence = 24). Different data collection methods and sites hindered the spatiotemporal analysis. An increase of one wet month in a year increased the odds of Ae. albopictus presence by 5.6 times. One unit of peri‐urban area increased the odds by 1.3 times. Using publicly available demographic and environmental data to better understand the key determinants of mosquito distributions may facilitate appropriately targeted public health messages and vector control strategies.


INTRODUCTION
The Aedes-transmitted arboviruses such as dengue, yellow fever, chikungunya and Zika have rapidly spread globally in the last decade (Kraemer et al., 2015;WHO, 2017WHO, , 2021. These arboviruses are of significant public health concern, causing widespread morbidity and mortality, and their expansion to the sub-Saharan Africa (SSA) region poses a major threat (Weetman et al., 2018). It is estimated that around 70% of the sub-Saharan African population is at risk of one of these arboviruses; however, the true burden is difficult to determine given the challenges of differential diagnosis in malaria co-endemic and resource-poor settings (Weetman et al., 2018).
The expansion of arboviral infections and disease is related to the distributions of the two main mosquito vectors, Aedes aegypti (=Stegomyia aegypti) and Aedes albopictus (=Stegomyia albopicta). Native to the African continent, Ae. aegypti is the main vector of these arboviruses globally (Moore et al., 2013). Ae. albopictus is native to the South-East Asian forests and was introduced to the African continent less than 30 years ago through the international commercialization of used tyres, and is now found in many countries across the continent (Bonizzoni et al., 2013;Brady & Hay, 2020;Weetman et al., 2018).
The geographical distribution of these two vectors is driven by human and physical geography, and a wide range of environmental factors (Brady et al., 2014;Cunze et al., 2016;Dickens et al., 2018;Kraemer et al., 2015Kraemer et al., , 2019. While these two vectors may coexist in the same areas (Brady & Hay, 2020;Paupy et al., 2010;Tedjou et al., 2020), they have been found to have different characteristics on a micro-level (Egid et al., 2021). For example, both species breed in containers, however Ae. aegypti prefers human-made containers in urban areas, while Ae. albopictus prefers containers surrounded by the presence of vegetation frequently found in peri-urban and rural areas (Brady et al., 2014;Brady & Hay, 2020;Egid et al., 2021;Kamgang et al., 2010;Kraemer et al., 2015;Tedjou et al., 2020).
Aedes albopictus is considered to be a secondary vector for dengue in Africa (Brady & Hay, 2020), but it is the main vector for chikungunya in Central Africa in recent years Paupy et al., 2010). However, its importance may increase with the rapidly expanding periÀ/urbanization taking place across SSA (United Nations, Department of Economic and Social Affairs, 2019).
Changes in the urban-rural interface, land use and land cover (LULC), and climate over time and space are important to monitor as they may influence vector distributions and risk (Ali et al., 2017;Burkett-Cadena & Vittor, 2018;Kalbus et al., 2021;Kamal et al., 2018;Ryan et al., 2018). For example, a lack of rainfall has been found to make larval and pupal stages of Ae. albopictus vulnerable to desiccation, except in areas where human-made breeding containers were available providing opportunities for breeding (Waldock et al., 2013).
Defining the human geography, land use patterns and climate envelope where the Ae. albopictus is present may help to identify key risk factors and target control measures. Recently in Cameroon, two main studies provided an update on the distribution of Ae.
aegypti and Ae. albopictus with geographical coordinates available for mapping (Simard et al., 2005;Tedjou et al., 2019). The later update in 2017 describes the distribution of Ae. albopictus being restricted to a certain latitude, that is, under 6 N, while Ae. aegypti was found to occur across the entire country .
However, a study in 2018 identified Ae. albopictus at a higher latitude near the city of Garoua at 9.3 N for the first time (Rodrigue Simonet et al., 2020), which is in line with global distribution predictions (Kraemer et al., 2015).
The factors influencing the geographical parameters of Ae. albopictus in Cameroon have not been examined but may be related to a combination of demographic (i.e., population and urbanization) and environmental (land cover, vegetation, temperature and rainfall) characteristics (Dickens et al., 2018;Kraemer et al., 2015), especially as their distribution may be affected in the future by a changing climate (Kamal et al., 2018). A better understanding of these risk factors may help predict new areas vulnerable to establishment of Ae.
albopictus and direct the management of vector control as a means to prevent arboviral diseases. Therefore, the aim of the paper was to extend the work on the recent update  and develop a geo-referenced vector database based on all publicly available records of Ae. albopictus, and to identify the potential demographic and environmental risk factors associated with the distribution in Cameroon.

Vector data sources
First, a geo-referenced database on Ae. albopictus in Cameroon was developed. Information from a review up to 2010 was collated (Kraemer et al., 2015) and then a comprehensive literature search was conducted using the online sources of PubMed and Web of Science to identify additional articles with data collected between 2000 and 2020. Search terms, and combinations thereof, included Cameroon, arbovirus, Aedes, Ae. albopictus, Ae. aegypti, dengue and chikungunya (Additional material 1). Further articles were sought from references listed within the identified articles, and then from references within those articles.
For each article the following information was collated into a database; name of collection site (i.e., village, town) as stated in the article; the local administrative area as defined by recent boundaries obtained from United Nations Office for the Coordination of Humanitarian Affairs office (OCHA) in Cameroon; the year of data collection (based on study start month); latitude and longitude of study site; vector presence or absence; and publication information.

Risk factor data sources
To examine demographic and environmental factors associated with distribution of Ae. albopictus, data from remote sensed satellitederived sources were used. Specifically, data were identified and examined in relation to four broad risk categories: human geography, landscape, vegetation, and climate, and the specific variables and data sources are listed in Table 1.

Mapping and analysis
All geo-referenced data were imported into the mapping software QGIS 3.16 (https://qgis.org/). First, all collection sites with information on the presence or absence of Ae. albopictus were mapped into two different decades (2000-2010 and 2011-2020) to examine patterns over time, using the available latitude and longitude coordinates.
If the paper did not report the sampling coordinates, the authors generated the centroid coordinates of the lowest level administrative boundary associated with the study areas in the paper. In Cameroon, the lowest level administrative boundaries were sourced from OCHA with a median area of 615 km 2 .
Due to the different spatial resolution and the probable inaccuracy of some data points, the authors created a 3 km buffer around each point using the QGIS geoprocessing spatial tool, and risk factor data were extracted within the buffer using the Zonal Statistics Raster Analysis tool. For each variable, the values were aggregated within the buffer for the analysis.
The remote sensed datasets were processed in Google Earth Engine. All environmental data were exported to statistical software R 4.0.1 (https://www.r-project.org/) for descriptive and statistical analysis. To explore which variables were associated with the presence/ absence of Ae. albopictus, the authors compared them using the Mann-Whitney non-parametric test (p ≤ 0.05 significance and corrected by Bonferroni). Then, a binomial logistic regression model was constructed using variance inflation factor (vif) to determine multicollinearity (excluded vif >5) of variables and stepwise analysis by AIC to select among the 13 variables collected for the final model.

Collation of Ae. albopictus data
In total, 12 articles reporting information on the presence or absence of Ae. albopictus from 111 geo-referenced sites were identified and summarized in Table 2

Comparisons of environmental measures
After correcting for Bonferroni, there were significant differences between Ae. albopictus presence and absence sites for all the climate variables and for canopy height and forest loss ( Table 3). The most dominant land cover classifications for Ae. albopictus presence sites were Savanna (n = 35) and Urban (n = 29) and for Ae. albopictus absence sites were Croplands (n = 15) and Savanna (n = 5) (data not shown).

Model
There was a high level of multicollinearity among the 13 original variables. Using the variance inflation factor (vif), the authors excluded five variables in the first step. Then five more variables were removed when the authors applied a stepwise process to select the best model based on the AIC. The best-fit model included number of wet months, the percent of peri-urban land and altitude as explanatory variables (   We found the number of wet months and percentage of periurban area around the sampling site to be most associated with Ae. albopictus presence in Cameroon. Similar variables have previously been found to be strong predictors of Ae. albopictus presence (Dickens et al., 2018;Juliano et al., 2002). More wet months provide more opportunities for breeding sites as this vector thrives in rainfilled containers, in particular used tyres and naturally formed cavities, such as tree holes (Bonizzoni et al., 2013;Egid et al., 2021;Paupy et al., 2009;Ryan et al., 2018). The northern part of Cameroon is drier than the rest of the country potentially making it more difficult for Ae.
albopictus to maintain populations. By contrast, Ae. aegypti has been reported in all the sites where Ae. albopictus was absent, indicating that Ae. aegypti might be more tolerant to drier environments (Juliano et al., 2002;Kamgang et al., 2010) or easily maintained by human water storage activities (Egid et al., 2021;Kraemer et al., 2015).
The present study's findings on the relationship between percentage of peri-urban areas and presence of Ae. albopictus agree with other studies (Kamgang et al., 2017;Mayi et al., 2020;Paupy et al., 2010). An increase of 1% of peri-urban area increases by 1.3Â the likelihood of the vector presence. Aedes albopictus is an ecological flexible and opportunistic vector (Bonizzoni et al., 2013). Consequently, peri-urban areas are ideal places for them to thrive as they originated at the fringe of forests but adapted to breed in artificial containers, and they normally feed on animals and humans, making the interface between urban and rural the perfect ecological habitat (Bonizzoni et al., 2013;Egid et al., 2021;Kamgang et al., 2012;Mayi et al., 2020;Paupy et al., 2009).
A limitation of this study is the variation in the study design of the data collated. Collection of vector data (immature stages or adults) was done using different methods, for instance, for adult mosquitoes the following methods were used: human landing catches (Akono  Simonet et al., 2020;Tedjou et al., 2020). The authors were able to conduct this analysis due to the advancement and availability of satellite data. However, the urbanization product, which classifies land into urban, peri-urban and rural, was available only for 2 years, 2000 and 2015. This could be important to understand the process of urbanization in the northern part of the country, where Ae. albopictus is expanding. In addition, authors (Kamgang et al., 2017;Tedjou et al., 2019Tedjou et al., , 2020) use peri-urban or suburbs to characterize the area of vector survey, but most did not adequately define the meaning or main characteristics of peri-urban areas, which may differ between sites and co-author definitions. In contrast to other studies (Brady et al., 2014;Kraemer et al., 2015), the authors did not find temperature to be a significant predictor. This is likely due to the restricted geographic range of this analysis compared with previous studies at the global level (Brady et al., 2014;Kamal et al., 2018;Kraemer et al., 2015;Ryan et al., 2018).
Aedes albopictus is well known to have a high tolerance for colder and mild temperatures (Juliano et al., 2002;Marini et al., 2020;Paupy et al., 2009;Thomas et al., 2012). This only becomes apparent when including sites in more extreme latitudes in the analysis, which allows Ae. albopictus to extend its range into Europe and North America (Hopperstad et al., 2021;Takumi et al., 2009) Social Affairs, 2019). However, as seen in this study, even in the same country, data collection is not performed in the same location over time.
Thus, evaluation of the impact of land use and land cover at fine scale at the same location over time is needed to understand the impact on vector distribution and helps to facilitate control strategies.

CONCLUSION
This is one of the first studies to quantify the influence of human and physical geography on Ae. albopictus in SSA. The length of wet months and the extent of peri-urban areas were found to be important risk factors in Cameroon, which provides some insights into the local vector ecology. However, for the national arbovirus disease control programmes to move forward, more standardized methodologies over time and space are needed to be able to better quantify the risk factors. This may facilitate appropriately targeted public health messages and vector control strategies.

CONFLICT OF INTEREST
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

DATA AVAILABILITY STATEMENT
All datasets used for this work are publicly available, in the manuscript or in the additional material.