Risk factors for occurrence and abundance of Aedes aegypti and Aedes bromeliae at hotel compounds in Zanzibar

A field survey was performed to investigate local environmental factors promoting occurrence and abundance of Aedes aegypti and Ae. bromeliae mosquitoes at hotel compounds in the south-east coastal region of Zanzibar Island. The potential risk factors were determined using generalized linear mixed models. Aedes (Stegomyia) spp. indices such as container index (CI) and pupae per container (PPC) index were also estimated. Aedes aegypti and Ae. bromeliae were the most abundant vector species, accounting for 70.8% of all Aedes mosquitoes collected. The highest CI was observed for plastic containers irrespective of the season, whereas the highest PPC was observed for coconut shells and aluminium containers in the rainy and dry seasons, respectively. The risk of Aedes mosquito occurrence and abundance were significantly associated with presence of plastic containers, coconut shells, used tyres and steel containers. These were discarded in shaded places, in the open and gardens, or found in plant nurseries. This study shows that Aedes species of global health significance occur at hotel compounds on this part of Zanzibar Island. The occurrence and abundance are sustained by the presence of abundant and poorly managed solid wastes and containers used for gardening tasks. This highlights an urgent need for the adoption of area-wide environmentally sustainable Aedes mosquito management interventions that also integrate solid waste management and ornamental plant production practices for reducing the risk of arboviral disease epidemics.

However, other Aedes species are highly efficient at transmitting life-threatening arboviruses between humans or between other primates and humans. This includes the more poorly investigated native species, such as Ae. bromeliae, Ae. africanus, Ae. furcifer, Ae. luteocephalus, Ae. metallicus, Ae. opok, Ae. vittatus and Ae. simpsoni sensu stricto [11]. Unprecedented urbanization coupled with poor solid waste management, movement of people and commodities and, not least, climate change are considered main drivers of vector and pathogen spread [4,12,13]. Moreover, it is argued that climate change could cause a shift in the burden of mosquito-borne diseases in Africa from malaria to arboviral diseases, as temperature increases may expand the environmental suitability for transmission of dengue and several other vector-borne arboviral diseases on the continent [14].
Mainland Tanzania has observed a surge in Aedesborne arboviral disease transmission, with six dengue outbreaks recorded within the past 10 years. The most recent outbreak, in 2019, spread from Dar-es-Salaam to Dodoma, Morogoro, Pwani, Singida and Tanga, resulting in 6670 confirmed cases and 13 deaths [15]. Similarly, active co-circulation of dengue and chikungunya viruses has been reported from other regions of the mainland [16][17][18] and neighbouring countries such as Kenya [19]. Currently, there is no entomological evidence of arboviral disease transmission in the Zanzibar Archipelago. However, recent serological surveys have suggested that dengue virus may be circulating in the local population [20][21][22]. The occurrence of abundant Ae. aegypti and Ae. bromeliae across urban and rural areas of Zanzibar Island (also known as Unguja) has also been recently confirmed [23][24][25]. Infestation of Zanzibar Archipelago by these vectors is a serious threat, as arboviral epidemics will have devastating impacts on the fragile public health system and the tourism-based economy. Tourism contributes 27% of the Zanzibar GDP and 80% of foreign revenue [26]. However, the tourism industry is also the most important generator of poorly managed non-biodegradable solid waste in Zanzibar [27]. Some of these solid wastes can sustain a higher diversity of epidemiologically relevant mosquito taxa [23]. In addition, hotels may concentrate visitors and staff moving to and from different highly endemic regions, which increases the likelihood of contact between local vectors and infectious hosts carrying new variants of mosquito-borne arboviruses. For instance, nearly 722,000 tourists arrived in Zanzibar Archipelago between 2019 and 2020 from mainland Tanzania [28], considered a high-risk pathway for pathogen importation into the archipelago [29,30]. Introduction of new variants of viruses may cause large and uncontrolled epidemics, as observed in recent outbreaks of CHIKV in Africa and Asia [3,31]. A similar situation was observed for YFV in Brazil in 2018 [32]. Apart from YFV, there is no widely available, safe and effective vaccine against the main mosquito-borne viruses, nor are there any specific antiviral treatments available for the management of disease cases [2,10]. Therefore, vector control remains the recommended measure to limit mosquito-borne diseases [10,33]. In Zanzibar, hotels implement periodic blanket spraying with residual-effect insecticide to reduce mosquito bite exposure. This approach, apart from causing widely known environmental repercussions [34], exerts selective pressure on local vector species to develop resistance to common classes of insecticides, as recently confirmed for Ae. aegypti populations found at different hotels in Zanzibar (Kampango et al. [23], unpublished). This indicates that non-chemical and environmentally sustainable control practices targeting mosquito sources should be implemented to reduce the risk of arbovirus infection exposure. The design and implementation of cost-effective environmental control approaches require accurate and thorough characterization of key environmental factors favouring the establishment and maintenance of vector populations in and around hotel settings. Hotels can be ideal locations to test the feasibility and efficacy of novel non-chemical environmental interventions before large-scale implementation as they have well delimited and fully managed environments.
The main goal of this study was to determine local environment-associated risk factors for occurrence and exposure to Aedes mosquito vectors found at hotel compounds on the eastern coastal region of Zanzibar Island. Findings from this study will expand the knowledge on the entomological profile of potential arbovirus vectors occurring in Zanzibar. Moreover, they may support the implementation of a comprehensive mosquito surveillance system and integrated arbovirus vector management, since hotels in the region are not currently covered by the public surveillance system and routine vector control campaigns.

Description of study sites
The study was carried out at four selected hotels in the south-eastern coastal region of Zanzibar Island (Fig. 1). The hotels, previously described in Kampango et al. [23], were selected according to compound size (total residential and non-residential area not less than one hectare), accessibility by local means of transportation during low and high tourism seasons, willingness to share data and willingness to accept publication of findings. Consent was obtained from hotel management to conduct the study on their properties. For privacy reasons, hotel names are anonymized, and Hotel A, B, C and D used as identifiers. The largest hotel (Hotel B) occupied an area of approximately 28.06 hectares, whilst the remaining hotels encompassed an approximate area of 6.6 hectares (Hotel A), 3.6 hectares (Hotel D), and 2 hectares (Hotel C). The rainfall regime of Zanzibar is divided into two main rainy seasons, the long rainy season (Masika) usually extends from mid-March to June, and the short rainy season (Vuli) from November to December. Average monthly precipitation ranges from 30 mm in the cold season (July) to 320 mm in the hot season (December) with the accumulated annual rainfall reaching 1600 mm [35]. The relative humidity is usually high, with a monthly average ranging from 87% during the long rainy season (March to June) to 76% in November and December (during the short rainy season), and reaching a minimum of 60% during the dry season (January to March and July to October) [36].

Mosquito collection
Aedes mosquito immature stages were sampled for a period of 10 months from September 2018 to October 2019, encompassing both the rainy and dry seasons. At each hotel, mosquitoes were surveyed over 4 to 6 days in each month using protocols proposed by Tun-Lin et al. [37] and Manrique-Saide et al. [38], with small adaptions as previously detailed in Kampango et al. [23]. For relatively small containers [water volume less than 1 l up to 5 l, e.g., plastic bottles and beer or soda cans (metal containers)], all larvae and pupae were sampled using pipettes. For large containers (water volume > 5 l, e.g., buckets, jerry cans and ceramic pottery) with relatively few specimens, all larvae and pupae were sampled using dippers, sweep nets, or bowls. For the largest containers (water volume > 20 l, e.g., water tanks, wells and septic tanks), specimens were collected by 10-15 random dips/ sweeps. If a container was too deep (e.g., wells) samples were collected using small, suspended buckets (approx. 5 l), filling ten buckets from different sites of the water container. Containers without mosquito larvae or pupae were considered as potential larval habitats for any local mosquito species if they had water and physical integrity capable of retaining collected water for at least three consecutive days, the average time required for larvae of several Afrotropical mosquito species to hatch from eggs in the wild [39]. Larval habitats were characterized according to type, function, location, size, sun exposure, presence of vegetation, presence of organic matter and season (Table 1).

Sample processing and identification
Samples of immature mosquitoes were reared at insectary environmental conditions of 27 ± 2 °C and 75 ± 10% relative humidity until they emerged as adults [40]. Adult mosquitoes were identified morphologically to species level using taxonomic keys for Afrotropical mosquito fauna [11,41].

Statistical analysis Aedes entomological indices
The following entomological indicators for Aedes (Stegomyia) spp. were estimated: container index (CI), pupae per container (PPC) and proportion of positive containers (PC). CI was estimated by dividing the total number of positive containers of a given type against the total number of all inspected containers. The PPC was estimated by dividing the total number of pupae collected from a given type of container by the total number of containers inspected, whereas the PC index was estimated as the quotient between the total number of a type of positive containers by the total number of all positive containers found.

Modelling the risk factors for Aedes mosquitos
Mosquito immature counts had excess zeros as the overall proportion of zero counts was over 30%. The counts were also over-dispersed, showing a positive meanvariance relationship [42]. Therefore, we applied zeroinflated negative binomial mixed models (ZINBM) with log link function, to determine local factors contributing to the abundance of Ae. aegypti and Ae. bromeliae at study sites, while accounting for excess zeros and overdispersion [43]. Dependent variables were Ae. aegypti and Ae. bromeliae counts. On the other hand, mixed effects logistic regression was applied to determine the risk factor for presence of any Aedes species in general. The dependent variable was the presence of Aedes immatures (larvae and pupae) (Yes = 1, No = 0). For this case, we assumed that the distribution of presence and absence of Aedes immature records followed a binomial distribution with logit function. For ZINBM and mixed effects logistic regression models, the potential risk factors considered were type of container, location, function, size, sun exposure, presence of vegetation and season (Table 1). Organic matter was always present in containers with immature Aedes, so it was not included in the model. Hotel and day of survey were considered random factors to account for dependence between repeated measurements across sites over time. A likelihood ratio test comparing the null model (model with dependent variable only) and model with the random factor was applied to determine whether adding a random factor was justifiable. Multicollinearity between predictors was evaluated using the variance inflation factor (VIF). VIF estimates less than or equal to three indicated low collinearity [44].
Variables showing high collinearity were fitted separately with other uncorrelated predictors. Model distributional assumptions were determined by examining model residuals for normality, dispersion and independence (no autocorrelation among residuals). Models were fitted using the package glmmTMB v. 1.0.2.9. [45]. To deal with convergence problems, we combined all underrepresented levels of predictor variables, comprising less than 1% of all level observations, and denoted them as "other". We assumed that these underrepresented levels would introduce more information to the model combined rather than alone. Model diagnostics were performed using the package DHARMa v. 0.3.3.0 [46]. Multicollinearity between covariates was determined using the package performance v.0.6.1 [47]. All data processing tasks and analysis were performed via the software R v. 4.2.0 [48].

Aedes populations and larval habitats
A total of 1466 water-holding containers were inspected out of which 763 were found positive for Aedes mosquitoes (larvae and pupae). A total of 19,533 Aedes mosquitoes were collected, comprising 16 (Table 2 and Additional file 1: Table S1).
Plastic containers (mostly bottles and yoghurt cups) were by far the most frequent type of habitat found with larvae and/or pupae. They constituted 69.9% (372/532) and 71.4% (165/231) of all positive habitats recorded in the rainy and dry seasons, respectively, followed by coconut shells, with 7.4% (38/532) versus 3.4% (8/231), metal containers at 4.7% (25/532) versus 4.3% (10/231), and Sun exposure Not exposed, exposed half the day, exposed more than half the day Amount of time the habitat has been exposed to sunlight   used tyres at 4.5% (24/532) versus 1.7% (4/231). There was also a group of underrepresented larval habitats, such as coconut flower spathes, broken appliances and bottle caps, denoted as "other", which made up nearly 3.6% (19/532) versus 3.5% (8/231) of all habitat types ( Table 2, Fig. 2). The highest relative CI was found for plastic containers irrespective of the season. The density of PPC was highest for steel containers (PPC = 5.42) and coconut shells (PPC = 5.38) in the rainy season. In the dry season, the highest PPC was observed for aluminium containers (PPC = 10.0) and used tyres (PPC = 5.10) (  Table S1).

Risk factors for occurrence of Aedes mosquitoes
The mixed effects logistic regression model indicated that natural and artificial containers such as used tyres, metal containers (e.g., soda cans), plastic containers (e.g., water bottles, jerry cans) and coconut shells were more likely to harbour at least one immature stage of any of the identified Aedes mosquitoes compared to other types of water-holding containers (Fig. 3). The likelihood of immature mosquito occurrence was not statistically different between locations within the compounds. It also did not differ according to container function or purpose (Fig. 3). Additionally, containers of medium size and containers that were less exposed to sunlight were less likely to be infested than larger containers or those receiving sunlight for more than half the day (Fig. 3). Figure 4 indicates that abundance of Ae. aegypti was, in order of magnitude, significantly associated with coconut shells, plastic containers, used tyres and metal containers when compared with other types of containers. The abundance was significantly higher at plant nurseries and garden/open spaces compared to other locations inside the premises. Larval habitats associated with the highest Ae. aegypti abundance were mostly discarded items and containers used in gardening tasks and exposed to sunlight for less than half of the day (Fig. 4). Habitats with and without vegetation produced statistically comparable mosquito abundances. In contrast, medium-and largesized containers and the dry season were associated with lower abundance of Ae. aegypti (Fig. 4). Equally, Ae. bromeliae was highly abundant in discarded plastic containers found in gardens/open spaces and plant nurseries, and those exposed to sunlight for nearly half of the day (Fig. 5).

Discussion
Understanding the contribution of environmental factors promoting occurrence and abundance of Aedes mosquitoes across space and time is pivotal for the design of cost-effective sustainable vector control interventions. This could involve raising local awareness about solid waste disposal to reduce the risk of mosquito-borne disease exposure and epidemics. Our data clearly indicate that some hotel compounds in Zanzibar provide suitable environmental conditions for establishment of Ae.
aegypti and Ae. bromeliae, and other less frequent Aedes mosquito vector species such as Ae. vittatus, Ae. africanus and Ae. metallicus. We found that occurrence and abundance of these vector species are significantly associated with the presence of improperly managed natural and artificial solid wastes and ornamental containers. A great number of studies have reported strong associations between the abundance of Aedes mosquitoes and their larval habitats and poorly managed disposables as well as decorative containers [49][50][51]. These materials can hold water and organic debris for long periods, providing relatively stable micro-environmental and microclimatic conditions for proliferation of diverse mosquito species [24,25,51,52]. We also found high productivity of Aedes mosquitoes, particularly Ae. aegypti and Ae. bromeliae, in natural and artificial containers such as coconut shells, plastic containers, steel containers and used tyres. These types of habitats contributed 92.6% and 93.9% of all Aedes larvae and pupae collected, respectively. This finding somewhat corroborates that of Saleh et al. [24,25], who reported higher risk of Ae. aegypti exposure associated with plastic containers and used tyres across rural and urban communities in Zanzibar. However, in contrast to Saleh et al. [24,25], concrete tanks were one of the least productive habitats in our study sites, contributing only 0.64% of all mosquitoes collected. Plastic containers had among the lowest density of pupae despite being the most common type of larval habitat found at the sites that we sampled. This finding suggests that plastic containers are less suitable for sustaining the development of mosquito larvae to the pupal stage. Some studies of Ae. aegypti have also reported lower productivity for plastic than other types of containers [53,54]. The factors determining Ae. aegypti female choice of certain larval habitat types remain poorly delineated. However, visual stimuli may play an important role, as females of Ae. aegypti are more attracted to lower reflectance colour targets [55]. As such, the likelihood of detecting Ae. aegypti immatures in darker-coloured containers may be high compared to lighter-coloured containers. Moreover, variation of physicochemical characteristics of the contained water can also affect mosquito detection, as these factors influence both development and survival of mosquitoes at earlier stages. For instance, it has been shown that the average diurnal water temperature in plastic containers can reach up to 35.6 ºC, exceeding the thermal optima range of 15-35 ºC for the development of Ae. aegypti immatures [56,57]. This suggests that plastic containers may become unsuitable larval sites depending on their daily amount of sunlight exposure. On the other hand, higher levels of pH and salinity may also be limiting factors [58]. However, Saleh et al. [24] found no clear association between variations of physicochemical factors and presence of Ae. aegypti in Zanzibar City. Further studies on ecological and physiological determinants of the occurrence and abundance of Aedes mosquito immature stages in larval habitats are encouraged. Mixed effects logistic regression and zero-inflated negative binomial mixed models revealed used tyres, plastic containers, coconut shells and metal containers (e.g., beer, soda and food cans) as the types of containers most associated with higher risk of Ae. aegypti and Ae. bromeliae presence and abundance, respectively. However, the likelihood of infestation was comparable between locations and type of container function or purpose. In contrast to the risk of infestation, the overall abundance of Aedes mosquitoes was up to nine times as high (e.g., An. bromeliae) in gardens/open spaces and plant nurseries compared to other locations. This finding suggests that these locations should be prioritized when implementing mosquito control interventions. Moreover, the result also indicates that mosquito management practices should be incorporated into ornamental plant production and trading activities, as they may contribute tremendously to maintenance and rapid dissemination of native and invasive vector species throughout the archipelago. Notably, invasive vector species such as Ae. albopictus have been successfully introduced and spread worldwide by the trade of ornamental flowers [59,60]. Higher mosquito abundance was consistently associated with discarded materials and containers used in gardening activities than for other purposes and/or functions. Similar findings were reported for community settings in Zanzibar [25] and Argentina [53]. Furthermore, containers less exposed to sunlight, usually found in densely vegetated places, were more likely to be infested and to have higher mosquito abundance. Similar findings have been reported in other parts of Zanzibar and elsewhere, suggesting that targeted environmental manipulation of shaded places should be prioritized in efforts to control Aedes mosquitoes. Virtually all containers found positive for either Aedes larvae or pupae contained organic matter detritus. It is well documented that dissolved organic matter provides conditions for development of microorganisms that can act as food for mosquito larvae. However, most recent evidence also shows that dissolved organic matter plays a key role in protecting mosquito immature stages from the lethal effects of solar ultraviolet radiation [61].
The association between poorly managed solid waste and high risk of Aedes infestation and abundance in public [24,25] and private settings (this study) provides evidence for the design and implementation of sustainable environmental mosquito control practices. This includes integrated and consistent solid waste management strategies, periodic inspection and substitution and drainage of ornamental plant water, and periodic waste removal campaigns, so as to reduce the risk of Aedes-borne arbovirus epidemics. Taking these actions is essential given recent evidence for resistance in Ae. aegypti from Zanzibar to some classes of insecticides approved for use in public health (Kampango et al., in progress). Additionally, this finding is also a call for implementation of strong legislation and regulation leading to environmentally safe waste disposal and treatment. This study was carried out in a relatively small number of hotels, which prevents us from making broader generalizations. Similar types of investigations involving hotels in other eco-geographical regions of Zanzibar may provide a more comprehensive understanding of risks factors for mosquito-borne arbovirus exposure at hotels in the archipelago.

Conclusion
We conclude that unmanaged solid waste from artificial and natural sources is the most important risk factor for the occurrence and abundance of the arbovirus vectors Ae. aegypti and Ae. bromeliae at hotel compounds investigated in this study. Notably, the most important sources of Aedes mosquitoes are disposable plastic containers (water bottles, yoghurt cups, jerry cans, etc.), used tyres and metal containers (e.g., soda, beer and food cans), and natural products such as coconut shells. Future vector control interventions should target these types of larval habitats, especially in shaded places in garden areas, plant nurseries and in the surrounding communities. Findings from this study underscore the need for development and implementation of area-wide mosquito control interventions that incorporate waste management practices, vegetation manipulation and community mobilization for environmental sanitation. Hotels present a unique opportunity to test the costeffectiveness of these integrated control approaches, given the human resources, the financial incentives and structured/systematic management of the compounds. We propose that future eco-certification/classification of tourism facilities could include implementation of area-wide non-chemical environmental management practices to reduce mosquito proliferation as an indicator, since mosquito infestation rates remain high in spite of periodic mosquito control activities involving the use of chemical insecticides.