Urban Green Spaces and Vector-Borne Disease Risk in Africa: The Case of an Unclean Forested Park in Libreville (Gabon, Central Africa)

In Africa, vector-borne diseases are a major public health issue, especially in cities. Urban greening is increasingly considered to promote inhabitants’ well-being. However, the impact of urban green spaces on vector risk remains poorly investigated, particularly urban forests in poor hygienic conditions. Therefore, using larval sampling and human landing catches, this study investigated the mosquito diversity and the vector risk in a forest patch and its inhabited surroundings in Libreville, Gabon, central Africa. Among the 104 water containers explored, 94 (90.4%) were artificial (gutters, used tires, plastic bottles) and 10 (9.6%) were natural (puddles, streams, tree holes). In total, 770 mosquitoes belonging to 14 species were collected from such water containers (73.1% outside the forested area). The mosquito community was dominated by Aedes albopictus (33.5%), Culex quinquefasciatus (30.4%), and Lutzia tigripes (16.5%). Although mosquito diversity was almost double outside compared to inside the forest (Shannon diversity index: 1.3 vs. 0.7, respectively), the species relative abundance (Morisita–Horn index = 0.7) was similar. Ae. albopictus (86.1%) was the most aggressive species, putting people at risk of Aedes-borne viruses. This study highlights the importance of waste pollution in urban forested ecosystems as a potential driver of mosquito-borne diseases.

This study was carried out from 14 July to 20 August 2020, during the long dry season (which lasts from June to September), in the Sibang arboretum (0 • 24 58 N and 9 • 29 23 E) and its inhabited surroundings ( Figure 1). The study was carried out during the dry season because we suspected that this wooded area may serve as a refuge for mosquitoes during the period of low rainfall, and we wanted to identify the mosquito communities and the vector risk associated with this ecosystem. The Sibang arboretum is a forest park that covers 160,000 m 2 . It is located in an urbanized area of the eastern part of Libreville and is crossed by the Adoung river. Cordier (2000) recorded 137 plant species and at least 40 bird species, and a poorly quantified diversity of vertebrate animals including reptiles and small mammals (e.g., squirrels) [28]. Despite its protected status, the Sibang arboretum is subject to an important anthropogenic pressure, as indicated by the presence of diverse traces of human frequentation and the accumulation of rubbish at some locations up to several tens of meters inside the arboretum, especially along one major border separating the arboretum from the surrounding households ( Figure 2).

Larval Sampling
Larval samplings were carried out during 13 non-consecutive days outside (ove radius of 800 m around the forest) and inside the forest. Mosquito breeding sites we investigated as exhaustively as possible, in the function of the field accessibility and p mission from residents to visit their properties. During the collection period, the sampli effort lasted 15 h and 63 h inside and outside the forest, respectively.
Water containers were explored at the different sites ( Figure 1C) in order to coll larvae and pupae using a dipper or a pipette and transfer them into vials labeled accordi

Larval Sampling
Larval samplings were carried out during 13 non-consecutive days outside (over a radius of 800 m around the forest) and inside the forest. Mosquito breeding sites were investigated as exhaustively as possible, according to the field accessibility and permission from residents to visit their properties. During the collection period, the sampling effort lasted 15 h and 63 h inside and outside the forest, respectively.
Water containers were explored at the different sites ( Figure 1C) in order to collect larvae and pupae using a dipper or a pipette and transfer them into vials labeled according to the container type, site location, and date. At the entomological laboratory of the Research Institute for Tropical Ecology (IRET), Libreville, larvae and pupae were placed into labeled trays covered with a mosquito net and maintained at room temperature until the emergence of adults. All samples were treated in the same conditions, from collection to rearing, to minimize the bias in abundance and diversity among sites. Upon emergence, adult mosquitoes were kept at −20 • C for 30 min to be euthanized, and then morphologically identified (species or genus) using a binocular microscope (Leica Microsystems©) and "customized" taxonomic keys based on the updates of the Edwards' identification keys for Ethiopian mosquitoes [29], and the Huang's key for the subgenus Stegomyia of Aedes mosquitoes from the Afrotropical region [30]. Species were named according to the online list of valid species (http://mosquito-taxonomic-inventory.info, accessed on 20 July 2020).

Adult Mosquito Collection
Adult female mosquitoes were collected using the human landing catch (HLC) technique during the daytime. The study was approved by the Gabon Ethics Committee (permit No. 016/2019/PR/SG/CNE). Three volunteers, posted at three fixed capture sites, collected adult females during two consecutive sampling sessions, one inside and one outside the forest. At each session, the three volunteers were separated by at least 50 m ( Figure 1C). Each sampling session was for three consecutive days, from 10:00 a.m. to 2:00 p.m. (4 h per day), representing a sampling effort of 12 h. Mosquitoes were captured with a mouth aspirator upon landing on the volunteer's bare legs and then transferred into a plastic jar covered with a net to prevent their escape. At the end of the day, mosquitoes were transported to the IRET entomological laboratory for identification, as described above.

Data Analysis
All statistical analyses were carried out using the R software v3.6.1 (https://www. r-project.org/, accessed on 1 September 2020). Spatial analyses were performed using Quantum GIS version 3.10.7 (https://www.qgis.org/, accessed on 15 September 2020). Species richness was determined as the number of mosquito species recovered. Species diversity (i.e., number of species and their abundance) was assessed using the Shannon-Weaver index (H) [31] and the "diversity" function of the vegan package. To investigate the similarity in terms of species composition and the density between mosquito communities inside and outside the forest, the Morisita-Horn similarity index (C) [32] was calculated using the "vegdist" function of the vegan package. Because "vegdist" is an analysis of dissimilarity (C ), C = 1 − C was used for this study. C ranged from 0 (0% of similarity between compartments) to 1 (100% of identity between compartments).
Environmental variables were collected to characterize the larval habitats exploited by mosquitoes in the Sibang arboretum and its surroundings. These variables included the substrate physical description, the type (artificial vs. natural), and the spatial location of larval habitats (inside vs. outside the forest). Multiple Correspondence Analysis (MCA) was used to assess the similarity of larval habitats according to the species composition and environmental characteristics. It allowed the assessment of the mosquito species' degree of specificity related to the larval habitat type and location. The MCA was also used to identify the most relevant biotic and environmental variables associated with the larval habitat segregation.
A k-means analysis based on Ward's method was performed using the fpc package [33] and the Calinski Harabasz index (CH index) [34] to determine the minimal parsimonious number of ecological species clusters. Based on the results of the Principal Component Analysis (PCA) using the FactoMineR package [35], Hierarchical Agglomerative Clustering (HAC) was used to determine and visualize species clusters within the same microecological niche.
The Wilcoxon's test based on the HLC data was used to assess the aggressiveness of bloodmeal-seeking female mosquitoes according to the sampling location (inside vs. outside the forest). An analysis of variance (ANOVA) was performed to determine the differences in the number of captured specimens per person among species in each sampling location.

Larval Habitat Typology, Similarity and Species Clustering
Ae. albopictus was mainly found in artificial breeding containers. Additionally, this species was associated with a higher relative abundances in both compartments compared with the other species (Table 2). Similarly, Cx. quinquefasciatus was mostly recovered in artificial breeding containers (13.8% and 35.1% inside and outside the forest, respectively). Conversely, Lu. tigripes was relatively more abundant in natural containers (22.7% and 28.6% inside and outside the forest, respectively) ( Table 2). Ae. aegypti was only recovered in artificial breeding containers and outside the forest. An. gambiae s. l. was almost exclusively found in artificial breeding sites outside the forest (6.5%) ( Table 2).
The MCA revealed that the best-correlated variables associated with the larval habitat distribution were the habitat type (natural or artificial) and spatial location (inside or outside the forest), and also the presence of the species Ae. albopictus, An. gambiae s. l., Ae. aegypti, Lu. tigripes, Cx. quinquefasciatus, Cx. trifilatus, Cx. duttoni, and Er. inornatus ( Figure S1). These variables explained 38.4% of the total variance associated with the species composition of larval habitats. This analysis showed a clear segregation of larval habitats according to their spatial location and type ( Figure 4A,B). In terms of larval habitat specificity, our results revealed that mosquito larval habitats were likely to be exploited by species that could be characterized as specialist (i.e., with a high level of habitat specificity: Ae. aegypti, An. gambiae s. l., Cx. trifilatus, Cx. duttoni, and Er. inornatus), opportunistic (i.e., species that, unlike specialist species, can adapt to a range of environmental conditions: Ae. albopictus and Lu. tigripes) or ubiquitous (i.e., species highly adapted to occupy and proliferate in varied ecological niches, possibly with a wide geographical distribution: Cx. quinquefasciatus) ( Figure 4C-J). The most exploited breeding sites by Ae. albopictus (opportunistic species) and by Cx. quinquefasciatus (ubiquitous species) were discarded plastic containers and worn tires/metallic or concrete containers (i.e., wash basin, see Table 1), respectively ( Figure 5). Lu. tigripes (opportunistic species) used discarded tires, puddles, plastic, metallic or concrete containers ( Figure 5). Ae. aegypti and An. gambiae s. l. (the main vectors of public health concern) exclusively exploited plastic containers and puddles (both natural and artificial), respectively, as breeding sites ( Figure 5).
To determine the degree of ecological niche similarity among mosquito species at the larval microhabitat scale, the k-means method based on the PCA (72.8% of the total explained inertia of count data over the first two principal components, see Figure S2) and the CH index revealed that the minimal parsimonious number of ecological clusters of species was four ( Figure 6A). Two of these clusters were mono-specific (cluster 2: Ae. albopictus; cluster 3: Cx. quinquefasciatus) and two were multi-specific (cluster 1: Ae. aegypti, Tx. evansae, Cx. umbripes, Cx. decens, Cx. trifilatus, Cx. cinerellus, Cx. univittatus, Cx. duttoni, Culex sp., Er. inornatus; cluster 4: Cx. tauffliebi and Lu. tigripes) ( Figure 6B). Our analysis showed that species in the same cluster were more likely to share the same type of larval ecological niche. quinquefasciatus) (Figure 4C-J). The most exploited breeding sites by Ae. albopictus (opportunistic species) and by Cx. quinquefasciatus (ubiquitous species) were discarded plastic containers and worn tires/metallic or concrete containers (i.e., wash basin, see Table 1), respectively ( Figure 5). Lu. tigripes (opportunistic species) used discarded tires, puddles, plastic, metallic or concrete containers ( Figure 5). Ae. aegypti and An. gambiae s. l. (the main vectors of public health concern) exclusively exploited plastic containers and puddles (both natural and artificial), respectively, as breeding sites ( Figure 5). The MCA explained 38.4% of the total variability of larval habitats in terms of species composition. These variables include spatial location (A) and type (B) of larval habitats, as well as the main species recovered (C-J). Intersection areas refer to habitats that tend to be similar in type (natural/artificial), location (inside/outside forest), and specific composition. The MCA explained 38.4% of the total variability of larval habitats in terms of species composition. These variables include spatial location (A) and type (B) of larval habitats, as well as the main species recovered (C-J). Intersection areas refer to habitats that tend to be similar in type (natural/artificial), location (inside/outside forest), and specific composition.

Mosquito Communities in the Urban Forested Area of Sibang
This entomological survey described the species diversity and the larval microhabitat typology of mosquito communities in an urban forested reserve and in its direct surroundings (within the city of Libreville, Gabon) to evaluate whether this environment full of waste influenced the VBD risk. Most of the surveyed potential breeding sites (≥95%) inside

Mosquito Communities in the Urban Forested Area of Sibang
This entomological survey described the species diversity and the larval microhabitat typology of mosquito communities in an urban forested reserve and in its direct surroundings (within the city of Libreville, Gabon) to evaluate whether this environment full of waste influenced the VBD risk. Most of the surveyed potential breeding sites (≥95%) inside and outside the forest area (i.e., tires, plastic containers, gutters, tree holes) contained mosquito larvae. The predominant mosquito genera were Aedes, Culex, and Lutzia, similar to what has previously been reported in previous investigations in Gabon (Aedes, Culex, and Lutzia mosquitoes) [36][37][38], central Africa [39][40][41], Asia [42,43], and South America (Aedes and Culex mosquitoes) [44][45][46].
Overall, Ae. albopictus, Cx. quinquefasciatus, and Lu. tigripes were the most frequently found species. The strong presence of Ae. albopictus can be explained by the availability of artificial water containers that are suitable habitats for this species [40]. Previous studies demonstrated Lu. tigripes' natural predatory ability over other mosquito species, including mosquitoes of the Aedes genus [47][48][49][50]. Thus, its abundance and distribution could follow the dynamics of the other species used as prey. The relatively high abundance of Cx. quinquefasciatus confirms its ubiquity in various habitats, especially artificial habitats mostly found in highly populated areas, such as urban settings [39]. An. gambiae s. l. (the major malaria vector in the world) was almost entirely found outside the forest in open puddles due to domestic wastewater runoff (a typical larval habitat created by human activity for this species) [51][52][53][54], and also in a discarded tire. This confirms the use of unusual microhabitats, including those exploited by Aedes mosquitoes, as described for Anopheles stephensi, an emerging malaria vector in Africa [55].
Although some sites, such as epiphytic plants, tree holes at high elevations, and underground animal burrows, were not explored because of access difficulties, our observations indicated that the mosquito diversity within the forested compartment was two-fold lower than that outside the forest. Moreover, in the forested compartment, no species presumed to be exclusively sylvatic was detected, whereas some species already observed in a natural sylvatic condition elsewhere in Africa were identified. Indeed, Diallo et al. [56] observed An. gambiae s. l., Cx. decens, and Cx. quinquefasciatus in a forest canopy in Senegal (Kédougou region), although these three species were rare in that forested habitat. Similarly, Pereira dos Santos et al. [57] found Ae. albopictus up to several hundred meters inside an urban forest in Brazil. The absence of exclusively sylvatic species could be due to the lack of suitable conditions for sylvatic species, such as natural breeding sites (e.g., leaf axils, tree holes, rock holes, fruit shells), or the absence of animal host species for their blood meals. Extending larval sampling to the rainy season could have increased the number of detected species, including potential forest specialist species. Lastly, despite differences in mosquito diversity, the two compartments (i.e., inside and outside the forest) showed quite an important similarity during the dry season, mostly due to the high relative abundance of common species (i.e., Ae. albopictus and Cx. quinquefasciatus), thus rendering their communities similar.
Our analyses showed a clear segregation of larval habitats based on their type and spatial location. This could be explained by the inherent ecological preference of the recovered species. This suggests that in the Sibang district, there are only few mosquito species with a limited, specific ecological niche. For instance, Ae. aegypti was exclusively found in plastic artificial containers outside the forest, and clustered with other species including mainly Culex spp. Similarly, An. gambiae s. l. was almost exclusively found in ground puddles. On the other hand, Ae. albopictus and Lu. tigripes exploited artificial and natural microhabitats inside and outside the forest. These two opportunistic species, characterized by higher ecological plasticity, did not cluster together, possibly because of different microhabitat preference. Lastly, Cx. quinquefasciatus (a ubiquitous species with a large ecological niche) was recovered mostly in metallic and concrete containers, but also in tires and plastic containers.

The Mosquito Proliferation Drivers in the Urban Forested Area of Sibang
In the field, the larval infestation level of artificial water collections was very high, even inside the forest that is obviously used as a waste dump (Figure 2). This abundance of human-sourced breeding sites tends to indicate a convergence of breeding site types (mostly associated with waste dumping), available hosts, and consequently mosquito communities, both inside and at the forest periphery. The high level of anthropogenic pressure from neighboring households (accumulation of domestic waste inside the forest) promotes the proliferation of major disease vector species, such as Ae. albopictus, Ae. aegypti, and Cx. quinquefasciatus, that also breed in artificial water containers.

Mosquito Aggressiveness in the Urban Forested Area of Sibang
During the HLC-based collection time (10:00 a.m.-2:00 p.m.), Ae. albopictus was the predominant species collected, with a peak of aggressiveness between 11:00 a.m. and 12:00 p.m. and a biting rate of 35.2 bph inside the forest. Kamgang et al. [58] in Cameroon and Delatte et al. [59] in La Reunion reported peaks of aggressiveness for Ae. albopictus later during the daytime, between 4:00 p.m. and 5:30 p.m., suggesting a higher aggressiveness of this species in the Sibang area.
Ae. albopictus is a worldwide invasive arboviral vector of major public health concern [60] that has already caused chikungunya outbreaks in Gabon [12,61,62]. To the best of our knowledge, Ae. albopictus biting rates inside and outside a forested compartment, both in anthropized and wild environments, are not well documented. Nevertheless, Ae. albopictus' aggressiveness level in this study was >2-fold higher than what had previously been observed in Libreville in suburban neighborhoods (15.7 bph) [37], some of which were wooded areas with chikungunya transmission records. Furthermore, in 2009, a study in the Central African Republic, which included forested peri-domestic areas among the sampling sites, reported a peak biting rate of 1.7 bph [63]. This low rate could be explained by the fact that this previous study was carried out during the early stage of the Ae. albopictus invasion in this country, when its density was still low. Alternatively, the sampled forested peri-domestic areas might not have been areas of waste dumping, which seems to be, based on our results, a driver of Ae. albopictus proliferation and aggressiveness. More surveys are needed to monitor the Ae. albopictus daytime biting rate over a longer period. In terms of public health, our results indicate the very high risk of diseases transmitted by Ae. albopictus for the human population in this area of Libreville, and the need for disease outbreak surveillance programs.
Ae. aegypti was the second most aggressive species, especially inside the forest. The low biting rate and the non-significant difference in the captured Ae. aegypti females between locations inside and outside the forest could be explained by its scarcity in the Sibang district. In addition, the relatively low proportion of Ae. aegypti larvae, compared with Ae. albopictus, suggests a population decline for this species due to the successful invasion of Ae. albopictus, as suggested in central Africa [37,41,64] and elsewhere in the world [65,66].
Ae. albopictus is well known for its opportunistic blood-feeding behavior and high vector competence for a number of disease-causing viruses [67]. Studies in the Sibang arboretum in the early 2000s reported a high biodiversity of vertebrate animals, including several species of small mammals, reptiles and birds [28]. Thus, the high density of Ae. albopictus might not only increase the risk of the inter-human transfer of Ae. albopictusborne pathogens, but also represent a risk for potential zoonotic pathogens that could be hosted by the vertebrate animals in this forested patch, including birds and rodents, and be transmissible by this vector. Both animal groups are recognized hosts (or potential hosts) for zoonotic arboviruses, including West Nile virus for birds [68].

Mosquito Aggressiveness and Urban Greening
This study showed the important aggressiveness of Ae. albopictus and the risk of arbovirus transmission associated with this urbanized and forested district of Libreville. Such a risk could be exacerbated by the fact that this forested area might constitute a human-maintained incubator ecosystem and resting place for vectors, and therefore might facilitate and sustain the disease spread during epidemic periods.
Our results also highlighted that the poor sanitation of such an urban green area might modulate the burden of vector-borne diseases by intensifying (due to a high density of mosquitoes) or diluting (due to the presence of alternative hosts for mosquitoes to feed on) disease transmission. Araujo et al. [68] showed that in a Brazilian city, dengue incidence was higher in heat islands where vegetation cover was low than in forested neighborhoods considered to be fresher. Therefore, urban forests could be "islands of coolness" that might mitigate the spread of arboviruses, for example by slowing down the replication rate of viruses in mosquito vectors that live under the forest cover, as hypothesized in previous studies [69][70][71]. However, this positive effect could be counteracted by the level of pollution (due to poor sanitation), as shown by the present study. This requires effective sanitation measures in urban green spaces to mitigate the risk of VBDs in such areas. Thus, developing clean green spaces (e.g., in temperate poorly forested countries) and managing them with appropriate sanitation measures (e.g., in tropical and highly forested countries) could improve biodiversity, climate warming mitigation, and also the inhabitants' well-being. Therefore, urban forest islands, such as the Sibang forest, should be well planned and managed to mitigate the risks of environmental-driven diseases in general, and VBDs in particular.

Limitations of the Study
The results of this study are based on data collected during a 1.5-month dry season, and may not be generalizable to what might be observed during other seasons of the year or over a full annual climate cycle. Therefore, additional studies are needed to obtain a finer insight into the mosquito distribution, abundance, population dynamics, and diversity in this urban forest area of Gabon, as well as into the associated VBD risk.
Moreover, potentially important larval habitats, and thus a potentially important portion of mosquito diversity and abundance, might not have been investigated, particularly because of the limited access to properties due to uncooperative residents.
Due to government restrictions associated with the COVID-19 pandemic, HLC-based collections were made only during the day, and thus the mosquitoes collected were only day-biting species. Thus, the results do not take into account exclusively night-biting mosquito species, which may also be an important component of the vector risk associated with this urban forest. In addition, the HLC-based collections inside and outside the forest were not made concomitantly but sequentially, because the same three volunteers were always used to minimize the bias due to attractiveness variability among volunteers. However, no apparent major environmental variation (e.g., population movements, rainfall or other meteorological shifts) that could affect the collection outcomes was noticed.

Conclusions
This study in the urban forested area of Sibang allowed the recording of approximately 12% of the currently known mosquito species in Gabon [37,[72][73][74][75]. The investigation revealed a variety of larval habitats exploited by the mosquito species in the study area. Most of these habitats were artificial and man-made. The study highlighted the importance of considering urban forested ecosystems, especially when associated with poor hygiene conditions, as potential drivers of disease emergence and spread in urban areas, which should be taken into account when designing vector control strategies. In Gabon, this study should contribute to guiding vector control strategies, particularly through the implementation of policies to ensure good environment management and the surveillance of vectors in urbanized areas.