Results show that mosquito populations in CU and TA form complex communities that can change between seasons. The highest species richness estimations were obtained in TA in the rainy season. During this period, more habitats may be available for larval development due to rainfall, not only in artificial containers but also in phytotelma. Although the sampling effort in this study was limited, most of the species present in each area were captured, based on the Coleman rarefaction curves and Chao2 values. The high standard deviation in TA’s Chao2 (Table 1) plus the incapacity of reaching a curve asymptote can be explained by the number of rare species and or singletons present in TA (Table 2, Supplementary data). This result is similar to other studies with comparable sampling efforts (44).In this regard, sampling arthropods in tropical areas usually requires a higher intensity than what is typically done for other taxa (45).
Mosquito diversity (Shannon Index) was higher in CU although the species estimation was higher in TA, which suggests that CU populations are more homogeneous, or contain fewer dominant species. The lower diversity (high evenness) values in TA may be due to the dominance of Cx. quinquefasciatus, which accounts the 47.6% of the total individuals in our samples (360/756) (Fig. 1). Interestingly, when the H’ of TA values were compared between sampling settings, the FO has the highest diversity. This can be further related to the anthropogenic pressure at the DO, PE, and PN settings because of the affinity of Cx. quinquefasciatus to human activity, domestic animals, and altered areas in the tropics (46). Low diversity has been extensively associated with a higher risk of vector-borne diseases (47), and functional diversity has been established as a good predictor for higher R0 in vector-borne diseases (48). Deforestation and changes in land use such as cattle ranching, which reduce local diversity, have been proposed as risk factor for vector-borne and emerging infectious diseases (49). Furthermore, recently productive landscapes such as oil palm and pineapple plantations have been associated with higher presence of disease vectors (50).
Changes in species composition (Chao-Sørensen Similarity Index) was different among sampling locations, nonetheless their degree of variation between seasons differ for each county. TA values were nested within seasons, showing small changes in their species composition. In contrast, CU had a more dissimilar population between seasons, with a high turnover in the presence of medically important species such as An. albimanus and De. pseudes. In this context, Cx. quinquefasciatus is a proven vector for several zoonotic arboviruses, including WNV and SLEV (51). This variation in species composition can explain the high seasonality of VEEV and WNV observed in the CU area, where neurological disease due to arboviruses has a higher incidence during the rainy season (28).
The Evenness in Cuajiniqul indicates that some species are dominant in this community yearlong (e.g. De. pseudes). It has been proposed that Shannon Evenness Index is a strong predictor of disease risk in multiple host communities (48). Historically, most cases of WVN and VEEV have been recorded in the northern pacific region of Costa Rica (52). Therefore, the vector community structure might be playing a fundamental role in viral activity due to the high variation in diversity index and species richness.
The variation in the seasonal NDVI on the local scale (buffered areas) did not show a strong correlation (R=-0.20) when compared with Chao- Sørensen values of the same sampling points between seasons. In contrast, the correlation of regional mean NDVI (all buffered areas per season) shows a stronger negative correlation (R=-0.53) with the difference in the NDVI values. The regional Chao- Sørensen index can represent a more representative change of the overall community since considers more subsets of the total population, giving a stronger correlation when compared with the regional NDVI. However, the NDVI has been broadly used to predict population changes in different environments (53). The arboviral encephalitis cases in horses have a high incidence during the rainy season in the region (28, 54). In CU and its surroundings, species turnover and its relation with the NDVI could be an important predictor of vector activity, furthermore, climate change and ENSO can extend rain seasons in the tropics which can therefore extend the period of vector patterns and regional viral activity (55–57).
These variations are also reflected in the larval abundance between seasons in CU, where several of the OV placed were completely dry upon later evaluation during the dry season (Supplementary Data, Table 4). The absence of water-filled tree-holes limits the availability of suitable larval habitats for species such as Haemagogus spp. and Sabethes spp. which can be vectors of YFV and Mayaro. Similarly, the availability of rice fields in CU was only present during the rainy season at CU. Since the absence/presence of an adult species is strongly related to the availability of a correct habitat, which in the case of An. albimanus are the rice fields (58, 59), the rice fields been proven to be of public health importance in other countries (60) since they favor its presence. In CU, the drought in the dry season eliminates water in rice fields and causes An. albimanus population to drop almost completely. Currently, there is no active transmission of Plasmodium spp. in CU, but it is present in other areas of Costa Rica due to human movement and anthropogenic landscape changes, including illegal gold mining (61). Therefore, anopheline larval habitat conditions and adult mosquito abundance at this site represent a potential risk for Plasmodium transmission in CU during wet seasons. Moreover, the severe dry season can also influence arboviral incidence in the region, considering that some mosquito populations (e.g. Culex pipiens and Anopheles quadrimaculatus) can increase drastically after severe droughts(62). Although no precipitation data was analyzed, NDVI values can reflect an increase/decrease in rain fall precipitations (63).
In contrast, species like De. pseudes have a similar abundance yearlong. As this species breeds in salty water-filled crab holes, their larval habitat does not depend on rainfall and is therefore present year-round. De. pseudes is a proven vector of VEEV, so the continuous presence of this species can help maintain enzootic VEEV transmission in the reservoir host population.
The dry and rainy seasons in CU are highly different because of the excessive difference in rainfall during both seasons (64). This can also reduce the population of tree hole breeders during this period. Previously, OV’s have been used in the tropical rainforest for sampling sylvatic enzootic vectors (30), but none have been used for sampling Culicidae in a tropical dry forest. Nonetheless, several studies in urban areas adjacent to tropical dry forests show a significant decrease in ovitrap capture success, since human activities can help maintain artificial containers filled with water (65, 66).
Most engorged females were caught in the animal pen and belong to the Culex genus. Broadly, Culex mosquitos have a wide range of feeding hosts, which can include, humans, domestic species, and wildlife. Nonetheless, mosquito feeding behavior can be aggregated, adapting to what hosts are available (67). Since our sampling was done in areas with a high presence of humans and domestic animals, it is expected for these to be the main feeding hosts. In addition to these, blood from a White-tipped dove (Leptotila verreauxi) was also detected, which is of interest given that neutralizing antibodies for WNV and SLEV have been detected before in this species (68). Despite of the potential role of this dove in the epidemiology of some arboviruses, the importance of these findings is that vector species, such as Cx. quinquefaciatus and De. pseudes, are feeding in putative enzootic hosts as well as dead end host, which is necessary for viral transmission into horse and human populations. The implications of reptile blood are probably of lesser importance than avian blood, although some species can serve as amplifiers for WNV (69).
The frequent detection of chicken blood in the mosquitoes collected at both sites can have repercussions on the epidemiological cycle. Chickens are refractory to WNV (70) infections and work as sentinels for WNV (71). This species has been proven to work as a zoo-prophylactic species for other vector-borne diseases (72–74). The possibility that chickens might be taking a zoo-prophylactic role for arborvirus transmission in Costa Rica needs to be further explored, since is a common backyard animal in rural areas. In contrast to other countries, cases of WNV in Costa Rica are rare in both humans and horses (28, 75). The most prevalent arbovirus in the country with a confirmed enzootic cycle is VEEV (76), which is still infrequent in comparison with other arboviruses such as DENV.
The detection of DENV and ZIKV RNA was not unexpected in these areas, since both viruses are prevalent in Costa Rica. DENV-3 has not been detected in humans in Costa Rica since 2016 (77). Although the cDNA amplified region is short, this positive sample had a 96,7% similarity with a 2016 Colombian isolate (MH544650.1). Considering that this DENV-3 positive sample was from a Cx quinquefasciatus that contained human blood, it is likely that the viral RNA was from a viremic human, since Cx. quinquefasciatus is not epidemiologically relevant as a DENV vector (78). Regarding ZIKV, this species has only been proven an efficient vector in a few vector competence studies, but the consensus is that it is not a major ZIKV vector(79).We do not consider that Cx. quinquefasciatus has a major role in DENV or ZIKV transmission to humans. Overall, viral detection was unexpectable low. Arboviruses usually circulate at low prevalence between vectors (80). Although our rarefaction curves indicate that we sample most of the species present in the area, the vector population captured may be considered low (e,g, Cx. Quinquefasciatus (514/1360)).