Relationships between traditional larval indices and meteorological factors with the adult density of Aedes albopictus captured by BG-mosquito trap

Objectives Larval indices have been used for Ae. albopictus surveillance for many years, while there is limited use in assessing dengue transmission risk and adult mosquito emergence. This study is aimed to explore the relationships between larval indices and the Ae. albopictus density captured by BG-mosquito trap (BG-trap) method, with considering the meteorological factors. Methods Data on larval density, adult mosquito density and meteorology factors were collected in an entomological survey carried out in Quzhou City, Zhejiang Province of China in 2018. The Spearman’s rank correlation and Pearson correlation were used for the analysis on the correlation of density indices. Generalized additive models were established to analyze the influencing factors of mosquito density. Results Breteau index (BI), House index (HI) and Container index (CI) were highly correlated with each other (r>0.7, p<0.05). The Ae. albopictus density was significantly correlated with CI (rs = 0.260, p<0.05), CI pre one week (rs = 0.259, p<0.05), and CI pre three weeks (rs = 0.329, p<0.05). BI was correlated with female Ae. albopictus density pre 4 weeks (r = -0.299, p<0.05). Female Ae. albopictus density was correlated with CI pre 3 weeks (rs = 0.303, p<0.05). The influencing factors of BI were average wind speed pre 1 week, average temperature and female Ae. albopictus density pre 4 weeks. The influencing factors of CI were average humidity pre 3 weeks and average temperature. The influencing factors of HI were average temperature and precipitation pre 4 weeks. The influencing factor of Ae. albopictus density and female Ae. albopictus density was temperature. Conclusions The adult Ae. albopictus density had low correlation with certain larval indices. Some of the meteorology factors played significant roles in the density of adult Ae. albopictus and larva with or without a time lag.

Introduction diet and density [20][21]. Dengue transmission cannot be explained by mosquito density alone, while infection rates and meteorological features should also be considered [15,22].
In 2018, an entomological survey was carried out in the Quzhou City, Zhejiang Province of China, which provided us an opportunity to study the relationships between traditional larval indices and the adult mosquito density monitored by the BG-trap, with considering the meteorological features.

Study sites and field work
This study was conducted in Quzhou City, Zhejiang Province, located in Southeast China. Considering the aspects of environment, coordination and operability, the Chongwen village (28˚53'46.68"N, 118˚54'44.02"E) and Songyuan village (28˚55'0.37"N, 118˚54'15.57"E) exhibited good representatives of the general rural areas in Zhejiang Province and were selected as the study sites. Besides, no major epidemics of dengue fever have occurred in this area during the study period, which could minimize the mosquito density fluctuation for dengue controlling. The study was conducted from April 26 to November 23, 2018 and lasted for 31 weeks. The larval density was monitored in about 50 households every week in Chongwen and Songyuan village, respectively. Trained field workers inspected and recorded household water containers and collected any pupae or larvae present for entomological examination. The water containers included any container with water in or around the households, such as flower pots, water storage containers, idle containers, waste tyres, garbage, rockery pool, open channel, bamboo or tree holes, stone holes, standing water in basement and parking lot, etc. A container was considered positive if it contained at least one larva or pupa.
The BG-trap (model: BG-Mosquitaire CO 2 , Biogents AG, Germany) baited with a steel cylinder filled with CO 2 emitted at a rate of 500g/24h. The trap was placed on the ground, the BG-Lure (Biogents AG, Germany) was placed in the pocket designated for the lure inside the trap, and the steel cylinder was set next to the trap. Each village placed three BG-traps at the peak time of Ae. albopictus with more than 200 meters away from each other, and lasted half an hour. All the captured mosquitoes were collected, and the species were identified morphologically.
The larval density and the adult mosquito density were defined as follows [7]. HI: the percentage of houses with containers positive for Ae. albopictus larvae. CI: the percentage of water-holding containers infested with Ae. albopictus larvae. BI: the number of positive containers per 100 houses inspected. The Ae. albopictus density: the number of Ae. albopictus including male and female trapped per trap in one hour. The female Ae. albopictus density: the number of female Ae. albopictus trapped per trap in one hour. Our filed work has been approved by the ethics committee of Zhejiang Provincial Center for Disease Control and Prevention (CDC). The ethics committee approved the procedure for verbal consent because Zhejiang CDC has the authority of the Zhejiang provincial government to collect the related information, which is part of the disease surveillance work in Zhejiang CDC. All the households were notified that they have the right to refuse or terminate the study at any point. Because we obtained verbal consent, documentation of consent was not required. However, the information collected from each household was kept confidential in Zhejiang CDC.

Statistical analyses
The statistical analyses were conducted with Statistical Program for Social Sciences 21.0 software (SPSS, Inc., Chicago, IL, USA) and R 3.6.2 software (The R Foundation for Statistical Computing Platform). A value of P<0.05 was considered as statistically significant. All the parameters were tested for normality. The Spearman's rank correlation and Pearson correlation with or without time-lag were used to analyze the correlation of the larval density, the adult mosquito density and the meteorological factors according to the data distribution. Generalized additive model (GAM) was used to analyze the influencing factors of the mosquito density.

The general description of the water containers
A total of 3109 households were investigated in the study, of which 1491 households had positive water containers, with a positive rate of 47.96%. 8911 water containers were inspected, 3350 was positive and the positive rate was 37.59%. In the positive containers, the highest percentage was seen in Ae. albopictus (2682, 80.06%), and followed by Culex pipiens pallens (631, 18.84%) and Armigeres obturbans (37, 1.10%). The BI ranged from 20.00 to 223.53 and the mean value of the two villages was 86.25. The mean value of CI was 30.46%, ranging from 5.52% to 66.13%. The mean value of HI was 42.79%, ranging from 18.00% to 76.00%.
Among all the water containers, the highest proportion was idle containers (6991, 78.45%), and followed by water storage containers (1510, 16.95%). Among different water containers, the highest positive rate was from tire water (48.34%), and followed by garbage water (47.62%) ( Table 1).

Correlation between larval density and adult mosquito density
The correlation between the larval density and the adult mosquito density of Ae. albopictus was analyzed. Considering the possible effect of time lag, 1~4 weeks were selected as the lag effect period. The Ae. albopictus density was correlated with CI (rs = 0.260, p = 0.041), CI pre 1 week (rs = 0.259, p = 0.046), and CI pre 3 weeks (rs = 0.329, p = 0.013). BI was correlated with female Ae. albopictus density pre 4 weeks (r = -0.299, p = 0.028). Female Ae. albopictus density was correlated with CI pre 3 weeks (rs = 0.303, p = 0.023). The three indices of larval density were highly correlated with each other (the r for BI and CI was 0.741, for BI and HI was 0.916, for CI and HI was 0.753, respectively, P<0.05), and were also correlated with a lag effect of 1~4weeks, with correlation coefficients decreased gradually over time.

Correlations between mosquito density and meteorological factors
The correlation analysis was carried out to explore the relationships between meteorological factors and mosquito density, and 1~4 weeks was selected as the lag effect period. The results showed that the meteorological factors such as precipitation, average air pressure, average humidity, sunshine hours, average temperature, and average wind speed were correlated with different indices of the mosquito density, with or without a lag effect. The significant parameters of the correlation were shown in Table 2.

The results of the GAM models
GAM models were used to analyze the influencing factors related to different density indices of Ae. albopictus. The significant variables in the correlation analysis were included in the models, and the best effect time of the same variable was selected with the highest correlation coefficient. Although there were high correlation among BI, CI and HI, they were different aspects of the larval density, and consequently the three indices were not included in the model of each other. As shown in Table 3, BI was significantly associated with average temperature, average wind speed pre 1 week and female Ae. albopictus density pre 4 weeks. BI increased to a peak value first, and then decreased with the increasing of the average temperature (Fig 1), decreased with the increasing of the average wind speed pre 1 week straightly (Fig  2), and decreased smoothly with the increasing of the female Ae. albopictus density pre 4 weeks (Fig 3). CI was significantly associated with average temperature and average humidity pre 3 weeks. CI increased to a peak value first, and then decreased with the increasing of the average temperature (Fig 4), and increased straightly with the increasing of the average humidity pre 3 weeks (Fig 5). HI was significantly associated with average temperature and precipitation pre 4 weeks. The relationship between HI and the temperature was similar to those with BI and CI (Fig 6), and with the increase of precipitation 4 week ago, HI increased first, then reached a plateau period (Fig 7). The Ae. albopictus density or female Ae. albopictus density had linear relationship with the average temperature with a time lag of two weeks (Figs 8 and 9).

Discussion
In the field survey, we found that the Ae. albopictus density had low correlation with CI or with a time lag of one or three weeks. BI had correlation with female Ae. albopictus density with a time lag of 4 weeks. The average temperature, precipitation, average humidity, and average wind speed played significant roles in the density of adult mosquito or larva with or without a time lag. BI is considered as a decision making parameter for mosquito control and dengue epidemic risk. Generally, the BI value of 5 serves as the lowest threshold. In a scenario where the BI value > 5 with reported dengue cases or BI > 20 even without any dengue case, control measures should be taken [18]. Three different risks of HI, with <0.1% as low, 0.1-5% as medium and >5% as high, were suggested by the Pan American Health Organization to prevent dengue transmission [23]. As for CI, one study found that 11.7 was the optimal cut-off value for discriminating outbreaks of dengue [24]. In this study, we found the average BI value was extremely high (86.25) in two villages, and similar values were also seen in HI (42.79%) and CI (30.46%). Although reasons for the high estimates in our study were complicated, there was possible explanation with respect to the breeding place for Ae. albopictus. As the Ae. albopictus generally breed in artificial water containers, any type of water-holding container with clean water would be a good larval habitat [3,8]. The two villages investigated in this study had good sanitation conditions, and vegetation was abundant in and around the villages. Besides, considerable idle containers and water cisterns with clean water were put in or around the yard (accounting for 95.4% of the total number of water containers), which would provide perfect breeding place for Ae. albopictus. Furthermore, consistent with a previous study [16], the positive rate for Aedes larval was found to be higher in discarded tires. As for adult Ae. albopictus monitoring, an effective trap would be less intrusive, labor saving, and more comprehensive coverage with an effective lure or attractant. The BG-trap, using CO 2 and the BG-lure to capture host-seeking female mosquitoes, is an effective mosquito monitoring method. Our entomological survey was conducted at the peak period of Ae. albopictus density [25], which were representative to a certain extent. The results showed that 86.18% of the adult mosquitoes captured by BG-traps were Ae. albopictus, indicating that the BG-traps were sensitive for Ae. albopictus. Consistent with a previous study [9], the BG-traps were more effective in capturing female rather than male Ae. Albopictus (82.42% vs. 17.58%). The thresholds of the classical larval indices for management of dengue epidemics were considered to be less effective and sometimes remained poor in predicting adult emergence [18]. Measuring adult mosquito density was the most representative quantitative estimate to obtain data about mosquito abundance, as larva needed to go through several developmental stages to become adult mosquitoes before they could transmit dengue virus [26]. Study had found that the household larval surveys and trap based surveillance systems were not interchangeable approaches [27]. In our study, the Ae. albopictus density and female Ae. albopictus density were calculated as two indices and the results were not exactly the same. The Ae. albopictus density, contained all the captured Ae. albopictus including male and female, while the female Ae. albopictus density, calculated the female Ae. albopictus only. The correlation analysis indicated that the two indices all were slightly correlated with CI with a certain time lag. While the female Ae. albopictus density pre 4 weeks was negatively correlated with BI, which was consistent with the results of the GAM model but contrary to our common sense. As only female Ae. albopictus was responsible for disease transmission, the indices would be more appropriate towards female mosquitoes directly. One interesting phenomenon found in our study was that, when the BG-traps were put in the grass or small bamboo grove, more mosquitoes would be caught and most of them were male. These findings may lead to bias of the result for different sites the traps placed, and the different emergence time of the male and female mosquito [9]. Consequently, regarding the correlation between the larval and adult mosquito density, it would be more appropriate towards Ae. albopictus density than female Ae. albopictus density.
Climatic factors, particularly the temperature, precipitation and humidity, could directly and indirectly affect the mosquito density and blood feeding behavior [8,[28][29]. In our study, the average temperature was the main influencing factor of the mosquito density, affecting all the study indices. Temperature is crucial for mosquitoes, not only for survival rate but also the lifecycle of the vector including oviposition, hatching, pupation, and emergence processes [16,[30][31]. Higher temperature could reduce the development time of mosquitoes, and increase the propagation speed of the virus [32][33][34]. Consistent with the above study, our results showed that the adult mosquito density increased straightly with the increase of the average temperature pre two weeks. Studies also found that the effects of temperature on the mortality rate of larvae, pupae and adult mosquitoes could be U-shaped with a lower mortality rate was seen when temperature ranged from 15 to 30˚C [20][21]35]. This probably explained the decrease of the BI, CI and HI from the peak value along with the increase of the average temperature in our study.
Precipitation played a crucial role in the transmission of mosquito borne diseases, due to the fact that mosquito required water for the aquatic larval and pupal breeding stages. Higher pupal productivity and entomological indices was found in the rainy season than dry season [3,16,26], and the effect of the precipitation to the larval density may have a time lag from 2 months to 1 month [33]. Precipitation could also influence the adult mosquito capturing. One study found that the BGS traps consistently captured nearly 20% of the marked female Aedes population in the wet season and about 30% in the hot and dry season [36]. Besides, BGS traps could increase the biting rate of mosquito via increasing the contact between humans and mosquitoes, as humans often stayed indoors when it rained [4]. Based on our results, with the increase of precipitation 4 weeks ago, HI increased at the beginning, and then reached a plateau period. Less precipitation reduced amount of water retained in containers which affect mosquito breeding. However, extremely heavy precipitation could lead to water containers saturation or even flush mosquito larvae away from breeding sites, eliminating habitats to decrease the vector population [26], which possible explained the plateaus of HI.
Relative humidity was an important meteorological factor in the life-cycle of mosquitoes [15], especially in lowland plains [16]. Humidity could also increase the transmission rate of human dengue fever infection in the context of imported dengue cases and mosquito density [4,30]. Relative humidity could affect larvae density by affecting adult mosquito survival, and also had a synergistic effect with the temperature [17]. While in our GAM models, CI increased with the rise of the average humidity pre 3 weeks. The wind speed could influence the effectiveness of the daily captures of mosquitoes [37]. Yin Q et al. suggested that the predicted hourly Ae. albopictus densities generally decreased with wind speed [25]. Endo N et al. found that wind direction and speed could influence the malaria vector populations by affecting the effect of CO 2 attraction and enable mosquitoes to identify village location [38]. In our models, the average wind speed was negatively correlated with BI and with one week lag effect. Higher wind speed may affect the dynamics of the mosquito population by affecting wave activity, advection of adult mosquitoes, and CO 2 attraction, resulting in a low density of larvae after a period of time.
Our study had several strengths. This is one of the few studies investigating the relationships between larval indices and the adult mosquito with BG-trap method in mainland China. Besides, during the analysis procedure, various meteorological factors were taken into

PLOS ONE
Larval indices and adult density of Aedes albopictus consideration in our study. Meanwhile, some limitations must be recognized in this study. Firstly, as the study sites and samples were only selected from rural area of Zhejiang Province, our results cannot be generalisable to broader national level. Secondly, the study relationships may be confounded by other factors such as socio-economic characteristics and human activity, which were not included in the current analysis.

Conclusions
Our findings suggested that the BG-trap was an effective adult trap for Ae. albopictus, especially for the female mosquitoes. The adult Ae. albopictus density was slightly correlated with certain larval indices. The average temperature, precipitation, average humidity, and average wind speed played significant roles in the density of adult mosquito or larva with or without a time lag. To prevent dengue fever, new monitoring method and thresholds should be developed based on adult mosquitoes, with considering meteorological factors.