(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



The relative importance of key meteorological factors affecting numbers of mosquito vectors of dengue fever [1]

['Yan Liu', 'School Of Mathematics', 'Statistics', 'Shaanxi Normal University', 'Xi An', 'Shaanxi', 'Xia Wang', 'Sanyi Tang', 'Robert A. Cheke', 'Natural Resources Institute']

Date: 2023-05

Although single factors such as rainfall are known to affect the population dynamics of Aedes albopictus, the main vector of dengue fever in Eurasia, the synergistic effects of different meteorological factors are not fully understood. To address this topic, we used meteorological data and mosquito-vector association data including Breteau and ovitrap indices in key areas of dengue outbreaks in Guangdong Province, China, to formulate a five-stage mathematical model for Aedes albopictus population dynamics by integrating multiple meteorological factors. Unknown parameters were estimated using a genetic algorithm, and the results were analyzed by k-Shape clustering, random forest and grey correlation analysis. In addition, the population density of mosquitoes in 2022 was predicted and used for evaluating the effectiveness of the model. We found that there is spatiotemporal heterogeneity in the effects of temperature and rainfall and their distribution characteristics on the diapause period, the numbers of peaks in mosquito densities in summer and the annual total numbers of adult mosquitoes. Moreover, we identified the key meteorological indicators of the mosquito quantity at each stage and that rainfall (seasonal rainfall and annual total rainfall) was more important than the temperature distribution (seasonal average temperature and temperature index) and the uniformity of rainfall annual distribution (coefficient of variation) for most of the areas studied. The peak rainfall during the summer is the best indicator of mosquito population development. The results provide important theoretical support for the future design of mosquito vector control strategies and early warnings of mosquito-borne diseases.

The study and control of the numbers of Aedes albopictus, the main vector of dengue fever in China, is crucial for the prevention and control of this mosquito-borne disease. Due to the major impact of rainfall and temperature on mosquito population sizes, we built a mathematical model based on the life cycle of mosquitoes and meteorological and mosquito vector monitoring data in Guangdong Province. Through this model, we explored the effects of temperature, rainfall and their distribution characteristics on the development of the mosquito population and found that they had different effects on the developing stages of mosquitoes in different periods and different locations. Moreover, according to the established meteorological indicators and their importance ranking, it was found that for most areas of Guangdong Province, rainfall was more important than the temperature distribution and the evenness of rainfall distribution, and that the peak rainfall in summer was the best predictor of mosquito population development. In addition, we also developed a method to utilize previous experience in mosquito control, which together with the above conclusions will be of great help in the design of future mosquito vector control strategies and the prediction and early warning of mosquito-borne diseases.

Funding: This work was supported by the National Natural Science Foundation of China (grant numbers 12171295 to XW., 12031010 to ST.) and the Project of Science and Technology Young Star in Shaanxi Province of China (2022KJXX-29 to XW.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The dynamic effects of a single factor on mosquito numbers can be characterized by simple experimental design and model characterization but taking account of the synergistic effect of multiple factors and determining their relative importance is more challenging. Therefore, we constructed a five-stage mosquito population dynamics model (described below) by collecting years of meteorological and mosquito-vector association data in key areas of dengue fever outbreaks in Guangdong Province. The model was then used for exploring the influence of climatic conditions on the development of mosquito populations, to identify the key meteorological indicators affecting mosquito numbers and ranking them in terms of importance. The model was then used to predict mosquito numbers in 2022 and to evaluate the effectiveness of the model and thus provide suggestions for future mosquito vector control strategies and early warnings to assist the control of mosquitoes and diseases.

The life cycle of A. albopictus is divided into four stages: the eggs, larvae and pupae are aquatic, while adult mosquitoes are terrestrial, but each stage is affected by climatic factors such as temperature and rainfall [ 6 , 7 ]. Experimental studies have found that under a constant temperature of 10°C, A. albopictus could not develop at all, and eggs failed to develop at 15°C, but with the temperature rising to 30°C, the development rate of each stage increased continuously, and the larvae and pupae developed fastest at about 30°C [ 8 ]. Moreover, rising temperatures created conditions allowing larvae or adult mosquitoes to overwinter [ 9 – 12 ]. A modelling study found that the quantity of mosquitoes in a specific area was significantly affected by season and the diapause process, and due to the tropical climate and relatively warm winter in Guangzhou, the mosquito numbers there could show single or double peaks [ 13 ]. Furthermore, in addition to temperature, the annual total rainfall and rainfall distribution are also important factors that cannot be ignored [ 14 ].

The warm and humid climate of several coastal provinces in southern China is particularly suitable for the breeding of Aedes albopictus, the main vector of dengue fever, which inevitably increases the risk of dengue outbreaks [ 1 , 2 ]. Therefore, the characterization and prediction of the population dynamics of A. albopictus, especially the analysis of the impact of key climate factors on mosquito numbers (based on proxy values derived from two entomological indices), are of practical significance for the prevention and early warning of dengue outbreaks [ 3 – 5 ].

Finally, in order to draw on the experience of mosquito control practice in previous years, we used the grey correlation analysis method again to analyze the meteorological indicators of each city from 2016 to 2022, taking the 11 meteorological indicators in 2022 as the reference sequence, the meteorological indicators of 2016 to 2021 as the comparison sequence, and the normalized results of the previously calculated grey correlation degree as weighted meteorological indicators. The grey correlation degree between meteorological indicators in 2016–2021 and 2022 was calculated. The greater the correlation degree, the more similar the climate distribution in a year was to that in 2022, and the closer the mosquito population development was without considering the influence of other factors.

We next used the random forest method to rank the importance of 11 meteorological indicators, and then used grey correlation analysis to verify the ranking results [ 31 – 33 ]. Due to the low risk of dengue fever outbreaks in winter in Guangdong Province, the government only publishes the mosquito vector monitoring results from March to October. Therefore, in order to maintain the comparability between subsequent verification results and random forest results, the total number of adult mosquitoes from March to October was selected as the dependent variable, and 11 meteorological indicators were used as independent variables. Based on the data of 84 years from 2015 to 2021 in the 12 cities, the importance of indicators was ranked by random forest. Then only 11 meteorological indicators in 2016, 2017 and 2019 were used for grey correlation analysis on the adult mosquito monitoring index (MOI) because the publication frequency of MOI in 2015 was inconsistent with those in 2016–2021, while MOI data in 2018 were too sparse, and the MOIs for 2020 and 2021 were affected by COVID-19. The indices were ranked according to the magnitude of correlation degree to verify the ranking results based on the random forest analyses.

In order to explore the key meteorological factors affecting the annual development of mosquito populations and rank their importance, we first clustered the 84 years’ data of the 12 cities from 2015 to 2021 by year using k-Shape clustering to extract the typical climate distribution, and finally the rainfall and temperature were divided into six categories [ 30 ]. Then we took the centroid of each category as the typical climate distribution. Combined with the annual total rainfall (1500mm, 2500mm, 3500mm, 5500mm, 7500mm, 9500mm), we arranged and combined the three meteorological variables (typical temperature distribution, typical rainfall distribution, annual total rainfall), and finally brought them into the model for simulation to explore the key meteorological factors affecting the development of mosquito populations throughout the year and indexed them.

For regional parameter estimations, 12 cities were classified. Firstly, according to the literature and the latitude of the 12 cities, in Yangjiang, Maoming and Zhuhai, eggs overwintered mainly, followed by larvae, while in other cities it is the overwintering of diapause eggs [ 21 ]. In addition, there is little difference in temperature distribution among the 12 cities ( Fig 1 ), so only the differences in rainfall data are considered in the classifications. Here, we used the rotating empirical orthogonal function method to classify the rainfall data of the 12 cities from 2015 to 2021 by region [ 27 , 28 ]. The 12 cities were finally divided into four categories by combining the two factors. When fitting parameters in each category, the sum of the Euclidean distances between the monthly relative numbers of BI and MOI and the monthly relative numbers of larvae and adult mosquitoes were used for the objective function, and then a genetic algorithm was used to find the optimal parameters [ 29 ].

See Tables 1 and 2 and Eqs ( 2 )–( 5 ) for the meaning and expression of variables in the model. The meanings of 28 parameters estimated in this paper are shown in Table 2 , and the parameter estimation results are shown in the Supplementary Information ( S1 Table ). Due to the large number of selected cities, before fitting, we first classified the 12 cities according to the insects’ overwintering mode as the main basis, supported as an auxillary by differences between the meteorological data. Thus, Yangjiang, Maoming and Zhuhai were placed in the first category; Guangzhou, Shantou, Shanwei, Zhaoqing, Shaoguan, Meizhou and Heyuan in the second category; Shenzhen in the third category; and Huizhou in the fourth category. Then, four groups of parameters were obtained by parameter estimation in Yangjiang, Guangzhou, Shenzhen and Huizhou selected as the "fitting cities" for fitting data to. All cities in each category shared a set of parameters, and the simulation results of the remaining cities except the fitting city were used for model verification. The analyses below under the four sets of parameters were carried out in the four fitting cities (Yangjiang, Guangzhou, Shenzhen and Huizhou).

Besides, the expressions for v(W), the development rate, the mortality rate, the emergence rate and the environmental capacity in the model are set as follows: (5) where T(t) is the daily mean temperature; R norm (t) represents the normalized value of the two-week total rainfall. Other parameters in Eq ( 5 ) and the references for upper and lower bound values about these parameters are shown in Table 2 .

See Eq ( 5 ) for the expression of v(W) and see Table 2 for the definition of b. Furthermore, according to the latitude of the 12 cities, adult mosquitoes in Yangjiang, Maoming and Zhuhai laid diapause eggs during the diapause period, while no diapause eggs were produced in other cities[ 21 ]. Therefore, in Yangjiang, Maoming and Zhuhai, (3) and in other cities, (4)

Considering the influence of meteorological factors on each stage of mosquito population growth, we established a stage-structured model. Here, we assume that the mosquito population is large enough in the study area and that the aquatic habitat has enough food for the developing stages of the mosquitoes. The model is as follows: (1) where E 0 are non-diapause eggs; E d are diapause eggs; L are larvae; P are pupae; A are adult mosquitoes and W is the moisture index. According to Shame et al., the impact of rainfall on the life cycle of mosquitoes is mainly reflected in two aspects: (1) rainfall can increase near-surface humidity, and the increase of humidity will enhance the flight activity and host-seeking behavior of mosquitoes, thus stimulating mosquitoes to lay eggs and accelerate the reproductive cycle of mosquitoes; (2) rainfall can alter the abundance and type of aquatic habitat in which eggs, larvae and pupae live [ 15 ]. Therefore, the moisture index W associated with rainfall was added into the model to depict the mosquito oviposition rate from the perspective of air humidity, and we described the environmental capacity of the aquatic stage according to rainfall. In the model, as shown in Table 1 , me 0 is the mortality rate at the non-diapause egg stage with a value of 0.05 according to the relevant literature [ 16 ], and me d is the mortality rate at the diapause egg stage with value a (defined in Table 2 ) times that of me 0 ; r(t) is the total daily rainfall at day t; ef(t) is the emergence rate at day t, and k is the environmental capacity of the aquatic stages, and they were set according to Tran et al. [ 16 ]; v 1 (T w ,W) and v 2 (T w ,W) are the oviposition rates of non-diapause and diapause eggs produced by adult mosquitoes at day t respectively, and see below in Eqs ( 2 )–( 4 ) for their settings, and here T w is the weekly mean temperature; de 0 (t), d(T w ), dl(t), dp(t) are the development rates at the stages of non-diapause eggs, diapause eggs, larvae and pupae at day t respectively, and ml(t), mp(t), ma(t) are the mortality rates at the stages of larvae, pupae and adults at day t, respectively, and these parameters were set according to Wang et al. [ 14 ]. δ is the evaporation rate, shown in Table 2 . In addition, the conditions related to the diapause period were set as follows: (1) when the weekly average temperature was lower than 21°C, adult mosquitoes produced diapause eggs; (2) when the weekly average temperature was higher than 15°C, diapause eggs hatched; (3) when the diapause rate was higher than 0.9, the diapause stage began, and the diapause stage ended when diapause eggs hatched (see Supplementary Information for details) [ 8 , 16 – 20 ]. Therefore, in the model, (2)

The temperature and rainfall data of the 12 cities are shown in Fig 1 . It can be seen from the figure that the average monthly temperature is between 10°C and 35°C, with the highest temperatures in June, July, August or September and the lowest temperatures in January or February. The winter of 2019 was the warmest, with the average monthly temperature above 15°C except in Shaoguan. In addition, from the perspective of the monthly average rainfall, most of the 12 cities’ rainfall during 2016–2021 was concentrated in summer, except in 2020 when it was mainly concentrated in summer and autumn. Yangjiang and Guangzhou had the most rainfall, while Huizhou had the least rainfall. Except for Huizhou and Zhuhai, the annual total rainfall showed a trend of low in the first four years and high in the last two years, ranging from 1000mm to 3200mm in the first four years and from 2800mm to 7500mm in the last two years.

Guangdong Province is located at 20°13 ’-25°31’ north latitude, 109°39 ’-117°19’ east longitude, located in the southern coastal zone of China, and belongs to the East Asian monsoon region, from north to south, with a subtropical and tropical climate. The province has jurisdiction over 21 prefecture-level cities. Since neighbouring cities share the same meteorological station data, we chose 12 cities to study according to the differences in their meteorological data including Guangzhou, Yangjiang, Shenzhen, Huizhou, Maoming, Shantou, Shanwei, Zhuhai, Heyuan, Meizhou, Zhaoqing and Shaoguan. Daily temperature and rainfall data for 12 municipalities in Guangdong Province during 2015–2021 were obtained from a weather website ( rp5.ru ), and the Breteau index (BI) and the ovitrap index (MOI), were obtained from the Guangdong Provincial Health Commission ( gd.gov.cn ).

Results

1 Estimation and fitting The model fitting results are shown in Fig 2, and the verification results are shown in the Supplementary Information (S1 and S2 Figs). According to the monitoring method published by the Guangzhou Center for Disease Control and Prevention, staff check the monitoring sites every four days, collect positive containers and add new containers in their place (gzcdc.org.cn). In addition, there are enough monitoring sites set in every city, so BI and MOI can reflect the change in trends of mosquito numbers, which is proportional to the actual mosquito population. Therefore, the goal of parameter estimation is that the fitted trends of adult mosquitoes and larvae are consistent with the change in trends of BI and MOI. Fig 2 shows that the fitted trends of adult mosquitoes and larvae in Yangjiang, Guangzhou, Shenzhen and Huizhou from 2016 to 2021 are mostly consistent with the actual MOI and BI trends, but there were overestimates for Guangzhou and Shenzhen in 2021, which may be related to the epidemic and control of COVID-19 in the past two years. Secondly, according to the Guangdong Provincial Health Commission, the monitoring efforts of BI and MOI are different in different cities and years, so the data have a certain degree of randomness and uncertainty, which leads to the overestimation in Yangjiang in 2016 and the underestimation in Huizhou in 2016 and Guangzhou in 2019. In addition, the implementation strength of mosquito control policies by the health commission is different in each city. Also, the BI and MOI are affected by mosquito control policies and monitoring efforts, so the validation results in some years will be poor due to the different implementation strengths of policies and monitoring efforts of BI and MOI in different cities. For example, no policy was implemented in Maoming in 2016, so there was an underestimate when the estimated parameters of Yangjiang were used to fit the mosquito numbers of Maoming. Moreover, the difference in the implementation strength of policies between the fitting city and the validating city also led to the overestimation phenomenon of Zhuhai in 2019 and the underestimation phenomenon of Shantou, Zhaoqing, Shaoguan and Meizhou in 2016. Ignoring these cases and the impact of COVID-19 in 2020 and 2021, it can be seen that the simulated trends for the 8 cities within 6 years are basically consistent with the actual MOI and BI trends under their respective parameters in the Supplementary Information (S1 and S2 Figs), which indicates that the model fitting results are basically consistent with the actual situation. Based on the parameter values estimated above, we can calculate the diapause duration of each city each year, and then find that adult mosquitoes and larvae overwinter in years with a diapause period less than 30 days by comparing the diapause duration with the fitted results of each city (S3 and S4 Figs). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Model fitting results for Yangjiang, Guangzhou, Shenzhen and Huizhou. (a)-(d) Fitting results for adult mosquitoes (MOI); (e)-(h) Fitting results for larvae (BI). Red circles represent the actual MOI and BI values. The blue asterisks represent the compressed fitting result. In order to clearly show the consistency of the changing trend in the simulation results and monitoring indicators in the figure, we compressed the simulation results. https://doi.org/10.1371/journal.pntd.0011247.g002

2 Factor analyses: Temperature and rainfall For exploring the influence of temperature and rainfall on the diapause period, double peaks of mosquito numbers in summer and annual total adult mosquito quantities, sensitivity analysis was carried out in the four fitting cities (Yangjiang, Guangzhou, Shenzhen and Huizhou) under four sets of parameters. Three different scenarios were simulated for each year and each season from 2016 to 2021 (2015 was omitted to avoid the impact of initial values): (a) increased seasonal rainfall; (b) simultaneously increased seasonal rainfall and temperature; (c) increased seasonal temperature. In the case of Shenzhen (Fig 3, the sensitivity analysis results in Yangjiang, Guangzhou and Huizhou are shown in the Supplementary Information (S5–S7 Figs)), the results reveal that increasing seasonal temperatures, seasonal rainfall or both can prolong the impacts on adult mosquito numbers. The aftermath of the impact of changing weather conditions in spring and summer can last for 1 to 2 months, while it can last for 1 to 6 months in winter and autumn (Table 3). However, the aftermath duration in Huizhou is significantly shorter than that in the other three cities, which may be related to the low rainfall in that city in previous years. In addition, as can be seen from the actual MOI in the Supplementary Information (S1 Fig), the adult mosquito numbers in Guangzhou, Shantou, Zhaoqing and other cities show double peaks in summer. According to the sensitivity analysis for these cities, the bimodal phenomenon will disappear with increasing summer rainfall, will strengthen with increasing summer temperature, and weaken when the two increase at the same time (S8–S12 Figs), which indicates that the bimodal phenomenon is mostly related to rainfall. By comparing the monthly average rainfall of the corresponding months in these years, it was found that when the monthly rainfall declines steeply, followed by a rise, the double peak phenomenon is highly likely. This condition occurs in summer and the greater the difference between early and late summer rainfall, the more pronounced is the double peak. For instance in the summer of 2020, the MOI data show that there was a double peak in Guangzhou, where the actual monthly average rainfall in each month from May to September was 68.15mm, 40.08mm, 8.61mm, 28.29mm and 46.03mm, respectively, which is more obvious than that in Zhaoqing, where the actual monthly average rainfall from May to August was 21.13mm, 14.3mm, 8.05mm and 22.58mm, respectively. This suggests that a sudden drop of mosquito numbers in one month during the summer could be followed by a new increase or outbreak. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Results of sensitivity analysis in Shenzhen from 2016 to 2021. (a)-(f) Variation of adult mosquito quantity with increasing winter temperature in Shenzhen. The red dashed line from left to right represents the start and end times of diapause, and the black dashed line from left to right represents the ends of winter, spring, summer and autumn. (g)-(l) Effects of seasonal climate change on annual total adult mosquito quantity in Shenzhen. The three sets of data for each season in the subfigure show the results of increasing seasonal rainfall, increasing both seasonal rainfall and seasonal temperature, and increasing seasonal temperature from left to right. The "winter" in figure (g)-(l) represents the winter of the previous year. For example, "winter" in figure (g) represents the winter of 2015, including December 2015 and January and February 2016. https://doi.org/10.1371/journal.pntd.0011247.g003 PPT PowerPoint slide

PNG larger image

TIFF original image Download: Table 3. The aftermath durations of effects of seasonal climate change. https://doi.org/10.1371/journal.pntd.0011247.t003 From Figs 3 and S5–S7, we find that except for Yangjiang in winter 2015 and spring 2018 and 2019, Guangzhou in summer, and Huizhou, increasing seasonal temperature and rainfall both increase the annual total numbers of adult mosquitoes. When both are increased at the same time, the effect on the increase of mosquitoes is the most pronounced. Since the temperature in Guangzhou is mostly between 26°C and 32°C in summer and the development of larvae will be slowed down at this time, as the temperature increases in summer, this situation will be aggravated, and the number of adult mosquitoes will decrease [8,12]. Therefore, increasing temperatures in summer or increases of temperature and rainfall at the same time in Guangzhou do not have a large positive impact on the annual total adult mosquito numbers. The reason for these abnormal results of the sensitivity analysis in spring of Yangjiang in 2018 and 2019 is shown in the Supplementary Information (S13(A) Fig), and the reason for the abnormal results in Huizhou are probably related to the low rainfall. In addition, although increasing winter temperature increases the mosquito development rate, it shortens the diapause period, and a short diapause period will lead to less accumulation of diapause eggs during this period than if the diapause period was normal. Thus, this will result in a decrease in the adult mosquito numbers in the following spring within a certain range (S13(B) Fig), which will in turn lead to a decrease in the adult mosquito quantity throughout the year if the climate remains unchanged in other seasons. For example, the annual total adult mosquito quantity in Yangjiang decreased significantly when the temperature increased in the winter of 2015 (S5 Fig). So, the mild winters may weaken the diapause’s protective effect on mosquito populations and only have a positive effect when the winter is warm enough [11].

[END]
---
[1] Url: https://journals.plos.org/plosntds/article?id=10.1371/journal.pntd.0011247

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.