Modeling and projection of dengue fever cases in Guangzhou based on variation of weather factors

Modeling and projection of dengue fever cases in Guangzhou based on variation of weather factors

Science of the Total Environment 605–606 (2017) 867–873 Contents lists available at ScienceDirect Science of the Total Environment journal homepage:...

909KB Sizes 0 Downloads 145 Views

Science of the Total Environment 605–606 (2017) 867–873

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Modeling and projection of dengue fever cases in Guangzhou based on variation of weather factors Chenlu Li a,1, Xiaofeng Wang b,1, Xiaoxu Wu a,⁎, Jianing Liu a, Duoying Ji a, Juan Du c a b c

College of Global Change and Earth System Science, Beijing Normal University, Beijing 100875, China Center for Disease Surveillance and Information Services, Chinese Center for Disease Control and Prevention, Beijing 102206, China Key Laboratory of Environmental Change and Natural Disaster, Ministry of Education of China, Beijing Normal University, Beijing 100875, China

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Three weather factors affecting dengue fever are determined using metaanalysis. • A dengue model covered a long period is developed only from climate perspective. • Modeling is validated against newly reported dengue. • Projected dengue cases based on climate model data show a clear seasonality. • Seasonal disease control and emission mitigation may help reduce dengue incidence.

a r t i c l e

i n f o

Article history: Received 22 March 2017 Received in revised form 22 June 2017 Accepted 22 June 2017 Available online xxxx Editor: D. Barcelo Keywords: Dengue fever Weather factors Model Projection Guangzhou

a b s t r a c t Dengue fever is one of the most serious vector-borne infectious diseases, especially in Guangzhou, China. Dengue viruses and their vectors Aedes albopictus are sensitive to climate change primarily in relation to weather factors. Previous research has mainly focused on identifying the relationship between climate factors and dengue cases, or developing dengue case models with some non-climate factors. However, there has been little research addressing the modeling and projection of dengue cases only from the perspective of climate change. This study considered this topic using long time series data (1998–2014). First, sensitive weather factors were identified through metaanalysis that included literature review screening, lagged analysis, and collinear analysis. Then, key factors that included monthly average temperature at a lag of two months, and monthly average relative humidity and monthly average precipitation at lags of three months were determined. Second, time series Poisson analysis was used with the generalized additive model approach to develop a dengue model based on key weather factors for January 1998 to December 2012. Data from January 2013 to July 2014 were used to validate that the model was reliable and reasonable. Finally, future weather data (January 2020 to December 2070) were input into the model to project the occurrence of dengue cases under different climate scenarios (RCP 2.6 and RCP 8.5). Longer time series analysis and scientifically selected weather variables were used to develop a dengue model to ensure reliability. The projections suggested that seasonal disease control (especially in summer and fall) and mitigation of greenhouse gas emissions could help reduce the incidence of dengue fever. The results of this study hope to provide a scientifically theoretical basis for the prevention and control of dengue fever in Guangzhou. © 2016 Elsevier B.V. All rights reserved.

⁎ Corresponding authors at: College of Global Change and Earth System Science, Beijing Normal University, No. 19 Xinjiekouwai Street, Haidian District, Beijing 100875, China. E-mail address: [email protected] (X. Wu). 1 Co-first authors.

http://dx.doi.org/10.1016/j.scitotenv.2017.06.181 0048-9697/© 2016 Elsevier B.V. All rights reserved.

868

C. Li et al. / Science of the Total Environment 605–606 (2017) 867–873

1. Introduction Dengue fever is one of the most important arboviral diseases transmitted to humans by mosquitoes, especially in tropical and subtropical regions. According to the World Health Organization (WHO, 2010), there were only nine countries affected by this disease in the 1970s. However, with the gradual expansion of its areas of prevalence during the latter 25 years of the 20th century, this disease is now present in over 100 countries and affects 50–100 million people annually (Azil et al., 2011; Gubler, 2002). In mainland China, dengue fever has occurred frequently since the 1980s, primarily affecting southern provinces such as Fujian, Guangdong, Hainan, and Guangxi. Dengue fever is able to cause serious morbidity and mortality, placing heavy burdens on both families and health care systems (Gubler, 2012). Climate change impacts on the transmission of infectious disease in terms of three essential aspects: pathogens, hosts (or vectors), and transmission (Wu et al., 2016). The impact of climate change on dengue fever is not different in this respect. Climate change can affect the development, survival, and reproduction of dengue pathogens. Dengue viruses are sensitive to climate condition and changes in temperature as well as humidity can affect their replication and extrinsic incubation period (Thai and Anders, 2011; Wu et al., 2016). Climate conditions and weather factors can also affect the vectors of dengue. The mosquito Aedes albopictus is distributed widely in southern China, and it is considered the primary dengue vector in some parts of China (Fan et al., 1989; Gratz, 2004). The larvae development of these mosquitoes can be influenced considerably by both temperature and precipitation (Hoshen and Morse, 2004). For example, increased precipitation is able to provide additional breeding places, suitably high temperatures help extend the life cycle of the mosquitoes, and increased humidity is considered an important factor in the environmental conditions suitable for dengue vectors (Descloux et al., 2012; Tun-Lin et al., 2000). In addition, the transmission of dengue fever is also influenced by climate factors. For example, temperature and precipitation affect the longevity and biting behaviors of the female adult mosquitoes (Patz et al., 1998; Yang et al., 2009). Several studies have addressed the establishment of early warning systems of dengue fever based on weather factors (Eastin et al., 2014; Gharbi et al., 2011; Hii et al., 2012). They found different lagged relationships between the occurrence of dengue cases and weather factors such as temperature, humidity, and rainfall. Based on this, some early warning models were developed to predict the occurrence of dengue fever; however, the vectors used in these studies were Aedes aegypti, which are different to the vectors in Guangzhou. Additionally, a few studies have investigated the projection of potential dengue fever epidemics using vectorial capacity based on climate model data (Liu-Helmersson et al., 2016). Although the abovementioned studies suggest that climate factors play an important role in dengue fever incidence, some studies have found socioeconomic and environmental factors also affect transmission of dengue fever. For example, a study in Colombia showed that key socioeconomic factors affecting incidence of dengue fever included population density and socioeconomic stratum; and environmental factors included plant nurseries, sewage system and tire shops (Delmelle et al., 2016). Since 1978, dengue fever has occurred frequently in Guangzhou, the capital city of Guangdong Province in China. From 1998 to 2014, there were 40,837 dengue cases reported in Chinese Center for Disease Control and Prevention. Research on dengue fever in terms of climate change has become an active area of work. On the one hand, some studies have investigated the impact of weather factors on dengue fever in Guangzhou. For example, one study showed that minimum temperature and minimum humidity were both positively associated with the incidence of dengue fever at a lag of one month, whereas wind velocity was negatively associated with dengue fever over the same period (Lu et al., 2009). Other studies have identified correlations between climate factors and total dengue cases (Lu et al., 2009; Wang et al., 2014). On the other hand, a few studies have modeled the occurrence of dengue fever

based on climate variation. One study modeled dengue fever cases, and the model was validated against data from a dengue outbreak in 2013–2014 based on imported cases and weather variables (S. Sang et al., 2015). Another study developed a dengue case model based on weather factors and mosquito density in 2005–2012, and the model was used to predict cases in 2013–2015 (Xu et al., 2016). However, until now, there has been little research addressing modeled dengue cases only from the perspective of climate change, and projected the future incidence in long-time scale. This study used times series Poisson analysis with the generalized additive model (GAM) approach to model the occurrence of dengue cases based on weather factors over a long period (1998–2014). Based on a climate model, the future long-term (2020–2070) incidence of dengue fever was projected, so as to provide a theoretical basis for scientific guidance on its prevention. 2. Method 2.1. Study setting and data 2.1.1. Study setting Guangzhou, the capital city of Guangdong Province in southern China, is located at 22°26′–23°56′N, 112°57′–114°3′E (Fig. 1), and its population was 8.42 million in 2014 (2014 National Census). Guangzhou has a subtropical monsoon climate that is characterized as warm and rainy with adequate sunshine and heat. Annually, the hottest month is July, with an average temperature of 28.7 °C. The rainy season is from April to June and hot weather extends from July to September. Obviously, Guangzhou provides suitable climatic conditions for dengue vectors and a large number of dengue fever cases are reported annually (Luo et al., 2012). Therefore Guangzhou was chosen as the study area based on its representativeness in China in terms of dengue fever and climate change. 2.1.2. Data Dengue cases Details of dengue fever cases in Guangzhou from January 1998 to July 2014 were obtained from the Public Health Science Data Center. Data aggregation is at the county level in space, and at the month level in time. The dengue data used in our study is the total dengue cases of all counties within the Guangzhou City. Weather data Monthly weather data from January 1998 to July 2014 were retrieved from the China Meteorological Data Sharing Service System. In this study, the weather factors considered included monthly average temperature (MeanT), maximum temperature (MaxT), minimum temperature (MinT), monthly average cumulative precipitation (Pre), monthly average wind speed (MeanWind), monthly average relative humidity (MeanRh) and minimum relative humidity (MinRh), monthly average vapor pressure (MeanP), monthly average sunshine hours (SunHour), average water vapor pressure (WaterVapor), and evaporation (Eva). There are two meteorological stations in Guangzhou city (Fig. 1), and the weather data used was obtained from meteorological station A, which is most close to the urban area of the Guangzhou city (S.W. Sang et al., 2015). Climate model data To predict the number of dengue fever cases that might occur in the future, long-term weather data derived from climate prediction models were used. The model employed in this study was the Beijing Normal University Earth System Model (BNU-ESM) developed at the Beijing Normal University (Ji et al., 2014). The derivation was performed under four greenhouse gas emission pathways (RCP 2.6, RCP 4.5, RCP 6.0, and RCP 8.5). The four RCP scenarios describe the possible range of radiative forcing of greenhouse gases in 2100 (i.e., + 2.6, + 4.5, + 6.0, + 8.5 W m−2, respectively) (Weyant, 2009).

C. Li et al. / Science of the Total Environment 605–606 (2017) 867–873

869

Fig. 1. The location of Guangzhou City and the distribution of meteorological stations in Guangzhou.

For each emission pathway, Coupled Model Intercomparison Project phase 5 datasets from each scenario were used in this study. 2.2. Data analysis Meta-analysis is a type of statistical analysis combining the results of multiple scientific studies. It is designed for collections of empirical findings from individual studies for the purposes of integrating, synthesizing, and understanding them (Wolf, 1986). In this study, the metaanalysis comprised two aspects: literature review screening and correlation analysis. Correlation analysis was conducted to check the correlation between two variables. Here, the Spearman rank coefficient and Pearson correlation coefficient were used to perform the correlation analysis. In this study, a Spearman correlation analysis was performed between dengue fever cases with different lag time and each weather factor to identify the optimal lagged effect. Furthermore, the Pearson correlation analysis was performed between the different weather factors to check their collinearity. Generalized additive model (GAM) is able to model some or all of the independent variables using a smoothing function, which is suitable for the analysis of many types of distribution and complex nonlinear relations. Modeling of dengue fever cases and estimates of model parameters were implemented using the GAM. The specific model is written as (Hastie and Tibshirani, 1987): g ðEðYÞÞ ¼ β0 þ f 1 ðx1 Þ þ f 2 ðx2 Þ þ … þ f m ðxm Þ

ð1Þ

The model relates dependent variable Y and independent variables xi, and g represents a link function. The fi may be functions with a

specified parametric form or specified non-parametrically or semiparametrically simply as “smoothing functions”. The GAM is widely used to model count data. In this study, it was used to develop a model between dengue fever cases and weather factors. 3. Results 3.1. Identifying key weather factors Meta-analysis was used to identify weather factors to which dengue fever incidence is sensitive. First, preliminary screening was performed based on literature reviews using 11 weather factors. Then, 8 weather factors were selected considering their association with dengue fever incidence: MeanP, MeanT, MaxT, MinT, MeanRh, MinRh, Pre, and MeanWind (Choi et al., 2016; Lu et al., 2009; Sang et al., 2014; Wang et al., 2014). Lagged analysis with lags of 0–3 months was performed to investigate the relationship between the incidence of dengue fever cases (1998–2014) and each of the weather factors. The results of the lagged analysis were shown in Table 1. It was found that MeanT, MaxT, and MinT at lags of 0–3 months were positively associated with monthly occurrence of dengue cases, especially at a lag of 2 months. MeanP was inversely correlated with dengue fever cases at lags of 0–3 months, MeanRh and Pre both showed relatively strong correlations with dengue fever cases at a lag of 3 months, and MeanWind had slight correlation with dengue fever cases but without statistical significance. Therefore, 7 weather factors (MeanT, MaxT, MinT, Pre, MeanRh, MinRh, and MeanP) were chosen for subsequent analysis. Some weather factors showed strong collinear relationships, e.g., MeanP and MeanT (Guo et al., 2016). To determine the key weather

870

C. Li et al. / Science of the Total Environment 605–606 (2017) 867–873

Table 1 Spearman correlation coefficients between dengue fever cases and weather factors in Guangzhou, China, 1978–2014. Lagged month

Lag 0

Lag 1

Lag 2

Lag 3

−0.336⁎⁎ 0.364⁎⁎ 0.403⁎⁎ 0.339⁎⁎

−0.570⁎⁎ 0.543⁎⁎ 0.567⁎⁎ 0.531⁎⁎ 0.185⁎⁎

−0.649⁎⁎ 0.566⁎⁎ 0.574⁎⁎ 0.571⁎⁎ 0.362⁎⁎

−0.024 0.295⁎⁎

0.224 0.400⁎⁎ 0.206

−0.563⁎⁎ 0.409⁎⁎ 0.400⁎⁎ 0.425⁎⁎ 0.452⁎⁎ 0.364⁎⁎ 0.415⁎⁎ 0.260⁎⁎

Weather factors MeanP MeanT MaxT MinT MeanRh MinRh Pre MeanWind

−0.011 −0.163⁎⁎ 0.093 0.096

0.137

⁎⁎ p b 0.05.

factors, correlation analysis was performed to check the collinearity between the seven selected weather factors (Table 2). It is evident that strong collinear correlations existed between some variables with correlation coefficients ranging from 0.84 to 0.995 for MeanT and MeanP, MeanT and MinT, and MeanT and MaxT. To avoid collinearity, MeanP, MaxT, and MinT were excluded from the final model. Although the correlation coefficients between MinRh and MeanRh was a little high (rw = 0.500), the correlation between MinRh and the occurrence of dengue fever cases was too low (rd = 0.364). Therefore, MinRh was also excluded from the key weather factors. Based on the optimal lagged effect and non-collinearity, 3 key weather factors were chosen for the modeling: MeanT (lag 2 months), MeanRh (lag 3 months), and Pre (lag 3 months). 3.2. Modeling of occurrence of dengue fever cases The monthly occurrence of dengue fever cases was modeled using time series Poisson analysis with the GAM approach. The model incorporated monthly dengue fever cases as the dependent variable, and MeanT (lag 2 months), MeanRh (lag 3 month), Pre (lag 3 month), and dengue fever cases in the previous month as the independent variables. The year and month were included to control the influence of long-term trends and potential seasonality, respectively. In this study, the quasiPoisson link function was adopted to correct for over-dispersion. Unbiased risk estimation was used to identify the most suitable model for differing degrees of freedom (Wood, 2000). The developed model, based on data from January 1998 to December 2012, can be written as: ln ðcaset Þ ¼ β0 þ sðMeanT t−2 ; df Þ þ sðMeanRht−3 ; df Þ þ sð ln ðPret−3 Þ; df Þ þ Yr þ Mon þ sð ln ðcaset−1 Þ; df Þ; where ln(case) denotes the monthly average occurrence of dengue fever cases with logarithmic transformation, function s() represents natural cubic splines to explain the potential nonlinear exposure– response relationship between the occurrence of dengue fever cases and the weather variables, df denotes the degree of freedom, ln(Pre) represents the monthly average cumulative precipitation with Table 2 Pearson correlation coefficients between the selected seven weather factors in Guangzhou, China, 1978–2014. Weather factor

MeanP

MeanT

MaxT

MinT

MeanRh

MinRh

Pre

MeanP MeanT MaxT MinT MeanRh MinRh Pre

1.000 −0.840 −0.820 −0.850 −0.490 −0.200 −0.590

−0.840 1.000 0.990 0.995 0.350 0.150 0.530

−0.820 0.990 1.000 0.970 0.250 0.089 0.470

−0.850 0.995 0.970 1.000 0.410 0.190 0.570

−0.490 0.350 0.250 0.410 1.000 0.500 0.570

−0.200 0.150 0.089 0.190 0.500 1.000 0.330

−0.590 0.530 0.470 0.570 0.570 0.330 1.000

logarithmic transformation, Yr denotes the year, and Mon represents a calendar month. The modeling results showed a good fit for which the value of R2 was 0.99 and 97.3% of the variance was explained. To ensure the scientific credibility of the developed model, data from January 2013 to July 2014 were used for validation. Fig. 2 shows a comparison between the reported and predicted incidences of dengue fever cases. Overall, the two data series have similar trends, and in some periods (January–May 2013, July–August 2013, November–December 2013, and February–April 2014), they exhibit excellent agreement. However, some discrepancies are evident. One noticeable difference occurs during September–October 2013 and May–July 2014, corresponding to high numbers of reported dengue fever cases. It is considered that this phenomenon might be related to the Artificial Lake Project in Guangzhou. The water area in the city has been expanding since the Artificial Lake Project was implemented in 2011. A previous study found that surface water area in Guangzhou was not associated with the amount of precipitation in 2013 and 2014; when precipitation was reduced, the surface water area even increased (Tian et al., 2016). As known to all, surface water is essential for the survival of mosquitoes because it provides a suitable breeding environment. An increase in the number of mosquitoes, as vectors of the dengue virus, contributes to enhanced transmission of dengue fever. The predicted occurrence of dengue fever cases considered precipitation as the only water source related to the vectors. These two factors might lead to lower predictions of dengue fever cases compared with reported results. Another difference exists for period January 2014. The model predicted three dengue fever cases in January 2014, whereas none was actually reported. In November 2013, typhoon “Swallow” swept across Guangzhou bringing heavy rain (Cai et al., 2014). Because of the lagged effect, rainfall is predicted by the model to have positive impact on the incidence of dengue fever in January 2014. However, the influence of rainfall on dengue fever incidence is not always the same depending on its intensity (Choi et al., 2016). Heavy rain can damage the habitats of mosquitoes and flush away their larvae and eggs (Koenraadt and Harrington, 2008), which could lead to a reduction in the number of actual dengue fever cases and cause some disagreement between the predicted and reported results. In general, the predicted results are considered to illustrate that the model is reliable and credible.

3.3. Projection of future dengue fever cases Future weather data for Guangzhou were input into the developed model as independent variables (MeanT (lag 2 months), MeanRh (lag 3 months), and Pre (lag 3 months)) to project the future occurrence of dengue fever cases. The future weather factors provided by the BNUESM comprised MeanT, MeanRh, and Pre from 2020 to 2070. The projected occurrences of dengue fever cases were calculated for four emission pathways: RCP 2.6, RCP 4.5, RCP 6.0, and RCP 8.5, where RCP 2.6 and RCP 8.5 reflect scenarios with relatively lower and higher CO2 emissions, respectively. A selection of outputs of projected dengue fever cases under the RCP 2.6 and RCP 8.5 scenarios is shown in Table 3 to illustrate the changes in potential occurrence of dengue fever cases. The frequency statistics of potential dengue fever cases (Table 3) illustrates that the incidence of dengue fever is distinct under the different climate scenarios. Under climate scenario RCP 2.6, the overall incidence of dengue fever is low and the occurrence of high numbers of cases is obviously less. Under climate scenario RCP 8.5, both the overall incidence and the occurrence of high numbers of cases increase. Therefore, greater numbers of dengue fever cases are likely to occur under RCP 8.5, which is consistent with the findings of a previous study in Europe (Liu-Helmersson et al., 2016). The reason for this is that the increase in CO2 emissions under scenario RCP 8.5 provides warmer living conditions for A. albopictus, and an increase in the number of A. albopictus increases the incidence of dengue fever cases.

C. Li et al. / Science of the Total Environment 605–606 (2017) 867–873

871

Fig. 2. Comparison of reported and predicted dengue fever cases (January 2013 to July 2014).

Table 3 also shows the obviously seasonal occurrence of potential dengue fever cases under future weather patterns. The seasons of high incidence of dengue fever under each scenario are summer and fall. The subtropical climatic conditions of Guangzhou lead to higher temperatures in summer, and the abundant rainfall and high humidity provide a good habitat for mosquito breeding and survival. As a vector of the dengue virus, mosquitoes are important for the transmission of dengue fever; thus, the incidence of projected dengue fever cases under RCP 8.5 is higher than under RCP 2.6 in most seasons. This study was most concerned about the potential number of dengue cases in the coming decade. Therefore, the occurrences of dengue fever cases from 2020 to 2030 were projected under the different emission scenarios (Fig. 3). The incidence of dengue fever exhibits obvious seasonality with a peak occurring mainly in summer and fall. Except for a few periods (i.e., fall 2027, and summer 2030) where the predicted numbers of dengue fever cases under scenarios RCP 2.6 and RCP 8.5 are the same, the two results show large differences. Overall, both have similar trends, but scenario RCP 8.5 cases have higher peaks and the overall incidence of dengue fever is greater. 4. Discussions and conclusions 4.1. Discussions By screening based on a literature review, this study selected 8 out of 11 weather factors to which the occurrence of dengue fever was found sensitive. Then, through lagged analysis, 7 climate factors were screened: MeanT, MaxT, MinT, Pre, MeanRh, MinRh, and MeanP. Based on this screening, collinear analysis was performed and 3 key weather factors determined (MeanT, MeanRh, and Pre). Supporting literature is abundant in this respect. For example, the results of one study indicated that the occurrence of local dengue fever cases was correlated positively with both temperature and precipitation but with different time lags (Sang et al., 2014). Another study argued that the incidences of dengue fever were correlated positively with a number of weather factors including relative humidity (Wang et al., 2014). The results of the current study were compared with related research in other areas. For example, monthly mean temperature and mean precipitation

were found significantly correlated with dengue fever incidence in three provinces of Cambodia, a dengue-endemic country in Southeast Asia (Choi et al., 2016). This finding is similar to the current study. However, there is also disagreement with other studies. Two previous studies indicated that temperature and rainfall affect dengue fever transmission independent of relative humidity (S. Sang et al., 2015; Xu et al., 2016). In fact, relative humidity is a very important weather factor regarding dengue fever cases. It influences the survival of mosquito eggs and adults, biting behaviors of female adult mosquitoes, and egg-laying behavior (Azil et al., 2010). A different study found that 75% relative humidity is optimal for preserving eggs, and that egg hatchability of A. albopictus is correlated positively with relative humidity at the same temperature (Wang et al., 2014). Compared with previous studies, the current study represents improvements in three aspects. First, this study covered a long-time span to determine the relationship between climate factors and dengue fever cases, making the analysis more reliable. Second, based on metaanalysis, this study adopted a stepwise procedure to select key weather factors to develop the model for dengue fever occurrence from a climate change perspective only. Previous studies on warning models have determined weather variables and their lagged effects according to their performance in modeling (S. Sang et al., 2015). To provide a comprehensive and scientific explanation of the incidence of dengue fever, the model developed in this study incorporated three weather factors (MeanT, MeanRh, and Pre), which is more than existing warning model systems (Xu et al., 2016). The validation based on data from January 2013 to July 2014 verified the credibility of the developed model. Third, predicted weather factors (2020–2070) provided by the BNUESM were input into the model to project the incidence of dengue fever in the future. Within the study area, the incidence of projected dengue fever cases exhibited obvious seasonality with peak occurrence in summer and fall. The projected incidence of dengue fever showed a similar seasonal pattern to a previous study (Wu et al., 2010). So, it is important that attention be given to seasonal prevention and control of dengue fever, especially in summer and fall. In addition, different frequencies of incidence were found for potential dengue fever cases under two emission scenarios. Scenarios RCP 2.6 and RCP 8.5 represent states with relatively lower and higher emissions of greenhouse gases,

Table 3 Statistics of frequency and seasonality of potential dengue fever cases under different emission pathways in Guangzhou, China (2020–2070). Cases

Frequency

Seasonality

Model

No case (ln(cases) = 0)

Low cases (0 b ln(cases) b 6)

High cases (ln(cases) ≥ 6)

Spring (Mar–May)

Summer (Jun–Aug)

Fall (Sep–Nov)

Winter (Dec–Feb)

RCP 2.6 RCP 8.5

318 281

170 176

150 185

25 25

124 125

119 139

52 72

872

C. Li et al. / Science of the Total Environment 605–606 (2017) 867–873

Fig. 3. Dengue fever cases for the coming decade (2021−2030) projected under greenhouse gas emission scenarios RCP 2.6 and RCP 8.5.

respectively. The projected incidence of dengue fever under scenario RCP 2.6 was less and with a lower peak than under scenario RCP 8.5. Under higher emission scenarios, the potential for the occurrence of dengue fever would increase. Our findings agree well with some previous studies. For example, a study in South China indicated that compared with the average in the period of 1997–2012, the annual mean days suitable for dengue fever transmission in 2020s, 2050s and 2080s will increase by 15, 25 and 40 days under RCP 8.5 scenario, and the areas of year-round dengue fever epidemic region will likely increase by 4536, 8780 and 20,680 km2 under RCP 8.5 scenario (Du et al., 2015). Therefore, it can be concluded that controlling greenhouse gas emissions could contribute to a reduction in the occurrence of dengue fever cases in the future. Even though the occurrence of dengue fever in Guangzhou is also correlated with other environmental factors (Tian et al., 2016), weather factors play particularly important roles in its transmission. Based on this study, the following two suggestions should be considered when regarding the prevention of dengue fever in Guangzhou. First, seasonal prevention is preferred. During the seasons of high incidence of dengue fever (i.e., summer and fall), people should pay greater attention especially to personal mosquito control, e.g., spraying mosquito repellent, and adopting proper garbage disposal and improved water storage measures (Gratz, 1991). Second, human activities that contribute to the mitigation of greenhouse gas emissions should be encouraged because they have great importance regarding the reduction of future potential dengue fever cases. For example, the development of public transport systems and the substitution of fossil fuels with bio-ethanol production should be encouraged (Bubeck et al., 2014; Balan et al., 2013). This study developed an early warning model for the occurrence of dengue fever cases based on weather factors. However, it should be noted that the model does have some limitations. For example, ecological factors such as mosquito abundance, land cover, and the proximity of water bodies were not considered. Furthermore, social factors such as population movement, importation of dengue fever cases, and human prevention measures are also important regarding dengue fever occurrence. Therefore, how to build a widely transferable model also considering socio-economic and ecological environment factors needs further research. This research used long-term dengue fever cases data which are mostly collected in the period with high incidence of dengue fever. The data covering longer historical period will make the model more representative. Besides, using a single climate model also introduces uncertainty in predicting future climate change. However, the quantitative results derived from this study will be helpful in advancing the understanding of how climate can influence infectious diseases and how best to control and prevent the occurrence of dengue fever in practice.

4.2. Conclusions Since the first outbreak of dengue fever in mainland China in 1978, its occurrence has been reported frequently. The climate of Guangzhou, which is typical of areas affected by dengue fever, is characterized as mild and dry in winter and relatively hot and wet in summer. The transmission of dengue fever is very sensitive to climate change, so the objective of this study was to model dengue fever incidence based on weather factors. Based on a literature review, lagged analysis, and collinear analysis, this study identified 3 key variables from 11 weather factors. Then, a model of dengue fever occurrence was developed based on long time series data (1998–2012). This model incorporated the monthly occurrence of dengue fever cases as the dependent variable and the key weather factors as the independent variables. The validity of the model was verified using data from January 2013 to July 2014. Finally, the future occurrence of dengue fever cases was projected based on the developed model and climate model data. The results of this research established that MeanT (lag 2 months), MeanRh (lag 3 months), and Pre (lag 3 months) were correlated positively with dengue fever cases. The dengue fever model developed on long-time scale was reliable and reasonable. The projected occurrence of dengue fever cases by the model demonstrated obvious seasonal characteristics and frequency variation under different climate scenarios (i.e., RCP 2.6 and RCP 8.5). Compared with previous studies, this study modeled the occurrence of dengue fever cases only from a climate perspective based on the long time series data. We modeled dengue cases based on determined key factors through meta-analysis. Furthermore, the developed model incorporated more climate factors than existing dengue fever warning models. This ensured the developed model had greater reliability in predicting the potential incidence of dengue fever. The projected occurrence of dengue fever cases (2020–2070) was calculated using the developed model. According to the projection of future dengue fever cases, two recommendations were offered regarding the prevention of dengue fever: seasonal disease control and mitigation of greenhouse gas emissions. Through this study addressed the occurrence of dengue fever, we hope to provide a theoretically scientific basis and guidance for government regarding the prevention and control of dengue fever in Guangzhou. Acknowledgements This work was supported by the National Key Research and Development Program of China (NO. 2016YFA0600104) and the Fundamental Research Funds for the Central Universities. The authors greatly appreciate Dr. Rui Mao at Beijing Normal University for help with obtaining

C. Li et al. / Science of the Total Environment 605–606 (2017) 867–873

climate data, and Dr. Qizhong Wu at Beijing Normal University for help with the climate model. References Azil, A.H., Long, S.A., Ritchie, S.A., Williams, C.R., 2010. The development of predictive tools for pre-emptive dengue vector control: a study of Aedes aegypti abundance and meteorological variables in North Queensland, Australia. Tropical Med. Int. Health 15, 1190–1197. Azil, A.H., Li, M., Williams, C.R., 2011. Dengue vector surveillance programs: a review of methodological diversity in some endemic and epidemic countries. Asia Pac. J. Public Health 23, 827–842. Balan, V., Chiaramonti, D., Kumar, S., 2013. Review of US and EU initiatives toward development, demonstration, and commercialization of lignocellulosic biofuels[J]. Biofuels, Bioprod. Biorefin. 7 (6), 732–759. Bubeck, S., Tomaschek, J., Fahl, U., 2014. Potential for mitigating greenhouse gases through expanding public transport services: A case study for Gauteng Province, South Africa[J]. Transp. Res. Part D: Transp. Environ. 32, 57–69. Cai, J., Tian, Y., Fu, W., 2014. “Haiyan” caused serious damage to the Philippines enlightens China's typhoon defense. China Flood Drought Manag. 76–78 (in Chinese). Choi, Y., Tang, C.S., McIver, L., Hashizume, M., Chan, V., Abeyasinghe, R.R., et al., 2016. Effects of weather factors on dengue fever incidence and implications for interventions in Cambodia. BMC Public Health 16, 241. Delmelle, E., Hagenlocher, M., Kienberger, S., Casas, I., 2016. A spatial model of socioeconomic and environmental determinants of dengue fever in Cali, Colombia. Acta Trop. 164, 169–176. Descloux, E., Mangeas, M., Menkes, C.E., Lengaigne, M., Leroy, A., Tehei, T., et al., 2012. Climate-based models for understanding and forecasting dengue epidemics. PLoS Negl. Trop. Dis. 6. Du, Y., Wu, X., Wang, H., 2015. Change in temperature in South China and its impact on the potential epidemic duration of dengue fever. Chin. J. Ecol. 34, 3174–3181 (in Chinese). Eastin, M.D., Delmelle, E., Casas, I., Wexler, J., Self, C., 2014. Intra- and interseasonal autoregressive prediction of dengue outbreaks using local weather and regional climate for a tropical environment in Colombia. Am. J. Trop. Med. Hyg. 91, 598–610. Fan, W.F., Yu, S.R., Cosgriff, T.M., 1989. The reemergence of dengue in China. Rev. Infect. Dis. 11, S847–S853. Gharbi, M., Quenel, P., Gustave, J., Cassadou, S., La Ruche, G., Girdary, L., et al., 2011. Time series analysis of dengue incidence in Guadeloupe, French West Indies: forecasting models using climate variables as predictors. BMC Infect. Dis. 11. Gratz, N.G., 1991. Emergency control of Aedes aegypti as a disease vector in urban areas. J. Am. Mosq. Control Assoc. 7, 353–365. Gratz, N.G., 2004. Critical review of the vector status of Aedes albopictus. Med. Vet. Entomol. 18, 215–227. Gubler, D.J., 2002. The global emergence/resurgence of arboviral diseases as public health problems. Arch. Med. Res. 33, 330–342. Gubler, D.J., 2012. The economic burden of dengue. Am. J. Trop. Med. Hyg. 86, 743–744. Guo, C., Yang, J., Guo, Y., Ou, Q.-Q., Shen, S.-Q., Ou, C.-Q., et al., 2016. Short-term effects of meteorological factors on pediatric hand, foot, and mouth disease in Guangdong, China: a multi-city time-series analysis. BMC Infect. Dis. 16. Hastie, T., Tibshirani, R., 1987. Generalized additive-models - some applications. J. Am. Stat. Assoc. 82, 371–386. Hii, Y.L., Zhu, H.P., Ng, N., Ng, L.C., Rocklov, J., 2012. Forecast of dengue incidence using temperature and rainfall. PLoS Negl. Trop. Dis. 6. Hoshen, M.B., Morse, A.P., 2004. A weather-driven model of malaria transmission. Malar. J. 3.

873

Ji, D., Wang, L., Feng, J., Wu, Q., Cheng, H., Zhang, Q., et al., 2014. Description and basic evaluation of Beijing Normal University Earth System Model (BNU-ESM) version 1. Geosci. Model Dev. 7, 2039–2064. Koenraadt, C.J.M., Harrington, L.C., 2008. Flushing effect of rain on container-inhabiting mosquitoes Aedes aegypti and Culex pipiens (Diptera: Culicidae). J. Med. Entomol. 45, 28–35. Liu-Helmersson, J., Quam, M., Wilder-Smith, A., Stenlund, H., Ebi, K., Massad, E., et al., 2016. Climate change and Aedes vectors: 21st century projections for dengue transmission in Europe. Ebiomedicine 7, 267–277. Lu, L., Lin, H., Tian, L., Yang, W., Sun, J., Liu, Q., 2009. Time series analysis of dengue fever and weather in Guangzhou, China. BMC Publ. Health 9, 395. Luo, L., Liang, H.Y., Hu, Y.S., et al., 2012. Epidemiological, virological, and entomological characteristics of dengue from 1978 to 2009 in Guangzhou, China[J]. J. Vector Ecol. 37 (1), 230–240. Patz, J.A., Martens, W.J.M., Focks, D.A., Jetten, T.H., 1998. Dengue fever epidemic potential as projected by general circulation models of global climate change. Environ. Health Perspect. 106, 147–153. Sang, S., Yin, W., Bi, P., Zhang, H., Wang, C., Liu, X., et al., 2014. Predicting local dengue transmission in Guangzhou, China, through the influence of imported cases, mosquito density and climate variability. PLoS One 9, e102755. Sang, S., Gu, S., Bi, P., Yang, W., Yang, Z., Xu, L., et al., 2015a. Predicting unprecedented dengue outbreak using imported cases and climatic factors in Guangzhou, 2014. PLoS Negl. Trop. Dis. 9, e0003808. Sang, S.W., Gu, S.H., Bi, P., Yang, W.Z., Yang, Z.C., Xu, L., et al., 2015b. Predicting unprecedented dengue outbreak using imported cases and climatic factors in Guangzhou, 2014. PLoS Negl. Trop. Dis. 9. Thai, K.T.D., Anders, K.L., 2011. The role of climate variability and change in the transmission dynamics and geographic distribution of dengue. Exp. Biol. Med. 236, 944–954. Tian, H., Huang, S., Zhou, S., Bi, P., Yang, Z., Li, X., et al., 2016. Surface water areas significantly impacted 2014 dengue outbreaks in Guangzhou, China. Environ. Res. 150, 299–305. Tun-Lin, W., Burkot, T.R., Kay, B.H., 2000. Effects of temperature and larval diet on development rates and survival of the dengue vector Aedes aegypti in north Queensland, Australia. Med. Vet. Entomol. 14, 31–37. Wang, C., Jiang, B., Fan, J., Wang, F., Liu, Q., 2014. A study of the dengue epidemic and meteorological factors in Guangzhou, China, by using a zero-inflated Poisson regression model. Asia Pac. J. Public Health 26, 48–57. Weyant, J., 2009. IPCC. Vol. 2015 Switzerland, IPCC Secretariat, Geneva. WHO, 2010. The World Health Assembly Resolution on Dengue Fever and Dengue Haemorrhagic Fever Prevention and Control. Wolf, F.M., 1986. Meta-Analysis: Quantitative Methods for Research Synthesis. Vol. 59. SAGE Publications, Inc. Wood, S.N., 2000. Modelling and smoothing parameter estimation with multiple quadratic penalties. J. R. Stat. Soc. Ser. B (Stat. Methodol.) 62, 16. Wu, J.Y., Lun, Z.R., James, A.A., Chen, X.G., 2010. Review: dengue fever in mainland China. Am. J. Trop. Med. Hyg. 83, 664–671. Wu, X.X., Lu, Y.M., Zhou, S., Chen, L.F., Xu, B., 2016. Impact of climate change on human infectious diseases: empirical evidence and human adaptation. Environ. Int. 86, 14–23. Xu, L., Stige, L.C., Chan, K.S., Zhou, J., Yang, J., Sang, S., et al., 2016. Climate variation drives dengue dynamics. Proc. Natl. Acad. Sci. U.S.A. Yang, H.M., Macoris, M.L.G., Galvani, K.C., Andrighetti, M.T.M., Wanderley, D.M.V., 2009. Assessing the effects of temperature on the population of Aedes aegypti, the vector of dengue. Epidemiol. Infect. 137, 1188–1202.