Climate variation drives dengue dynamics

Abstract

Dengue, a viral infection transmitted between people by mosquitoes, is one of the most rapidly spreading diseases in the world. Here, we report the analyses covering 11 y (2005-2015) from the city of Guangzhou in southern China. Using the first 8 y of data to develop an ecologically based model for the dengue system, we reliably predict the following 3 y of dengue dynamics-years with exceptionally extensive dengue outbreaks. We demonstrate that climate conditions, through the effects of rainfall and temperature on mosquito abundance and dengue transmission rate, play key roles in explaining the temporal dynamics of dengue incidence in the human population. Our study thus contributes to a better understanding of dengue dynamics and provides a predictive tool for preventive dengue reduction strategies.

Keywords: climate; dengue; prediction; structural equation model; zero-inflated generalized additive models.

Fig. 1.

Spatiotemporal dynamics of dengue in mainland China from the beginning of 2005 to the end of 2015. (A) Geographical distribution of dengue incidence in China. Radiuses of circles indicate incidence of dengue. Red circles represent dengue in Guangzhou, black circles represent dengue in other cities in Guangdong province, and gray circles represent dengue in other provinces. (B) Monthly time series of dengue incidence in Guangzhou. (C) Monthly time series of adult population density of the dengue vector A. albopictus. (D) Monthly time series of average maximum temperature in Guangzhou. (E) Monthly time series of number of days with rainfall in Guangzhou.

Fig. 2.

SEM analysis revealed direct and indirect climate effects on dengue incidence in Guangzhou from 2005 to 2015. Arrows with numbers indicate ecological effects and standardized coefficients. Temperature and precipitation are correlated (correlation coefficient = 0.39). Asterisks indicate statistically significant pathways (P < 0.05).

Fig. S1.

Test of zero inflation. Histogram plots of monthly mosquito density and dengue incidence data.

Fig. 3.

Analysis of potentially nonlinear influences on dengue incidence in Guangzhou based on data from years 2005−2012, i.e., excluding the three last years of data. To account for zero inflation, (A−D) a separate binomial submodel quantifies predictor effects on outbreak risk (logit scale probability of incidence > 0) and (E−H) a lognormal submodel quantifies predictor effects on outbreak intensity when an outbreak occurs [ln(incidence)].

Fig. 4.

Observations and 1-mo-ahead predictions. (A) Adult mosquito density. (B) Dengue incidence. The vertical lines separate the years 2005−2012 from the years 2013−2015 over which out-of-sample predictions were made.

Fig. 5.

Analysis of potentially nonlinear influences on dengue incidence in Guangzhou based on data from all years 2005−2015. To account for zero inflation, (A−D) a separate binomial submodel quantifies predictor effects on outbreak risk (logit scale probability of incidence > 0) and (E−H) a lognormal submodel quantifies predictor effects on outbreak intensity when an outbreak occurs [ln(incidence)].

Fig. S2.

The initially selected ZIGAM of dengue incidence based on data from 2005 to 2012. To account for zero inflation, (A–D) a separate binomial submodel quantifies predictor effects on outbreak risk (logit scale probability of incidence > 0) and (E–H) a lognormal submodel quantifies predictor effects on outbreak intensity when an outbreak occurs [ln(incidence)]. The effect of average maximum temperature in G is likely overfitted.