Temperature impacts on dengue emergence in the United States: Investigating the role of seasonality and climate change

Abstract

Tropical mosquito-borne viruses have been expanding into more temperate regions in recent decades. This is partly due to the coupled effects of temperature on mosquito life history traits and viral infection dynamics and warming surface temperatures, resulting in more suitable conditions for vectors and virus transmission. In this study, we use a deterministic ordinary differential equations model to investigate how seasonal and diurnal temperature fluctuations affect the potential for dengue transmission in six U.S. cities. We specifically consider temperature-dependent mosquito larval development, adult mosquito mortality, and the extrinsic incubation period of the virus. We show that the ability of introductions to lead to outbreaks depends upon the relationship between a city's temperature profile and the time of year at which the initial case is introduced. We also investigate how the potential for outbreaks changes with predicted future increases in mean temperatures due to climate change. We find that climate change will likely lead to increases in suitability for dengue transmission and will increase the periods of the year in which introductions may lead to outbreaks, particularly in cities that typically have mild winters and warm summers, such as New Orleans, Louisiana, and El Paso, Texas. We discuss our results in the context of temperature heterogeneity within and across cities and how these differences may impact the potential for dengue emergence given present day and predicted future temperatures.

Keywords: Aedes aegypti; Arbovirus; Climate change; Dengue; Emergence; Puerto Rico; Temperature; United States.

Fig. A.8.

The fraction of each serotype of cases among all cases reported in (NOAA, 2018) between 1999–2008.

Fig. A.9.

Model fits (black curves) to reported dengue case data (all serotypes) each week of the dengue season in San Juan (blue dots).

Fig. A.10.

Model fits (black curves) to reported dengue case data (DENV-3) each week of the dengue season in San Juan (blue dots).

Fig. A.11.

Histograms of parameters from the parameter selection process. (A) The density dependent parameter a, (B) The initial Susceptible population, S_h(0), and (C) The product of a and S_h(0). The vertical dashed black line in each figure indicates the values used throughout this modeling study.

Fig. B.17.

Cumulative number of cases reported following a single introduction on the day given on the horizontal axis under current average yearly temperatures (solid line) and when average yearly temperatures are increased by 0.5 °C (dashed line), 0.5 °C (dotted line), or 1.5 °C (dash-dotted line). All parameters are the default values listed in Tables 1–2. Average yearly temperatures (M₁ in the top half of Table 2) are increased by the values given in the legend of this figure.

Fig. B.18.

Cumulative number of cases reported following a single introduction on the day given on the horizontal axis under current average yearly temperatures (solid line) and when variation in seasonal temperatures is changed by a factor of 0.8 (dotted line) and 1.2 (dashed line). These values correspond to a 20% decrease and 20% increase in variation, respectively. All parameters are the default values listed in Tables 1–2. Variation in seasonal temperatures (A₁ in Table 2, Eq. (1)) are changed by multiplying A₁ by the values given in the legend of this figure.

Fig. C.19.

Two potential characterizations of average adult vector lifespan as they relate to temperature. The solid curve is obtained from (Brady et al., 2014) and is used throughout this work. The dashed curve is obtained from (Yang et al., 2009).

Fig. C.20.

Cumulative number of cases reported following a single introduction on the day given on the horizontal axis when adult vector survival was desribed by (Brady et al., 2014) (solid line) or (Yang et al., 2009) (dashed line). All parameters are the default values listed in Tables 1–2. For these simulations, temperature varies seasonally only.

Fig. 1.

Temperature-dependent life history traits of Ae. aegypti. (A) Average larval development time as obtained from (Rueda et al., 1990). (B) Average adult vector lifespan obtained from (Brady et al., 2014).

Fig. 2.

Dengue infection dissemination and average EIP at different temperatures. The fraction of total mosquitoes censored at each time point that have disseminated infection (black circles and error bars that represent mean and standard error) for three different temperatures (A) 26°C, (B) 28°C, and (C) 30°C. In each panel (A)-(C), colored lines represent 100 trajectories generated by bootstrap sampling of the means. (D) The average EIP as a function of temperature. Circles represent the means obtained from each fit of the gamma distribution to the data, and the blue curve is Eq. (4) fit to the average EIP data points.

Fig. 3.

The impacts of seasonal fluctuations of temperature on demographic, viral, and epidemiological measures. Fluctuations in the average daily temperature (T(t)) (A), average EIP (ω(T)) (B), vector-host ratio (Eq. 2.8) (C), average adult vector lifespan (L_v (T)) (D), and the instantaneous probability of surviving the EIP (Eq.(5)) (E). All parameters are the default values given in Tables 1–2.

Fig. 4.

Cumulative number of cases reported within one year following an initial introduction on the day listed on the horizontal axis. A log₁₀ scale is used to emphasize comparisons among cities. Note that the curves in this figure are shown in a linear scale below as part of Fig. 5. All parameters are the default values given in Tables 1–2.

Fig. 5.

Cumulative number of cases reported following a single introduction on the day given on the horizontal axis in the presence of seasonal temperature fluctuations (solid line) and seasonal and diurnal fluctuations (dashed line). All parameters are the default values listed in Tables 1. Parameter values for the different temperature characterizations are given in Table 2.

Fig. 6.

Cumulative number of cases reported following a single introduction on the day given on the horizontal axis under current average yearly temperatures (solid line) and when average yearly temperatures are increased by 0.5 °C (dashed line), 1 °C (dotted line), or 1.5 °C (dash-dotted line). All parameters are the default values listed in Tables 1–2. Average yearly temperatures (M₁ in the top half of Table 2) are increased by the values given in the legend of this figure.

Fig. 7.

Timing of the initial introduction that leads to reported outbreak sizes of greater than (A) 1, (B) 10, or (C) 100 total cases under three different average temperature assumptions: current temperatures, and when average yearly temperatures are increased by 0.5 °C, 1 °C, or1.5 °C. All parameters are the default values listed in Tables 1–2. Average yearly temperatures (M₁ in the top half of Table 2) are increased by the values given in the legend of this figure. Note that Los Angeles was excluded from this analysis since none of these scenarios led to more than one case.