We tested the effect of temperature on a number of life-history traits of Ae. aegypti over a series of six experiments. Each experiment assessed larval development time and survival, and a number of female reproductive traits, including the pre-blood meal period, length of the gonotrophic cycle, and the number of eggs in the first gonotrophic cycle (assays are described in detail below). Temperatures tested ranged between 12°C and 40°C, with 4–6°C intervals. After first identifying upper and lower thermal limits of development and reproduction under constant temperatures, we then explored the effect of diurnal fluctuations around these minimum and maximum temperatures. Low temperatures were tested together (Experiment 1∶12°C, 16°C and 20°C), as were higher temperatures (Experiment 2∶30°C, 35°C and 40°C). Between the temperatures of 35°C and 40°C, we needed greater resolution to identify the upper survival threshold for this population, so we tested 39°C (Experiment 3), and 37°C and 38°C (Experiment 4). Finally, we tested the effect of diurnal fluctuations around low (16°C) and high (35–37°C) mean temperatures (Experiments 5 and 6, respectively), with the constant temperature re-tested as an internal experimental control.

The amplitudes of the fluctuations used in the experiments were either small (7.6°C) or large (18.6°C). These profiles represent the shape and magnitude of DTRs observed in the high and low DENV transmission seasons in central west Thailand. The profiles followed a truncated sinusoidal progression during the day and exponential decrease at night, because this more accurately represents the profile of daily temperature variation compared to a symmetrical sinusoidal profile [16], [20]. The shapes of the profiles were the same for each of the fluctuating temperature experiments, but the profile was raised or lowered in order to adjust the mean temperature as necessary. Temperatures were maintained in Binder KBF115 environmental chambers (Tuttlingen, Germany). Data loggers, both internal in the chamber and independent HOBO data loggers (Onset, Cape Cod, MA), recorded temperature and humidity variation. Actual temperatures were within 0.3°C of the programmed temperatures throughout the duration of the experiments. The relative humidity level was maintained between 70% and 80% across all experiments.

Using this empirical development time data, we calculated the required DDs for Ae. aegypti larval development, and constructed a stage-structured model to estimate the impacts of the temperature regimes on Ae. aegypti population dynamics. We also re-analyzed data from Carrington et al. [17], which were generated using the same laboratory protocols, to include in the latter portion of this study. In this report, we make the distinction between the effect of temperature and DTR. We use temperature to refer to changes in mean constant temperatures (e.g.; 16°C, 20°C, 30°C etc) and DTR to refer to differences between the magnitudes of fluctuations around a given mean temperature (e.g.; constant [where the DTR = 0°C], small DTR or large DTR).

We used Ae. aegypti collected from Kamphaeng Phet (KPP), Thailand in our experiments. Originally collected from the field in January 2011, the mosquitoes were maintained under standard laboratory conditions until they were used for experiments. Eggs were hatched in water using a vacuum manifold and reared under a controlled density (≈200 larvae per tray) in containers (≈33×29×6 cm) with 1.5L of deionized water. Colonies were maintained with a population size of >500 individuals per generation. Larvae were fed as previously described in Styer et al. [21]. The F₄ generation of mosquitoes was assessed in our life-history trait assays.

For each experiment, estimates of egg to pupa development time were made by rearing mosquitoes in replicate rearing cups, with 20 individuals per cup with 100 ml deionized water, hatched under vacuum to ensure minimal variation of hatching time. We tested between 13 and 20 replicate cups per temperature treatment. For each temperature, pre-warmed water was used to replenish the evaporated water every day at the same time that food was administered. We assessed survival daily (except for the first day, due to the fragility of the newly-hatched first instar larvae) by counting the number of remaining larvae left in each cup. Mean development time from egg hatch to pupation was recorded in hours or days. High temperature experiments (a mean of 26° and above) were scored every 6 hours from first pupation to the last and low temperature experiments (≤20°C) were scored once daily due to slower development rates. Median time from pupation to emergence was recorded to enable complete population dynamic estimates for subsequent modeling.

Based on a linear regression constructed from the rates of development for each rearing cup at constant temperatures between 16°C and 35°C, we calculated the minimum development threshold temperature, and subsequently determined the DD required for the KPP population Ae. aegypti to pupate.

Mosquitoes assessed in experiments were all reared under the temperatures at which they were tested, with controlled densities in experimental cages. Between 10 and 18 females reared per temperature were each placed with a single male within 24 hours of emergence. We measured the length of time from emergence until the first blood meal (termed the pre-blood meal period), the time it took from the first blood meal to the first day eggs were observed (length of the first gonotrophic cycle) and the number of eggs that were laid (clutch size). Human blood (L.B.C.) was offered for 15 minutes every day from the start of the experiment, until two days after the first blood meal of each female, which allowed females two subsequent opportunities to feed to repletion. If a female had still not fed on the third day after the experiment began (or fifth day at the lower temperatures), she was removed from the analysis. Females at higher temperatures were allowed a minimum of seven days to lay eggs. At the cooler temperature treatments females were allowed 3 weeks.

The University of California at Davis Institutional Review Board determined that this experiment (allowing mosquitoes to take blood meals from people) did not meet the criteria for human subjects research and thus, did not require human subjects approval.

We constructed a simple stage-structure matrix model [22] for Ae. aegypti population dynamics to allow us to translate experimental differences in life-history parameters into real-world expectations for potential population growth in the absence of other limiting influences such as density dependence, predation or inter-specific competition. This allowed us to identify the temperatures and DTRs that maximize reproductive potential for Ae. aegypti. The model, written in R version 2.15 [23] used median development times to pupation and adult emergence, and fecundity estimates for adult females at each temperature. Parameter estimates used in the model were based on estimates taken directly from our KPP population in this study. Remaining estimates, for those traits we did not measure, were based on the following published measures and assumptions. Because Ae. aegypti eggs can maintain viability for up to one year [24], we standardized hatch time between treatment groups to five days, with a 99% chance of survival each day. We assume the female:male ratio is 1∶1 for offspring, and that mating did not limit population dynamics. Based on our initial measurements of components of the gonotrophic cycle at 26°C constant temperature [17], we reduced the clutch size by 5% each subsequent cycle after the first, and extended the duration of each gonotrophic cycle by 15% [17], [25]. A 90% survival rate per day was assumed for adult females [25], [26]. Because we had not observed females to complete more than four gonotrophic cycles in our previous study with this population at 26°C [17], we assumed that females die after completion of this fourth cycle. Temperature-DTR combinations that did not result in viable offspring in our experiments (e.g., if females did not lay eggs or immatures did not survive to adult emergence) were not considered for modeling.

Using empirical data from this study, our deterministic model allowed us to collectively evaluate the impacts of all life-table parameters of Ae. aegypti. Our experimental data showed that the KPP population may be sustained under constant temperatures ranging between 20°C and 35°C, but reproduction was not successful at 16°C (both with and without fluctuations) or at 35°C and 37°C with small fluctuations. As a result, population dynamics could not be modeled for these temperature treatments.

A temperature of 26°C with both small and large fluctuations, however, sustained population growth and our model predicts that the previously reported ‘negligible or slightly positive’ effects on mosquito life-history traits observed under small fluctuations [17] would lead to a population size more than double that of the constant 26°C treatment after one month (Figure 6). Daily population growth rates were 33.3% and 27.9% respectively. Conversely, a large DTR at the same temperature results in a 40% reduction in the adult mosquito population size compared to the constant temperature control, resulting from a slightly lower growth rate of 25.6%. According to the model, the optimum temperature for population growth is 30°C. At this temperature, the model predicts a growth rate of 41.5% each day, leading to more than 60,000 eggs and 3,525 adults in the population after 30 days. The least optimal temperature for population growth (15.3% per day) is 20°C, with 16 adults after 30 days,

Development time estimates were highly temperature sensitive, and inversely related to temperature up to a maximum of 35°C. Beyond this threshold, the rate of development declined. The addition of a small DTR around 16°C and 35°C did not alter development time relative to the constant temperature control, but at 16°C a large DTR significantly accelerated development. A similar result demonstrating the importance of DTR magnitude was reported for a large DTR at 26°C [17]. Richardson et al. [7] report conflicting results to these. They reported that the development time of mosquitoes reared under constant temperatures accurately predicted development time of mosquitoes under fluctuating conditions. While the temperature profiles of their experimental fluctuating temperature regimes were not reported, our results suggest that DTR in their cyclical temperature treatments were similar to the magnitude of our ‘small DTR’ (i.e.; <10°C). Tun-Lin et al. [4] explored Ae. aegypti development directly under field conditions, in various sized containers and with different water volumes. They found that field-reared mosquitoes had greater variability in development rates compared to those reared in the laboratory, especially in smaller containers where water temperature fluctuated more extensively than in larger containers. Only 50% of comparisons between survival data from the laboratory and the field were similar, demonstrating that the DTR of water can influence immature mosquito development and survival. Other laboratory experiments have shown that fluctuating temperatures can modify the dependence of traits on temperature. Ae. aegypti from Trinidad reared at constant temperatures had a significant temperature-dependence of wing length, but under fluctuating water temperatures, this effect disappeared [6].

Although the mean temperature a mosquito is exposed to throughout the day may be the same across DTR treatments, as indicated by significant effects of DTR on our DD estimates, there appears to be an underlying physiological response of the mosquito to the daily variation in temperature that contributes to changes in the speed of development that is not observed under equivalent constant temperatures. Directionality of the change is dependent on mean temperature. At intermediate and high temperatures, both large and small fluctuations (at 26°C and 35°C respectively) resulted in mosquitoes needing a greater amount of thermal energy to reach pupation than under the constant temperature control. At low temperatures, however, large fluctuations significantly reduce this number of DDs, despite the exposure temperatures dropping below the minimum for development for portions of the day. We are unaware of any studies that have performed comparable experiments with realistic fluctuations in temperature in other species.

These results underline the importance of using appropriate temperature regimes when measuring population parameters in order to accurately describe mosquito life-history. Without considering this variation, it is possible that development rates of mosquitoes (and other insects) in nature are either under- or overestimated, dependent upon mean temperature.

High mean temperatures with fluctuations were detrimental for mosquito reproduction. The addition of even a small DTR to a 35°C mean temperature terminated reproduction, presumably because the maximum temperature reached during the day (∼39.7°C) was damaging to the mosquitoes. Initial testing demonstrated that mosquitoes did not reproduce at 16°C, but we hypothesised that the addition of fluctuations might rescue female reproduction. Despite an increase in blood feeding activity, there were no eggs laid by any female exposed to fluctuating temperatures around of mean of 16°C, suggesting that either this mean temperature or the extremely low overnight temperatures are detrimental to reproduction.

It is possible that the females that did not lay eggs in these experiments were in fact unmated; we did not physically observe all females copulate and we did not dissect ovaries to confirm insemination and egg development in the blood fed females. Future experiments would benefit from allowing females more time to lay eggs, and subsequently dissecting individuals to determine stages of egg development. Another possibility is that these low temperatures negatively impacted male fertility. This issue needs further examination. Although we cannot offer an explanation to the cause of failed reproduction, this result is biologically meaningful. In our experiment, we reproduced field conditions by rearing both sexes and allowing mating to occur at the temperatures at which they were tested.

Unfortunately, we were unable to test for statistical significance between temperature-DTR treatments at high and low temperatures for reproduction because no estimates were obtained. These descriptive results, however, strongly suggest that DTR at both upper and lower ends of the temperature scale can alter estimates of life-history traits relative to constant temperatures. Predictions of mosquito responses to temperature in the wild based on laboratory estimates should thus be made with caution. Despite a lack of reproduction it is important to note that the continued blood feeding of mosquitoes at these high and low temperatures may still allow for the transmission of mosquito-borne pathogens, including DENV which can be transmitted at temperatures as low as 13°C [27].

Previous studies have shown that both temperature and humidity have the ability to alter fecundity estimates. Costa [5] considered small temperature fluctuations (DTR = 4°C) around each of three mean temperatures, with two humidity profiles. Fecundity was reduced with increasing mean temperatures, as was the length of the gonotrophic cycle. Increasing humidity also intensified the effects of temperature. Constant temperature controls were not included so the influence of fluctuations could not be teased apart.

Our model predicts population growth for Ae. aegypti exposed to various temperature-DTR combinations and demonstrates that constant temperatures around 30°C are optimal for the KPP population. Our simple model does not consider influences of inter-specific competition, limited access to nutrition or predation, and, therefore, our results inform on growth rate under optimal conditions. Under field conditions, each of these factors can influence the development and/or fecundity of mosquitoes and would be expected to limit population growth. The 26°C small and large DTR temperature treatments are representative of conditions observed in the high and low DENV transmission seasons in central west Thailand [17]. Relative to a constant 26°C, our model predicts that the potential for mosquito vector population growth under small daily fluctuations is markedly higher (+5.4% difference in daily population growth rates), whereas under large fluctuations, population growth is reduced (–2.3%). Entomological surveys in nearby areas support seasonal mosquito density changes [28] and these changes are positively associated with DENV transmission patterns.

Our population dynamic model could only be run for each temperature when we obtained mosquito trait values after they completed each stage in the model. The lack of experimental data we obtained at particular high and low temperature-DTR combinations, which is a result in itself, meant that we were unable to obtain output at those temperature regimes. We did not assess the effect of fluctuations at constant 30°C and constant 20°C, which the model ranked as the most and least optimal conditions for population growth, respectively.

Modeling of Ae. aegypti population dynamics under constant and cyclic temperatures indicates that more in depth studies are required to fully understand the complexities of environmental variation of life-history traits. Predictions of population growth and decline based on environmental temperature in the wild may be inaccurate due to differences between the mean temperatures observed and the DTR experienced by the mosquitoes. For example, our study population of KPP Ae. aegypti have a calculated thermal minimum for development of 11.78°C. We know, however, that development under a 16°C+large DTR regime still occurs, despite the fact that temperature drops ∼4°C below this thermal minimum for around 1/3 of the day. In fact, development occurs at a rate faster than that of mosquitoes reared at a 16°C constant temperature regime. It might be expected that the time to development would be longer for a mosquito experiencing sub-optimal temperatures, given that a portion of the day is spent below the thermal minimum for development, but in fact it takes less time and fewer degree-days to develop. The thermal minimum for development, therefore, must be interpreted cautiously.

Fluctuating temperatures alter estimates of life-history traits in immature and adult Ae. aegypti with the direction of the change dependent upon the magnitude of the DTR and mean temperature around which the fluctuations oscillate. Differences in DD estimates between fluctuating and constant temperature treatments of the same mean reveal a physiological response by the mosquitoes that modify their normal developmental rate. Results from this study demonstrate the sensitivity of life-history traits to temperature fluctuations at the upper and lower thermal limits and suggests that standard constant temperature laboratory-based experiments do not fully reflect what is happening in nature. Vector control efforts and population dynamic models will be improved when realistic parameter estimates of mosquito populations are incorporated in future surveillance activities and research projects.