We hypothesized, first, that development rate is significantly influenced by several environmental factors and that the interaction of these factors is an important predictor of development rate variation. The results of both meta-analytic approaches suggest that temperature is the main fixed factor driving development rate, to the exclusion of other factors of known importance such as diet and density. This bolsters the contention that temperature is the most important ecological determinant and, when modeling development, sufficient to predict development rate . When larvae experience nutritional deprivation or high densities, this can dampen  or exacerbate  the impacts of temperature. Thus, while research suggests that diet [73, 113] and larval rearing density [84, 114] do matter, these results underscore that they should not be considered to the exclusion of temperature. Based on model selection, the relative importance of these factors can be ranked as temperature followed by temperature variability, larval rearing density, then diet, and lastly photoperiod (Table 2, Table 3). The relative importance of factors is consistent between the periods of hatch to pupation and hatch to emergence. While this analysis shows other variables such as latitude were not significant in explaining development rate variation, they may impact other important life history traits including survival, body size, fecundity , and morphology .
The relationship between temperature and development rate is linear within a median temperature range [116–119], and features of this linear relationship, such as slope and intercept, have biological interpretations. The slope of this relationship is considered the cumulative effect of temperature on the rate of development, and the intercept can be interpreted as the theoretical temperature at which development can no longer occur [5, 109], also called the developmental zero. Although at extremes of low temperature the development curve is non-linear, the linear portion is extrapolated to the intersection with the temperature axis to estimate the developmental zero . This extrapolation based on slope may, in part, explain the large variation in the estimates reported in Tables 4 and 5. This may also explain estimates that were less than zero, which is biologically implausible. Meta-analysis these parameters across many studies allows for outliers to be more easily identified.
Despite these limitations, the developmental zero is often considered a fixed characteristic of a species for the purposes of modeling and predicting population abundance [39, 109, 121, 122]. Thus, we also sought to test the hypothesis that the effect of temperature and the developmental zero are fixed characteristics of Ae. aegypti strains. While the meta-analytic results are consistent with a positive, linear relationship between temperature and development rate, tests for heterogeneity suggest a significant amount of variation in response to temperatures within this range. These data do not support the hypothesis that the developmental zero and the effect of temperature are fixed constants. Both the effect of temperature and the developmental zero are heterogeneous across studies considered in the meta-analysis. These results have implications for the modeling of development rate as well as population abundance, which often relies on development times of larval populations [22, 123]. These compiled data may be used as the basis for modeling these parameters as a distribution rather than choosing one value from a single study. Variation in development time (i.e. the inverse of development rate) has been modeled as a continuous random variable with a distribution of frequencies, such as the normal distribution  or with a heterogeneity factor . Other modeling approaches to incorporate development rate variation stochastically by treating development rate as a random variable dependent on the variability in the level of catalytic enzymes [126–128], positing a biophysical basis for variability.
There are several hypotheses to address why the response to temperature may be heterogeneous. Our results indicate that factors of larval rearing density, diet, latitude, and photoperiod were not factors that could explain heterogeneity of the effect of temperature. A limitation of this analysis was the narrow range of reported values of diet and initial larval rearing density. While many studies reported at least one level of different factors such as temperature, diet, and larval rearing density, few studies in Ae. aegypti examined development across gradients of multiple environmental conditions. Such experiments are needed in order to establish the relative importance of environmental factors in the variation of development rates. Assessing the impact of varied environmental conditions on the developmental phenotypes of mosquito larvae can be complex with interactive effects [18, 24, 129]. For example, Padhmanhaba et al. 2011  show that increased the rearing temperature for starved Ae. aegypti larvae impacts development rate, and this impact changes depending on the larval stage and the temperature.
Publication author was adequate to explain heterogeneity in the effect of temperature on development rates. It is difficult to identify the aspects of this factor to describe its significance in explaining development rate variation. We evaluated the dichotomy of laboratory versus field experiments, which generally corresponded to constant versus variable temperatures. Mosquito response to variable rather than constant temperatures has been a recent focus both for life history traits and vectorial capacity [123, 130–134]. Variable temperatures have been shown to increase , decrease , and have no impact  on development rates of mosquitoes and other insects. Inconsistency in the relationship between temperature and development rate has been attributed to field conditions versus laboratory conditions . To test this, we compared development rates estimated under constant versus variable temperature conditions, which corresponded to laboratory versus field conditions. This comparison showed no significant difference overall in the relationship between development rate and temperature based on temperature variability for either larval stages or to hatch to emergence (Figure 2). This finding is consistent with recent reports that Ae. aegypti life-history traits depend not only on variability but also the magnitude of temperature fluctuations .
The factor of publication may be a proxy for methodological differences such as diet composition (i.e. ingredients of diet). Of the 49 studies, almost all reported information on diet composition. However, few used the same diet preparations, and this prevented this factor from being included in meta-analysis. Some diets were created from detritus of the larval habitat in order to mimic natural conditions [36, 61, 137] or incorporated detritus . The majority however provided no explanation for the choice of diets. Diet choice can influence development rate as well as interspecific larval competition [138, 139] and adult wing length . To facilitate comparison of larval performance across populations, these findings support a need for standardization of diet composition for laboratory colonies. This is especially important in the context of transgenics. Our literature search yielded only two studies with estimating development rate of Ae. aegypti transgenic strains. The low sample size impeded statistical comparison of transgenic versus wild-type development rate estimates, leading to singularity errors in the linear mixed effects modeling. Future comparisons of transgenic and wild strains in other important life-history traits such as body size, fecundity, and longevity may also be informative. Estimating and evaluating life-history traits across different environmental conditions is critical to provide a basis for comparison between wild and transgenic strains and may guide future transgenic release programs [29, 58, 69].
Other factors not considered in this analysis may also impact the effect of temperature, and perhaps contribute to heterogeneity. Examples include genetic variation, microbial symbiotic partners, and maternal effects. Population differences in larval survival and body size in response to different temperatures have been demonstrated in other insects  but such differences have also been attributed to adaptive phenotypic plasticity through a hormonal cascade that stops growth . Inclusion of latitude as a variable was one proxy for comparing populations broadly. Latitude has been suggested as a potential gradient for local adaptation to thermal stress in mosquitoes . However, our results suggest latitude does not explain heterogeneity of the effect of temperature. The strain origin/study location was included as a random effect as another indirect proxy for genetic differences in population, but we found no associations with strain origin. There is evidence of genetic structure across geographic space  and seasons , but examples of strong local adaptation in development rate is lacking in Ae. aegypti populations . Richardson et al.  suggested that the lack of strong local adaptation may be evidence of a limited capacity to evolve in response to thermal stress. More studies are needed to evaluate the potential for adaptive phenotypic plasticity in response to temperature in Ae. aegpyti that could explain the heterogeneity of responses characterized here. Further, in natural conditions other ecological factors not considered here such as interspecific competition, such as between Ae. aegypti and Ae. albopictus[34, 137], and predation  may impact development rate and warrant further investigation.
Recent work compares life-history traits such as body size and fecundity across multiple environmental conditions [145, 146]. More empirical estimates of these traits across environments have been recently made available since the preparation of this work [123, 134, 147, 148], a limitation of conducting a meta-analysis in a rapidly developing field of research. Recent advances suggest variation in these traits has been attributed to responses to environmental conditions during development [20, 130, 149] as well as adaptive genetic responses due to selection at different temperatures [7, 150]. Developmental life-history traits are of particular epidemiological importance for arboviral disease dynamics as they have been associated with critical aspects of vectorial capacity such as changes in bite rate, dispersal  and virus infection and dissemination .