|
nif:isString
|
-
In this study we used the Skeeter Buster model [16], a stochastic, biologically detailed, spatially-explicit model of Ae. aegypti populations, based on biological elements of the CIMSiM model [49]. Because of the complexity of the model, we do not detail all of its features here. Instead we provide a succinct description of model components critical to understanding this study. For further description of these specific components, see Text S2. For a complete description of the model, and the empirical data upon which it is based, see reference [16]. The Skeeter Buster model simulates the biological development of four life stages of Ae. aegypti: eggs, larvae, pupae in individual water-holding containers, and adults in individual houses. Houses are laid out on a rectangular grid, and containers are assigned to their specific house on the grid. The model includes a detailed description of temperature-dependent development rates of all life stages. Density-dependent effects of intraspecific competition in the larval stages are described using the model by Gilpin and McClelland [50] that tracks the parallel changes in food amounts in each breeding site and larval weight. Larval weight (and potential starving) in turn affects survival and development time. Adult females are assumed to be strictly monogamous [51] and for each unmated female, a mate is selected at random (with a probability proportional to its weight) among all males present in the same house; if no males are present, that female will remain unmated until the next day. Adults can move from one house to one of its immediate neighbors with a daily probability of 30%, a value obtained from field dispersal studies (see reference (32)). This probability is identical for adult males and females. We assume that humans are present in all simulated premises, and consequently that the dispersal probability is not affected by the status of the female (host-seeking or resting). The direction of this short range dispersal is chosen at random. Therefore, while daily dispersal is limited to nearest neighbors, the potential lifetime dispersal distance is much larger. The model also contains a function for long-distance movement, and this was used to simulate the initial movement of released males.
We choose to simulate mosquito populations in a part of the city of Iquitos, Peru, because the application of Skeeter Buster to this location has been the object of a specific case study [28]. Iquitos is located in the northeast portion of the Amazon Basin of Peru (location: 73.2W, 3.7S) and has a tropical climate with no marked seasonality. The city is relatively isolated, with no road connection to other urban centers in Peru or in neighboring countries. Detailed descriptions of the study area and its Ae. aegypti population are available from earlier published studies [30], [31]. We simulate a 2448-house area (68×36) parameterized according to data collected in a heavily surveyed region of the city [28].
Simulating FK Strain Release in Skeeter Buster: FK was simulated in Skeeter Buster as a female-specific lethal allele carried on a single locus with two alleles. When strains with several insertions of the same FK construct were considered, the insertions were assumed to occur on separate linkage groups, and, therefore, segregate independently. Lethality occurred on the first day that an adult female with one or more copies of the FK allele emerges from pupation. These conditions match the characteristics of the FK strain of Ae. aegypti developed by Fu et al. [17] in which emerging females cannot fly and are effectively removed from the population at emergence due to their inability to locate mates or take blood meals. Each run of the model was initiated with 20 eggs in every container, and there was a 1-year burn-in period before releases to allow the mosquito population dynamics to stabilize. For each scenario the model was run 30 times.
The release of adults was modeled by adding cohorts of the same number of homozygous FK males to each desired grid location (i.e. individual premises) in the model space. The same locations were used for each weekly release. To simulate release of eggs, following the general scenario put forward by Fu et al. [17], additional containers were incorporated into the model every week in the corresponding release houses. These containers had an initial supply of nutritional resources that ensures favorable larval development of the released eggs, but were shielded from ovipositing females. An identical number of male and female eggs were input into the container, however, because we consider only the case of 100% efficiency of the female-specific lethal elements, only male adults emerged from these containers. They were then removed from the simulated area as soon as all released individuals had either died or emerged as adults. We simulated three distinct spatial release approaches to examine the impact of dispersion of the transgenic mosquitoes on outcomes. The spatial patterns we consider represent ideal release scenarios, and while they might be operationally impossible to replicate exactly, comparison between them is informative as to what actual release programs should aim for. Homogeneous releases consisted of releasing the same number of transgenic mosquitoes in every house in the simulated area. Because coverage approaching this pattern appears achievable only through aerial coverage of a city, we only applied this spatial pattern to adult releases. Point source releases corresponded to selection of a discrete subset of 10% of houses as release sites. We define two distinct patterns of point source releases depending on the process of selection of these release sites: random, in which 242 release sites were randomly chosen among the 2448 simulated houses and uniform in which 242 sites were laid out at regular intervals on the grid. In both cases, release sites remained fixed throughout the release period.
When a large number of adult males are simultaneously released in the same site, we simulate scenarios where only a (deterministically set) fraction of these released males will remain in the release site on the day of release. Each remaining male in the released cohort is then immediately allocated to another house in the simulated area, chosen at random (and is therefore never assigned to the release site). We consider cases where 0%, 10%, 20% or 50% of the released cohort are such early dispersers (and therefore only 100%, 90%, 80% or 50%, respectively, of the released mosquitoes are placed upon release in the designated release site).
We simulated releases in Iquitos, where the spatial heterogeneity in mosquito production among houses is relatively limited [30], [31], [52]. In more heterogeneous environments (e.g. Tapachula, Mexico [53], [54]), however, spatially homogeneous mosquito releases could result in heterogeneous ratios of transgenic to local mosquitoes among houses [55]. We model situations where we artificially increased the variation in pupal productivity among houses by transforming each house into either a high-producing house (with probability 1/Φ, where Φ≥2) or into a low-producing house (with probability 1−1/Φ). In the former case, every container found in that house was included Φ−1 times in the corresponding house in the simulation, while in the latter case, every container found in that house had a probability 1/(Φ−1) to be kept in the simulation, and is otherwise removed. With this method the average total number of containers in the city remained identical to the default situation. The case leaves the original distribution unchanged, while integers values of or larger increase the heterogeneity in container distribution among houses. Applied to a perfectly homogeneous setting, this method would result in a coefficient of variation in container distribution among houses equal to , i.e. or approx. 71% for , and 150% for . We simulated pre-release control programs based on traditional methods, consisting of adulticidal control applied periodically in premises across the modeled area. Two weeks of insecticide application were simulated, causing an assumed 90% additional mortality among adults every day, ending on the day before the onset of FK releases. Throughout that period, we considered that each premise had a 10% probability of not receiving insecticide on any given day, based on the assumption that coverage cannot reach every premise in the targeted urban area. Supplementary runs were run with an increased (20%) probability of not receiving insecticide (Figure S7), showing little to no difference compared to the default value. Additional methods (community participation and uncertainty analysis) are presented in Text S1.
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}