Date Published: February 9, 2010
Publisher: Public Library of Science
Author(s): Simon I. Hay, Marianne E. Sinka, Robi M. Okara, Caroline W. Kabaria, Philip M. Mbithi, Carolynn C. Tago, David Benz, Peter W. Gething, Rosalind E. Howes, Anand P. Patil, William H. Temperley, Michael J. Bangs, Theeraphap Chareonviriyaphap, Iqbal R. F. Elyazar, Ralph E. Harbach, Janet Hemingway, Sylvie Manguin, Charles M. Mbogo, Yasmin Rubio-Palis, H. Charles J Godfray
Abstract: Simon Hay and colleagues describe how the Malaria Atlas Project has collated anopheline occurrence data to map the geographic distributions of the dominant mosquito vectors of human malaria.
Partial Text: Despite advances in mapping the geographical distribution and intensity of malaria transmission ,, the ability to provide strategic, evidence-based advice for malaria control programmes remains constrained by the lack of range maps of the dominant Anopheles vectors of human malaria. This is because appropriate vector control depends on knowing both the distribution and epidemiological significance of Anopheles vectors . Substantial investments by major donors in the distribution of long-lasting insecticide-treated nets and indoor residual spraying campaigns  are, therefore, not always fully informed by the basic biology of local anophelines.
There are 462 formally named Anopheles species, with a further 50 provisionally designated and awaiting description –. Of these, approximately 70 have been shown to be competent vectors of human malaria  and from this set, 52 candidate dominant vector species (DVS) were initially chosen for inclusion in the MAP vector distribution mapping project. These DVS are species (or species complexes) that transmit the majority of human malaria parasites in an area by virtue of their abundance, their propensity for feeding on humans, their mean adult longevity (only old individuals incubate the parasite long enough to transmit the disease), or any combination of these and other factors that increase overall vectorial capacity . The DVS were the inclusive set of those species identified as “main” ,, “dominant” , or “principal” , in major reviews of Anopheles distribution and biology. The list was then further refined by anopheline experts from the Americas, Europe, Africa, Asia, and the Pacific, who co-author this article, to exclude 11 species that were not considered important vectors either because few recent data had implicated them in transmission or because they acted as vectors in only restricted geographical areas (Text S1). Following the convention of the major reviews in this area –,,, the DVS of the Anopheles (Cellia) gambiae complex are listed separately. We hope also to map at species level three other complexes, where examination of the primary literature has indicated sufficient species-specific data (the An. (Nyssorhynchus) albitarsis, An. (Cellia) culicifacies, and An. (Cellia) dirus complexes). Further details are provided in the legend of the maps of each complex in Text S3 (for the An. (Nyssorhynchus) albitarsis complex) and Text S5 (for the An. (Cellia) culicifacies and An. (Cellia) dirus complexes).
An exhaustive and systematic search of formal and informal literature was conducted, mirroring the approaches developed by the MAP in building a global database of malaria parasite prevalence . Only information collected after 31 December 1984 was searched. This criterion ensured that the data collected were representative of the contemporary distribution of the DVS and that the DVS occurrence records included only data collected using modern taxonomic species concepts ,. Following the introduction of cytological and then molecular methods to mosquito systematics, the taxonomy of the Anopheles changed radically, making many earlier species determinations potentially unreliable ,,–. This date restriction also served to focus finite literature retrieval and abstracting resources on newer references, that are easier to retrieve from libraries, have sites that are less problematic to geo-position, and have authors that can often still be contacted with queries.
Many initiatives are being developed to provide information on the geographical distribution of disease vectors, including the Anopheles (Table 1; for example surveys of the geographical distribution of different forms of insecticide resistance –). These initiatives will be a significant help in data acquisition. Duplication of search effort will be minimized by ensuring compatibility between different data abstraction ontologies (e.g.,  and Text S2), so that where possible, data exchange can be automated. Where this cannot be achieved, data will be incorporated manually into the MAP archives with its provenance clearly recorded.
Recent years have seen the development of a number of new techniques to predict species ranges –, of which the most promising include methods based on boosted regression trees ,, generalised additive models , and maximum entropy approaches . In addition, Bayesian statistical approaches –, which have been widely used in mapping malaria prevalence –, have recently begun to be applied to mapping the relative frequency of Anopheles species . Bayesian models are able to integrate information from disparate sources and allow the comprehensive quantification of prediction uncertainty, something that is often overlooked in species mapping exercises .
The statistical techniques we shall employ in future mapping efforts will model species occurrence as a function of environmental variables. We can then predict species distributions as a function of environmental conditions that can be obtained from Earth-observing satellite imagery . During model formulation and validation we shall use coarse spatial resolution (∼8×8 km) multitemporal remotely sensed imagery  to reduce computational demand. Once the particular mapping technique is chosen, we will move to more contemporary Moderate Resolution Imaging Spectroradiometer (MODIS) satellite imagery, available globally at ∼1×1 km spatial resolution , to improve the spatial resolution of the predictions. Adapting temporal Fourier analyses techniques, which ordinate seasonal environmental data ,, to cope with the irregular compositing periods of MODIS data, has been completed and the data has already been made available in the public domain .
The usefulness of the species range maps when available online , can be improved by combining them with summaries of the species-specific life history characteristics or “bionomics” of the DVS. Anopheline vector bionomics are critical in defining the appropriate (and inappropriate) modes of control at the national and local level –. For example, indoor residual spraying of houses for the control of a vector that is predominantly an outdoor resting species and prefers biting animals (e.g., An. (Cellia) arabiensis) is unlikely to be an optimal control strategy . Conversely, if the vector feeds predominantly indoors and at night (e.g., An. (Cellia) gambiae), insecticide-treated nets are likely to be a very appropriate intervention ,. Information on characteristics of specific larval habitats and range will also be informative. Public health and education measures aimed at larval reduction may be feasible across large parts of the Middle East and Asia , where An. (Cellia) stephensi is the major DVS. This species readily breeds in urban areas, often using human-made water containers as its preferred larval habitat. Conversely, environmental management techniques such as installing tidal gates or constructing drainage systems are likely to be more effective as a permanent means of reducing or eliminating suitable coastal habitats of members of the An. (Cellia) sundaicus complex across substantial areas of South East Asia .
The completed DVS databases and predictive maps will be made available online once generated, alongside the wider portfolio of MAP products, including spatial limits and endemicity maps for the human malaria parasites ,. This juxtaposition of information should represent an important cartographic resource for those engaged in malaria control and where feasible, its elimination. The success and long-term sustainability of this DVS mapping initiative depends critically on its continued support, development, and refinement in the malaria vector control and research communities. We hope that the information on the aims and objectives provided here, and the commitment to providing data in an open access venue, will help ensure that support.