In sub-Saharan Africa, Plasmodium falciparum malaria is primarily transmitted by mosquito species belonging to the Anopheles gambiae and Anopheles funestus complexes 1 2 3 4 . The intensity of malaria transmission is heterogeneous across the continent, and influenced by mosquito species' compositions, vector competence, and underlying demographic and environmental factors 5 . High levels of transmission frequently occur where both An. gambiae sensu lato and An. funestus are present, as they tend to exploit different breeding habitats and peak at different times, thereby prolonging the transmission period. Generally, Anopheles gambiae s.l. are most abundant during the rainy season, and An. funestus is predominant at the end of the rains and beginning of the dry season 1 2 3 . The extent to which these species are influenced by the same environmental factors is largely unknown, as very few studies have examined them simultaneously over a wide geographical range. One of the most comprehensive studies undertaken in Kenya, by Mbogo et al 6 , provides an opportunity to retrospectively analyse the spatial abundance, distribution and transmission data on An. gambiae s.l. and An. funestus, and compare climate, vegetation and elevation data derived from remote-sensed satellite sources in key locations.

The study by Mbogo et al 6 provides information on the numbers and transmission intensities, i.e. entomological inoculation rates (EIRs), of An. gambiae s.l. and An. funestus at 30 villages in the Malindi, Kilifi and Kwale Districts along the south-eastern coast of Kenya. Mosquito collections between June 1997 and May 1998 indicated that An. gambiae sensu stricto, Anopheles arabiensis and An. funestus were the main malaria vectors, with differing geographical abundance and transmission patterns over the 200 km study area. Interestingly, An. gambiae s.s. was found to be widespread, whereas An. arabiensis was mostly confined to Malindi in the north and An. funestus to Kwale in the south. Preliminary climate analyses by Mbogo et al 6 , found positive correlations between rainfall and the temporal distributions of An. gambiae s.l. and An. funestus, however, these varied by species and between districts, and climate data were limited to one meteorological station in each district.

The recent advances in space technology and increased public access to remote-sensed satellite data provide a cost-effective and efficient alternative to examine relationships between climate, the environment and mosquito vectors of human disease 7 8 . This is important in poorly resourced regions of the world where the collection of reliable data over large geographical areas is not possible. As a follow-up to the Mbogo et al 6 study, comparisons of satellite-derived precipitation, temperature, humidity, vegetation and elevation measures at each study site in Malindi, Kilifi and Kwale Districts, and in An. gambiae s.s., An. arabiensis and An. funestus clustered locations were carried out.

First, the average number and daily EIRs of An. gambiae s.l and An. funestus (data from Table 1 in Mbogo et al 6 ), and climate, vegetation and elevation data for each district were summarized. Second, the relationship between the relative contribution (%) of the three main species, i.e. An. gambiae s.s, An. arabiensis and An. funestus, to annual EIR (data from Table 2 in Mbogo et al 6 ), and each environmental variable was examined using bivariate correlations, and Pearson's correlation coefficient (2-tailed P value ≤ 0.05 significance). Third, the spatial patterns of An. gambiae s.s, An. arabiensis and An. funestus transmission were examined in ArcGIS using Spatial Analyst tools (ESRI, Redland, CA). The Moran's I statistic was used to determine spatial autocorrelation patterns i.e. clustered, dispersed, random, and the Getis-Ord Gi* statistic was to identify the locations with high and low clustering (Z scores, 95% confidence levels (CI) -1.96 and +1.96 standard deviations). The distributions of clustering across the study area were highlighted in relation to elevation, using a 3D wireframe map created in the surface mapping programme Surfer 7.0 (Golden Software Inc., Golden, CO). Mean environmental measures between high and low clustering trends were compared using the Mann-Whitney U test with Bonferroni correction for multiple comparisons. All statistical analyses were performed in Microsoft Excel and SPSS 16.0 (SPSS Inc., Chicago, IL).

Climate and vegetation data corresponding to the 30 mosquito collection sites (i.e. latitude and longitude), and original time period (i.e 1997-1998) were obtained from the best available sources, accessed via the IRI/LDEO Climate Data Library of the International Research Institute for Climate and Society 9 . Average daily precipitation (mm), monthly temperature (C°) and daily specific humidity (qa) measures for each month were extracted from satellite data from the National Oceanic and Atmospheric Administration (NOAA) 10 11 12 . Vegetation cover was based on Normalized Difference Vegetation Index (NDVI) satellite data extracted from monthly maximum NDVI data available from U.S Geological Survey's (USGS), Africa Data Dissemination Service 13 . Elevation data were derived from the USGS ETOPO2 Digital Elevation Model available from ESRI (Redlands, CA).

These simple comparative analyses of 30 sites across three districts in Kenya indicate that An. gambiae s.l and An. funestus can have distinct ecological niches and requirements within a relatively small geographical area. This is supported by other entomological studies carried out in the region, which highlight the heterogeneous nature of these species' seasonality, host feeding preferences 14 15 16 , body size 17 and the distribution and type of breeding sites 18 19 20 . For example, An. gambiae s.s larvae mostly occur in open shallow sunlit puddles and pools close to homesteads, whereas An. funestus larvae prevail in permanent vegetated aquatic habitats such as stream pools of rivers. In general, malaria transmission by An. funestus predominantly occurs in rural areas of sub-Saharan Africa, and the fact that Kwale District was less urbanized than the other districts 21 22 , may also explain why An. funestus prevailed in this region. Furthermore, the presence of both An. gambiae s.s. and An. funestus, whose ecological requirements may be complementary to each other 23 , may also account for the overall higher EIRs found in Kwale District 6 .

Changes in the local environment are important to understand because they can create, or reduce the number of, suitable breeding sites for local vectors, thereby affecting their abundance and transmission patterns. In central Kenya, the introduction of irrigated rice cultivation appeared to reduce the risk of malaria transmission by An. funestus but not by An. arabiensis 24 , and in Lake Victoria, a recent reduction in water level has created newly emerged land and habitats more suitable for An. funestus than for An. gambiae 25 . Along the Kenyan coast, information on the impact of urbanisation 22 , agricultural activities and changes in climate on malaria transmission is limited, but becoming increasingly important. Currently, the prolonged drought affecting Kwale District and other Kenyan communities, has resulted in changes to human food security, population movement, cattle density, grazing and water storage practices 26 27 , which will almost certainly alter vector abundance distributions and the risk of malaria.

Similarly, the impact of interventions such as insecticide-treated bed nets (ITNs), long-lasting insecticide-treated nets (LLINs) and indoor residual spraying (IRS) should be considered as they may affect species differently, especially if distributed widely over large geographical areas. The introduction of ITNs in Kilifi and Kwale District during the 1990s significantly reduced the number of indoor-resting Anopheles species, and a change in mosquito composition and biting times of An. gambiae s.l. 28 , and in feeding preference of An. funestus with a shift among the outdoor resting females from endophagy on humans to exophagy on animals 15 . However, these ITNs were restricted to selected areas and are unlikely to have affected the overall relative abundance of the difference species in the study region. Other studies in East Africa 29 30 and elsewhere 31 32 33 have shown that An. funestus can be readily eliminated from an entire area by IRS programmes. However, this vector can reappear and become widespread again, sometimes with resistance to the insecticides used in the spray campaign 34 35 36 . This poses a further complication for vector control programmes. It also emphasizes the need for on-going mosquito and insecticide resistance surveillance 37 , especially given the mass distribution of LLINs and IRS programmes currently taking place across sub-Saharan Africa, which could alter mosquito compositions and transmission dynamics over time 38 .

Although there was considerable overlap between An. gambiae s.s. and An. arabiensis, An. gambiae s.s had no significant clustering or environmental differences between high and low transmission locations. The reasons for this may be related to its wide distribution and ability to exploit a range of habitats 1 2 3 18 , but may also be because this species may comprise different molecular or chromosomal forms which are not well defined in this region compared with other regions of sub-Saharan Africa 1 . In West Africa, the chromosomal forms of An. gambiae s.s have shown to have differing spatial distributions and environmental parameters 39 40 41 , and distinct differences between the M and S molecular forms have been described in Mali 42 . In this coastal region of Kenya, only the An. gambiae S form has been detected in two locations 43 , therefore, a better understanding of the speciation and transmission patterns of the An. gambiae s.s forms is crucial, especially as An. gambiae s.l appears to be the main vector of both malaria and lymphatic filariasis in Kilifi and Kwale Districts 15 44 45 .

The study by Mbogo et al 6 collected mosquitoes using pyrethroid spray catches (PSC) inside houses, which could potentially underestimate the abundance of exophilic mosquito species such as An. arabiensis, as shown in other East African countries 46 . In general, measuring the population dynamics of An. gambiae s.l and An. funestus is difficult, and studies have shown great variability depending on the sampling technique used, and whether interventions such as ITNs are present and acting as a deterrent 16 28 47 48 49 . The presence of cattle for An. arabiensis is also an important consideration as they prefer to feed on these animals over humans and other livestock 46 50 51 . Although there are limitations to using the PSC method to estimate abundance and transmission patterns, the study by Mbogo et al 6 is one of largest datasets available for East Africa, which compares the abundance and transmission potential of An. gambiae s.l and An. funestus across a diverse ecological range using a standard sampling technique.

Malaria transmission is complex, and more knowledge on the relationship between the environment, mosquito vectors, human disease and demography in sub-Saharan Africa will help implement appropriate control measures in a rapidly changing landscape. This is particularly important in areas already reporting changes in transmission intensity 52 53 , and may be additional factors to include in future malaria models. This small follow-up study to Mbogo et al 6 aimed to elucidate environmental factors associated with the An. gambiae and An. funestus complexes in one region of Kenya. It exemplifies what can be done with existing entomological data contained in the literature and elsewhere, and how modern satellite and GIS technologies in public health research may be exploited, especially for climate sensitive diseases in developing countries, such as malaria 54 55 .

The authors declare that they have no competing interests.

LKH identified data sources, designed the study, carried out the data analysis and wrote the first draft of the manuscript. JH participated in the interpretation of the results and editing of the manuscript. EM conceived the idea for the study, and contributed to the writing and editing of the manuscript. All authors have read and approved the final manuscript.