This study constitutes the first published cross-seasonal anopheline survey in the northern Kruger National Park. Nine different Anopheles species were collected during the sampling period. PCR identification of members of the An. gambiae complex showed that An. quadriannulatus has the widest distribution, occurring at all five sites sampled, while An. arabiensis was the predominant species. The wide distribution of An. quadriannulatus observed in this survey tallies with previous studies conducted in the Kruger National Park . Anopheles arabiensis was mainly confined to Malahlapanga except in a few instances during the peak of the rainy season when it was found at Sirheni Dam and Louis se gat. Species-specific identification of the An. funestus group showed the presence of An. rivulorum at Malahlapanga and Sirheni Dam.
Species distribution between collection sites
Malahlapanga showed the highest species richness with all nine species recorded there. Sirheni bush camp showed the next highest species diversity (six out nine species collected) while Matiovila and Mafayeni showed the lowest species diversity. The prevalence and distribution of anophelines in the northern Kruger National Park could be explained by the ecological conditions at each collection site. Breeding site availability, animal host availability and the presence of suitable vegetation as a source of carbohydrate (nectar) affects the presence and abundance of mosquitoes [19–21]. Of the sites sampled in this study, Malahlapanga offers the best ecological conditions for mosquito breeding. Malahlapanga contains numerous suitable breeding ponds formed by water flowing downstream from the eye of a natural hot spring. In addition, there are abundant ruminant and antelope herds that use the spring as a water source, providing a blood source for host seeking female mosquitoes. There are also perennial mosquito breeding sites at Sirheni. At Louis se gat mosquito breeding seems to occur in temporary rain puddles formed around the Mphongolo River, although the river was dry during most of the sampling period. Mosquito collections at Louis se gat were only productive during the rainy season from November through to February due to the nature of these temporary breeding pools. Limited species diversity at Matiovila and Mafayeni can be attributed to unfavorable breeding conditions. Results of water tests from these two pools showed that the water was brackish. These salinity levels are lower than those reported by others [14, 22] and might explain the presence of An. quadriannulatus at these two sites. However, it is unclear if the salinity changes during the year as this was not measured in the current study.
Seasonal changes in Anopheles gambiae complex density
Results of these surveys showed that anopheline density in the northern Kruger National Park is seasonal, with the abundance of mosquitoes peaking at the beginning of summer (rainy season). There was seasonal variation in An. arabiensis abundance where numbers increased dramatically following the first rains. The population number then stabilized and then significantly decreased during the dry months. During the collection period there was a second peak in abundance in late summer. This can be attributed to fluctuation in rainfall that decreased prior to a second rainy period in April. These seasonal dynamics changed during November and December 2012. In these months there was a dramatic and unexpected reduction in An. arabiensis abundance at Malahlapanga and an increase in An. quadriannulatus abundance. In addition, three other anopheline species (An. squamosus, An. rufipes and An. merus) were also recorded from this site.
Reasons for this sharp change in species composition at Malahlapanga are unclear, but may have been caused by delayed rains experienced during 2012, resulting in the migration of mosquitoes in search of favourable breeding sites. It is also possible that unusually dry conditions in November-December 2012 resulted in unfavorable saline conditions at the spring. This could have been caused by a lack of groundwater recharge from rainfall at the eye of the spring and evaporative water loss at the surface. During November-December 2012, An. merus were recorded from Malahlapanga for the first time. The presence of An. merus during this time further supports the notion that a change in the salinity of the breeding sites made them unsuitable for An. arabiensis, but still suitable for An. quadriannulatus and other species recorded during this time. It will also be interesting to establish whether this sudden change in anopheline species composition and population dynamic is a permanent change or if the An. arabiensis population will recover over time. This highlights the importance of long-term mosquito surveillance before implementation of an intervention programme.
Another interesting phenomenon observed during this two-year mosquito survey was that collections were highly influenced by prevailing climatic conditions. Three environmental factors (humidity, temperature and wind speed) determined the number of anopheline specimens collected. Generally, humidity above 65%, temperatures above 24°C and wind speeds below 2 m/sec offered the best collection conditions. However, high humidity (85%) provided the most conducive conditions for mosquito collections. This observation is supported by other studies that show that mosquito activity is disrupted by changes in environmental conditions. Snow  showed that biting activity of An. melas and Culex thalassius ceases at wind speeds above 1.2 m/s and, in a similar study from South Africa, it was shown that activity of An. merus is greatly affected by environmental factors such as temperature, wind speed and rain . Gilles and Wilkes,  also showed that wind has a direct effect on mosquito flight. During these collections mosquito activity decreased as a result of an increase in rain drizzle intensity. These weather conditions should be taken into consideration in order to maximize surveillance activities when vector numbers are low, especially in South Africa.
Relationship between species composition and collection method
Analysis of the relationship between collection method and species composition was limited to Malahlapanga where all three sampling methods were productive in collecting mosquitoes. It was generally established that CO2-baited net traps were the most effective adult mosquito sampling technique accounting for the majority of total mosquitoes collected. However, its main disadvantage was that it was not selective for the malaria vector An. arabiensis and collected any host seeking female mosquito regardless of taxon. Large numbers of untargeted culicines and Aedes specimens were collected in the CO2 traps. This method is therefore most suitable for studying species diversity in an area rather than for the collection of specific taxa. Human landing collections were highly effective for collecting An. arabiensis females. This was an interesting observation as this population does not normally interact with humans and mainly feeds on game animals, indicating the opportunistic feeding behavior of this species. Anopheles quadriannulatus are generally not attracted to humans as was evident in these data. Relative species abundances based on adult collection methods do not necessarily compare with those from larval collections. Larval collections are however invaluable in terms of obtaining large samples for routine surveillance.
Field site selection for a pilot SIT feasibility study directed against Anopheles arabiensis
There are a number of factors that need to be considered when choosing an appropriate site for SIT . As the primary objective of this study was to choose an appropriate site to assess population reduction of An. arabiensis, only limited factors were considered. The first determinant was the presence of an An. arabiensis population in significant numbers. Ideally, the population should be geographically isolated to avoid confounding factors such as reinvasion from surrounding populations. It has been shown that an isolated mosquito population can be controlled by SIT [26, 27] unless invasion from surrounding populations causes a reduction in efficacy [28, 29]. The other prerequisite investigated was easy accessibility to the site. The logistics of transporting irradiated males for release and the frequency of site visits to monitor progress in population reduction are cited as important factors to consider for successful implementation of the control programmes .
Five areas in the northern Kruger National Park were evaluated for the presence of An. arabiensis populations. Of the five sites investigated, An. arabiensis were consistently found at Malahlapanga making it an attractive site for a SIT pilot study in South Africa. Furthermore, Malahlapanga is geographically isolated and inaccessible to tourists visiting the Kruger National Park. The nearest human habitation is approximately 9 km away making this mosquito population relatively free from human intervention. Insecticide susceptibility studies carried out on samples collected from this site showed that the population is still susceptible to all classes of insecticides [Munhenga, unpublished data]. Due to the successful malaria control programme in South Africa, there is only one other relatively large An. arabiensis population to be found in the country (at Mamfene in northern KwaZulu-Natal). However, extensive sampling in 2005  showed a relatively small An. arabiensis population at Mamfene compared to the population at Malahlapanga. The greatest drawback of Malahlapanga is that it is not easily accessible throughout the year and indications from this long term surveillance showed that the An. arabiensis population fluctuates dramatically with no warning. Although it is only 62 km from Shingwedzi Research Camp, the greater part of the road leading to the site is smectite clay soil that makes accessibility to the site more challenging in periods of heavy rain.