Dry season survival of Anopheles funestus, Anopheles gambiae and Anopheles arabiensis in the Kilombero valley a dry savannah zone of east Africa, was investigated with over 400 collections from 23 areas, covering 300 sq km of the valley. Anopheles gambiae was found only in association with humans, in forested areas of high annual rainfall, while An. funestus occurred at high densities at the valley edge where large non-moving bodies of water remained. A large population of An. arabiensis was present along the river system throughout the middle of the valley, and mosquitoes probably derived from this population were occasionally caught in villages bordering the valley. No evidence was obtained of aestivation in any mosquito species. Anopheles gambiae was the most long lived, 6.3% compared to 2.0% of the An. arabiensis and 4% of the An. funestus surviving for four or more gonotrophic cycles, the approximate duration of the extrinsic cycle of most malaria parasites. Oocysts of malaria parasites were found in 5.4% of An. funestus and 2.3% of An. arabiensis from villages. Oocyst rates in An. funestus differed significantly between areas but not between houses within areas. Anopheles funestus is the most important dry season malaria vector in the valley, and remains in foci closely associated with groups of houses. All three species survive at high densities but as otherwise hidden refugia populations.

Non‐radioactively labelled DNA probes were tested for their ability to identify dried museum specimens of Anopheles gambiae and its sibling species An.arabiensis. The specimens were the progeny of wild‐caught females collected in 1991 from villages in western Kenya. Three years later, specimens whose identity was known to the second author were provided ‘blind’ to the first author for identification with oligonucleotide probes (SH 5 and SH 4 derived from pAngss and pAngsl, respectively) using a simplified squash‐blotting protocol for non‐radioactive probes. All specimens were successfully identified with whole‐body squashes, and the results agreed with previous identifications of parents or siblings based on rDNA‐polymerase chain reaction . The amounts of DNA released by squash‐blotting were just sufficient for identification by those experienced in the technique, but not for squashes of heads or thoraces alone. An aim of the study was to determine whether squash‐blot methods of identification might be useful for establishing the genetic identity of name‐bearing type specimens of sibling species held in museum collections.