Contrasting field observations of the collapse of An. gambiae with the predictions of population dynamics models assuming no density-dependence of vector reproduction so that emergence rates are directly proportional to mean longevity. A: Simulated declining biting exposure to Anopheles gambiae s.s. and An. arabiensis as insecticide treated net usage (ITNs) increases. B: Corresponding impact of increasing ITN use upon predicted proportion of bites upon humans by An. gambiae. C: Direct comparison of observed near-disappearance of An. gambiae from Kilombero Valley in southern Tanzania  with simulations based upon observed ITN usage rates estimated as described previously, with only 4.7% of nets treated within the previous 6 months up to 2004 , following which long-lasting net retreatment kits were introduced so all nets reported as treated are considered ITNs . All simulations were executed  assuming equal baseline emergence rates (E0 = 2 × 107 mosquitoes per year) for both vector species and a ratio of cattle to humans consistent with livestock census results in Kilombero Valley (Nh = 1000, Nc =140). All ITN-induced mortality was assumed to occur before feeding so the excess proportion of mosquitoes killed after attempting to attack a protected human was assumed to be negligible (θu, post = 0). Simulated An. gambiae s.s. and An. arabiensis populations differed only in their parameter values for the proportion of human exposure to bites that occurs indoors (πi = 0.9 versus 0.4, respectively [37, 50]), the attack availability rates of cattle (ac = 2.5 × 10-5 versus 1.9 × 10-3 attacks per host per host-seeking mosquito per night ) and the excess proportions of mosquitoes which are diverted (θΔ = 0.2 versus 0.6, respectively) or killed before feeding (θμ,pre = 0.8 versus 0.6, respectively) while attempting to attack a human while using an ITN [52–54].