Simulated predictions of the proportion of emerging mosquitoes that will take a given number of blood meals from humans over their lifetimes, depending on their natural preference for humans and the protection of those humans with interventions that either repel or kill them, taking Anopheles culicifacies (A), An. arabiensis (B) and An. gambiae (C) as examples of vectors with weak, intermediate and strong preferences for feeding on humans, respectively. All simulations were implemented exactly as described previously , assuming that these mosquitoes differ only in their preferences for human and cattle hosts (parameterized as per ), and that high demographic coverage (C
= 0.8) and protective efficacy (ρ = 0.8) of the intervention measures are maintained at all times of the day (π
= 1). All toxicity is assumed to act on contact before mosquitoes feed so that products with toxic (θ
=0) and repellent (θ
=0.8) profiles confer equivalent personal protection (ρ = 0.8) and differ only in the level of community-level protection achieved [14–16]. The proportional frequency of emerging mosquitoes which take a given number of human blood meals per lifetime (F
) is calculated as product of the mean probability of survival per feeding cycle (p
) and the human blood index (Q
) to the power of the number of blood meals (i) divided by the sum of the values for this term for all possible numbers of blood meals: . Parameter values for the relative availability of humans, compared to cattle, were estimated based on published field observations of variations in human blood index with local host abundance, exactly as previously described for Anopheles gambiae and Anopheles arabiensis, and by direct comparison of observed attack rates upon cattle and humans for Anopheles culicifacies.