The extrinsic incubation period (EIP) of malaria is the time required for Plasmodium parasites to undergo sexual reproduction and asexual replication within the mosquito (sporogony). The EIP is long relative to mosquito life expectancy meaning most mosquitoes do not live long enough to transmit malaria and small changes in mosquito survival strongly affect malaria transmission. The EIP is typically estimated in a population of experimentally infected mosquitoes, which are dissected on different days post infection for the presence of infectious, salivary gland sporozoites. I developed a novel mathematical model to estimate the EIP distribution from these data (Chapter 2). This model accounted for heterogeneity in mosquito infection and infection-dependent differences in mosquito survival, which could affect the EIP estimated using historical methods.

Existing malaria transmission models that incorporate temperature-dependent changes in the EIP mostly use an existing degree-day model, which was parameterised using the time for the first Anopheles maculipennis mosquitoes (a species complex native to Europe) to be dissected with sporozoites. I therefore estimated the effects of constant temperature on the EIP distribution among Anopheles gambiae sensu stricto mosquitoes (a dominant Plasmodium falciparum vector) (Chapter 3). These EIP estimates were similar to the degree day model except at low temperatures.

Theoretically, accounting for the complete EIP distribution marginally affects malaria transmission compared to assuming that all mosquitoes have the mean EIP. If vector control and mosquito mortality are high, however, accounting for heterogeneity in the EIP is important because mosquitoes with a short EIP become more important for transmission (Chapter 4).

In the wild, temperature is not constant. I therefore developed a model to predict changes in the EIP of mosquitoes exposed to fine scale diurnal and seasonal fluctuations in temperature using the constant-temperature EIP relationship (Chapter 5). Model performance was assessed using laboratory data from mosquitoes that were exposed to a range of different temperature profiles. Incorporating fluctuations in temperature improved the model performance.

Accounting for changes in the mosquito age distribution due to seasonality in mosquito emergence is theoretically important when measuring sporozoite prevalence (Chapter 6). I incorporated seasonality in mosquito emergence and temperature-dependent changes in the EIP into a mechanistic malaria transmission model and assessed the model’s predictive accuracy using seven years of monthly sporozoite prevalence samples from a holoendemic area of Senegal. Seasonality in mosquito emergence caused a small increase in predictive accuracy, and temperature-dependent changes in the EIP had a very small effect if mosquito emergence was allowed to vary and no effect if mosquito emergence was assumed to be constant (Chapter 7).