Several attempts have been made to map the distribution of Anopheles gambiae s.s. and Anopheles arabiensis[1–5], two of the most important vectors of human malaria in sub-Saharan Africa. MacDonald  showed that limiting the human-vector contact reduces malaria transmission, and that the most efficient control measure is to increase the mortality rate of the involved mosquitoes. His thinking has been adopted in current malaria control efforts. Two of the most common interventions today are indoor residual spraying (IRS)  and insecticide-treated bed nets (ITNs) . Often, there is no detailed understanding of the life history, behaviour and species composition where the interventions are applied .
Anopheles arabiensis inhabits areas from South Africa in the south to Mauritania and Sudan in the north. In Central-West Africa there is a pocket with very few observations of An. arabiensis. The border of this pocket is formed by Angola, Zambia, Burundi, Rwanda, Uganda, South-Sudan, Central African Republic, Congo, Gabon, and Equatorial Guinea. Anopheles gambiae s.s. is currently separated into five chromosomal forms: Forest, Bamako, Savanna, Mopti and Bissau , and two molecular forms: M and S [10, 11]. It is distributed from South Africa to Mauritania and northern Mali, but is absent in Ethiopia and Northern Sudan. The species is considered the most efficient malaria vector in Africa .
Recent studies have shown that interventions aimed to prevent malaria has an impact on balance between An. gambiae s.s. and An. arabiensis. The relative fraction of each species can vary from month to month, and year to year . In Tanzania it has been shown that multi-decadal changes in the species composition can influence malaria transmission . Given the observed changes in species composition, and their different capacity as vectors of malaria, it is highly relevant to have models which include several species when assessing the impact of climate variability and climate change.
This paper is the second of two describing and validating a new model of the dynamics of An. gambiae s.s. and An. arabiensis The model, which is described in part one , is a biophysical model driven by output from a climate model. Biophysical models seek to understand what drives a certain biological process, and to describe this with mathematical equations. Unlike statistical models, which often rely on observations to predict species presence and absence, biophysical models can be run with no information with respect to observed distribution and densities, and base the model equations on laboratory studies aiming to isolate different aspects of the life history of the mosquitoes. The role of field observations on the presence or absence of a species in the case of biophysical models, is to validate the model after an experiment has been completed. In some studies observations are used to reduce the uncertainty of unknown parameters .
In addition to predicting the current distribution, these type of models can be used to project changes in the historical and future density and distribution of these species. They can describe changes from day-to-day, month-to-month, year-to-year, and decade-to-decade. The model, named Open Malaria Warning (OMaWa) , includes several components, describing the mosquito’s life from the aquatic stages to adult. In the aquatic stages, life history varies for eggs, larvae and pupae. As adults the life history changes with age. OMaWa is driven with air temperature, relative humidity of the air, wind speed and direction, soil temperature, relative soil moisture, and runoff from a climate model. These variables are used to parametrize mortality, rate at which eggs are laid, biting rate, development rate in the aquatic stages, and dispersion (spread) of mosquitoes. In part one, it was shown how the model responded to different forcings, and focused on its sensitivity to temperature, humidity, mosquito size, the probability of finding blood, and dispersion. Thus the results presented here should be seen in light of the sensitivity analysis. A full description of the model used here can be found in part one .
This is the first time a biophysical model has been used to model the relative density of An. gambiae s.s. and An. arabiensis, with simulations covering an entire continent. It is also the first time age dependent life history and mosquito dispersion (spread of mosquitoes) has been included in a continental analysis. The model is validated against 6,927 presence/absence points of the two species, and a more detailed analysis is carried out for Madagascar. The data is freely available to the public . This study has also evaluated the ability to model the temporal variability, using case studies for Ethiopia.