Background: Multi-model ensembles could overcome challenges resulting from uncertainties in models' initial conditions, parameterization and structural imperfections. They could also quantify in a probabilistic way uncertainties in future climatic conditions and their impacts.
Methods: A four-malaria-model ensemble was implemented to assess the impact of long-term changes in climatic conditions on Plasmodium falciparum malaria morbidity observed in Kericho, in the highlands of Western Kenya, over the period 1979-2009. Input data included quality controlled temperature and rainfall records gathered at a nearby weather station over the historical periods 1979-2009 and 1980-2009, respectively. Simulations included models' sensitivities to changes in sets of parameters and analysis of non-linear changes in the mean duration of host's infectivity to vectors due to increased resistance to anti-malarial drugs.
Results: The ensemble explained from 32 to 38% of the variance of the observed P. falciparum malaria incidence. Obtained R²-values were above the results achieved with individual model simulation outputs. Up to 18.6% of the variance of malaria incidence could be attributed to the +0.19 to +0.25°C per decade significant long-term linear trend in near-surface air temperatures. On top of this 18.6%, at least 6% of the variance of malaria incidence could be related to the increased resistance to anti-malarial drugs. Ensemble simulations also suggest that climatic conditions have likely been less favourable to malaria transmission in Kericho in recent years.
Conclusions: Long-term changes in climatic conditions and non-linear changes in the mean duration of host's infectivity are synergistically driving the increasing incidence of P. falciparum malaria in the Kenyan highlands. User-friendly, online-downloadable, open source mathematical tools, such as the one presented here, could improve decision-making processes of local and regional health authorities.