The Highlands of Madagascar were malaria free until 1878, when a severe epidemic occurred, following the development of irrigated rice farming. Then, the disease became endemic. Between 1949 and 1962, malaria was "eradicated" on the Highlands by joint house spraying and chemoprophylaxis measures. The main vector An. funestus disappeared. In 1986-1988, a very severe epidemic with high lethality rate devastated the Highlands. It is now under control. Thanks to the data of a religious dispensary, we could follow the evolution of malaria on the Highlands from 1971 to 1995. The number of cases begin to grow in 1975 when the surveillance was neglected. A second step was observed in 1979, when chemoprophylaxis/chemotherapy centres were closed. Then, the increase of malaria became exponential up to 1988. At the time, the prevalence had became similar to that of 1948, before the eradication. The epidemic is not due to global warming because the temperature has been stable for the last 30 years. The malaria rise was due the cancellation of control measures. When control was reactivated, the epidemic ceased. In Swaziland, Zimbabwe and South Africa, malaria epidemics were also due to control failure. In Uganda Highlands, above 1500 m, malaria rise seems linked to the environmental changes, e.g. the cultures which replace papyrus swamp in the valley. But malaria did not overcame the altitude of 1900 which it had already reached in 1960. Rainfall should also be considered as a key factor in the epidemics. In the Sahel West Africa, temperature increased from 0.5 degree C to 01 degree C degree in the last 25 years, but rainfall decreased from 30%. As a result, one of the vector, Anopheles funestus disappeared and malaria prevalence dropped by 60 to 80%. It is not acceptable to predict the future evolution of malaria in taking in account only one parameter: the temperature. The whole factors involved in the epidemiology should be taken into account. The predictions based only on the temperature increase (global warming) can be totally wrong if the rainfall, for example, decreases.