Apple scab, caused by Venturia inaequalis, and peacoq leaf spot, cause by Fusicladiumoleagineum, are key diseases respectively for apple and olive growers. These disease usuallyrequire a large number of treatments. Adequate protection need a good evaluation ofquantitative disponibility of inoculum, and estimation of infection conditions.In this thesis, these two problems were studied using only field data. This is an originalaprroach to obtain knowledge about biology of fungi : most previous works were realised byregression of laboratory data obtained under controled conditions.In the case of Venturia inaequalis, primary inoculum consists of pseudothecia present in leaflitter. Treatments agains ascospore release period is the cornerstone of the strategy againstapple scab. However, the existing forecasting models are not reliable, and are all based ondegree-day time scale, proposed in 1982.Here, using a corpus of data acquired between 1996 and 2013, including observations ofascospore release and weather data, we assessed the daily rate of development of primaryinoculum by fitting generic new time scale functions. Further improvements were then studiedto take into account elements reported in litterature about the incidence of rain or wetness.Different methods were tested and adapted for the parameterisation of models by numericaloptimisation. Some forcasting models were proposed and adapted to the area where the studywas conducted, with parameters including rain and temperature. The validity was tested, andfurther developements of the forecasting tool was then proposed.In the case of Fusicladium oleagineum, a field-operational model predicting disease outbreakswas established as a function of temperature and relative humidity. First with the help ofpoints selected by experts, Secondly automaticaly using a neural network. A model defininglatent period as a function of average temperature after contamination was then realised.