The FERRUGEM simulation model of bean rust epidemics, based on infecti
ons of Uromyces appendiculatus that occur on daily cohorts of bean lea
f growth, was validated with data from 35 bean rust epidemics monitore
d in the southeastern region of Brazil. Subjective and goodness-of-fit
tests of data were used. Good agreement was obtained between model ou
tput and real data in the majority of epidemics. Linear regression lin
es of Gompertz-transformed proportions of pustular area predicted by t
he model did not differ for both slope and intercept from regression l
ines of observed pustular areas in 60.0% of the epidemics. In 71.4% of
the epidemics, there was no significant difference in intercepts. Lac
k of agreement between model-predicted outcomes and observed epidemics
occurred in two situations: when the initial proportion of pustular a
rea observed in the field was very low (<0.000005) and when there were
long periods of weather very favorable for infection. In the first si
tuation, it may be possible to improve the fit of the model by more ri
gorous sampling to obtain accurate estimates of initial pustular area,
because solitary pustules and small foci are easily missed by chance.
In the second situation, the model overestimated the final pustular a
rea, possibly because long periods of leaf wetness may have had a nega
tive influence on dispersal of the urediniospores. This latter factor
was not considered in the model routines. Specific experiments to eval
uate the effect of weather variables on the dispersion of U. appendicu
latus should be conducted before the model is implemented.