Decisions for sustainable forest resources management require a better unde
rstanding of forest dynamics as a result of disturbances. Forest fires are
one of the major types of natural disturbances in many forest landscapes. E
xisting fire regime models can be used in evaluating the influence of a fir
e regime on landscape dynamics, but these models require users to define fi
re frequency and its size distribution as separate characteristics of a fir
e regime before running the models. By using a long-term spatial fire regim
e model, we tested the hypothesis that fire frequency and its size distribu
tion are correlated with each other under natural conditions. Our study dem
onstrates that the hypothesis can not be rejected and the correlation betwe
en fire frequency and its size distribution was robust. Thus, the natural f
ire size distribution can be estimated when the fire return interval for a
given forest landscape is known, if the natural fire size distribution can
be approximated by a negative exponential probability distribution. This re
sult simplifies the description of a fire regime from two parameters to one
in some existing fire regime models. This simplification is limited to a '
let burn' scenario.