Abstract Two cirrus cloud systems observed during the winter of 2001 at the Southern Great Plains site of the Atmospheric Radiation Measurements program in Oklahoma, USA are studied because of the distinct neutrally stratified layers formed within the clouds. Observations are obtained with 35 GHz millimeter-wave radar and backscattering cross-section η ( t ) signals within radar-reflectivity restricted sublayers of the clouds are analyzed. The neutrally stratified layers of cirrus clouds are known to be associated with the existence of generating cells. The statistics of radiative properties within the neutrally stratified layers is obtained to be non-Gaussian and time-dependent. The purpose of this research is to derive a model of the cloud-generating cells layer in cirrus based on the statistics of observations. The Fokker–Planck equation approach provides suitable framework to treat non-Gaussian, time-dependent probability density functions (pdfs) such as those found for the η ( t ) signals. It is shown that the deviations from Gaussianity of radiative properties of the neutrally stratified generating cells layer in cirrus can be modeled by linear stochastically perturbed dynamics with multiplicative noise statistics. Because the multiplicative noise is often identified with state-dependent variations of stochastic feedbacks from unresolved system components it is expected that derived stochastic model will be useful for parameterization of cirrus in global circulation models (GCMs).