A new lightning parameterization has been developed to enable cloud models
to simulate the location and structure of individual lightning flashes more
realistically. To do this, three aspects of previous parameterizations hav
e been modified: 1) To account for subgrid-scale variations, the initiation
point is chosen randomly from among grid points at which the electric fiel
d magnitude is above a threshold value, instead of being assigned always to
the grid point having the maximum electric field magnitude. 2) The thresho
ld value for initiation can either be constant, as in previous parameteriza
tions, or can vary with height to allow different flash initiation hypothes
es to be tested. 3) Instead of stopping at larger ambient electric field ma
gnitudes, extensive flash development can continue in regions having a weak
ambient electric field but a substantial charge density. This behavior is
based on lightning observations and conceptual models of lightning physics.
However, like previous parameterizations for cloud models, the new paramet
erization attempts to mimic only the gross structure of flashes, not the de
tailed development of lightning channels, the physics of which is only poor
ly understood. Though the choice of parameter values affects the dimensions
of a flash, the qualitative features of simulated flash structure are simi
lar to those of observed lightning as long as the parameter values are cons
istent with the larger electric field magnitudes measured in storms and wit
h simulated charge densities produced over reasonably large regions. Initia
l simulations show that, by permitting development of flashes in regions of
substantial charge density and weak ambient electric field, the new parame
terization produces flash structure much like that of observed flashes, as
would be expected from the inferred correlation between observed horizontal
lightning structure and thunderstorm charge.