The standard E - epsilon model generates a planetary boundary layer th
at appears to be much too deep. The cause of the problem is traced to
the equation for the dissipation rate (epsilon) of turbulent kinetic e
nergy (E), specifically the parameterization of dissipation production
and destruction. In the context of atmospheric boundary-layer modelli
ng, we argue that a part of the dissipation production should be model
led as the input to the spectral cascade from the energy-containing pa
rt of the spectrum, with a characteristic length l(epsilon), while the
'equilibrium' imbalance between local production and destruction of d
issipation is modelled as proportional to E-2/E, as in the standard mo
del. We propose an E - epsilon - l turbulence closure scheme, in which
both the mixing length, l(m), and l(epsilon) are prescribed. The impo
rtance of the a equation is diminished, though it still determines the
dissipation rate in the E equation.