Atmosphere Ocean Science Student Seminar

Isentropic Analysis of Convection’s Response to Numerics and Thermal Forcing

Speaker: Michael Lever

Location: Warren Weaver Hall 1314

Date: Friday, February 14, 2020, 4 p.m.


Thermodynamic relationships for moist air are highly
nonlinear, and therefore only truly hold on the length-scale of an
individual air parcel. Additionally, under the Eulerian framework of
integration, deep convection is a rare event in space-time with
approximately one percent of the domain involved at any one time. These
two facts couple to make analysis difficult. In practice, one petabyte
of floating point data may be generated for an experiment using a
numerical model of convection. This amount of data is infeasible to work
with and so averaging must be performed, but the rarity of events and
the non-linearity of the dynamics makes simple averages in space or time

Instead, I will use an isentropic streamfunction and the Mean Air Flow
as Lagrangian Dynamics Approximation (MAFALDA). These methods use
averages of observed quantities conditioned on height and an adiabatic
invariant -- like entropy. This reduces the dimensionality of the data
from 4D to 2D and can reduce the size of the data from terabytes to a
few megabytes. Even though the data has been so strongly averaged, this
physically motivated averaging scheme preserves the information of
interest. Furthermore, the 2D, steady state nature of the average allows
construction of a streamfunction. Closed streamlines are then
interpreted as representative parcel paths, and classical thermodynamic
analysis -- which usually is accurate only for a parcel -- succeeds in
extracting important bulk properties of the flow.

Namely, the response of convection to vertical resolution and the change
in kinetic energy generation with warming sea surface temperatures will
be discussed.