Multicolumn and global mode

Motivation

Most codes, AGNI included, approach atmospheres as a single 1D column. They thereby represent the full planet with a single zenith angle and stellar flux scale factor (see Stellar irradiation). This naturally cannot capture how climate states vary with longitude and latitude; e.g. the day-night temperature contrast expected on a tidally locked rocky exoplanet. AGNI's multicolumn (globe) mode addresses this by solving several 1D columns, each placed at a different point on the planet's surface, and coupling them through a horizontal heat-redistribution flux. This retains the numerical efficiency of the 1D column solver, without the cost of a fully 3D general circulation model.

Constructing the globe

A globe is built from a single, already-configured Atmos_t instance, which acts as the template ("worker") atmosphere. Multicolumn mode is enabled in the configuration file by adding a [planet.globe] section that specifies the longitude and latitude of each column:

\[\{(\lambda_i, \phi_i)\}_{i=1}^{N_\text{col}}\]

where $\lambda_i \in [0^\circ, 360^\circ]$ and $\phi_i \in [-90^\circ, 90^\circ]$. AGNI constructs one independent copy of the atmosphere for each column, following the pattern described in multicol.construct!. Each column's zenith angle is derived from its longitude and latitude (atmosphere.calc_zenith_angle), and its stellar flux is set accordingly, so that columns nearer the substellar point receive more instellation than those nearer the terminator or nightside. The underlying SOCRATES Fortran library is not thread-safe, so columns are solved sequentially.

Heat redistribution

Each column receives a heat-redistribution flux $F^\text{redist}_i$, deposited in log-pressure: an advective heating term. This allows a column to act as either a net heat source or a net heat sink [60]. Physically-motivated scaling laws can drive the redistribution strength and profile quantities self-consistently from each column's temperature, composition, planetary radius, and rotation rate.

Solving the globe

solver.solve_globe! iterates over the set of columns, calling the standard Newton-Raphson column solver independently on each one, then checks whether all columns have converged to a consistent total flux. Depending on solution_type (see Solution types). Each outer iteration re-solves every column to full convergence, multicolumn runs are correspondingly more computationally expensive than single-column runs, scaling roughly with the number of columns.

Bibliography for this page