A Genetic Algorithm Approach to Infer Jupiter's Rossby Wave Structure from JunoCam Images

We use a selected pair of JunoCam images taken during the inbound or outbound branch of some of Juno's perijoves to derive a polar azimuthal vorticity map. Our goal is the inference of a Rossby wave structure from such vorticity map; Rossby waves are observed in one or more jets in Jupiter's south polar region (Rogers et al., 2022, Icarus 372, 114742). We implement a genetic algorithm to approach this goal. A genetic algorithm is a computer model inspired by Darwinian evolution.

We describe the phenotypical aspect of a Rossby wave structure by means of meridionally Gauss-weighted Fourier terms. The sum of those terms distort circles of latitudes meridionally into more general fibers. The standard deviation of the vorticity values along such a fiber provides a measure of the fitness of a modelled Rossby wave structure with respect to the observed vorticity map.

The genotypical aspect encodes each meridionally Gauss-weighted Fourier term by a gene. Such a gene encodes each parameter of the term by an integer number, which itself is encoded by a string of bits. A genome consists of a set of such genes. It represents the set of terms needed to be summed up into the meridional distortions approximating the Rossby wave structure. The genome describes and represents a member of a population. Our algorithm evolves such a population of genomes. The population starts with genomes initiated with parameters set to zero or to random values, which are then evolved through rounds of mutation and recombination. The basic evolution steps are

• the creation of a new genome by recombining two randomly selected genomes of the population,
• mutation of the new genome,
• the calculation of the fitness of the new genome, and
• the survival of the fittest genomes.

Recombination of two genomes selects randomly about half the genes from each of the two genomes to be recombined. Single bits of the parameters of the genome flip with a low probability to introduce random point mutations. New genes can form that way. Genes are deleted with a low probability after recombination in order to keep the genomes and hence the approximation of the Rossby structure simple.

Several populations can be run with different pseudo-random number seeds in order to investigate the reproducibility of the results.

The development of the algorithm is motivated by our intention to observe changes of the Rossby wave structure over time on the basis of JunoCam images, but also to define reasonable global initial conditions for simulation runs of the polar regions.