Objective characterization of mesoscale cloud patterns from graph theory and an Ising-like model (BICIM).

The organization of mesoscale cloud fields is crucial for understanding atmospheric dynamics and their modeling. 
The role played by these cloud structures on their direct environment and, more generally, on the climate remains challenging to incorporate in climate models. 
In this contribution, we propose a methodology for automatically identifying and studying these cloud organizations, combining two innovative approaches: 
(1) an Ising-like model (called BICIM) capable of reproducing cloud fields with specific organization patterns, and 
(2) graph theory, applied to the outputs of this model and satellite observations, which allows us to derive distributions of surfaces and perimeters of clouds as well as inter-cloud distance distributions. 
For the first point, sensitivity tests on input data and fluxes revealed that BICIM consistently responds to changes, produces realistic results, and highlights humidity and wind as key factors in the formation of cloud organizations.
From the second point, we propose a new quantity, denoted as M, as a Metric for Assessing Similarity between Cloud Organization Layouts (MASCOL). 
As its name suggests, M quantifies the similarity between two cloud fields. 
We apply this methodology to satellite data, identifying cloud structures, with MASCOL aiding in classifying cloud fields relative to reference organizations (from BICIM).
This approach therefore enables a faster and more objective identification of each structure compared to visual methods. 
The reason is that it is based on graph theory, which is an efficient mathematical tool.
Furthermore, the methodology's reliance on graph theory and robust pattern recognition metrics makes it particularly well-suited for integration with machine learning and artificial intelligence techniques, opening avenues for automated analysis and large-scale applications.