Mutual Neighbour Analysis of Fog Transition Patterns in SYNOP Observations from Italian Stations

Understanding how fog forms and dissipates is important in aviation because rapid changes in visibility affect operational decisions, airport capacity, and safety. For this purpose, we tested whether a symmetric mutual neighbour rule reveals asymmetric class participation and transition structure in 2,890,458 surface synoptic (SYNOP) records from 72 Italian stations, including many stations located at airports. FOG and NO FOG denote operational classes defined by visibility below and at least 1000 m, respectively. Within each station, a fog and no-fog record were paired when each was the other’s nearest member of the opposite class in a meteorological–temporal representation. The analysis identified 75,050 pairs. The median station coverage was 84.6% for fog and 1.73% for no fog. After stratification by station, visibility band, month, and reporting hour, no-fog transition records had higher odds of mutual set membership (OR = 2.52, 95% CI 2.11–2.87), with positive associations in 68 of 72 stations. The fog-side association was weaker and heterogeneous (OR = 1.42; 37 positive stations). Historical out-of-fold errors were also more frequent inside than outside the mutual set across classifier families. Thus, a formally symmetric relation produced markedly asymmetric participation and consistently isolated no-fog observations near the observed onset and dissipation. These findings show that mutual cross-class analysis isolates a small and interpretable set of observations within the adopted representation that helps to characterise transitions between fog and no-fog states in aviation weather data.