Evaluation of preventive cancellations: discrepancy between aeronautical forecast (TAF) and observation (METAR) using temporal models

Abstract

This study evaluates in the ten Peruvian airports with the highest movement between 2015 and 2025, to what extent the discrepancy between the TAF aeronautical forecast and the METAR observation is associated with preventive cancellations, operationalized through a monthly proxy based on commercial operations and observed meteorological severity. Historical TAF/METAR series downloaded from OGIMET and commercial operations by airport-month were integrated from CORPAC's National Airport General Movement report. The proxy label was defined as the conjunction between a strange drop in operations and observed adverse weather determined by global percentiles of METAR variables (FG/TS phenomena, visibility, wind and ceiling). A base model with METAR variables was compared with models incorporating TAF-METAR discrepancies, using temporal validation (train ≤2022; test ≥2023). The results show incremental improvement by including robust discrepancies in phenomena such as fog and storm. In addition, airports with a higher frequency of falls without adverse weather observed were identified, providing evidence to focus operational audits and future studies with flight-level data.