Climate influence on Panama Canal operations: Predicting canal water times with integrated environmental and operational data

In this study, we examine the extent to which Panama Canal Water Times (CWT) can be predicted using a combination of historical values and external features, including vessel activity, climate indicators, and operational constraints. With the use of high-frequency Automatic Identification System (AIS) data, Gatun Lake water levels, and the Oceanic Niño Index (ONI), we reconstruct a weekly panel of vessel transits and evaluate a range of prediction models, including statistical time series methods and supervised machine learning algorithms. Among these, Long Short-Term Memory (LSTM) networks demonstrate better performance for short-term predictions, while Random Forests perform best at longer horizons. Our results show that lagged CWT values and vessel behaviour, particularly prearrival speed, are strong predictors of future congestion. Importantly, we identify a consistent negative relationship between ONI (lagged six weeks) and CWT, suggesting that El Niño triggers prompt preemptive actions such as early booking of transit slots, which may reduce CWT. These climate-related effects also appear to influence the composition of vessels, with bulk carriers reducing throughput, while more time-sensitive vessels, such as container vessels, remain largely unaffected. These findings have important implications for voyage planning through the Panama Canal and contribute to broader discussions on how climate variability affects sensitive maritime chokepoints. • El Niño events, on average, reduced Panama Canal Water Times–contrary to belief. • LSTM forecasts short-term Canal Water Times best; Random Forest leads long-term. • El Niño reduces bulk transits more than container or other high-value segments. • Findings inform debate on climate risks and their impact on maritime operations.