Mathematical Modeling and Prediction of Heatwaves Using Time Series and Satellite Data

Source abstract

Urban heatwaves require methods capable of capturing seasonality, persistence, and extreme temperature events. A retrospective modeling framework was developed for Guayaquil using reanalysis-based air temperature, oceanic anomalies, and satellite-derived land surface temperature. Temporal dynamics were represented using dynamic harmonic regression with autoregressive integrated moving average (ARIMA) errors and compared against a persistence baseline through independent chronological validation. A probabilistic forecasting approach and an alternative heatwave definition were also evaluated. Temporal modeling reduced the forecasting error for ERA5-Land maximum temperature relative to the naive baseline, whereas oceanic anomalies provided only marginal improvements for maximum temperature and degraded minimum temperature predictions. Probabilistic forecasting improved heatwave event detection, although it also increased the frequency of false alarms. MODIS observations revealed greater nighttime land surface te