Modeling and prediction of tropospheric ozone based on singular spectrum analysis and linear recurrent formula in Campo Grande, Brazil

This study applies Singular Spectrum Analysis (SSA) and the Linear Recurrent Formula (LRF) to model and forecast daily tropospheric ozone (O₃) concentrations in Campo Grande, Brazil, using satellite data from the SISAM/INPE system for the period 2000–2018. The original O₃ series was decomposed into interpretable components that revealed annual, semi-annual, and quarterly oscillations associated with regional meteorological dynamics and photochemical processes. SSA efficiently separated the signal from noise, while the selection of window length (L = 30, 60, 90) influenced the identification of dominant periodic structures. Forecasting with the LRF model yielded satisfactory performance (RMSE = 0.789 ppb; MAE = 0.636 ppb). Although a statistically significant upward trend was detected (p < 0.01), its low explanatory power (R² = 0.009) indicates that O₃ variability is mainly controlled by seasonal and meteorological factors. The results confirm the robustness of the SSA–LRF approach as a non-parametric framework for detecting temporal structures and predicting short-term ozone fluctuations in tropical urban environments, contributing to air quality monitoring and environmental management strategies.