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.
Souza et al. (Tue,) studied this question.