ABSTRACT This study identified four hidden rainfall states in Northeast Brazil (NEB) during the austral autumn (March–May) from 1981 to 2015 using a Hidden Markov Model (HMM). Two states correspond to widespread dry (State 1) and wet (State 2) conditions, both exhibiting strong daily persistence and associations with large‐scale atmospheric features, including the Intertropical Convergence Zone (ITCZ), the South American Monsoon System (SAMS), and the South Atlantic Subtropical High (SASH). The remaining two states are characterized by localized rainfall over the northwestern sector (State 3) and the coastal areas (State 4) of the NEB. Running correlation analyses revealed that the interannual frequency of these states is modulated by oceanic modes, whose influences exhibit non‐stationary behaviour over time. The Extreme Gradient Boosting (XGBoost) algorithm showed a good fit for predicting state frequencies, with particular emphasis on State 1. SHapley Additive exPlanations (SHAP) values highlighted the dominant role of Atlantic variability modes, with a strong negative Atlantic Meridional Mode (AMM) phase and positive South Atlantic Ocean Dipole (SAOD) phase tending to modulate the frequency of wet and dry conditions. The El Niño‐Southern Oscillation (ENSO) shows a stronger association with State 3 and acts as a conditional modulator of State 1, becoming dominant during strong El Niño events. State 4 reflects a more complex inter‐basin structure, with alternating Atlantic and Pacific influences and a secondary contribution from the Indian Ocean. These findings enhance our understanding of the non‐stationary and non‐linear ocean–atmosphere interactions that shape seasonal precipitation variability in the NEB and provide valuable insights for improving seasonal climate assessments and climate risk management in the region.
Silva et al. (Fri,) studied this question.