To investigate how different training modes of salt stress priming affect the dynamic variation of the salt tolerance threshold (STT) in summer maize, a micro-plot experiment with staged brackish water irrigation was conducted. Based on physiological and biochemical parameters, along with shoot and root traits, a dynamic salt tolerance coefficient (α_STT) was defined to quantify STT across growth stages. The results revealed a clear two-stage adaptive response to salt stress, consisting of an initial physiological adaptation phase followed by a phenotypic adaptation phase. Different training modes induced distinct salt stress memory effects by regulating the coordination between these two stages. Among treatments, the S_1-2-3 regime—corresponding to mild (2.0 g·L⁻¹), moderate (4.0 g·L⁻¹), and severe (6.0 g·L⁻¹) salinity applied sequentially at the six-leaf, ten-leaf, and tasseling stages—exhibited the most favorable adaptive outcome, with α_STT gradually recovering to 1.0 at later stages and a concomitantly higher STT. Furthermore, a unified predictive framework was established to estimate STT dynamics, within which the process-constrained PCR-STP pathway outperformed purely data-driven pathways. Overall, our study elucidates the dynamic nature of salt tolerance in summer maize and provides a scientific basis for optimizing brackish water irrigation regimes and refining salt stress modules in crop models.

To investigate how different training modes of salt stress priming affect the dynamic variation of the salt tolerance threshold (STT) in summer maize, and to enable the accurate quantification and prediction of STT, a micro-plot experiment utilizing diverse regimes of brackish water irrigation was conducted. Utilizing physiological, shoot, and root indicators, a comprehensive evaluation framework was developed to define a dynamic salt tolerance coefficient (αSTT), enabling the precise quantification of STT across growth stages. Building on this, the study established a unified predictive framework to systematically evaluate the performance of diverse modeling pathways and machine learning algorithms. The results revealed a distinct two-stage stress response pattern of summer maize to salt stress, characterized by an initial physiological adaptation phase dominated by regulatory adjustments, followed by a phenotypic adaptation phase associated primarily with improvements in growth performance. Different training modes led to distinct salt stress memory effects by modulating the coordination between these two adaptive stages. Among all modes, the S1-2-3 training regime exhibited the most favorable adaptive outcome, with the αSTT gradually recovering to 1.0 during later growth stages, indicating full adaptation to saline stress, and concomitantly exhibiting a relatively high STT. Regarding predictive performance, the PCR-STP modeling pathway incorporating process constraints outperformed purely data-driven pathways, and its combination with CatBoost achieved the highest accuracy (R² = 0.910, RMSE = 0.241). Overall, our study elucidates the dynamic nature of salt tolerance in summer maize, while the proposed STT quantification and prediction method provides a scientific basis for refining salt stress modules in crop models, optimizing brackish water irrigation regimes, and improving precision water resource management in arid and semi-arid regions.