Predictive Modeling of Rice Germination Sensitivity to Salinity Using Literature-Derived Data

Soil salinity remained a major constraint on crop establishment, and rice (Oryza sativa L.) was recognized as particularly sensitive during germination. The present study aimed to quantify the concentration-dependent response of rice germination to NaCl stress using a literature-derived dataset and regression-based modeling. Published data on rice germination under different NaCl concentrations were compiled from peer-reviewed studies identified through Google Scholar. Germination percentage was used as the primary response variable, and linear regression as well as polynomial regression models of degree 2 and degree 3 were implemented in Python using the scikit-learn library. The compiled dataset showed uneven representation across salinity treatments, with concentrations of 75, 125, and 150 mM being reported more requently than several low and intermediate–high levels. Overall, germination declined with increasing NaCl concentration, although moderate-range variability was observed in the aggregated data. Regression analysis showed that all models captured the general negative trend, while their predictive behavior differed according to the evaluation metric. The second-degree polynomial model produced lower mean absolute error, whereas linear-type models showed stronger performance for error variance-related metrics in the current dataset. The study demonstrated that regression-based approaches were useful for integrating fragmented salinity experiments and for approximating the germination response of rice across tested and intermediate NaCl levels.