Effects of straw return on soil salinity and crop yield in salt-affected soils across Asia: A meta-analysis combined with machine learning

Soil salinization threatens agricultural productivity and sustainability worldwide, particularly in Asia, where salt-affected croplands are widespread. Straw return is a potentially sustainable strategy for mitigating salinity stress and improving crop production, but its effectiveness varies with environmental and management conditions. We synthesized 75 field studies comprising 754 paired observations for soil electrical conductivity of the saturated paste extract (ECe) and 417 paired observations for crop yield. A multilevel meta-analysis was combined with moderator analyses, continuous meta-regression, treatment-level ECe–yield analysis, and interpretable machine learning. Overall, straw return reduced soil ECe by 28.86% (95% CI: −36.90% to −19.77%) and increased crop yield by 20.15% (95% CI: 17.97–22.36%). Treatment-level analysis showed a significant positive association between ECe reduction and crop-yield response ( β = 0.477, 95% CI: 0.272–0.683, P < 0.01). Moderator and meta-regression analyses indicated that straw-return effects varied with straw application rate, straw type, incorporation depth and duration, initial soil ECe and pH, crop species, and climatic conditions, highlighting their context-dependent effects on soil salinity and crop yield. XGBoost showed the highest predictive accuracy, with R 2 values of 0.82 and 0.72 for soil ECe and 0.66 and 0.62 for crop yield in the training and testing datasets, respectively. SHAP analysis identified straw application rate, initial soil ECe, MAT, straw type, and MAP as major predictors of the soil ECe response, whereas MAP, initial soil pH and ECe, incorporation depth, and MAE were most important for crop-yield prediction. These findings highlight the context dependence of straw-return effects and support site-specific management to jointly improve soil salinity and crop productivity.