Optimization and Validation of a Multitrait Physiological Drought Mitigation Index (PDMI) for Screening PGPR-Induced Drought Tolerance in Strawberry

Abstract

Drought mitigation by plant growth-promoting rhizobacteria (PGPR) is commonly assessed using individual physiological traits, although plant responses to water deficit are intrinsically multivariate. We developed a Multitrait Physiological Drought Mitigation Index (PDMI) to summarize PGPR-specific physiological responses in strawberry (Fragaria × ananassa) while separating drought-specific effects from general performance under optimal moisture. The balanced two-season experiment comprised 240 plant-level observations and 10 complete matched replication sets per season. The primary PGPR analysis excluded the non-bacterial magnesium-sulfate comparator (CMg) and included four PGPR strains, the uninoculated control, 200 plant-level observations, and 80 replication-level difference-in-differences (DiD) profiles. Nine nonredundant physiological traits were direction-aligned and standardized using the 2021 development season and aggregated with equal weights; all parameters were then applied unchanged to the 2022 holdout season. Block-adjusted MANOVA confirmed a joint inoculation variant × moisture interaction (Pillai’s trace = 0.533, F(36,664) = 2.835, p < 0.001). Block-restricted PERMANOVA detected a small multivariate strain effect (pseudo-F = 0.995, R2 = 0.0378, p = 0.0289), whereas PERMDISP was nonsignificant (p = 0.273). In 2021, DLGB 2 obtained the highest PDMI (0.160) and had a 66.3% bootstrap probability of rank 1 and an 88.3% probability of inclusion in the top two. In the 2022 holdout, AJ 1.2 ranked first; cross-season rank agreement was positive but uncertain (Spearman’s ρ = 0.60, p = 0.40). The best individual trait, water-use efficiency, showed greater within-season rank stability than PDMI (P(rank 1) = 0.998), demonstrating that the composite did not universally outperform single measurements. PDMI was not significantly associated with holdout yield mitigation (R2 = 0.0458; block-clustered p = 0.127; within-block permutation p = 0.341). PDMI should, therefore, be interpreted as a transparent integrative descriptor of coordinated physiological response rather than a definitive classifier, a universally superior screening metric, or a standalone yield predictor.