What Determines the Success of Restoring the Threatened African Savanna Mahogany ( <i>Khaya senegalensis</i> ) Beyond Seed-Tree Size? A Multi-Model Test of Site and Maternal-Origin Effects

<jats:title>Abstract</jats:title> <jats:p> <jats:italic>Khaya senegalensis</jats:italic> (Desr.) A. Juss., the West African savanna mahogany, is listed as Vulnerable on the IUCN Red List due to prolonged commercial and medicinal overexploitation, compounded by failing natural regeneration amid growing deforestation and climate change. Restoration strategies generally rest on the implicit assumption that rigorous selection of seed trees based on their size improves offspring quality. Yet this assumption has never been confronted with the competing hypothesis of a determinism dominated by the planting site, using statistical approaches of adequate power. This study tests the relative contribution of seed-tree dendrometric characteristics (diameter, height) and planting site to germination and juvenile growth (6-24 months) of <jats:italic>K. senegalensis</jats:italic> , using six complementary analytical approaches: continuous regression, linear mixed models with site × seed-tree interaction, seed-mass mediation analysis, non-linear growth-trajectory modelling, between-site phenotypic plasticity indices, and multi-predictor allometric equations, across two contrasting bioclimatic sites in Côte d’Ivoire (Korhogo: Sudanian savanna, and Daloa: humid forest) and six seed-tree categories. Converging across all these approaches, the results show that the planting site exerts a decisive influence on germination (71.1% at Korhogo versus 63.9% at Daloa; χ <jats:italic>²</jats:italic> = 6.09; <jats:italic>p</jats:italic> = 0.014) and on the entire growth trajectory ( <jats:italic>p</jats:italic> &lt; 0.001), whereas no seed-tree characteristic affects offspring performance, whether in continuous regression ( <jats:italic>p</jats:italic> &gt; 0.37), mixed models with interaction ( <jats:italic>p</jats:italic> &gt; 0.29), or through a seed-mass mediation pathway (Sobel test, <jats:italic>p</jats:italic> = 0.46). An exponential growth model fits better than linear or power-law models (Δ <jats:italic>AIC</jats:italic> = 13.6-14.7 depending on site), and multi-predictor allometric equations (diameter and height) explain 97.4% of the variance in total biomass, versus 94.8% for the single-predictor model. These convergent results indicate that, contrary to the widespread practice of selecting seed trees on size criteria, it is the choice of planting site that truly determines the success of restoring this vulnerable species. A conclusion calling for a reorientation of priorities in reforestation and agroforestry programmes across West Africa. </jats:p>