Nonlinear interactions between fractional vegetation coverage and plant diversity across urban land-use types: insights from the Grand Canal region

• Four response types of indices to FVC, including U and fluctuation observed. • 3 asynchronous reactions among indices: simultaneous increases, decreases in Simpson’s only, mixture. • Both localized and overall synergies between native and non-native detected. • Anthropogenic disturbance can be only compared in similar FVC interval. • Results refine urban biodiversity theory and improve diversity prediction models. Rapid urbanization along the Grand Canal, a UNESCO World Heritage Site, highlights the need for systematic large-scale biodiversity assessments. This study focuses on the highly urbanized Wuxi monitoring zone (168.69 km2), integrating Sentinel-2 remote sensing with plot-based surveys in 81 sites to examine how fractional vegetation cover (FVC) reflects plant diversity across 12 land-use types, with distinctions drawn between native and non-native species. A framework includes Species richness, Simpson, Shannon–Wiener diversity, and Pielou’s evenness are established and analyzed with totally 137 species. Overall, all four indices increase with FVC (Spearman ρ ≈ 1, p < 0.05). Yet within individual land-use types, heterogeneous relationships (saturation, inverted U-shaped, U-shaped and fluctuating) are identified, as well as 3 asynchronous reactions among indices (simultaneous increases, decreases in Simpson’s only, mixture). Both localized and overall synergies between native and non-native communities are detected. Anthropogenic disturbance displays negative correlation only with certain native-specific indices (ρ = –0.62 to –0.92, p < 0.05) within comparable FVC intervals. Regression analyses (R2 = 0.24–0.62) identify FVC as the most robust predictor, whereas disturbance and interaction terms are significant for only a subset of indices. These results demonstrate that FVC–diversity relationships are highly nonlinear and context-sensitive. Beyond natural processes, site conditions, historical trajectories, socioeconomic factors, management regimes, and aesthetic preferences further modify their relationship, leading to deviations from linear FVC or urban–rural gradient assumptions. The findings provide refined evidence for biodiversity prediction using remote sensing in highly urbanized landscapes and offer important theoretical and practical insights for biodiversity research under global urbanization.