Corridor–barrier effects of urban road networks on native spontaneous plant richness: Nonlinear thresholds and regulation by green space patterns

Urban road networks can function as either dispersal corridors or ecological barriers, but existing studies have not fully elucidated the complex effects of road networks on plant diversity. On one hand, many studies assume linear relationships, neglecting nonlinear responses and ecological thresholds; on the other hand, research typically focuses on road morphology while insufficiently considering the ecological significance of road network topology. Moreover, road networks and green space patterns are often analyzed independently, without fully revealing the regulatory role of green spaces in the ecological effects of roads. Currently, there is a lack of a framework capable of quantitatively integrating the nonlinear effects of road networks, the interactions of green space patterns, and their ecological significance. This study, using Shenzhen, China, as a high-density urban case, constructed an XGBoost model and integrated SHAP, interaction analysis, segmented regression, and individual conditional expectation curves to investigate the nonlinear effects of road network characteristics on native spontaneous plant richness and explore how green space coverage and cohesion regulate these effects. On this basis, an indicator, total effect of road network (TRSHAP) was developed to represent the net contributions of road metrics, thereby diagnosing corridor-like and barrier-like tendencies of road effects. In the Shenzhen case, when arterial road density exceeds 2.83 km/km 2 ; intersection density exceeds 63.18 intersections per km 2 , and mean road width exceeds 5.35 m, their effects on spontaneous plant species richness turn from positive to negative, while road mean connectivity must exceed 2.94 to maintain its positive effect. Segmented regression analysis identified marginal-response breakpoints reflecting shifts in response intensity, including 5.64 km/km 2 for arterial road density, 5.01 m for mean road width, and 2.48 for road mean connectivity. Together, these values indicate that the nonlinear effect of road networks on native spontaneous plant richness involves both directional transitions and changes in marginal response intensity. Furthermore, green space patterns regulated the direction and magnitude of road effects. Green space coverage primarily determined whether roads showed corridor-like or barrier-like tendencies, while green space cohesion regulated the strength of these tendencies in a coverage-dependent manner. This study developed a quantitative framework centered on TRSHAP for diagnosing the corridor-like and barrier-like tendencies of road networks and provided actionable evidence for biodiversity-oriented coordinated planning of roads and green spaces, with applicability to new road planning, urban renewal, and ecological corridor design. • Road metrics showed nonlinear threshold effects on native species richness. • TRSHAP diagnosed net corridor-like or barrier-like road contributions. • Roads shifted between corridor or barrier roles across green space contexts. • Green space coverage determined total road corridor–barriers tendencies. • Green space cohesion regulated road effects in a coverage-dependent manner.