Protected-area-based nature-based governance and green-blue infrastructure-to-built conversion: spatial regression discontinuity evidence from Kinmen National Park, Taiwan

Nature-based solutions (NbS) and green-blue infrastructure (GBI) are increasingly promoted for sustainable built environments, yet boundary-based evidence on specific land-conversion mechanisms remains limited. This study evaluates Kinmen National Park, Taiwan, using a spatial regression discontinuity design (Spatial RDD) on 30 m grid cells for 2014–2024. GBI-to-built conversion (G2B) is the primary outcome; strict built-up change, developed-land change, Normalized Difference Vegetation Index (NDVI) change, and GBI fragmentation are contrasting or secondary outcomes. The main G2B coefficient is −0.00188, equivalent to a 0.19-percentage-point lower conversion share immediately inside the boundary and about 19% of the raw outside-sample mean as a descriptive benchmark. Heteroskedasticity-robust and robust bias-corrected RD inference support a negative discontinuity, but the 60 m donut specification and Conley spatial-HAC corrections weaken statistical precision. No robust discontinuity is found for strict built-up expansion. External distance bands show positive dBuilt and G2B signals in the outer 600–1,000 m belt, interpreted as spatially associated peripheral development pressure rather than park-caused displacement. The findings therefore support a boundary-local GBI loss-reduction pattern while limiting stronger causal claims.