Urban landscape patterns and residents’ perceptions of safety under extreme city flood disasters

• Urban landscape pattern (ULP) holds a substantial sway over residents’ perceptions of safety (RPS) during floods, with a notable importance of 44.8%. • The relationship between ULP indices and RPS is intricate, showcasing varied nonlinear patterns like “S-shaped,” “inverted U-shaped,” and “inverted V-shaped,” with distinct threshold effects. • The interaction effect reveals that the interplay of ULP indicators and their interactions with flood-relevant contextual variables significantly enhances RPS. • Emphasized the necessity for flood response strategies tailored to ULP variations, enhancing safety perceptions and community resilience in flood-prone regions. Urban areas are increasingly susceptible to extreme flooding events, posing significant challenges to residents’ safety and well-being. In this context, understanding the relationship between urban landscape patterns (ULP) and residents’ perceptions of safety (RPS) is crucial for effective flood disaster management. This study explores the relationship between ULP and RPS during extreme flooding events, using Zhengzhou city as a case study. The goal is to offer strategic insights for managing extreme flood disasters. Surveying 1329 residents impacted by urban flooding in Zhengzhou, this study employed the Gradient Boosting Decision Tree (GBDT) model to examine the nonlinear relationship between ULP and RPS amidst extreme flood disasters. Results indicate: (1) ULP significantly impacts RPS, with a relative importance of 44.8%, particularly influenced by the Largest Patch Index (LPI), Landscape Connectivity Index (CONTAG), and Patch Density (PD); (2) The effect of landscape pattern indices on RPS reveals a complex nonlinear relationship, exhibiting various patterns such as ’stage-wise negative correlation,’ ’U-shaped,’ and ’inverted V-shaped’ with noticeable threshold effects. The interaction effect also shows that the internal interplay of the ULP indicator, along with its interactions with flood-relevant contextual variables, significantly enhances RPS. (3) As covariates, flood impact and emergency resource accessibility also influence RPS with nonlinear trends. This study offers a new perspective on the relationship between ULP and RPS, emphasizing the critical role of flood response strategies tailored to ULP differences in enhancing RPS.