关键词: 共享单车骑行目的地, 随机森林, 建成环境, 非线性影响, 厦门岛

Abstract:  As an important short-distance traveling mode, bicycle sharing plays a positive role in establishing a green and low-carbon travel system. However, it also brings many problems, such as illegal parking, blind increase of shared bicycles, and spatiotemporal supply-demand mismatching. Studying nonlinear influences of built environment on travel with shared bicycles is conducive to solving these problems and building a high-efficiency, green urban traffic system. This study examines the case of Principal Island in Xiamen. The study area was divided into 300m×300m grids. Influences of land use factors, public transit facilities, road design, and natural environmental factors on arrival density of shared bicycles in different periods were investigated based on the random forest algorithm. Results show that traveling with shared bicycles in Xiamen has obvious rush hours in the morning and evening. Land use can significantly influence the direction of shared bicycle flows during rush hours. During morning rush hours, working, studying, and financial facilities are major destinations. During evening rush hours, shopping, restaurant, and accomodation facilities are major destinations. The influences of land use intensity on travel intensity with shared bicycles under different scales have a threshold. For example, the influences of public transport stations on arrival density of shared bicycles decays with distance. The accessibility of the nearest subway stations is more important than accessibility of the nearest bus stations. Influences of subway stations on arrival density of shared bicycles disappear after reaching a distance of 140m to the travel unit. The same distance threshold for bus stations ranges from 75m to 100m. There's a U-shaped influencing relationship between road design factors and arrival density of shared bicycles. Based on these results, some suggestions are proposed: (1) Important urban areas should be recognized, including rush hours and directions of shared bicycles during dispatch and infrastructure planning of bicycles, to improve spatiotemporal imbalance. Planners should layout service facilities at all levels to form a service facility system with diversified functions and compact development, optimize combintions of the traffic network, optimize biking infrastructure and service facilities at all levels, and promote low-carbon travel. (2) According to local public transport development conditions, attention should be paid to the connection distance range threshold between different types of public transit stations and shared bicycles. This will increase density of traffic stations and designated parking places in the reasonable threshold range, and improve connections between public traffic and shared bicycles in order to solve the "last mile" problem more effectively. (3) It is recommended to build a high-efficiency biking system. For different types of blocks, comprehensive consideration shall be given to street scale, biking safety, and traffic efficiency factors.

Key words: destinations of shared bicycles, Random Forest, built environment, nonlinear effects, Xiamen Island