关键词: 公众参与地理信息系统, 跑步频率, 街景图像, 绿视率, 回归模型, 深度学习

Abstract:  Community health issues are increasing due to the imbalance between the social and economic functions of cities, the dramatic reductions in urban land, and the advent of motorised transportation during rapid urbanisation. Physical activities such as running are an effective exercise to maintain public health and prevent chronic diseases. Studies on the correlation between street environment and physical activities often focus on geographical units close to home locations. However, linear activities like running and biking cover an extensive spatial range. Hence, the traditional method of generating buffer zones by centring the home does not accurately reflect the actual locations of running activities. Selection of green environmental indicators is limited by difficulties in street image data acquisition, and it is difficult to perform a large-scale and refined calculation of the Green View Index. Hence, previous studies hardly consider the influences of the Green View Index. which can indicate the impact of the three-dimensional spatial greening effect on physical activities. The running routes of Helsinki residents (N=545) were acquired by public participation in GIS, thus generating varying scales of linear buffer zones. Street environmental indicators, such as NDVI, Green View Index, green space density, hydrophilic space density, pedestrian density, road intersection density, traffic light density, public transport station density, population density, regional slope, and land use hybrid entropy were all quantified. A multi-term logistic regression model was built by using running frequency as the dependent variable to analyse street environmental characteristics and the influences of individual factors on running frequency. Results showed that streetscape greenery was the dominant factor influencing running frequency and had a positive impact on running behaviours. The positive correlation between street greening rate and running frequency further indirectly verified the positive effects of urban greenland on public health. Higher slopes in a region tended to increase running frequency. Furthermore, people with different motivations for running may have various environmental needs. For example, casual joggers enjoy environments full of vitality, such as social communication activities similar to pedestrians. Conversely, athletic runners prefer paths which have fewer interactions with other road users, higher terrain slopes, and more challenging conditions. Intersection density may decrease enthusiasm for running. Appropriate running environments help runners maintain a speed, improve sport performances, and decrease the probability of injuries. Given a higher intersection density, runners are more easily influenced by other traffic users, thus decreasing their running speed and destroying motion continuity. Gender, age, and education background also have significantly positive effects on running frequency. Running frequency of people with lower willingness to exercise is mainly influenced by individual factors, while running frequency of runners with moderate willingness to exercise can be improved significantly by improving green environment and traffic environmental quality. To create cities friendly to physical activities like running, urban environmental renewal should not be limited within greenland areas, green coverage, and other traditionally-indicated areas in the future. Instead, it should focus on improving street greening quality. For example, streets with low green coverage can be improved by strengthening vertical greening. Moreover, attention must be paid to permeability and accessibility of greenways and green corridors which connect these public greenlands, because isolated urban greenlands are less likely to encourage running activities. Ultimately, running behaviours can be increased effectively by establishing comprehensive running path networks with appropriate infrastructure and sufficient three-dimensional greening.

Key words: public participation GIS, running frequency, streetscape images, Green View Index, regression model, deep learning