作为人类经济和社会活动的空间载体，城市的可持续发展需要全面掌握城市空间中人的活动。公园、绿地、广场等城市开放空间由于开阔的空间属性和多样化的服务功能，其中的人群活动具有复杂性，如果管理不善将会带来人群聚集和公共卫生等方面的风险。因此，有必要深入解析城市开放空间的人群活动特征及其与建成环境的相互作用。

近年来，位置感知技术，特别是Wi-Fi技术的普及使得获取精细化的人群活动数据得以实现。Wi-Fi技术通过采集手机等智能设备发送的Wi-Fi探针请求信号记录设备携带者活动的位置和时间信息。即使在未连接Wi-Fi网络或是未开启Wi-Fi功能的情况下，大部分手机也会持续发送Wi-Fi探针请求信号。因而，相比GPS和手机信令等定位技术，Wi-Fi技术具有较高的定位精度和较强的采样代表性，在刻画城市开放空间中的人群活动特征方面具有优势。

本研究以北京什刹海风景区为典型研究区，以到访人群为研究对象，通过在主要道路上布设28个Wi-Fi探针采集了2018年10月至2019年3月超过290万人的活动数据，包含2180余万条Wi-Fi探针请求信号。经预实验验证，所采集的Wi-Fi数据对到访人群的检测率可达70%以上。在此基础上，依据时间地理学的理论框架构建Wi-Fi数据与个体活动轨迹的时空映射关系，表征了个体在城市开放空间中的移动和停留。通过引入气象与气候学研究领域常用的经验正交函数（empirical orthogonal function，EOF）方法，并提出语义时空矩阵（semantic space-time matrix，SSTM）和基于人群停留特征的吸引力指数（stay characteristics-based attractiveness index，SCAI）方法，全面解析了城市开放空间人群到访的时空模式、到访人群类型、到访人群的停留特征以及基于人群停留特征的建成环境吸引力。本研究的核心内容和主要结论如下：

第一，将EOF方法与Wi-Fi数据大规模、长时间、高精度的特点相结合，可以从时间和空间同步变化的角度揭示城市开放空间人群活动的热点及其随时间的变化规律。从什刹海风景区的研究案例来看，EOF₁和EOF₂作为主要和次要的空间模态分别揭示了人群活动的一般特征和局部热点，PC₁和PC₂反映了其随时间的变化规律，主要受到节假日和天气状况的影响。

第二，通过对电子地图兴趣点（point of interest，POI）数据进行语义分析可以解析轨迹点的活动属性，弥补Wi-Fi数据缺乏属性信息的不足。在此基础上，SSTM方法基于轨迹点的频率分布识别了个体重复且有规律的活动特征，进一步通过聚类分析识别了具有相似活动特征的到访人群类型。研究区的到访人群包括本地居民、附近居民、上班族、常访游客、工作日游客和节假日游客等六个类型，其中工作日游客和节假日游客所构成的单日游客是主要类型。通过与实地调查结果对比，验证了SSTM方法用于识别到访人群类型的有效性。

第三，与研究区的其他人群类型相比，单日游客具有最高的到访人群流量，并且其活动受到周围建成环境的显著影响。因此，从进行人群活动常态化管理的实际需求出发，选取单日游客为研究对象，进一步解析其在不同位置的停留特征。采用停留时间和停留人数测度城市开放空间到访人群的停留活动，揭示出平均停留时间-停留人数关联特征，可以反映不同位置吸引到访人群停留的空间差异。基于单日游客的停留活动，研究区的活动站点可以分为高-高型、低-高型、高-低型和低-低型, 其中高-高型是指停留人数和平均停留时间都较高的活动站点，其它三个类型的含义以此类推。

第四，SCAI方法从城市开放空间到访人群与建成环境相互作用的视角，反映了建成环境吸引力的空间差异和时间变化。通过对SCAI的结果进行回归分析，可以识别对建成环境吸引力具有影响的关键因素。研究区的活动站点可以分为高吸引力、中吸引力和低吸引力三个等级，其空间差异主要受到餐厅、商店和酒吧的空间分布影响，时间变化主要受到周末和游客游览时间影响。

本研究验证了将Wi-Fi技术应用于精细尺度个体活动研究的可行性和有效性，集成了多种研究方法，拓展了城市空间人群活动的研究视角。研究结果能够为城市空间的规划、建设与管理的相关决策提供参考。

As the spatial carrier of human economic and social activities, the sustainable development of cities cannot achieve without a comprehensive understanding of crowd activities in urban spaces. Urban open spaces, such as parks, green areas, and squares, are characterized by complex crowd activities due to their spacious attributes and multiple functions, which can result in risks of crowd gathering and public health if not properly managed. Therefore, it is imperative to analyze in depth the characteristics of crowd activities and their interactions with the built environment in urban open spaces.

In recent years, the popularity of location-aware technologies, especially Wi-Fi technology, has enabled the acquisition of fine-grained crowd activity data. Specifically, Wi-Fi technology collects the space-time information of individual activities by capturing Wi-Fi probe requests sent by smart devices, such as cell phones. Most phones continue to send Wi-Fi probe request signals even when they are not connected to a Wi-Fi network or when the Wi-Fi function is turned off. Therefore, compared to GPS and cell phone signaling, Wi-Fi technology has higher positioning accuracy and better sampling representativeness, making it advantageous in characterizing crowd activities in urban open spaces.

In this study, we take the Shichahai scenic area in Beijing, China as a typical study area and crowd activities in it as the research object. Through deploying 28 Wi-Fi probes near arterial roads in the study area, more than 21.8 million Wi-Fi probe requests from over 2.9 million individuals were collected from October 2018 to March 2019. Verified by the pre-experiment, the detection rate of Wi-Fi probes of crowd activites can reach 70%. Based on the theoretical framework of time geography, we achieved a mapping from Wi-Fi data to individual activity trajectory and modeled crowd movement and stay in urban open space. Through introducing the empirical orthogonal function (EOF) method, which is commonly used in the field of meteorology and climatology, and proposing the semantic space-time matrix (SSTM) method and the stay characteristics-based attractiveness index (SCAI), we comprehensively explored the space-time patterns of crowd activities, the traveler types of crowds, and the attractiveness of the built environment in urban open space. The core contents and main findings of this study are as follows:

 (1) By applying the EOF method to the large-scale, long-time, and high-precision Wi-Fi data, we captured crowd activity hotspots and their temporalities in urban open space from the perspective of synchronous changes in time and space. The case study of the Shichahai scenic area showed that as the predominant and secondary spatial modes, EOF₁ and EOF₂ revealed the general and local patterns of crowd activities. Also, PC₁ and PC₂ captured the temporalities of spatial modes and were mostly influenced by the holiday and weather conditions.

 (2) Through semantic analysis of Point of Interest (POI) data, we identified the activity types of track points, which are lacking in Wi-Fi data. Based on this, we proposed an SSTM method to derive space-time patterns of individual repetitive and regular activities by analyzing the frequency distribution of track points. Different traveler types composed of individuals with similar space-time patterns were further identified. The identified traveler types in the Shichahai scenic area included residents, nearby residents, commuters, frequent visitors, weekday tourists, and holiday tourists. Among them, single-day tourists (consisting of weekday tourists and holiday tourists) had the highest flow volume. The identified traveler types have been further validated by a field survey, justifying the feasibility of the SSTM method.

(3) Compared with other traveler types, single-day tourists had the highest crowd flow volume. Besides, their activities were significantly influenced by the surrounding built environment. To fulfill the practical need for crowd management in the study area, we selected single-day tourists to further analyze their stay characteristics in different locations. We employed the stay time and stay flow volume to reveal crowd stay characteristics in urban open spaces. On this basis, we revealed the associated characteristics between average stay time and stay flow volume, which can be used to reflect the abilities different locations had to attract crowds to stay. Based on the stay characteristics of single-day tourists, we clustered the activity stations in the Shichahai scenic area into four groups, namely H-H, H-L, L-L, and L-H groups. The H-H group includes the activity stations with longer average stay time associated with higher stay flow volume, and the other groups are in the same way for interpretation.

 (4) The SCAI reflects the spatial differences and temporal changes in the attractiveness of the built environment from the perspective of the interaction between the crowd and the built environment of urban open space. Based on the results of SCAI, the activity stations in the Shichahai scenic area could be clustered into three levels: high-level attractiveness, moderate attractiveness, and low-level attractiveness. The most influencing factors on the spatial differences of SCAI were restaurants, shops, and bars. For temporal changes, the most influencing factors were weekend and visiting time.

The study validates the feasibility and effectiveness of applying Wi-Fi technology to fine-scale individual activity research. Also, it integrates multiple research methods and expands the perspective of crowd activity research in urban space. The research findings can provide references for decision-making in the planning, construction, and management of urban spaces.