Understanding population movement patterns is crucial for various applications, including economic analysis, traffic demand prediction, and epidemic control. China's vast size and diverse transportation systems make analyzing population flow patterns particularly important. Traditional data sources for such analysis often lack national scale, uniformity, or sufficient sample size. This study leverages Baidu migration data, a large-scale, relatively uniform dataset derived from a popular Chinese map service. This data provides a novel opportunity to analyze the national population flow network. The 2020 Spring Festival (Chunyun) period, coinciding with the early COVID-19 outbreak in Wuhan, offers a unique context to examine the impact of population movement on epidemic spread. The study aims to analyze the hierarchical structure of population flow, identify subnetworks and their spatial distribution, and explore the role of population flow in the COVID-19 spread using Baidu migration data from January 1st to 22nd, 2020, dividing the period into 'non-Chunyun' (January 1st-9th) and 'Chunyun' (January 10th-22nd) phases. This will be achieved through the use of network analysis techniques, including weighted degree centrality (WDC), betweenness centrality (BC), and community detection methods.

Previous research has explored population movement patterns using diverse data sources, including mobile phone data, train and flight frequency data, and geo-tagged social media data. While mobile phone data offers high accuracy and large sample sizes, concerns regarding cost and privacy limit its widespread use. Other sources, such as train and flight data, provide insights into inter-city movement, while geo-tagged social media data can offer fine-grained temporal and spatial resolution. Research on Chinese population flow networks has faced challenges regarding data availability and uniformity. This study addresses these limitations by using Baidu migration data, offering advantages in scale, uniformity, and reliability. Existing studies have utilized complex network theory to analyze population flow, treating cities as nodes and population flow as edge weights. This approach allows for the identification of key nodes and subnetworks, crucial for understanding epidemic transmission.

The study utilized Baidu Map migration data for 367 major Chinese cities from January 1st to 22nd, 2020. Data pre-processing involved cleaning and aggregating migration scale indices (OD records) between cities. Each city was represented as a node in a complex network, with the OD record between cities representing edge weight. The data was divided into non-Chunyun (January 1-9) and Chunyun (January 10-22) periods. Confirmed COVID-19 cases from the Tianditu map website (as of February 6th, 2020) were used to assess the correlation between population flow and epidemic spread. Network analysis techniques were employed: * **Weighted Degree Centrality (WDC):** Calculated using the formula Sᵢ = ΣWᵢⱼ, where Wᵢⱼ is the weight of the edge between nodes i and j. This measure identifies cities with high overall population flow. * **Betweenness Centrality (BC):** Defined as the number of shortest paths between any two nodes that pass through a given node (formula provided in the paper). This measure identifies cities playing a crucial role in connecting different parts of the network. * **Community Detection:** This method partitioned the network into subnetworks (communities) with high internal connectivity and low external connectivity, providing insights into regional population flow patterns. The Louvain algorithm was likely used to optimize the modularity of the partition. The correlation between migration volume from Wuhan and the number of confirmed COVID-19 cases in other cities was assessed using correlation analysis. Chord diagrams were used to visualize migration flow from Wuhan to other cities, particularly within Hubei province. Spatial distribution maps visualized migration patterns and network centrality.

The study revealed several key findings: 1. **Uneven Spatial Distribution:** Population flow during the non-Chunyun period was concentrated east of the Hu Huanyong Line, with several regional hotspots centered around provincial capitals. Provincial capitals played a central role in inter-provincial population flow. 2. **Hierarchical Network Structure:** The analysis of WDC and BC revealed a hierarchical structure in the population flow network, with some core cities emerging as important hubs. Chengdu, Zhengzhou, and Xi'an were identified as significant destination cities despite not being located in China's three most developed urban agglomerations, suggesting industrial shifts toward central China. Wuhan ranked highly in both WDC and BC, indicating its importance in the network even before the outbreak. 3. **Subnetwork Structure Aligned with Provincial Boundaries:** Community detection revealed a strong alignment between population flow subnetworks and provincial administrative divisions. Major inter-provincial population movement occurred between major metropolitan areas (Beijing, Shanghai, Guangzhou, Chengdu-Chongqing) and their surrounding areas, with northern cities (Changchun, Harbin, Shenyang, Hohhot) also playing significant roles. The findings suggest that transport planning should consider both aviation and high-speed rail networks. 4. **Strong Correlation between Wuhan Migration and COVID-19 Cases:** A significant positive correlation was found between migration flow from Wuhan and the number of confirmed COVID-19 cases in other cities (r = 0.943, p < 0.01). The majority (68.93%) of Wuhan's outbound migration during the Chunyun period was within Hubei Province itself, with Xiaogan and Huanggang being particularly affected. Spatial proximity to Wuhan was a crucial factor in the early spread of the virus. 5. **Exceptions and Nuances:** Wenzhou, despite having relatively low migration flow from Wuhan, experienced a high number of COVID-19 cases, highlighting the influence of business travel and social interaction patterns on disease transmission.

The findings demonstrate the significant role of population flow networks in the spread of infectious diseases. The close alignment of subnetworks with provincial boundaries underscores the effectiveness of provincial and municipal lockdowns in containing the COVID-19 outbreak. The hierarchical network structure identifies key cities and regions that require increased surveillance and intervention measures during outbreaks. The strong correlation between Wuhan's outbound migration and COVID-19 case numbers validates the use of population flow data in epidemic forecasting and response planning. The case of Wenzhou highlights the need to consider factors beyond simple population flow, such as business travel patterns and social interactions, in predicting disease transmission. The study’s insights are relevant for evidence-based policymaking related to population management, transportation planning, and public health emergency response.

This study provides valuable insights into the spatial patterns of population flow in China and its crucial role in epidemic control. The integration of Baidu migration data and network analysis techniques offers a powerful approach for understanding disease spread dynamics. The findings highlight the importance of considering population mobility in designing effective strategies for epidemic preparedness and response. Future research should investigate the longitudinal patterns of population flow, the relationship between population flow and various transportation modes, and the development of more sophisticated models to predict disease spread incorporating social interaction data and transportation patterns. Validating Baidu migration data against other sources and incorporating other demographic factors could further improve the reliability and generalizability of the findings.

The study relies on Baidu migration data, which may not capture the entire population due to smartphone ownership disparities. This could lead to underestimation of population movement, particularly for children and seniors. The analysis focuses on a specific time period, and longitudinal data are needed to understand the dynamics of population flow over longer periods. Further research is needed to incorporate other factors that could affect the spread of COVID-19, such as demographic characteristics, social interactions, and individual behaviors. The reliance on only one data source (Baidu) limits the generalizability and robustness of the findings, and triangulation with other data sources would strengthen the conclusion.