Genetic makeup of insect pest is informative for source-sink dynamics, spreading of resistant genes, and effective management. However, collecting samples from geographical populations without considering temporal resolution and calculating parameters related to historical gene flow may not capture contemporary genetic pattern and metapopulation dynamics of highly dispersive pests. Plutella xylostella (L.), the most widely distributed Lepidopteran pest that developed resistance to almost all current insecticides, migrates heterogeneously across space and time. To investigate its real-time genetic pattern and dynamics, we executed four samplings over two consecutive years across Southern China and Southeast Asia, and constructed population network based on contemporary gene flow. Across 48 populations, genetic structure analysis identified two differentiated insect swarms, of which the one with higher genetic variation was replaced by the other over time. We further inferred gene flow by estimation of kinship relationship and constructed migration network in each sampling time. Interestingly, we found mean migration distance at around 1000 km. Such distance might have contributed to the formation of step-stone migration and migration circuit over large geographical scale. Probing network clustering across sampling times, we found a dynamic metapopulation of P. xylostella with more active migration in spring than in winter, and identified some regions are consistent sources (e.g., Yunnan in China, Myanmar and Vietnam) while several others are persistent sinks (e.g., Guangdong and Fujian in China) in its overwintering regions. Rapid turnover of insect swarms and highly dynamic metapopulation highlight the importance of temporal sampling and network analysis in investigation of source-sink relationships and thus effective pest management.