Microbiomes in their natural environments vary dynamically with changing environmental conditions. The detection of these dynamic changes in microbial populations is critical for understanding the impact of environmental changes on the microbial community. Here, we propose a novel method to detect time-series changes in the microbiome, based on multivariate statistical process control. By focusing on the interspecies structures, this approach enables the robust detection of time-series changes in a microbiome composed of a large number of microbial species. Applying this approach to empirical human gut microbiome data, we accurately traced time-series changes in microbiota composition induced by a dietary intervention trial. This method was also excellent for tracking the recovery process after the intervention. Our approach can be useful for monitoring dynamic changes in complex microbial communities.