Abstract: Scientific identification of the spatial sources of mainstream flood events in the middle and lower Yangtze River basin is of great significance to the design of a regional flood control strategy and its engineering application. This study uses an improved method of principal component analysis (PCA) to screen significant flood fluctuation samples, based on the 2003-2022 daily average discharge series from this basin’s mainstream and major tributaries. Major flood events are determined using warning water levels, through combining a three-zone Flood Combination Index (FCI), high-FCI years, and primary discharge combination patterns for different zones. The results indicate that the improved PCA extracts discharge fluctuating characteristics effectively, and significant samples with the first principal component (PC1) score above the threshold 1.36 are in good agreement with the real cases in the high range of flow discharge. Inter-annual variations occur in the distribution of consecutive significant-sample periods. In 2021, significant flood periods occurred quite late under the influence of autumn floods from the Han River, while in 2022, early-season inflow impacts were evident due to a "dry flood season". The year of 2020 was marked by typical basin-wide flood events, and regional floods dominated 2010, 2012, 2016, and 2017, consistent with historical flood records.

Key words: middle and lower Yangtze River, warning water level, mainstream-tributary combination, improved principal component analysis, flood combination index