The water quality of large rivers is influenced by complex interactions between natural factors and human activities. Understanding the interactive effects of landscape patterns of different land use types on the migration and transport pathways of pollutants under extreme rainfall conditions presents a significant challenge in interdisciplinary research. This study focuses on the Yellow River basin (YRB), a typical large-scale arid and semi-arid region globally, and develops a theoretical framework for landscape pattern and water quality responses to extreme rainfall based on the "source-sink" landscape theory. By coupling random forest and partial least squares structural equation modeling, we quantitatively reveal the transmission pathways and mechanisms among extreme rainfall, landscape patterns, and water quality. The results indicate that: (1) Over the past 70 years, the rainfall pattern in the YRB has undergone significant changes, with a continuous increase in the threshold for extreme rainfall. Spatially, high-value areas of extreme rainfall have formed in the middle reaches of the YRB, exhibiting spatial dependence with water quality; (2) The expansion of the largest agricultural land patches and built-up areas are the primary sources contributing to the deterioration of water quality in the YRB, with extreme rainfall significantly increasing the output of organic pollutants from these source landscapes; (3) Grassland aggregation (AI.G) and percentage of forest land (PLAND.F) serve as key sinks for water quality degradation, with extreme rainfall driving the dominant sink for water quality degradation in the YRB from forest to grassland. During extreme rainfall events, AI.G > 91 % significantly enhances its capacity to filter and adsorb runoff pollutants, while during non-extreme rainfall events, increases in PLAND.F more effectively mitigate the scouring effects of rainfall. Therefore, in controlling water pollution in large-scale arid and semi-arid basins, it is crucial not only to manage anthropogenic organic pollutant emissions but also to recognize the importance of increasing AI.G in addressing watershed non-point source pollution caused by extreme rainfall.