Under the combined effects of climate change and human activities, hydrological changes of large freshwater lakes in complex river–lake systems have attracted worldwide attention, mainly because of the high value of lakes in socio-economic development and environmental conservation. Poyang Lake in the Yangtze River basin of China was used as a case study to investigate the hydrological characteristics and their associated driving forces of influencing lake water level variability. Observed and simulated data and the composite multi-scale sample entropy to measure water level and streamflow complexity were used. The back-propagation neural network, combined with scenario testing analysis, was applied to determine the impact of streamflow and topographic changes on lake water level variability. Study results indicated that the water level complexity of Poyang Lake was higher than the streamflow complexity of the tributaries in the Poyang Lake basin and the main stem of the Yangtze River and that the complexity of both water level and streamflow was enhanced as the time scale increased. Comparing the condition from 2004 to 2016 with that from 1990 to 2002, Poyang Lake experienced a remarkable recession and an enhancement of water level complexity, as a coupling result of the streamflow and topographic changes. Streamflow reductions resulted in a water level drop throughout most of the year, particularly during the flood period. The impact of the deepened bottom topography was marginal during the flood period, but it was significant during the dry period. The strengthened water level complexity could be partially attributed to the deepened bottom topography, which enhanced the disorderliness of drought occurrence, the degree of drought, and the uncertainty in the prediction of drought events. Results from this study shed light on the impact of topographic changes on lake water level complexity and enhanced understanding of the characteristics and mechanisms behind the hydrological changes of large freshwater lakes.