To comprehensively understand the response of climate change and human activities impact of the river hydrological health, the precipitation-runoff double accumulation curve method and Pettitt method were used to analyze the abrupt change of runoff of Yong’an River, and the long short-term memory (LSTM) neural network model was developed to reconstruct the natural flow series based on meteorological and runoff datasets. Various optimal ecological flow calculation methods were compared and filtered to reveal the change trend in the river hydrological health degree ( R HD), and contribution of each driving factor was quantitatively evaluated. The results show the mutation in the runoff series of Yong’an River occurred in 2003, which is consistent with the operating time of the Xia’an reservoir. Compared with the five probability density function methods, the Q50 flow is recognized as the optimal ecological flow calculation metho for the Yong’an River Basin. Compared to the contribution of human activity (approximately by 34.74%), climate change is the main factor (approximately by 65.26%) contributing to the decline in the hydrological health of Yong’an River. The R HD of the measured runoff after the operation of the Xia’an reservoir showed a significant increase compared with the R HD of the restored natural runoff, especially in March, April, and October.