Baseflow is a component of streamflow derived from shallow and deep subsurface flows that is concurrently controlled by multiple factors. Rational estimation of baseflow is critical for understanding its spatiotemporal dynamics and influencing factors within a river basin. To address this, different filtering parameters were applied to separate the baseflow of the Heihe River Basin (HRB) in Northwest China using digital filtering methods. Moreover, using bivariate and multivariate wavelet coherences, multivariate relationships between baseflow and meteorological factors/large-scale circulation indices were identified for several factors, which explained most of the variations. Results showed annual average baseflow was 10.3–91.1 mm and that the baseflow index (BFI) varied between 0.50 and 0.72 (average: 0.62). This indicates that 62% of long-term streamflow likely originates from groundwater discharge and other delayed sources. Positive/negative Spearman correlation coefficients between baseflow and extreme climate indices were more significant at upstream (Yingluoxia, Liyuanbao-and Wafangcheng) stations in comparison with midstream (Suyukou, Shunhua) and downstream (Yangyangchi) stations. Correlation for the BFI was relatively weaker than for baseflow. Furthermore, bivariate wavelet coherences revealed that precipitation (six stations) and the Atlantic Multidecadal Oscillation (four stations) were the individual factors that best explained baseflow variations. Multiple wavelet coherence demonstrated that all meteorological factors/large-scale circulation indices had the highest percentage of the numbers of power significant at the 95% significance level that could best explain baseflow variations. However, the average power of wavelet coherence was not increased. Differences likely attributable to consideration of additional variables were diminished by collinearity effects among factors. Furthermore, baseflow at the midstream Zhengyxia and downstream Yangyangchi stations had significant positive and negative correlation with population and effective irrigation area, respectively. The findings indicate that development of regional hydrometeorological models should primarily consider the impact of climate change in the upstream HRB, whereas the effects of both climate change and human activities should be considered in the midstream and downstream HRB.