Large dams are a leading cause of river ecosystem degradation. Although dams have cumulative effects as water flows downstream in a river network, most flow alteration research has focused on local impacts of single dams. Here we examined the highly regulated Colorado River Basin (CRB) to understand how flow alteration propagates in river networks, as influenced by the location and characteristics of dams as well as the structure of the river network—including the presence of tributaries. We used a spatial Markov network model informed by 117 upstream-downstream pairs of monthly flow series (2003–2017) to estimate flow alteration from 84 intermediate-to-large dams representing >83% of the total storage in the CRB. Using Least Absolute Shrinkage and Selection Operator regression, we then investigated how flow alteration was influenced by local dam properties (e.g., purpose, storage capacity) and network-level attributes (e.g., position, upstream cumulative storage). Flow alteration was highly variable across the network, but tended to accumulate downstream and remained high in the main stem. Dam impacts were explained by network-level attributes (63%) more than by local dam properties (37%), underscoring the need to consider network context when assessing dam impacts. High-impact dams were often located in sub-watersheds with high levels of native fish biodiversity, fish imperilment, or species requiring seasonal flows that are no longer present. These three biodiversity dimensions, as well as the amount of dam-free downstream habitat, indicate potential to restore river ecosystems via controlled flow releases. Our methods are transferrable and could guide screening for dam reoperation in other highly regulated basins.