The Upper Geyser Basin at Yellowstone National Park (Wyoming, USA) harbors the greatest concentration of geysers worldwide. Research suggests that individual geysers are not isolated but rather are hydraulically connected in the subsurface with other geysers and thermal springs. To quantify such connections, we combined techniques from machine learning, causal inference, and dynamical systems to characterize the collective eruptive behavior of a set of 10 geysers over 18 months (April 2007 – September 2008) focusing on geyser-geyser interactions. Model predictions were up to 15 times more accurate when we sought to predict a geyser's eruption time series based on outflow channel temperatures from the network than based on its own time series alone, suggesting the existence of a complex interconnected subsurface groundwater system. On average, cone-type geysers had larger impacts on other geysers than did fountain-type geysers. Similarly, cone-type geysers were on average more insulated from other geysers. However, substantial unexplained variation remained after considering the cone versus fountain dichotomy. Distance between geysers also affected interactions: nearby geysers had stronger effects on focal geysers than did geysers located farther away. Collectively, results support the hypothesis of geyser interdependence at timescales of 5 min–10 days. Our analyses highlight the existence of quantifiable geyser-to-geyser interactions that can be resolved through pairwise and system-level analyses. These findings emphasize the subsurface interconnectedness of thermal features, provide information relevant to visitor experiences in Yellowstone National Park, and suggest strategies for exploring patterns of interdependence that may exist among other episodic geological phenomena.