Several structural theories for effective network governance of environmental systems have been proposed, with the system conceived as a network of collaborating organizations. Following the ground-breaking work of Ostrom (2007) and the precise network formulation of Bodin and Tengo (2012), I argue, however, that the system may be best represented as a multilevel network involving both human and environmental nodes: a social ecological system. This new conceptualization fits well with recent parallel methodological development, including Exponential random graph models for multilevel network structure. I show how these new methods may be applied to the empirical example of Bodin and Tengo on the management of forests by clans in Madagascar. I discuss the structural preconditions for sustainability in this social ecological system.