There is broad concern that humans are transforming our environment. This transformation has potential to impact humanity as we depend on the environment ecosystem services. According to the Millennium Assessment (2005), degradation and unsustainable exploitation presently threaten over 60% of ecosystem services with real implications for health and standards of living. Furthermore, both the exploitation of ecosystem services and the growth rate of that exploitation have been far higher in recent decades than ever before due to population growth and rising standards of living, i.e. consumption. Increasing pressure on ecosystem services has driven thinking on mitigation strategies. Payment for ecosystem services (PES) has emerged as a strategy to encourage provision of services or, often, to discourage activities that reduce provision. In economic terms, the inability of agents to capture the full rents of service provision results in divergent private and social values. By creating markets for these services, PES arrangements can correct this disincentive and bring provision closer to the socially optimal level. While private sector PES schemes have been envisioned and in some cases implemented, most large-scale PES programs to date have been implemented by governments. We consider optimal design of policy aimed at increasing the provision of services from private land. In particular, we examine on a theoretical level the possibility of optimal decision-making hierarchies among government agencies targeting ecosystem service provision. Should we have multiple agencies focusing on separate services or one agency coordinating efforts across services? Should policy be implemented nationally, regionally, or locally? Under what conditions and assumptions does one organizational structure stand out as optimal? To answer these questions, we adapt the model of hierarchy design developed by Hart and Moore (JPE, 2005). We develop a two period model of decision-making with n agencies and m assets. The assets are parcels available for targeting under PES schemes. Each agency is tasked with thinking about how to use a subset of the m assets to enhance service provision, according to its mission. Also associated with each agency is some probability of success in its task - i.e., the probability of thinking of a productive use for the assigned subset of assets - and a value generated for society if the task is completed. There is some rivality among assets; the use of an asset by one agency may preclude its use by another. Determination of seniority and assignment of tasks occur in period 0. In period 1, agencies with access to all assets they require carry out the tasks and generate value for society. Assets are unavailable if put to a conflicting use by a senior agency. Altering the hierarchy structure alters the set of completed tasks and thus the total value. We optimize across hierarchies by assigning tasks and seniority in period 0 to maximize total expected value in period 1. Following the development of the model, we explore implications and results. To demonstrate the driving intuition, we provide results in the two-agency, three-agency, and general case. Our results shed light on the optimal design of hierarchies, including the optimal relationship between coordinators (those considering how to use many assets simultaneously) and specialists (those considering a narrower subset of assets). The model relies on a number of assumptions - some of which are more restrictive than others - and we examine the implications for our results of relaxing assumptions. Two preliminary results stand out as generally applicable. First, in an optimal hierarchy an agency's seniority should be inversely related to its probability of success. So agencies with a low probability of having an idea about how to use the assets assigned to them should have high seniority. This seems counter-intuitive as the value of an idea is not considered, but it becomes clearer considering that tasks and seniority are assigned in period 0 to maximize period 1 total expected value. With this endogenized task selection, no agency would be assigned a task with low value and a low probability of success. Second, crisscross hierarchies are never optimal. This result, which states that agency a should never be senior to agency b on one asset and junior on another, is more intuitive. The central contribution of this paper is the adaptation of a theoretical model of hierarchy design to the context of programs targeting ecosystem services. The nature of interaction between various government entities involved in encouraging service provision necessitated an alternative representation of rivality between agencies. Assumptions were evaluated and revised based on their applicability to behavior in this context. Future research may involve further modification of the model to account for issues like threshold effects, joint production, and uncertainty.