Understanding the interactions between the multiple ecosystem services (ES) which can be delivered from a single landscape is essential. Most studies on ES relationships use spatial or temporal statistical analysis (for example: correlations between services). Methods from microeconomic theory have recently received attention for describing ES relationships. The nature and intensity of ES relationships can be assessed by fitting a production possibility frontier that indicates the maximum amount of one ES that can be produced by landscape, for different levels of another ES. This study estimates production frontiers empirically, and compares the ES relationships insights gained this way with those inferred from correlation approaches. InVEST software was used to model and map the provision of six ES in the Reventazón watershed in Costa Rica. Spatial and temporal ES correlation patterns were analyzed for four observed land uses/land covers (LULC). Production frontiers were constructed using a set of 32 simulated scenarios. Production frontier was the most sensitive method for detecting ES relationships. The nature and intensity of ES relationships revealed depended on the analytic methods used. In comparison with correlations, the production frontier approach provided additional information relating to tradeoff intensity and Pareto efficient LULC configurations.