The concept of the adaptive landscape has been invaluable to evolutionary biologists for visualizing the dynamics of selection and adaptation; and is increasingly being used to study morpho-functional data. Here we construct adaptive landscapes to explore functional trade-offs of humerus morphology in turtles adapted to three different locomotor environments: marine, semiaquatic, and terrestrial. Humerus shape from 40 species of cryptodire turtles was quantified using a pseudolandmark approach. Hypothetical shapes were extracted in a grid across morphospace and four functional traits measured on those shapes that represent strength, stride length, mechanical advantage, and hydrodynamics. Quantitative trait modeling was used to construct adaptive landscapes that optimize the functional traits for each of the three locomotor ecologies. Our data show that turtles living in different environments have statistically different humeral shapes. The optimum adaptive landscape for each ecology is defined by a different combination of performance trade-offs, with turtle species clustering around their respective adaptive peak. Further, species adhere to pareto fronts between marine-semiaquatic and semiaquatic-terrestrial optima, but not between marine-terrestrial. Our study demonstrates the utility of adaptive landscapes in informing the link between form, function, and ecological adaptation; and establishes a framework for reconstructing turtle ecological evolution using isolated humeri from the fossil record.