We propose a network-based method for determining basins and barriers of complex free energy surfaces (e.g., the protein folding landscape) from the time series of a single intramolecular distance. First, a network of transitions is constructed by clustering the points of the time series according to the short-time distribution of the signal. The transition network, which reflects the short-time kinetics, is then used for the iterative determination of individual basins by the minimum-cut-based free energy profile, a barrier-preserving one-dimensional projection of the free energy surface. The method is tested using the time series of a single C-beta-C-beta distance extracted from equilibrium molecular dynamics (MD) simulations of a structured peptide (20 residue three-stranded antiparallel beta-sheet). Although the information of only one distance is employed to describe a system with 645 degrees of freedom, both the native state and the unfolding barrier of about 10 kJ/mol are determined with remarkable accuracy. Moreover, non-native conformers are identified by comparing long-time distributions of the same distance. To examine the applicability to single-molecule Forster resonance energy transfer (FRET) experiments, a time series of donor and acceptor photons is generated using the MD trajectory. The native state of the beta-sheet peptide is determined accurately from the emulated FRET signal. Applied to real single-molecule FRET measurements on a monomeric variant of lambda-repressor, the network-based method correctly identifies the folded and unfolded populations, which are clearly separated in the minimum-cut-based free energy profile.