We investigate the distributions of length and velocity scales in a Mach 0.9 isothermal round jet for the purpose of developing a linear surface-based model for the noise source that can be informed by the time-averaged flow field. Large eddy simulation of the flow enables the computation of two-point space-time correlations throughout the jet and its near acoustic field. The resulting time, length, and convective-velocity scales are examined on the surface of peak Reynolds stress (SPS), representing the location of the most energetic eddies, and on a "radiator surface" at the boundary between the rotational and irrotational fields. The radiator surface is defined such that the convective velocity distribution on it matches that on the surface of peak Reynolds stress. The location of the radiator surface, and distributions of flow scales on it, are critical elements of the noise source model and are examined in detail. The nature of the space-time correlations, and resulting scales, differ significantly for axial velocity fluctuations and pressure fluctuations. Velocity-based correlations appear to capture localized turbulent events, while pressure-based correlations appear dominated by the interaction of large eddies with the surrounding potential flow. Consequently the axial and radial distributions of the corresponding length scales exhibit different trends. Correlation scales on the radiator surface are larger than on the SPS thus indicating that fine-scale vortical motions do not make a significant imprint on the radiator surface. Scales associated with the Reynolds-averaged flow field, relevant to low-cost modeling, are emulated from LES data and compared to the LES-based scales. Simple relationships are established that may aid the development of rapid predictive models.