Richter’s local magnitude distance correction terms may have a significant bias with respect to distance if applied in a region with attenuation properties different than southern California. Using data from a network of broadband seismometers installed in northern Oklahoma, we empirically constrain the attenuation properties based on the amplitudes of observed seismograms to determine a formula for local magnitude for this region that is unbiased with respect to distance. We use both the classic definition of local magnitude as proposed by Richter and a second method that calibrates local magnitude so that it agrees with moment magnitude where a regional moment tensor can be computed. For both methods, the new formula results in magnitudes systematically lower than the existing local magnitude model used by the Oklahoma Geological Survey. We compare the resulting magnitudes and discuss the benefits and drawbacks of each method. Our results highlight the importance of determining accurate distance correction terms for magnitude computation with regional network data.