A case study of GOES-15 imager bias characterization with a numerical weather prediction model