A statistical algorithm for calibrating the LFM model

Steve Sain, NCAR
Derek Bingham, Simon Frasier and Shane Reese, BYU

A common statistical approach to calibrating numerical models involves building a statistical representation or an emulator of the model output based on model parameters and inputs. The complex, high-dimensional nature of the LFM model outputs makes this task computationally difficult. We propose an alternative method that dramatically simplifies the computation while still yielding point estimates of calibration parameters as well as some measure of the uncertainty in these estimates. Issues concerning this method and the calibration of the LFM model will be discussed and an example will be shown based on some legacy runs of the LFM model.

Back to agenda