Dynamic Bayesian Nonlinear Calibration

TL;DR

This paper introduces a dynamic Bayesian approach for nonlinear calibration that accounts for time-varying instrument characteristics, improving calibration accuracy in fields like radiometry and spectroscopy.

Contribution

It presents a novel dynamic Bayesian calibration method that models drifting regression parameters in nonlinear relationships, addressing a key challenge in ongoing instrument calibration.

Findings

Effective in modeling time-dependent calibration parameters

Applied successfully to microwave radiometer data

Validated with simulated spectroscopy data

Abstract

Statistical calibration where the curve is nonlinear is important in many areas, such as analytical chemistry and radiometry. Especially in radiometry, instrument characteristics change over time, thus calibration is a process that must be conducted as long as the instrument is in use. We propose a dynamic Bayesian method to perform calibration in the presence of a curvilinear relationship between the reference measurements and the response variable. The dynamic calibration approach adequately derives time dependent calibration distributions in the presence of drifting regression parameters. The method is applied to microwave radiometer data and simulated spectroscopy data based on work by Lundberg and de Mar\'{e} (1980).