Abstract:
|
Spectroradiometers found on earth observing satellites can be used to measure averaged trace gas mass fractions, such as carbon dioxide, in columns of atmosphere. However, trace gas measurements typically require near linear instrument response, i.e., as light flux increases, the instrument response should increase proportionally. A novel method for estimating the non-linear behavior of an instrument and quantifying uncertainty in that estimate is presented. A probability model for instrument responses observed under many distinct levels of light flux forms the basis of the approach. The likelihood function is maximized to produce an estimate, and the technique for regression models known as bootstrapping pairs is leveraged for uncertainty. A simulation study shows that confidence intervals maintain nominal coverage for underlying true parameter values that are physically plausible and consistent with existing methodology. A real data set from the NIST Beam Conjoiner is also examined, and it is found that the uncertainty contribution from estimation of the non-linear detector response is less than 0.02% for most values of light flux.
|