Using neural networks to describe tracer correlations

[摘要] Neural networks are ideally suited to describe the spatial andtemporal dependence of tracer-tracer correlations. The neural network performs well even in regions where the correlations areless compact and normally a family of correlation curves would be required. For example, theCH₄-N₂O correlation can be well described using a neural network trained with the latitude,pressure, time of year, and \methane\ volume mixing ratio (v.m.r.). In this study a neural network using Quickprop learningand one hidden layer with eight nodes was able to reproduce the CH₄-N₂Ocorrelation with a correlation coefficient between simulated and training values of 0.9995. Such an accuraterepresentation of tracer-tracer correlations allows more use to be made of long-term datasets to constrain chemical models. Such asthe dataset from the Halogen Occultation Experiment (HALOE) which has continuously observedCH₄ (but not N₂O) from 1991 till the present. The neural networkFortran code used is available for download.

[学科分类] 大气科学