^{1}

^{1}

^{1}

We propose an innovative approach for analysing ground-based FTIR spectra which allows us to detect variabilities of lower and middle/upper tropospheric HDO/H<sub>2</sub>O ratios. We show that the proposed method is superior to common approaches. We estimate that lower tropospheric HDO/H<sub>2</sub>O ratios can be detected with a noise to signal ratio of 15% and middle/upper tropospheric ratios with a noise to signal ratio of 50%. The method requires the inversion to be performed on a logarithmic scale and to introduce an inter-species constraint. While common methods calculate the isotope ratio posterior to an independent, optimal estimation of the HDO and H<sub>2</sub>O profile, the proposed approach is an optimal estimator for the ratio itself. We apply the innovative approach to spectra measured continuously during 15 months and present, for the first time, an annual cycle of tropospheric HDO/H<sub>2</sub>O ratio profiles as detected by ground-based measurements. Outliers in the detected middle/upper tropospheric ratios are interpreted by backward trajectories.