Assessing the potential of free tropospheric water vapour isotopologue satellite observations for improving the analyses of latent heating events

Abstract. Satellite-based observations of free tropospheric water vapour isotopologue ratios (δD) with good global and temporal coverage have become recently available. We investigate the potential of these observations for constraining the uncertainties of the atmospheric analyses fields of specific humidity (q), temperature (T), and δD and of variables that capture important properties of the atmospheric water cycle, namely the vertical velocity (ω), the latent heating rate (Q₂), and the precipitation rate (Prcp). Our focus is on the impact of the δD observations if used in addition to the observation of q and T , which are much easier to be observed by satellites and routinely in use for atmospheric analyses. For our investigations we use an Observing System Simulation Experiment, i.e. simulate the satellite observations of q, T , and δD with known uncertainties, then use them within a Kalman filter based assimilation framework in order to evaluate their potential for improving the quality of atmospheric analyses. The study is made for low latitudes (30° S to 30° N) and for 40 days between mid-July and end of August 2016. We find that the assimilation of q and T observations alone well constrains the atmospheric q and T fields (analyses skills in the free troposphere of up to 60 %), and moderately constrains the fields of δD, ω, Q₂, and Prcp (analyses skills of 20 %–40 %). The additional assimilation of δD observations further improves the quality of the analyses of all variables. We use Q₂ as proxy for the presence of condensation and evaporation processes, and we show that the additional improvement is rather weak when evaporation or condensation are negligible (additional analyses skills of generally below 5 %), and strongest for high condensation rates (additional skills of about 15 % and above). The very high condensation rates (identified by large positive Q₂ values) are rare, but related to extreme events (very high ω and Prcp) that are not well captured in the analyses (for these extreme events also the analyses uncertainties of ω, Q₂, and Prcp are very large), i.e. the additional assimilation of δD observations significantly improves the analyses of the water cycle related variables for the events when an improvement is most important. In real world satellite datasets δD observations affected by such strong latent heating events are frequently available, suggesting that the here demonstrated additional δD impact for the simulated world is also a realistic scenario for a real world data assimilation.