Abstract. We have developed a Lagrangian air-parcel cirrus model (LACM), to diagnose the processes controlling water in the tropical tropopause layer (TTL). LACM applies parameterised microphysics to air parcel trajectories. The parameterisation includes the homogeneous freezing of aerosol droplets, the growth/sublimation of ice particles, and sedimentation of ice particles, so capturing the main dehydration mechanism for air in the TTL. Rehydration is also considered by resetting the water vapour mixing ratio in an air parcel to the value at the point in the 4-D analysis/forecast data used to generate the trajectories, but only when certain conditions, indicative of convection, are satisfied. The conditions act to restrict rehydration of the Lagrangian air parcels to regions where convective transport of water vapour from below is significant, at least to the extent that the analysis/forecast captures this process. The inclusion of hydration and dehydration mechanisms in LACM results in total water fields near tropical convection that have more of the "stripy" character of satellite observations of high cloud, than do either the ECMWF analysis or trajectories without microphysics.

The mixing ratios of total water in the TTL, measured by a high-altitude aircraft over Brazil (during the TROCCINOX campaign), have been reconstructed by LACM using trajectories generated from ECMWF analysis. Two other Lagrangian reconstructions are also tested: linear interpolation of ECMWF analysed specific humidity onto the aircraft flight track, and instantaneous dehydration to the saturation vapour pressure over ice along trajectories. The reconstructed total water mixing ratios along aircraft flight tracks are compared with observations from the FISH total water hygrometer. Process-oriented analysis shows that modelled cirrus cloud events are responsible for dehydrating the air parcels coming from lower levels, resulting in total water mixing ratios as low as 2 μmol/mol. Without adding water back to some of the trajectories, the LACM and instantaneous-dehydration reconstructions have a dry bias. The interpolated-ECMWF reconstruction does not suffer this dry bias, because convection in the ECMWF model moistens air parcels dramatically, by pumping moist air upwards. This indicates that the ECMWF model captures the gross features of the rehydration of air in the TTL by convection. Overall, the ECMWF models captures well the exponential decrease in total water mixing ratio with height above 250 hPa, so that all the reconstruction techniques capture more than 75% of the standard deviation in the measured total water mixing ratios over the depth of the TTL. By picking up the main contributing processes to dehydration and rehydration in the TTL, LACM reconstructs total water mixing ratios at the top of the TTL, close to the cold point, that are always in substantially better agreement with observations than instantaneous-dehydration reconstructions, and better than the ECMWF analysis for regions of high relative humidity and cloud.