Evaluation of ECMWF water vapour fields by airborne differential absorption lidar measurements: a case study between Brazil and Europe
1Deutscher Wetterdienst, Hohenpeißenberg, Germany
2DLR Institut für Physik der Atmosphäre, Oberpfaffenhofen, Germany
3European Centre For Medium Range Weather Forecasts, Reading, UK
Abstract. Three extended airborne Differential Absorption Lidar (DIAL) sections of tropospheric water vapour across the tropical and sub-tropical Atlantic in March 2004 are compared to short-term forecasts of the European Centre for Medium Range Weather Forecasts (ECMWF). The humidity fields between 28° S and 36° N exhibit large inter air-mass gradients and reflect typical transport patterns of low- and mid-latitudes like convection (e.g. Hadley circulation), subsidence and baroclinic development with stratospheric intrusion. These processes re-distribute water vapour vertically such that locations with extraordinary dry/moist air-masses are observed in the lower/upper troposphere, respectively. The mixing ratios range over 3 orders of magnitude. Back-trajectories are used to trace and characterize the observed air-masses.
Overall, the observed water vapour distributions are largely reproduced by the short-term forecasts at 0.25° resolution (T799/L91), the correlation ranges from 0.69 to 0.92. Locally, large differences occur due to comparably small spatial shifts in presence of strong gradients. Systematic deviations are found associated with specific atmospheric domains. The planetary boundary layer in the forecast is too moist and to shallow. Convective transport of humidity to the middle and upper troposphere tends to be overestimated. Potential impacts arising from data assimilation and model physics are considered. The matching of air-mass boundaries (transport) is discussed with repect to scales and the representativity of the 2-D sections for the 3-D humidity field. The normalized bias of the model with respect to the observations is 6%, 11% and 0% (moist model biases) for the three along-flight sections, whereby however the lowest levels are excluded.