Assimilation experiments with ground-based GPS observations in the Environment Canada Global and Regional Deterministic Prediction Systems
This paper presents results of recent data assimilation experiments involving addition of NOAA/GSD GPS-IPW network ZTD observations to new experimental versions of the EC global and regional deterministic prediction systems (GDPS and RDPS respectively). In the new GDPS and RDPS, analyses that provide initial conditions for forecasts are obtained with a hybrid Ensemble-Variational (EnVar) approach, as opposed to the current operational 4D-Variational (4D-Var) method. In EnVar, a variational analysis is done using background error covariances that are, in part, derived from a 4-dimensional ensemble of model forecasts produced by the EC Ensemble Kalman Filter. As ZTD is an integrated quantity, background error covariances are critical in distributing the humidity analysis increments from ZTD data assimilation optimally in the vertical. The flow-dependent background error covariances for humidity in the new EnVar data assimilation systems (DAS) are better suited for this purpose than the static “NMC-method” variances applied in the operational 4D-Var systems. In addition, a higher analysis increment grid resolution in the new DAS compared to the operational 4D-Var systems is better able to capture the smaller scale variations in humidity sampled by the GB-GPS ZTD data.
The impact of GB-GPS data assimilation in the experimental EnVar-based versions of the EC GDPS and RDPS is evaluated through verifications of forecasts for a two month period in the summer of 2011 using radiosonde observations, GB-GPS observations, rain gauge data and analyses from the European Centre for Medium-range Weather Forecasts (ECMWF).