Semi-Lagrangian moisture transport in the NMC spectral model

Abstract

The moisture advection component of the National Meteorological Center global spectral model is replaced with a semi-Lagrangian transport scheme. Several interpolation schemes, both monotonic and nonmonotonic, which were not eliminated from contention in earlier studies with simple test systems, are compared for moisture transport in five-day forecasts. The observed conservation properties of the schemes are considered first. Many of the schemes are found to be unsatisfactory with respect to conservation although several are very good. The best nonmonotonic interpolants are Hermite cubic and a rational cubic both coupled with a cubic derivative estimate. The best monotonic interpolants are these same two with the cubic derivative estimate modified to satisfy the appropriate C° monotonicity condition. The conservation aspects of these four schemes are better than the original spectral scheme if their computational sources are compared to the negative mixing ratios introduced by the spectral method. The skill of the semi-Lagrangian forecasts as measured by the traditional global skill scores (anomaly correlation and rms) is the same as that of the original spectral model. The characteristics of the precipitation in the semi-Lagrangian forecasts are very different from those of the spectral forecast. While the major precipitation events are similar in the two in both magnitude and location, the patterns of light precipitation are very different. The spectral model produces regions of light precipitation which are attributable to spectral truncation. The semi-Lagrangian method shows no similar spurious precipitation.