Variations in Earth rotation are measured by comparing a time based on Earth’s variable rotation rate about its axis to a time standard based on an internationally coordinated ensemble of atomic clocks that provide a uniform time scale. The variability of Earth’s rotation is partly due to the changes in angular momentum that occur in the atmosphere and ocean as weather patterns and ocean features develop, propagate, and dissipate.
The NAVGEM Effective Atmospheric Angular Momentum Functions (EAAMF) and their predictions are computed following Barnes et al. (1983), and provided to the U.S. Naval Observatory daily. These along with similar data from the NOAA GFS model are used to calculate and predict the Earth orientation parameters (Stamatakos et al., 2016).
The Navy’s high-resolution global weather prediction system consists of the Navy Global Environmental Model (NAVGEM; Hogan et al., 2014) and a four-dimensional variational data assimilation system (4DVar) (Rosmond and Xu 2006, Xu et al., 2005). NAVGEM includes a semi-lagrangian/semi-implicit dynamical core, along with cloud liquid water, cloud ice water, and ozone as fully predicted constituents. NAVGEM is run with a 6-hr update cycle and a 6-hr assimilation window centered about the analysis times of 00, 06, 12 and 18 UTC.
An essential component for data assimilation is the estimation and specification of the background and observation error covariances. NAVGEM 4DVar uses a “static” background error covariance formulation that does not vary over time. In reality, the background errors are highly dependent on the actual weather regime. The NAVGEM hybrid data assimilation system (Kuhl et al., 2013) explicitly includes this flow-dependent background error information or “errors of the day”. We use the NAVGEM ensemble forecast system (McLay et al., 2008 and 2010) along with the Ensemble Transform (Bishop and Toth, 1999) to generate an eighty-member ensemble. The ensemble forecast error covariance generated from the 3-hr ensemble forecasts are linearly combined with the static error covariance to produce a hybrid covariance that more accurately represents the “errors of the day”. This allows the data assimilation system to more effectively use the observations to correct the short-term forecast errors.
An important component of NAVGEM is the Forecast Sensitivity Observation Impact (FSOI). FSOI is a mathematical method to quantify the contribution of individual observations or sets of observations to the reduction in the 24-hr forecast error (Baker and Daley, 2000; Langland and Baker, 2004). The measure of error used in the NRL FSOI is a moist energy norm, which is sensitive to such factors as errors in position and intensity of mid-latitude cyclones, high-pressure centers, and jet streams. The FSOI allows for dynamic monitoring of the relative quality and value of the observations assimilated by NAVGEM, and the relative ability of the data assimilation system to effectively use the observation information to generate an improved forecast.
For these experiments, NAVGEM was run with a resolution of T425L60, with an effective horizontal resolution of ~31 km, and 60 vertical levels with a model top at 0.04 mb (~ 72 km). The global data assimilation system using Hybrid 4DVar was run with an inner loop resolution of T119L60 (~ 110 km). The global assimilation runs were initialized on December 1, 2015 at 00 UTC, using the NAVGEM initial model and satellite radiance bias correction history files from previous NAVDAS-AR (4DVar) experiments. The EAAMF forecasts were initialized from the 00 UTC analyses, and the entire system was run through January 31, 2015 at 00 UTC. The FSOI was computed every analysis cycle, e.g. 00, 06, 12, 18 UTC. For this study, along with the FSOI based on the global moist energy error norm, we computed the FSOI using an error norm based on the Effective Angular Momentum Functions. This modification allowed us to assess which observations were most beneficial in reducing the 24-hr forecast error for the atmospheric angular momentum.