Skip Navigation
JCSDA Logo and banner

JCSDA News and Newsletters


Satellite Data Assimilation Updates in Navy's New Global Prediction System

From the December 2013 issue of the JCSDA Quarterly

The Navy's high-resolution global weather prediction system, run operationally at the Fleet Numerical Meteorological and Oceanographic Center (FNMOC), is the critical component of the Navy's atmospheric prediction capabilities, providing 180-hour forecasts every six hours and 16-day guidance twice-daily using a 20 member global ensemble. On February 13, 2013 a significant milestone in the U.S. Navy's weather forecast system development was achieved as the Navy Global Environmental Model (NAVGEM) replaced the Navy Operational Global Atmospheric Prediction System (NOGAPS) for operational global weather prediction (Hogan et al. 2013). This new operational system, NAVGEM 1.1, combines a semi-Lagrangian/semi- implicit dynamical core with advanced physical parameterization schemes for moisture, convection, ozone, and radiation. Model physics improvements in the transition to NAVGEM include the representation of cloud liquid water, of cloud ice water, and of ozone as fully predicted constituents. Following the successful testing of a new eddy diffusivity mass flux (EDMF) scheme, developed under the Office of Naval Research "Unified Parameterization for Extended Range Prediction" Departmental Research Initiative, a second transition to the NAVGEM (NAVGEM 1.2) occurred on November 6, 2013 (Hogan et al. 2014, submitted). The implementation of NAVGEM 1.2 at FNMOC has brought with it updates to satellite data processing by the 4D-Var assimilation component, NAVDAS-AR. These updates include the assimilation of MetOp-B sensors AMSU-A and MHS, as well as the activation of the Suomi-NPP ATMS.

Since the Suomi-NPP ATMS is a new technology, we will detail some of the procedures used in the assimilation of these data. The NAVGEM 1.2 system, which includes NAVDAS-AR, has active assimilation of ATMS channels 04-15, and 18-22. A Gaussian spatial filter is used to increase the signal-to-noise ratio for channel 03-22. An average brightness temperature at scan position, s, and beam position, b, can be computed using a weighted mean as follows:

Average brightness temperature at scan position

We have chosen to use pre-computed Gaussian weights, wi, using σ = 36 km, corresponding to a Full Width at Half Maximum (FWHM) of ~ 85 km. The Gaussian weights are defined as follows:

Gaussian weights,

where the scene separation distances, ri , are computed for the ATMS scan geometry using the distances between the scene at (s,b) and the closest 100 points (i=1 to 100) in the domain as follows:

Scene separation distances.

The quality control applied to ATMS consists of checks for sea-ice, cloud liquid water, and a scattering index. These checks mostly target the imaging and sounding channels that have sensitivities to the surface, cloud liquid water, and to rain, and do not apply to channels 09-15. The cloud liquid water and scattering index checks apply successively lower thresholds for higher peaking channels, as was described in more detail in the JCSDA seminar presented by Ruston on 12 Dec, 2012. A final 1.25° thinning is applied before assimilation into the system. After the innovation, defined as the departure of the bias corrected observation from the background simulated brightness temperature, is computed, an observation is rejected if its innovation exceeds three times the square root of the sum of the assigned channel-dependent observation and background errors (σo2 + σb2)1/2. An example of the resulting standard deviation of the global innovations in the NAVGEM/NAVDAS-AR system for temperature sounding channels, computed using these quality control procedures, is shown in Figure 1. It can be seen that the global standard deviation of the innovations for the ATMS channels compares very well with that of the various AMSU-A sensors with channels having similar weighting functions.

Global innovation statistics for AMSU-A and ATMS

Figure 1. Global innovation statistics from AMSU-A and ATMS for 08Dec2013 at 00UTC. Each channel's standard deviation of the bias corrected innovation is shown as a horizontal bar aligned with the peak of the channel temperature weighting function.

Another widely-used method to gauge the relative impact of the various observing systems employs the adjoints of the forecast model and the assimilation system, NAVGEM and NAVDAS-AR. This adjoint methodology allows for the simultaneous determination of the impact of any or all individual observation types on the 24-hour forecast total moist static energy error norm. This can be shown for a single observation, but is often shown as a cumulative measure with single channels or entire sensors aggregated over both space and time. An example of the ATMS impacts on a channel-by-channel basis is shown in the left panel of Figure 2, while the right panel illustrates the impact of the assimilated radiances from the various infrared and microwave satellite sensors. Most of the monitoring for innovation statistics and observation sensitivity can be found in near-real-time from the NAVDAS-AR monitoring page (www.nrlmry.navy.mil/metoc/ar_monitor). The Suomi-NPP ATMS sensor has been showing an impact similar to that of an AMSU-A sensor. Since the ATMS includes moisture sounding channels, however, we hope to eventually gain more impact from the ATMS if it becomes possible to lower its observation errors or to include error correlation terms. At this time, correlated error can be diagnosed from the ATMS innovation statistics using the Desrozier technique (Desrozier et al. 2005). Analysis shows a strong correlation in humidity channels, as was expected, but a correlation in temperature channels also arises in ATMS due to 1/f noise in the sensor receiver electronics. Similar findings have been shown in technical reports by both the ECMWF (Bormann et al. 2012) and the UK MetOffice (Doherty et al. 2012). In NAVDAS-AR, each assimilated ATMS channel is assumed to be un-correlated. To compensate for this neglected channel correlation, the observation errors are inflated.

Reduction in the 24-hour forecast total moist static energy error norm

Figure 2. Global cumulative reduction in the 24-hour forecast total moist static energy error norm due to satellite observations. On the left is the channel-by-channel breakdown for the Suomi-NPP sensor, while the right shows the various infrared and microwave sensors. The time period is from 09Nov - 09Dec 2013.

Overall, the performance of ATMS has been stable and on an observation-by-observation basis, and is similar to the performance of observations from AMSU-A, MHS, and SSMIS. The wider swath width and additional channels in the water vapor band, however, more than double the observations from the heritage AMSU-A/MHS sensors, thereby increasing the cumulative impact of ATMS. In summary, the ATMS sensor has proven to be an asset to the Navy NAVGEM/NAVDAS-AR system, and will be relied on to provide temperature and moisture sounding data throughout its lifetime.

(Benjamin Ruston, Steve Swadley, Nancy Baker, and Rolf Langland: Naval Research Laboratory, Monterey, CA)

Newsletters supplied as PDFs


 
 
Modified January 28, 2014 11:01 PM
NOAA | NWS | NASA | Navy Weather
Air Force Weather | Website owner: JCSDA
Level A conformance icon, W3C-WAI Web Content Accessibility Guidelines 1.0Valid HTML 4.01