Global AVHRR Winds Assimilation at Fleet Numerical Meteorology
and Oceanography Center/Naval Research Laboratory, Monterey, CA
From the Spring 2017 issue of the JCSDA Quarterly, DOI: 10.7289/V5V98648
Atmospheric Motion Vectors (AMVs) from polar-orbiting satellites have
been in operational use at Fleet Numerical Meteorology and Oceanography
Center (FNMOC) since winds from the Moderate-resolution Imaging Spectro-
radiometer (MODIS) produced at the University of Wisconsin were
introduced in October 2004. Traditionally, such polar winds are based on
imagery from overlapping swaths in successive orbits from a single
satellite. Their dependence on overlapping swaths also limits the data
to polar regions, typically poleward of 60 degrees. However, the
European Organisation for the Exploitation of Meteorological Satel lites
(EUMETSAT) is taking advantage of having Metop-A and Metop-B in the same
orbit (separated by half an orbit) and began operational production of
two-satellite AMVs using image pairs in February 2015 (Borde et al.
2016), and using image triplets (A-B-A or B-A-B) in January 2016
(EUMETSAT 2016). The "dual Metop" AMVs are available globally, while the
"triplet Metop" join "single Metop" AMVs in polar regions. (Note that
at FNMOC, we form superobs for these data without differentiating among
single Metop, dual Metop, and triplet Metop wind vectors, instead
treating them as a single observation type.) This article describes
results from tests of Global AVHRR AMVs in the U.S. Navy's
global modeling system, where they have been used operationally since
Atmospheric motion vectors have had large beneficial forecast impact
in the U.S. Navy's global operational numerical weather prediction
system for many years, so the U.S. Naval Research Laboratory (NRL)
and FNMOC continue to aggressively pursue testing and
assimilation of new satellite winds datasets. The U.S. Navy's
global forecast system is composed of NAVDAS-AR (NRL Atmospheric
Variational Data Assimilation System-Accelerated Representer), a
hybrid ensemble/4DVAR (four-dimensional variational) global data
assimilation system in observation space (Xu et al. 2005; Rosmond and
Xu 2006; Chua et al. 2009; Kuhl et al. 2013), and NAVGEM (Navy
Global Environmental Model), a global atmospheric model currently run
with a resolution of 425 spectral waves with triangular truncation and 60
levels (Hogan et al. 2014). Global AVHRR testing by NRL/FNMOC
began in November 2015 and immediately showed beneficial impacts, so
these winds were introduced into operations in February 2016.
Soon thereafter it was noted (Stone et al. 2016) that some of
the data excluded by standard QC procedures looked as if it might also be
beneficial, motivating a fresh look at some of the QC measures and a set
of test runs which relaxed some of the QC measures.
The control run for our experiment emulated operations as closely
as possible. In our tests, we relaxed two routine QC checks that are
used in operations. One is a check which screens out incoming AMVs
(prior to being superobbed) based on observation-minus-background
(OmB) vector differences; the OmB limit ranges from 8–12 m/s,
depending on the pressure level of the observation. The other is a
check which screens out observations based on their pressure
level; all AMVs above 175 hPa, below 975 hPa, and almost all AMVs
between 425 and 675 hPa are excluded from the assimilation. Our test run
Hnorejvec bypassed the routine which screens based on OmB vector
difference, while the test run Hnocutout also bypassed the check against
the background, but, in addition, allowed through observations above
175 hPa and between 425 and 675 hPa.
We applied the relaxed QC measures to all sources of AMVs, but Global
AVHRR and JMA's Himawari-8 winds were responsible for the great majority
of newly admitted data. Most of the mid-level data excluded from the
control is in polar and near-polar regions, while most of the upper-
level data excluded from the control is in the tropics. Fig. 1 shows
data distributions for a typical six-hour data window.
Figure 1. Meridional slice of zonally averaged data counts of
superobbed Global AVHRR winds in the control (a, upper left) and the
experiment with relaxed quality control screening (b, upper right).
Plotted geographic positions of AMVs assimilated in the experiment but
not the control above 175 hPa (c, lower left) and 425-675 hPa
(d, lower right) for a single 6-hour data window.
The mean vector difference (MVD) between the observations and the
background, one indication of data quality, is plotted in profile in
Fig. 2a. The vector differences of AMVs with heights between 425–675 hPa
indicate that the data in these mid-levels are comparable in accuracy to
the AMVs in the levels above and below. Fig. 2b shows the impact of the
data using the Forecast Sensitivity Observation Impact (FSOI) method of
Langland and Baker, 2004. Again, the data in mid-levels is comparable in
impact, and perhaps even more beneficial than the data above and below
because the mid-levels were a relative data void. At upper levels, the
newly admitted data has significantly larger MVDs, and while we cannot
determine how much of the increase is due to data quality as opposed to
background quality, we do see that the counts above 175 hPa are small
enough that the beneficial impact at those levels is quite small.
Figure 2. Mean Vector Difference (a, upper left) and Forecast
Sensitivity Observation Impact (b, upper right) of the control and
both experiments, binned by pressure levels. Time series of total FSOI
for all vertical levels (c, lower left). Ranking of observation types
by total contribution to FSOI during July 2016 (d, lower right).
Fig. 2c shows the FSOI due to Global AVHRR AMVs (all levels)
for each six-hour analysis data window during the test period July 2016.
There were no instances of non-beneficial impacts (as occasionally
happens with other instruments, particularly when data counts are low),
and the beneficial impact in the Hnocutout run was greater than in the
control in all but two six-hour windows. In our tests, Global AVHRR's
contribution to total FSOI was greater than the contributions from all
but one of the geostationary satellite sources. Global AVHRR winds
provide approximately 2.3 percent of the total FSOI, which is more than
the combined surface satellite-derived winds from ASCAT,
SSMIS, and WindSat (Figure 2d).
Because of these positive results, the mid- level cutout for Global
AVHRR AMVs was eliminated from the operational suite, allowing these
data into the operational analysis, beginning with the update that went
in on January 25, 2017. Global AVHRR's upper level cutout and its
screening against background values remain in place pending further
Rebecca Stone (SAIC, U.S. Naval Research Laboratory, Monterey, CA),
Patricia Pauley (U.S. Naval Research Laboratory, Monterey, CA), Nancy Baker
(U.S. Naval Research Laboratory, Monterey, CA), Randal Pauley (Fleet
Numerical Meteorology and Oceanography Center, Monterey, CA), Bryan
Karpowicz (Devine Consulting, U.S. Naval Research Laboratory, Monterey,
The authors gratefully acknowledge the support from the Office of
Naval Research through Program Element PE0603207N. This is NRL
contribution NRL/PU/7530-17-074, which is approved for public release and
distribution is unlimited.
Borde, R., O. Hautecoeur, and M. Carranza, 2016.
EUMETSAT Global AVHRR Wind Product.
J. Atmos. Oceanic Technol., 33, 429–438,
Chua, B., L. Xu, T. Rosmond, and E. Zaron, 2009. Preconditioning
representer-based variational data assimilation systems: Application to
NAVDAS-AR. Data Assimilation for Atmospheric, Oceanic and Hydrologic
Applications. Springer-Verlag, pp. 307–320.
EUMETSAT, 2016. The new AVHRR Triplet
Mode Polar Winds are available on EUMETCast and the GTS from today (21 January).
Accessed February 14, 2017. [Available online at
Hogan, T.F., M. Liu, J.A. Ridout, M.S. Peng, T.R. Whitcomb, B.C. Ruston,
C.A. Reynolds, S.D. Eckermann, J.R. Moskaitis, N.L. Baker, J.P. McCormack,
K.C. Viner, J. G. McLay, M.K. Flatau, L. Xu, C. Chen, and S.W. Chang,
2014. The Navy Global Environmental Model. Oceanography, 27(3), 116–125,
Kuhl, D.D, T.E. Rosmond, C.H. Bishop, N.L. Baker, and J. McLay, 2013.
Comparison of hybrid ensemble/4D-Var and 4D-Var within the NAVDAS-AR data
assimilation framework. Mon. Wea. Rev., 141, 2740-2758.
Langland, R.H., and Baker, N.L. (2004). Estimation of observation impact
using the NRL atmospheric variational data assimilation adjoint system.
Tellus A, 56: 189–201.
Rosmond, T., and L. Xu, 2006. Development
of NAVDAS-AR: Non-linear formulation
and outer loop tests. Tellus A., 58, 45–58.
Stone, R.E., N. Baker, P. Pauley and B. Karpowicz, 2016. Comparison and Impact of
Newly Available Atmospheric Motion Vectors in NAVGEM.
14th JCSDA Technical Review Meeting & Science Workshop on Satellite
Data Assimilation, Moss Landing, CA, Joint
Center for Satellite Data Assimilation, Poster
Session. [Available online at
Xu, L., T. Rosmond, and R. Daley, 2005. Development of NAVDAS-AR:
Formulation and initial tests of the linear problem. Tellus A, 57, 546–559.