All-sky Microwave Imager Data Assimilation at NASA GMAO
From the Summer 2017 issue of the JCSDA Quarterly, DOI: 10.7289/V54J0C9F
Satellite radiance observations combine global coverage with high
temporal and spatial resolution and bring vital information to numerical
weather prediction (NWP) analyses, especially in areas where
conventional data are sparse. However, most satellite observations that
are actively assimilated have been limited to clear-sky conditions due
to difficulties associated with accounting for non-Gaussian error
characteristics, nonlinearity, and the development of appropriate
observation operators. To expand existing capabilities in satellite
radiance assimilation, operational centers including the European Centre
for Medium-range Weather Forecasts (ECMWF), United Kingdom's MetOffice,
Japan Meteorological Agency (JMA), and National Centers for
Environmental Prediction (NCEP) have been pursuing efforts to assimilate
radiances affected by clouds and precipitation from microwave sensors.
The expectation is that these data can provide critical constraints on
meteorological parameters in dynamically sensitive regions and have
positive impact on forecasts of precipitation.
As described in the previous issue of the JCSDA Quarterly Newsletter,
NCEP's efforts to assimilate all-sky data in the Gridpoint Statistical
Interpolation (GSI) system have been focused on temperature sounding
data from the Advanced Microwave Sounding Unit-A (AMSU-A) and Advanced
Technology Microwave Sounder (ATMS) in non-precipitating cloudy
conditions. Efforts in all-sky satellite data assimilation at the Global
Modeling and Assimilation Office (GMAO) at NASA Goddard Space Flight
Center have been focused on the development of GSI configurations to
assimilate all-sky data from microwave imagers such as the GPM Microwave
Imager (GMI) and Global Change Observation Mission- Water (GCOM-W)
Advanced Microwave Scanning Radiometer 2 (AMSR-2). Electromagnetic
characteristics associated with their wavelengths allow microwave imager
data to be relatively transparent to atmospheric gases and thin ice
clouds, and highly sensitive to precipitation. Therefore, GMAO's all-sky
data assimilation efforts are primarily focused on utilizing these data
in precipitating regions. The all-sky framework being tested at GMAO
employs the GSI in a hybrid 4D-EnVar configuration of the Goddard Earth
Observing System (GEOS) data assimilation system, which will be included
in the next formal update of GEOS. This article provides an overview of
the development of all-sky radiance assimilation in GEOS, including some
performance metrics. In addition, various projects underway at GMAO
designed to enhance the all-sky implementation will be introduced.
Highlights of all-sky satellite data configuration in GEOS
Table 1: Comparison of clear-sky and all-sky microwave TB assimilation
framework in GEOS-5 ADAS. (T: atmospheric temperature, q: specific
humidity, Tskin: skin temperature, Ps: surface pressure, oz: ozone
mixing ratio, u: u-wind, v: v-wind, ql: liquid cloud mixing ratio,
qi: ice cloud mixing ratio, qr: rain water mixing ratio, and qs:
snow water mixing ratio, Ψ: stramfunction, Χunblanced:
unbalanced velocity potential, Tunblanced: unbalanced
temperature, Psunbalanced: unbalanced surface pressure,
RH: relative humidity)
Various components of the GEOS system have been modified to
assimilate cloudand precipitation-affected microwave radiance data
(Table 1). To utilize data in cloudy and precipitating regions, state
and analysis variables have been added for ice cloud (qi), liquid cloud
(ql), rain (qr) and snow (qs). This required enhancing the observation
operator to simulate radiances in heavy precipitation, including frozen
precipitation. Background error covariances in both the central analysis
and EnKF analysis in hybrid 4D-EnVAR system have been expanded to
include hydrometeors. In addition, the bias correction scheme was
enhanced to reduce biases associated with thick clouds and
a. Cloud scattering coefficients in CRTM
Figure 1. Microwave scattering properties as a function of
frequency for a snow water content of 0.1 g m-3. In actual CRTM lookup
table, extinction coefficients are stored in [m2kg-1] unit.
The observation operator for satellite radiances in GSI consists of
spatial interpolation and the Community Radiative Transfer Model (CRTM),
version 2.2.3. Scattering and extinction coefficients, asymmetry factor,
and phase functions associated with hydrometeors for microwave wavelengths
are read from a lookup table built using the Mie calculation for various
cloud types (i.e., cloud ice, cloud liquid, rain, snow, graupel, and
hail), and for various effective radius assuming a Gamma size distribution
(Yang et al. 2005). Due to the known limitations of Mie scattering
parameters for frozen hydrometeors, especially for high frequency (> 85
GHz) microwave channels (Kim 2006, Liu 2008, Geer and Baordo 2014), new
parameters were calculated using the Discrete Dipole Approximation (DDA)
method for non-spherical frozen precipitation from Liu (2008). Eleven
different non-spherical ice crystal shapes in Liu's database, in addition
to scattering properties of spherical ice crystals calculated with the Mie
method, are examined to find an optimal choice of ice crystal shape to
reconstruct the cloudscattering coefficients for CRTM (Figure 1). For each
shape of ice crystal, a CRTM cloud coefficient lookup table was generated
for 33 microwave frequencies between 10.65 GHz and 190.31 GHz, seven
atmospheric temperatures between 243 K and 303 K, and 405 effective radius
sizes starting from 0.005 mm. The maximum effective radius considered for
rain in the new CRTM coefficients is 1.191 mm. For snow crystals, the
maximum effective radius considered ranges from 0.664 mm to 1.278 mm,
depending on snow crystal shape. Field et al. (2007) particle size
distribution was assumed for frozen hydrometeors and Marshall-Palmer size
distribution (Marshall and Palmer 1948) was assumed for liquid
hydrometeors. After replacing original cloud coefficients with new cloud
coefficients constructed with DDA scattering parameters, Simulated GMI
brightness temperatures based on the new cloud coefficients are found to
be closer to the observations and exhibit less first-guess departure bias
in precipitating regions than those based on the original coefficients
b. Enhanced bias correction
Figure 2. Comparisons of simulated GMI 166 GHz vertically polarized
brightness temperatures (TB) with the observations near Hurricane Celia
on July 12, 2016 00Z: (a) Observed TBs, (b) CRTM simulated TBs with
original scattering coefficients based on Mie method, and (c) CRTM
simulated TBs with the DDA method calculated scattering properties of 3-
bullet rosette snow crystals. The color bar shown in (c) works for (a)
and (b) as well.
As with clear-sky radiances in the GEOS, bias correction for all-sky
microwave radiances is performed using a variational bias correction
scheme (VarBC, Dee 2004, Auligné et al. 2007) which estimates bias
correction coefficients as part of the variational assimilation. For
clear-sky microwave radiance data from microwave sensors such as AMSU-A,
SSMIS, and ATMS, the bias predictors include a constant, the scan angle,
a second-order polynomial of the atmospheric temperature lapse rate
weighted by the radiance weighting function, and the retrieved
cloud water path.
For the all-sky implementation, three changes were made to the
original VarBC: First, the retrieved cloud liquid water path was removed
as a predictor. Second, only nearclear sky observations with near-clear
sky background are used in updating bias correction coefficients for
pre-existing predictors. Third, the mean of the observed and calculated
cloud index based on 37-GHz brightness temperatures, (CIavg), and its
square, are used as two additional bias-correction predictors to correct
the cloud amount-dependent first-guess biases. Results indicate that the
modified VarBC scheme removes most of the bias in the first-guess
departures, as indicated in Figure 3. The magnitude of remaining biases
associated with thick cloud and heavy precipitation are reduced to less
than 2 K in all CI ranges. Similar results are obtained for all GMI
channels (not shown).
Figure 3. Bias of first-guess departures of GMI channel 13 as a
function of CIavg. Thin solid line shows the biases before, while thick
solid line shows biases after, using CIavg as additional predictors in
VarBC. All assimilated data points between 12/01-12/31/2015 were used.
Results only in the bins that have the number of data points greater
than 5 are shown in this figure.
c. Background error covariance matrix
The analysis-control vector in the current GEOS analysis scheme
includes stream function, unbalanced velocity potential, unbalanced
virtual temperature, unbalanced surface pressure, relative humidity,
ozone mixing ratio, and skin temperature (Rienecker et al. 2008). With
the newly added control variables, the corresponding static and
flowdependent background error covariances must be generated.
Climatological statistics were estimated following the NMC method
(Parrish and Derber 1992) using pairs of 24- hour and 48-hour GEOS
forecasts between June 1, 2016 and January 16, 2017. Ensemble
covariances are based on the spread of the 32 ensemble forecasts from
the GEOS hybrid scheme during each analysis cycle.
Figure 4: Comparisons of static background errors (left figures) and
ensemble background errors (right figures) as a function of latitude and
vertical level. (a) and (b): liquid cloud, (c) and (d): ice cloud, (e)
and (f): rain, and (g) and (h): snow.
The panels on the left side of Figure 4 show the vertical
distribution of the static background errors for cloud liquid, cloud
ice, rain, and snow water. Aside from the fact that the estimated errors
are by construction zonally invariant, they have generally smooth
spatial structure. Relatively large errors for liquid clouds are seen in
storm tracks in midlatitudes. Static errors in the Southern Hemisphere
are slightly larger than in the Northern Hemisphere. Generally speaking,
the maximum errors for cloud liquid water occur in the layer between 900
hPa and 850 hPa. Static background errors for cloud ice water show large
values near the tropical tropopause, where large amounts of cloud ice
exist in the anvils of convective clouds. Static background errors for
rain and snow are larger in the tropics than other latitudes. Large
background errors for rain occur in the tropical lower troposphere
between sea level and 600 hPa, while large errors for snow occur in the
tropical middle troposphere between 600 hPa and 450 hPa.
The panels on the right side of Figure 4 show cross-sections of
ensemble background errors for hydrometeors taken from GEOS on 12
December 2015 12Z. The results indicate that the magnitudes of the
ensemble background errors are similar to those of the static
background errors, although ensemble- based estimates show more
detailed flow-dependent structures. Note that there are regions with
nearly zero ensemble error corresponding to areas where the ensemble
members forecasted nearly zero clouds (clear sky). In contrast, the
static background errors show nonzero values over broad ranges of latitude.
Cycled data assimilation experiments were conducted to examine the
impact of all-sky GMI radiances on GEOS analyses and forecasts. GEOS was
run in a hybrid 4D-EnVar configuration with a horizontal resolution of
0.5 degrees for the analysis and 0.25 degrees for the forecast. The
control run assimilated all the data used routinely in GEOS
(conventional data, AMSU-A, ATMS, MHS, IASI, AIRS, GPSRO, and satellite
wind data), while the experimental run assimilated allsky GMI data
additionally. It was found that the all-sky GMI data generally have a
significant impact on the lower tropospheric humidity and temperature
analyses, especially in the tropics, which leads to improved forecasts
of these quantities (Figure 5). Similar results were obtained for all
seasons (not shown). In addition, a noticeable positive impact of all-
sky GMI assimilation on hurricane track forecasts was identified for
Hurricane Melor, which occurred in the western
Pacific during December 2015 (Figure 6).
Figure 5: RMS error differences between the GEOS control and all-sky
GMI experiment in the tropics for December 2015: (a) 850hPa specific
humidity (b) 850hPa temperature, (c) 850hPa V-wind.
Figure 6: GEOS forecasts of the track of hurricane Melor
(December 2015): (a) without and (b) with the assimilation of
all-sky GMI data.
Currently, static and ensemble background errors have the same weight
(0.5) for all analysis variables in GEOS. Climatological background
errors for highly nonlinear and situation-dependent clouds and
precipitation may be less meaningful compared with other dynamical
variables. To assign much larger weight to the ensemble-based background
errors for hydrometeors, the capability to assign different weights for
hydrometeors versus other dynamic variables is under development. The
current all-sky framework will be enhanced by various updates both in
the forecast model and analysis scheme. For example, the inclusion of a
two-moment microphysics scheme (Barahona et al. 2014) in the GEOS
forecast model will provide estimates of cloud particle size
distributions to the all-sky observation operator. Future versions of
the CRTM will account for cloud fraction in calculating radiances. This
should improve the simulation of brightness temperature compared with
the current version of CRTM, which considers only clear-sky or
completely overcast conditions. In addition, we are testing various
dynamic thinning approaches in order to use more data in cloudy and
precipitating regions. All these enhancements are expected to extend the
scope of all-sky radiance assimilation to include more microwave
measurements and lead, in turn, to improved analyses and forecasts.
Min-Jeong Kim, Jianjun Jin, Will McCarty, Amal El Akkaroui,
Ricardo Todling, Wei Gu, and Ron Gelaro (NASA Global Modeling and
Auligné, T., A.P. McNally, and D.P. Dee, 2007: Adaptive bias correction
for satellite data in a numerical weather prediction system. Q.J.R.
Meteorol. Soc., 133, 631–642.
Barahona, D., A. Molod, J. Bacmeister, A. Nenes, A. Gettelman, H.
Morrison, V. Phillips, and A. Eichmann, 2014: Development of two-moment
cloud microphysics for liquid and ice within the NASA Goddard Earth
Observing System Model (GEOS-5), Geosci. Model Dev., 7, 1733–1766.
Dee, D.P., 2004: Variational bias correction of radiance data in the
ECMWF system. Proceedings of the ECMWF Workshop on Assimilation of High
Spectral Resolution Sounders in NWP, Reading, UK, 28 June to 1 July 2004.
Field, P.R., A.J. Heymsfield, and A. Bansemer, A., 2007: Snow size
distribution parameterization for midlatitude and tropical ice clouds, J.
Atmos. Sci., 64, 4346–4365.
Geer, A.J., and F. Baordo, 2014: Improved scattering radiative transfer
for frozen hydrometeors at microwave frequencies. Atmos. Meas. Tch. 7,
Geer, A. J., and P. Bauer, 2011: Observation errors in all-sky data
assimilation. Q.J.R. Meterol. Soc., 137, 2024–2037. DOI: 10.1002/qj.830
Kim, M.-J., 2006: Comparisons of single scattering approximations of
randomly oriented ice crystals at microwave frequencies. J. Geophys.
Research, 111 (D14201): DOI: 10.1029/2005JD006892.
Liu, G., 2008: A database of microwave single- scattering properties for
nonspherical ice particles. Bulletin of the American Meteor. Soc., Vol. 89,
Marshall, J.S., and W.M.K. Palmer, 1948: The distribution of raindrops
with size. J. Meteor., 5, 165–166.
Parrish, D.F., and J.C. Derber, 1992: The National Meteorological
Center's spectral statistical interpolation analysis system. Mon. Wea. Rev.,
Rienecker, M.M., M. J. Suarez, R. Todling, J. Bacmeister, L. Takacs, H.-
C. Liu, W. Gu, M. Sienkiewicz, R. D. Koster, R. Gelaro, I. Stajner, and J.
E. Nielsen, 2008. The GEOS-5 Data Assimilation System—documentation of
versions 5.0.1, 5.1.0, and 5.2.0. Technical Report Series on Global Modeling
and Data Assimilation, Vol. 27, 1–118 pp.
Yang, P., H. Wei, H. Huang, B.A. Baum, Y.X. Hu, G.W. Kattawar, M.I.
Mishchenko, and Q. Fu, 2005: Scattering and absorption property database for
nonspherical ice particles in the near- through far-infrared spectral
region. Applied Optics, 44, 5512–5523.