gamo(c), taichu tanaka(a), koji yamashita(d), takuji...
TRANSCRIPT
Kozo Okamoto 19th Coherent Laser Radar Conference
CLRC 2018, June 18 – 21 1
Evaluation of potential impacts of future Japan’s space-based
Doppler Wind Lidar (DWL) on polar- and tropical-orbiting
satellites
Kozo Okamoto(a,b), Toshiyuki Ishibashi(a), Shoken Ishii(b), Philippe Baron(b), Kyoka
Gamo(c), Taichu Tanaka(a), Koji Yamashita(d), Takuji Kubota(e)
(a) Meteorological Research Institute of Japan Meteorological Agency, Tsukuba, Japan
(b) National Institute of Information and Communications Technology, Kobanei, Japan
(c) FUJITSU FIP Corporation, Tokyo, Japan
(d) Meteorological Satellite Center of JMA, Tokyo, Japan
(e) Japan Aerospace Exploration Agency, Tsukuba, Japan
Abstract: The feasibility of coherent Doppler wind lidars (DWLs) has been
investigated based on a sensitivity observation system simulation experiment (OSSE).
The pseudo-truth atmospheric field is generated from the Sensitivity Observing System
Experiment (SOSE) approach. Hourly aerosols are produced to simulate DWL by a
global aerosol chemical transport model in which wind field is nudged with pseudo-
truth. Simulated measurement of horizontal line-of-sight wind speeds is assimilated by
using the four-dimensional variational (4D-Var) scheme of the operational global data
assimilation system at Japan Meteorological Agency. We evaluated potential impacts
of DWLs onboard polar and low-inclination orbiting satellites in January and August in
2010. We found positive impacts of DWLs on either satellites and greater impacts in
the January experiment. The results also showed seasonal dependency of impacts and
importance of quality control and observation error setting.
Keywords: Data assimilation, OSSE, satellite
1. Introduction
The 3D global wind observations are essential for numerical weather prediction (NWP) but current
global observation network does not satisfy the NWP needs. A space-based Doppler wind lidar (DWL)
is one of good candidates and its feasibility study and development have been underway in many studies.
The Japanese scientific community is also investigating the impact of DWLs on NWP by using an
observing system simulation experiment (OSSE). Compared with many previous studies on the OSSE
for DWLs, this study simulated hour-by-hour aerosol fields and then realistic DWL data distributions
and quality information, which allowed us to discuss the importance of quality control and observation
error setting in a data assimilation system. The paper gives a brief description on our OSSE approach
and results of DWL assimilation. The detail is found in [1].
2. OSSE configuration
We adopted the OSSE based on a sensitivity observing system experiment (SOSE) approach [2] because
it could use all existing observations as opposed to a traditional nature-run OSSE that requires simulating
them. Our SOSE-OSSE configuration is shown in Fig.1. First, we simulated DWL winds and their
quality information such as the signal-to-noise ratio from a pseudo-truth (PT) atmospheric field, and
clouds and aerosols. The PT field was constructed by correcting an original initial (analysis) field using
the adjoint sensitivity structure and by merging it with existing observations. Clouds were generated
from the global forecast model through data assimilation cycle of the PT generation process. Aerosols
were calculated using a global chemical transport model [3] nudged with PT winds. PT atmospheric
field, clouds and aerosols were generated hourly and consistent each other.
Mo4
Kozo Okamoto 19th Coherent Laser Radar Conference
CLRC 2018, June 18 – 21 2
The end-to-end lidar simulator, named the Integrated Satellite Observation Simulator for a Space-Borne
Coherent Doppler Lidar (ISOSIM-L) calculates backscattered power, background noise power, signal-
to-noise ratio (SNR), noisy Doppler-shifted signal, retrieved winds in the direction of line-of-sight, and
retrieval quality information. With respect to DWL, we assumed two coherent receivers pointing toward
the ground at 35° off-nadir with azimuthal angles of 45° and 135° along the satellite track. Laser
wavelength, pulse energy and repetition frequency are 2.0 μm, 125 mJ, and 30 Hz, respectively. Target
vertical resolution and wind speed accuracy are set to 0.5 km and 1 m s−1 at an altitude between 0 and 3
km, 1 km and 3 m s−1 between 3 and 8 km, and 2 km and 3 m s−1 between 8 and 20 km. A candidate
platform for the DWL is a super-low-altitude satellite flying at 220 km or lower. The detail of ISOSIM-
L and DWL instruments are described in [4,5], respectively.
3. Data assimilation experiments
We assimilated simulated observations of horizontal line-of-sight (HLOS) wind speed. In the
preprocessing procedures of data assimilation, we removed the anomalous data flagged by ISOSIM-L,
low SNR data and data substantially departed from first-guess (short-range forecast). The quality control
(QC) procedures with rigid thresholds successfully removed DWL winds that disagreed from PT: 85 %
of all data were excluded (mainly data above 10 km and near the surface), of which 60 % by SNR-based
QC. Observation error R was defined with 220.1 mECR : E is measurement error estimated from
ISOSIM-L according to altitude and season. Other error components such as observation operator and
representativeness error were assumed to be 1.0 m s-1. Finally, we inflated the observation error by
multiplying an experimental factor of C (now set to 4.0) to avoid degradation of forecast skills through
data assimilation experiment. This degradation was probably attributed to inappropriate treatment of
oversampling of DWLs and PT error.
We conducted analysis–forecast cycle experiments to assess the impacts of the DWL data simulated by
ISOSIM-L in two one-month assimilation experiments in January and August 2010. The data
assimilation system was the low-resolution version of the operational global data assimilation system of
JMA as of 2010 [6]. The analysis system was an incremental 4D-Var method with an inner loop of
horizontal resolution of 120 km and outer loop of 60 km. We performed 6 h data assimilation cycles
from December 20, 2009, to February 7, 2010, (January experiment) and from July 20, 2010, to
September 9, 2010 (August experiment). We also ran 120 h forecasts at 12:00 UTC from January 1 to
31, 2010, for the January experiment and from August 1 to 31, 2010, for the August experiment. The
results are shown from three experiments with different observation configurations. One is a reference
experiment, denoted as CNTL, where all of the observations used in the operational system were
assimilated. The second experiment, (TESTP) assimilated DWL HLOS winds from the polar-orbiting
Figure 1. SOSE-OSSE scheme. The upper block is an offline simulation process that produces PT
atmospheric fields, clouds, and aerosols and then simulates DWL winds by using ISOSIM-L. The
lower block is a regular assimilation process, except for adding DWL simulations.
SOSE pseudo-truthand
forecasted cloud
Lidar simulator(ISOSIM-L)
Simulated WLOS, quality info
aerosol model (MASINGAR)
DWL wind simulation
wind aerosol
Existing observation
Simulated WLOS, quality info
assimilationfirst guess analysis forecast model
assimilationfirst guess
data assimilation cycle
Kozo Okamoto 19th Coherent Laser Radar Conference
CLRC 2018, June 18 – 21 3
satellite in addition to CNTL observations. The third experiment (TESTT) assimilated DWL HLOS
winds from tropical satellite with low-inclination of 35.1° in addition to CNTL observations.
Figure 2 shows the relative forecast error reduction of zonal wind for TESTP and TESTT in the January
experiment. The relative forecast error is defined with root mean square (RMS) difference from PT field
and normalized by CNTL RMS forecast error. Positive impact (error reduction) is evident in both
hemispheres and in the tropics through wide tropospheric layers over a broad forecast range up to day 5
for TESTP and TESTT experiments. Improvement is clear, particularly in the upper and lower
troposphere in the tropics, with statistical significance. Figure 3 shows that impacts for the August
experiment are also generally positive with some exceptions of negative impact in short-range forecasts
in the southern hemisphere for TESTP. Because we found simulated DWL quality was a little lower in
August (not shown), those data probably contaminated analyses and then forecasts even after the rigid
QC. This suggests that more appropriate QCs, such as having more situation dependency, should be
developed. We found clear positive impacts on temperature forecasts for TESTP and TESTT in the
January and August experiments (not shown). The track forecasts of tropical cyclones (TCs) was slightly
improved for both TESTP and TEST when their center positions of 18 cases for 5 TCs were verified
against best track dataset. Finally, we directly compared the impacts between TESTP and TESTT. The
direct comparison of TESTT and TESTP in Fig. 4 shows that relative impacts are mixed and,
interestingly, that TESTT is superior to TESTP up to Day 2 but is inferior beyond Day 2 in the tropics
for both the January and August experiments. We need more investigation to make a firm conclusion
on whether TESTT or TESTP are superior.
Figure 2. Relative forecast error reduction (%) of zonal wind as a function of forecasts up to 120
h for (upper) TESTP and (lower) TESTT for the January experiments. It is verified against PT in
(a, d) the northern hemisphere (20°N–90°N), (b, e) tropics (20°N–20°S), (c, f), and southern
hemisphere (20°S–90°S). Positive values correspond to forecast improvement in TESTs. The
contour lines indicate the statistical significance at the 90% and 95% confidence levels.
(a) TESTP in NH (b) TESTP in TR (c) TESTP in SH
(d) TESTT in NH (e) TESTT in TR (f) TESTT in SH
Figure 3. Similar to Fig. 2 but for the August experiment.
(a) TESTP in NH (b) TESTP in TR (c) TESTP in SH
(d) TESTT in NH (e) TESTT in TR (f) TESTT in SH
Kozo Okamoto 19th Coherent Laser Radar Conference
CLRC 2018, June 18 – 21 4
4. Summary and final comments
We found that a space-based DWL gave us significant positive forecast impacts from the SOSE-based
OSSE study. The similar positive impacts were achieved by DWLs on polar- and tropical-orbiting
satellites and decision of the superiority of DWL on either satellite could not be made at this stage of
the study. We also found positive impacts of DWL in both August and January and greater in January.
Also, the results demonstrated the importance of appropriate QC and observation error assignment and
that suggested more positive impacts would be obtained by refining them. For example, the simple
observation error inflation setting in this study could be replaced with an adaptive inflation according
to the density of existing observations. Furthermore, some observation errors were not taken into
account such as neglecting local effect of vertical wind component. Future study will include these
developments. We will investigate the complementarity between DWL and atmospheric motion vectors
(AMVs) derived from tracking successive imageries of passive infrared and visible imagers [7].
5. Acknowledgements
A part of this research was supported by JSPS KAKENHI under Grant Numbers 15K05293 and
15K06129.
6. References
[1] Okamoto, K., T. Ishibashi, S. Ishii, P. Baron, K. Gamo, T. Y. Tanaka, K. Yamashita, and T. Kubota:
“Feasibility study for future space-borne coherent Doppler wind lidar. part 3: Impact assessment using sensitivity
observing system simulation experiments”, J. Meteor. Soc. Japan, 96, 179-199, (2018).
[2] Marseille, G. J., A. Stoffelen, and J. Barkmeijer: “Sensitivity observing system experiment (SOSE): A new
effective NWP‐based tool in designing the global observing system,” Tellus A, 60, 216–233 (2008).
[3] Tanaka, T. Y., and M. Chiba, “Global simulation of dust aerosol with a chemical transport model,
MASINGAR”, J. Meteor. Soc. Japan, 83A, 255–278 (2005).
[4] Baron, P., S. Ishii, K. Okamoto, K. Gamo, K. Mizutani, C. Takahashi, T. Itabe, T. Iwasaki, T. Kubota, T. Maki,
R. Oki, S. Ochiai, D. Sakaizawa, M. Satoh, Y. Satoh, T. Y. Tanaka, and M. Yasui: “Feasibility study for future
spaceborne coherent Doppler Wind Lidar. Part 2: Measurement simulation algorithms and retrieval error
characterization,” J. Meteor. Soc. Japan, 95, 319–342 (2017).
[5] Ishii, S., K. Mizutani, P. Baron, M. Aoki, K. Mizutani, M. Yasui, S. Ochiai, A. Sato, Y. Satoh, T. Kubota, D.
Sakaizawa, R. Oki, K. Okamoto, T. Ishibashi, T. Y. Tanaka, T. T. Sekiyama, T. Maki, K. Yamashita, T. Nishizawa,
M. Satoh, and T. Iwasaki: “Feasibility study for future spaceborne coherent doppler wind lidar, Part 1: Global
Wind Profile Observing System,” J. Meteor. Soc. Japan, 95, 301–317 (2017) .
[6] Japan Meteorological Agency, Outline of the operational numerical weather prediction at the Japan
Meteorological Agency [Available at http://www.jma.go.jp/jma/jma-eng/jma-center/nwp/outline-nwp/index.htm]
(2007).
[7] Ishii, S., K. Okamoto, P. Baron, T. Kubota, Y. Satoh, D. Sakaizawa, T. Ishibashi, T. Y. Tanaka, K. Yamashita,
S. Ochiai, K. Gamo, M. Yasui, R. Oki, M. Satoh, and T. Iwasaki, “Measurement performance assessment of future
space-borne Doppler wind lidar,” SOLA, 12, 55–59 (2016).
Figure 4. Relative forecast error reduction (%) of zonal wind speed for TESTT against TESTP.
A positive value indicates that TEST has a smaller RMSE than TESTP.
(a) U TR for January exp (b) U TR for August exp