Observing-System Research and Ensemble Data Assimilation at JAMSTEC - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 507

archive at UCAR 1 . Stream 2008 was integrated to 0 UTC 25 February 2009 to
cover the periods of the following field campaigns: PALAU 2008 in the western
Pacific, the Mirai Arctic Ocean Cruise 2008, and the summer and winter T-PARC
(THORPEX Pacific Asia Regional Campaign; THORPEX: the observing-system
research and predictability experiment). Stream 2010 was initiated separately to
quickly evaluate field campaigns conducted in 2010: the Mirai Arctic Ocean
Cruise 2010 and the Vietnam-Philippines Rain Fall Experiment (VPREX) 2010.
Stream 2010 continues to integrate toward the present. In this section, preliminary
results are shown for ALERA2, and OSEs are shown for the latter two campaigns.

21.3.1 ALERA2

ALERA2 provides smoother analysis ensemble spreads than those in ALERA.
Figure 21.3 depicts the analysis ensemble spread of sea-level pressure in the Arc-
tic. Note that the date is arbitrarily chosen to be the same, but the year is different.
Discontinuities found in ALERA are absent in ALERA2. Another advantage of
ALERA2 over ALERA is the richness of diagnostic variables produced from 6-h
forecast. Here, we compare ALERA2 precipitation with the Global Precipitation
Climatology Project (GPCP) analysis ( Adler et al. 2003 ) as an example. The forecast
(Fig. 21.4 a) agrees well with the satellite-based analysis (Fig. 21.4 b) at peaks that
represent disturbances both in the mid-latitudes and in the tropics. Figure 21.4 c
shows the ensemble spread of the precipitation. Convective precipitation contributes
most of the variability, causing weak precipitation in the subtropics and tropics.
Large-scale precipitation is responsible for the variability in the higher latitudes.
A large ensemble spread may also be interpreted as uncertainty of the mean.
ALERA2 even represents some fine details (Fig. 21.5 ). On 18 July 2003 torrential
rainfalls occurred in northern Kyushu. As a result of this disaster, 23 human
lives were lost; 51 houses were destroyed; and thousands of houses were flooded.
ALERA2 reproduces features found in the GPCP analysis: a local maximum over
northern Kyushu owing to a precipitation band running from southwest to northeast,
and a secondary band in the south of Shikoku and the Kii peninsula. Both GPCP
and ALERA2, however, fail to reproduce the intensity observed by radar and gauge
observations, likely because the precipitation bands organized at the meso scales
cannot be resolved at the coarse resolution.
The forecast has excessive weak precipitation. It is known that reanalysis
produces a larger global average than satellite-based estimates ( Onogi et al 2007 ).
Uncertainties remain in the satellite-based estimates despite recent improvements in
rain/no rain classification methods ( Kida et al. 2009 ). In addition, larger biases exist
in forecast model models. Precipitation is unrealistically predicted in the convective
parameterization over the subtropical ocean, where convective inhibition should act

1 Currently PREPBUFR observations are available continuously from 1 May 1997.