Info

The vorticity plots verify our assessment. With the assimilation of GPSRO soundings, DARTNB produced a storm with a maximum vorticity of 24 x 10"5 s"1. The corresponding WRF 3D-Var experiment has a value of 9 x 10"5 s"1, less than 50% of that of the WRF/DART experiment. Moreover, the assimilation of COSMIC GPSRO soundings produces an increase of 3 x 10"5 s"1 in WRF/DART, which extends from 900 mb to 600 mb (Fig. 10). It is clear that the assimilation of COSMIC

GPSRO soundings produces a stronger and more robust typhoon vortex. On the other hand, the impact of COSMIC GPSRO assimilation with WRF 3D-Var is much weaker, and visible only in the lowest 1 km.

Similar results are found for the moisture, temperature, and height fields. The WRF/DART system produces a stronger typhoon and a well-defined warm core. Also, the assimilation of GPSRO soundings from COSMIC increases the temperature at the

Figure 7. Integrated cloud water forecasts from (a) CYCLNB, (b) CYCLALL, and (c) DARTNB experiments valid at 0000 UTC 16 September 2006. The verifying observed IR image at 0000 UTC 16 September 2006 is shown in (d).

Figure 7. Integrated cloud water forecasts from (a) CYCLNB, (b) CYCLALL, and (c) DARTNB experiments valid at 0000 UTC 16 September 2006. The verifying observed IR image at 0000 UTC 16 September 2006 is shown in (d).

Figure 8. Zonal wind (solid fine) and potential temperature (dashed line) analysis along 125.8°E from (a) DARTNB, (b) CYCLNB (3D-Var), (c) DARTNBNG, and (d) CYCLNBNG (3D-Var) at 0000 UTC 14 September 2006. Units: zonal wind (m/s), temperature (K). The location of the typhoon is marked with the typhoon symbol.

Figure 8. Zonal wind (solid fine) and potential temperature (dashed line) analysis along 125.8°E from (a) DARTNB, (b) CYCLNB (3D-Var), (c) DARTNBNG, and (d) CYCLNBNG (3D-Var) at 0000 UTC 14 September 2006. Units: zonal wind (m/s), temperature (K). The location of the typhoon is marked with the typhoon symbol.

Figure 9. Relative vorticity analysis along 125.8° E from (a) DARTNB and (b) CYCLNB (3D-Var) at 0000 UTC 14 September 2006. Units: 1.0 X 10"5 s"1. The location of the typhoon is marked with the typhoon symbol.

Figure 10. Relative vorticity difference analysis from (a) DARTNB-DARTNBNG and (b) CYCLNB-CYCLNBNG (3D-Var) at 0000 UTC 14 September 2006. Units: 1, 0 x 10"5 s"1. The location of the typhoon is marked with the typhoon symbol.

center of the typhoon by about 0.5°C with WRF/DART, while the impact is not apparent in WRF 3D-Var (not shown). The WRF/DART assimilation of GPSRO soundings produces profound changes of water vapor over the western Pacific, with amounts varying from 1.5 to 2gkg_1 [Fig. 11(a)]. The corresponding changes in WRF 3D-Var are much more modest, with amounts on the order of 0.5 gkg-1 or less

The track of typhoons over the western Pacific is strongly affected by the subtropical high. Naturally, we would be interested in seeing how the analysis of the Western Pacific Subtropical High is influenced by the dataassimilation systems. Figure 12 shows the 500 mb geopotential height for both the DARTNB experiment and the CYCLNB (WRF 3D-Var) experiment. We see a much stronger high-pressure system in the vicinity of the typhoon. For example, the height field to the northeast of Typhoon Shanshan has a geopotential height contour of 5920 m in the WRF/DART experiment [Fig. 12(a)], while that in the WRF 3D-Var experiment is only 5900 m [Fig. 12(b)]. Moreover, the assimilation of GPSRO soundings produces a 500 mb potential height difference of 30m with WRF/DART [Fig. 12(c)], while the impact of COSMIC GPSRO data assimilation is barely visible with WRF 3D-Var [Fig. 12(d)].

One may ask: Why should WRF/DART perform better than WRF 3D-Var? This is a

Figure 11. Water vapor difference fields from (a) DARTNB-DARTNBNG and (b) CYCLNB-CYCLNBNG (3D-Var) at 0000 UTC 14 September 2006. Units: 1.0 X 10~3 gkg~1. The location of the typhoon is marked with the typhoon symbol.

very important question and will require considerably more analysis before we can fully answer it. However, we make the following observations. First of all, the ensemble dataassimilation system (i.e. WRF/DART) uses flow-dependent background error covariances, while the background-error covariances used in WRF 3D-Var are not flow dependent. Second, the WRF/DART system takes into account the forecast multivariate error correlations between specific humidity and temperature, as well as surface pressures, while they are not taken into consideration in WRF 3D-Var. Of course, one advantage of WRF 3D-Var is the significantly reduced computational cost. For one-day assimilation with 1h cycling, the WRF/DART system with 32 ensemble members takes 5.5 h of wall-clock time on a machine with 32 IBM Power

5 processors. The corresponding cost for WRF 3D-Var is about 0.4 h. So, the WRF/DART system is more than one order of magnitude more expensive than WRF 3D-Var. Therefore, the improved analysis comes with an increased computational cost.

3. Impact of COSMIC GPSRO

Soundings on Mei-yu Prediction

The Western Pacific Subtropical High (WPSH) has a profound influence on weather systems over East Asia, in particular the East Asia monsoons and the tropical cyclones. In late spring and early summer, Taiwan and southern China are significantly influenced by a quasi-stationary Mei-yu front. Mesoscale convective systems embedded within the Mei-yu front travel eastward along the front, and can produce heavy precipitation. The location and intensity of the Mei-yu front and the formation and development of mesoscale convective systems are strongly affected by the intensity and position of the WPSH, as well as the southwesterly monsoon flows that originate from the Indian Ocean and the South China Sea. Because of the lack of observations over the Pacific Ocean, the South China Sea, and the Indian Ocean, weather analysis, particularly moisture analysis, is often subject to significant uncertainty. Because of this lack of observations over the Pacific Ocean, the intensity and location of the WPSH are often not accurately analyzed by global models. With the availability of GPSRO soundings, COSMIC provides an opportunity to improve the analysis of the WPSH, the southwesterly monsoons, and the Mei-yu front.

In this study, we assimilate GPSRO soundings from COSMIC over a two-week period, from 1 to 14 June 2007, using the WRF/DART ensemble filter data-assimilation system. For this study, we use a WRF/DART system at a horizontal resolution of 36 km, with 35 vertical levels. The number of ensemble members is 32. Figure 13 shows the distribution of COSMIC

Wrf Typhoon
Figure 12. 500 hPa height analysis from (a) DARTNB and (b) CYCLNB at 0000 UTC 14 September 2006; difference fields of (c) DARTNB-DATNBNG and (d) CYCLNB-CYCLNBNG at the same time. Unit: m. The location of the typhoon is marked with the typhoon symbol.

GPS RO soundings during this period and the experimental domain. There are a total of 1,567 GPSRO soundings uniformly distributed over the model domain. In addition to GPSRO soundings, we assimilate upper-air soundings, surface reports, satellite winds, and the cloud-free AIRS-retrieved temperature data. Two experiments are performed. The first is a NoGPS experiment, which assimilates the conventional operational data from NCEP, which include radiosondes, satellite cloud motion winds, and QuikScat surface winds. The AIRS-standard-retrieved temperature profiles (at 50 km resolution) from NASA/JPL are also used. The other experiment is the GPS experiment, which assimilates COSMIC GPSRO soundings in addition to all the aforementioned data. The assimilation experiments are done with 3 h cycling for the entire two-week period.

Figure 14 shows the 850 mb wind fields averaged over the two-week period of 1-14 June 2007 for the NoGPS and GPS experiments. At first glance, they look almost identical, aside from some subtle differences in the flow fields over the western Pacific to the east of Taiwan. The 850 mb wind fields show that the WPSH is extended to about 110°E over the South China Sea. During this period, Taiwan and southern China are under the influence of two confluent flows: one is the southwesterly monsoon flow originating from the Bay of Bengal, and the other is the southerly returning flow associated with the WPSH. The difference fields between the NoGPS and GPS experiments show

Figure 13. Distribution of 1,567 COSMIC GPSRO soundings from 1 to 14 June 2007 over the experiment domain.

an anticyclonic gyre located at about 145°E and 25° N. This gyre is of a scale of about 2,500 km, and has a northeast and southwest orientation. The difference fields suggest that the assimilation of COSMIC GPSRO soundings has enhanced the WPSH over the western Pacific. Note that there is little difference over the northern and eastern lateral boundaries of the model domain. This is because identical lateral boundary conditions, obtained from the NCEP AVN global analysis, are used for the GPS and NoGPS experiments. It is possible that if we use a much larger model domain that covers the entire Pacific Ocean, the COSMIC GPSRO soundings will have an even bigger impact on the entire Pacific subtropical high.

The intensity of the WPSH will have a profound influence on the moisture fluxes, which could have a significant impact on clouds and precipitation. Figure 15 shows the mean moisture fluxes at 850 mb averaged over the two-week period of 1-14 June 2007. The 850 mb moisture flux indicates that a significant amount of moisture originates from the

Bay of Bengal, climbs over the Indochina peninsula, and converges with the returning moisture flow associated with the WPSH. Taiwan and southern China are under the strong influence of the southwesterly moist flow, after these two air streams converge. The impact of COSMIC GPSRO simulation can be visualized by examining the differences in moisture flux between the NoGPS and GPS experiments. Again, an anticyclonic gyre is clearly visible. Over the western part of this gyre, moisture is being transported toward Taiwan. This suggests that the assimilation of COSMIC GPSRO soundings produces an improved analysis, with more moisture being transported to the Taiwan area.

During the period of 6-9 June, Taiwan was under the influence of a Mei-yu front. Mesoscale convective systems propagated from west to east, and produced a significant amount of precipitation. Heavy precipitation took place at first on the west coast of Taiwan on 6 June 2007 [Fig. 16(a)]. It then migrated over northwestern Taiwan, with 24 h accumulated rainfall exceeding 150 mm on 7 June

Figure 14. 850 mb wind fields for (a) the NoGPS experiment and (b) the GPS experiment, and (c) their differences.

[Fig. 16(b)]. This continued into the day of 8 June [Fig. 16(c)], with significant precipitation over the Taichung area, as well as northern Taiwan and southern Taiwan immediately to the west of the Central Mountain Range. The maximum 24 h accumulated rainfall ending at 0000 UTC 9 June exceeded 300 mm over the Taichung area [Fig. 16(c)]. By 0000 UTC 10 June, most of the precipitation, with weaker amounts, had fallen over the Central Mountain Range

[Fig. 16(d)]. An interesting question is: Would the assimilation of COSMIC GPSRO soundings help improve rainfall forecasts?

To answer this question, we show in Fig. 17 the 850 mb moisture flux at 0000 UTC 8 June 2007 for the GPS run, and the differences between the GPS and NoGPS experiments. The basic flow pattern and moisture flux pattern are essentially the same as those of the two-week averages. The difference field of the 850 mb moisture flux shows an interesting structure. The anticyclonic gyre is already established at this time, although the center of the gyre is located further to the west from its two-week mean position. More interestingly, significant eastward moisture fluxes are found to the west of Taiwan and southern China. There is also enhanced moisture flux convergence over southern China and the Taiwan Strait. This should contribute to increased precipitation over this region.

Figure 18 shows the 850 mb moisture analysis for the NoGPS and GPS experiments, and their differences (GPS - NoGPS) at 0000 UTC 8 June. The moisture content is much larger and robust in the GPS experiment. For example, the 16gkg_1 contour is found on the east coast of China near Fujiang province in the GPS experiment, while a weaker amount is found in the NoGPS experiment at the same location. The difference in moisture over the southeastern China coast exceeds 1.0 gkg-1 on the east coast of China, which is directly related to the precipitation event [Fig. 18(c)]. Figures 17 and 18 suggest that the assimilation of COSMIC GPSRO soundings over a one-week period (1-8 June 2007) has produced noticeable changes in moisture distribution and moisture fluxes associated with the Mei-yu system. One should expect that such changes would have an influence on precipitant forecasts.

Indeed, this is the case. Figure 19 shows the 24h precipitation forecast from the WRF model at 12 km, which is initialized with the WRF/DART 36 km analysis at 0000 UTC 8 June from the NoGPS and GPS experiments.

SSO hPo Q Flu* GPS

SSO hPo Q Flu* GPS

1OOE 120E 1+OE 160E

1 oo"

S50 hPa Q Flux, GPS-NoGPS

S50 hPa Q Flux, GPS-NoGPS

0 0

Post a comment