Our laboratory has performed daily high-resolution weather simulation around the Japan Islands by using the CReSS (Cloud Resolving Storm Simulator) since December 2004. During this simulation period, we have accumulated distributions of pressure, wind, precipitation and other meteorological parameters of the simulation, except for the interruption of the simulation. These simulation results are presented in the web-site of our laboratory ( http://www.rain.hyarc.nagoya-u.ac.jp/CReSS/fcst_exp.html). We utilize the JMA (Japan Meteorological Agency)-RSM (Regional Spectral Model) data distributed from Japan Meteorological Business Support Center as the initial and boundary conditions of these simulations. We have performed the daily weather simulation with 5 km horizontal grid resolution after the renewal RSM data.
We show the result of the daily weather simulation of the snowfall event around the Central Japan Region on February 1 to 2, 2005. Figure 1 shows the distribution of the precipitation obtained by the JMA radar. A snow cloud band intrudes from the Sea of Japan to Nagoya area. There was a heavy snowfall event at Nagoya city on this day. The daily weather simulation result retrieves these snow cloud bands well. We utilized the daily weather simulation result in order to decide the aircraft observation across the Baiu front over the East China Sea in June 2005. Thus, the daily weather simulation is a good agreement with the weather phenomena qualitatively.
We will continue to the daily weather simulation by using the CReSS and analyze these results statistically in order to improve the CReSS, especially its parameterization schemes. We also improve the cumulus and saturation adjustment parameterizations in the GCM (General Circulation Model) by analyzing the accumulated daily weather simulation results statistically. We have a purpose to contribute the disaster prevention by utilizing the daily weather simulation with 1 km grid resolution.
Fig. 1: Horizontal distribution of precipitation rates observed by JMA radar at 03 JST (9 hours ahead of UTC) on February 2, 2005.
Fig. 2: Same as Fig. 1 but of the daily weather simulation result by the CReSS at the same time.
Fig. 3: Horizontal distribution of convergence of water vapor (shaded) and horizontal wind field (vector) at the height of the 10 m (around the surface) over the East China Sea at 12 JST on June 23, 2005. The Aircraft observation was carried out along the prescribed line A-A' across the convergence zone of water vapor.
When a typhoon is located over the low-latitude ocean, it is composed of active cumulonimbus clouds. The eye-wall and spiral rainbands are conspicuous within the typhoon. On the other hand, they are often modified and become indistinct when the typhoon comes to the mid-latitude around the Japanese Islands or the Korean Peninsula. Although the latent heat from the sea is reduced, a heavy rainfall occasionally occurs over the land and it results in floods and landslides. One of significant cases of heavy rainfall caused by typhoons in the mid-latitude occurred in the central part of Japan on 20 October 2004. The typhoon 0423 brought the heavy rainfall and caused a severe disaster due to the flood. Even though the eye-wall and spiral rainbands were indistinct when T0423 approached the area, the intense rainfall more than 30 mm hr-1 was brought to the area. We studied the heavy rainfall associated with T0423 using the cloud-resolving model. The purpose of the present study is to clarify the formation process of the heavy rainfall caused by the typhoon when it came to the mid-latitude.
In order to perform simulation and numerical experiments of high-impact weather systems, we have been developing a cloud-resolving numerical model named the Cloud Resolving Storm Simulator (CReSS), which is a non-hydrostatic and compressible equation model with detailed cloud microphysics. In order to simulate both the overall structure of T0423 and the detailed structure of the heavy rainfall, the domain of calculation is as large as 1536 times 1408 km in horizontal and the horizontal resolution is as high as 1 km. We performed 30-hour simulation of T0423 from 1200UTC, 19 October 2004. The initial and boundary conditions were provided by the JMA (Japan Meteorological Agency)-RSM (Regional Spectral Model) output.
The result of the prediction experiment shows that the typhoon-track, total rain distribution (Fig. 4) and rainfall intensity were quantitatively simulated. The heavy rainfall in the northern Kinki District was simulated successfully (Fig. 5). This was associated with the intrusion of the intense upper-level rainband. The prediction of precipitation by CReSS was compared with JMA surface observations and its accuracy was evaluated statistically using parameters of root mean square error, correlation coefficient, threat score, and bias score. The evaluation of the prediction experiment using these parameters shows that the CReSS model simulated the heavy rain with sufficiently high scores during the period of the prediction experiment. The successful results indicate that the cloud-resolving model is useful and effective for the accurate and quantitative prediction of heavy rainfall.
Fig. 4: Total rainfall (mm) obtained from the prediction experiment using the CReSS model for the period from 15 UTC, 19 to 15 UTC, 20 October 2004. The cross in the figure indicates Fukuchiyama.
Fig. 5: Time series of rainfall rates (mm hr-1) of AMeDAS observation (bars), predictions by CReSS (solid line) and RSM (dashed line) at Fukuchiyama in the northern Kinki District.
In East Asia, heavy rain is often caused in association with the Baiu/Meiyu front. Understanding and numerical prediction of heavy rain are one of the most important objectives of the mesoscale meteorology. Since intense convective systems are usually composed of cumulonimbus clouds, a cloud resolving numerical model is necessary for simulation of heavy rain. In order to perform simulations and numerical experiments of cloud and precipitation systems, we have been developing a cloud-resolving numerical model named ``the Cloud Resolving Storm Simulator'' (CReSS).
Since heavy rainfall systems often have a multi-scale structure ranging from the cloud-scale to the synoptic-scale, a large computational domain and very high resolution grid to resolve individual classes of the multi-scale structure are necessary to simulate evolution of convective systems. The purpose of this research is simulation of a localized heavy rain which occurred to the south of the Baiu front. Using the result of the simulation, we studied the complex processes of the heavy rainfall.
The heavy rain event occurred on 19-20 July 2003 in Kyushu, which is the western Japan. During this period, the Baiu front was located to the north of Kyushu. The most intense rain occurred during the period of 16-22 UTC, 20 July. The total amount of rain of this period reached about 210 mm at Minamata City which is located in the western Kyushu. The heavy rain caused a flood and 21 people were killed. Radar data provided by JMA (the Japan Meteorological Agency) shows that intense echo systems developed within a mesoscale convective system and moved into the west coast of Kyushu. An intense rainband extended in the east-west direction and some orographic rainfall echoes developed in the western Kyushu.
The initial time of the simulation experiment with a horizontal resolution of 1 km was 00 UTC, 19 July 2003 and 24-hour integration was performed. The experiment successfully simulated the localized heavy rain. The rainfall pattern in the simulation is quite similar to the observation with regard to the pattern and intensity of rain. The result showed that an intense rainband developed to the west of Kyushu in association with the cloud cluster. The simulation also shows that another rainband is present to the south of the main rainband. It forms on the lee side of small islands over the sea, which is named the Koshiki islands (Fig. 6). If the small islands are removed in a sensitivity experiment, the rainband completely disappeared (Fig. 7). This indicates that the islands are essentially important for the formation of the rainband. We consider that both the two rainbands caused the localized heavy rainfall.
Fig. 6: Simulation experiment of the Minamata heavy rainfall event on 19 July 2003. Cloud and rain mixing ratio (g kg^{-1}) and horizontal velocity at a height of 1500 m at 15 hours from the initial time.
Fig. 7: Same as Fig. 6 but for the sensitivity experiment without the Koshiki islands.
In summer, a cumulonimbus cloud develops over the slope of mountains with a valley wind and often brings heavy rainfall to surface. The cumulonimbus cloud is one of the dominant elements of the water circulation process associated to the valley wind circulation. In this study, to investigate the role of a cumulonimbus cloud in the water circulation process associated to the valley wind circulation, I examine the evolution of cumulonimbus clouds which developed over the slope of mountains with the valley wind and the time variation of precipitable water vapor over the slope. As the example of mountains, I select the mountains where includes the Ryoupaku Mountains and the Hida Highland on the north of the Noubi Plain.
I used the radar data of Japan Meteorological Agency (JMA), AMeDAS data of JMA and precipitable water vapor (PWV) retrieved from dense network of global positioning system (GPS) operated by Geographical Survey Institute (GSI) of Japan. The case is on July 4, 2000 when a high at surface was present over Japan and a cold and dry air was present over Japan. Local Standard Time (LST = UTC + 9 hours) is used in the present paper.
Figure 8 shows the convective echo observed by JMA radar which correlated with a cumulonimbus cloud and the wind speed and direction at each AMeDAS point. At 1230 LST, the convective echo P1 developed over the peak of mountains with the valley wind (Fig. 8a). At 1410 LST, the outflow from the weakened echo P1 extended over the peak (Fig. 8b). On the south and north slopes of mountains, convective echoes N1 and H1 developed, respectively. To notice the echo N1, it had the rainfall intensity over 64 mm h-1. At the Hachiman of AMeDAS observation point, rainfall of 14 mm in 60 minutes and the outflow from the convective echo N1 were observed when the echo N1 passed. Valley wind blew on the downslope side of the echo N1. At 1530 LST, the convective echo N2 which had the rainfall intensity over 64 mm h-1 developed on the downslope side of the weakened echo N1 (Fig. 8c). On the peak of mountains, namely the upslope side of the echo N1, the convective echo P2 appeared.
Figure 9 shows the time deviation of PWV and the wind speed and direction at each AMeDAS point. At 1230 LST, PWV increased over the peak of mountains with the development of the convective echo P1 (Fig. 9a). On the foot and slope of mountains with valley wind, PWV increase a little or decreased. At 1410 LST, PWV decreased abruptly with the weaken echo P1 (Fig. 9b). On the slope, PWV increased abruptly with the development of the echo N1. PWV also increased on the downslope side of the echo N1 where valley wind blew. At 1530 LST, PWV decreased at the weakened echo N1 (Fig. 9c). The large gradient of time deviation of PWV existed at the echo N2. Over the peak of mountains, PWV increased with the appearance of the echo P2.
After the cumulonimbus cloud caused by the transportation of water vapor to the peak of mountains by valley wind, the new cumulonimbus cloud occurred over the slope where the outflow from the old cumulonimbus cloud and the valley wind made the horizontal convergence and PWV increased. The cumulonimbus cloud developed on the region where PWV increased adruptly and had large rainfall intensity. I consider that the cumulonimbus cloud played the roles of the accumulation of water vapor transported by valley wind, the formation of rain from the water vapor and the precipitation on the surface. Furthermore, it is considered that the outflow from the cumulonimbus cloud played the roles not only the formation of new horizontal convergence with the valley wind on the downslope side but also the retransportation of water vapor to the peak of mountains.
Acknowledgment: The GPS data was analyzed by Meteorological Research Institute of JMA with use of the GIPSY/OASIS II software package developed by JPL/NASA.
Fig. 8: The distribution of radar echo observed by JMA radar and the direction and speed of wind at each AMeDAS observation point at (a) 1230 LST (b) 1410 LST (c) 1530 LST. Black contours with shades are radar echo at 1, 4, 16, 32 and 64 mm h-1. Barbs are the direction and speed of winds. Thin black contours indicate the terrain at 100, 500, 800, 1000, 1500, 2000, 2500 m ASL.
Fig. 9: The time deviation of PWV and the direction and speed of wind at each AMeDAS observation point at (a) 0900 LST - 1230 LST (b) 1230 LST - 1410 LST (c) 1410 LST - 1530 LST. White contours with shades are the time deviation of PWV every 2 mm. Barbs are the direction and speed of winds. Thin black contours indicate the terrain at 100, 500, 800, 1000, 1500, 2000, 2500 m ASL. Black shaded areas are the terrain over 1500 m ASL.
A purpose of this study is to clarify characteristics of convective boundary layer (CBL) and vertical circulation in the CBL over the Huaihe River Basin during early summer in 2004. The Huaihe River Basin has a large plain in which farmland is nearly uniform and flat, and double cropping (wheat and rice) is cultivated annually. In early summer, the vegetation change from mature wheat fields to paddy fields.
Firstly, the data observed by 1290 MHz wind profiler radar (WPR) and 30 m flux tower was analyzed. In the former period which vegetation was mature wheat fields or bare fields, sensible heat flux (SHF) from land surface was nearly equal to latent heat flux (LHF). In the latter period which vegetation changed to paddy fields, LHF was much larger than SHF. ``Dry-case'' as a representative of the former period and ``wet-case'' as that of the latter period are defined, and two fine days are selected for my analyses. On the dry-case, a deep CBL corresponding to large SHF developed rapidly from early morning. Thermal updrafts in the CBL were strong. On the wet-case, a shallow CBL corresponding to small SHF developed slowly from late morning. Thermal updrafts in the CBL were weak. The circulation formed by thermal updrafts and inter-thermal downdrafts was detected only below the height of the CBL top, therefore the circulation on the dry-case was higher than that in the wet-case.
Secondly, numerical simulation using the CReSS (Cloud Resolving Storm Simulator) is conducted in order to investigate three-dimensional structure including temperature and humidity, and to evaluate vertical flux caused by the circulation. The surface heat flux and the development of the CBL on both cases were reproduced adequately. Vertical heat and moisture fluxes in the CBL were estimated (Fig. 10). In addition, contribution to buoyancy flux of these fluxes was evaluated. On the dry-case (Fig. 10a), the heat flux contributes to almost all of the buoyancy flux. On the wet-case (Fig. 10b), in contrast, the contribution of the moisture flux was equaled to that of the heat flux. Large contribution of the moisture flux to the buoyancy is the characteristics of the CBL in humid continental area.
Fig. 10: Profiles of the vertical fluxes at 5 hours later from the starting of the simulation of the dry-case (a) and the wet-case (b). The solid line indicates the buoyancy flux. The dotted line indicates the heat flux (the contribution to the buoyancy flux). The dash-dotted line indicates moisture flux. The broken line indicates the contribution of the moisture flux to the buoyancy flux. The unit of the moisture flux is [10-3 kg kg -1 m s-1], those of others are [K m s-1].
In order to confirm the structure of convective circulation in the atmospheric boundary layer (ABL) over the western Pacific Ocean around the Southwest Islands of Japan under the subtropical high, we performed an intensive observation using radiosonde and Aerosonde in August 2002. From these Aerosonde observations, there is a negative-correlation type eddy; that is, positive (negative) potential temperature and negative (positive) mixing ratio of water vapor anomalies exist simultaneously, even in the lower subcloud layer below the height of 0.2 km.
We examined the negative correlation type eddy observed by the Aerosonde in the lower subcloud layer by using the CReSS (Cloud resolving Strom Simulator) with 100 m horizontal grid resolution in order to resolve thermals in the subcloud layer. A horizontal domain of this 3-dimensional simulation is 10×10 km2. We carried out numerical simulation by using the vertical profile observed by the radiosonde observation at 00 UTC on August 24, 2002, as the initial condition.
Some thermals recognized by the updraft regions develop in the subcloud layer in this simulation (Fig. 11). The horizontal scale of the thermals is about 0.5 to 1.5 km, and maximum vertical velocity in the thermals exceeds 1.5 m s-1. A cool and moist air mass appears in the updraft regions not only in the upper subcloud layer but also in the lower one. Figure 12 shows vertical profiles of buoyant, heat and moisture fluxes, and the contribution of heat and moisture to the buoyant flux. A positive buoyant flux appears in the lower subcloud layer, therefore, thermals in this layer are driven by the positive buoyancy flux. This positive buoyant flux is driven by the contribution of moisture, not that of heat. This positive buoyant flux contributed by moisture is attributed to small density of water vapor compared with the dry air. This should be attributed to the supply of small sensible and abundant latent heat fluxes from the sea surface under the small air-sea temperature difference. We consider that the convective circulation driven by moisture should be the robust process of the convective circulation over the warm ocean: that is, over the subtropical and tropical western Pacific Ocean.
Fig. 11: Vertical cross sections of potential temperature (a), and mixing ratio of water vapor (b), with vertical velocity (contour) 3 hours later from the starting of the simulation. Contours for the updraft (solid line) and downdraft (broken line) are every 0.2 m s -1.
Fig. 12: Vertical profiles of buoyant flux (solid line), heat flux (long dashed line) and moisture flux (dotted line) 3 hours later from the starting of the simulation. Vertical profiles of the contribution of heat to the buoyant flux is nearly equal to that of heat flux. Vertical profile of the contribution of moisture to the buoyant flux is shown by "short dashed line."