A precipitation system brought heavy rainfall around Aichi Prefecture on August 28-29, 2008. Total rainfall amounts on these two days were 304.5 mm at Okazaki, 240.0 mm at Ichinomiya, and 202.0 mm at Nagoya. Hourly rainfall amounted to 146.5 mm at Okazaki from 01 to 02 Japan Standard Time (JST) on August 29. In this study, we examine the structure of the precipitation system using a new X-band polarimetric radar situated at Nagoya University and the Cloud Resolving Storm Simulator (CReSS).
From observations using the polarimetric radar, a line-shaped precipitation system aligned from south-southwest to north-northeast propagates southeastward slowly (at about 7 m/s). Southeastern and northwestern regions of the precipitation system in the lower troposphere are occupied by southeasterly and northwesterly winds, respectively; thus, the low-level convergence between these low-level winds should maintain the precipitation system (Fig. 1). The differential reflectivity (ZDR) observed by the polarimetric radar shows the existence of large raindrops whose median volume diameter (D0) is greater than 2.7 mm. The observation of the ground-based drop size distribution (DSD) using the Disdrometer also shows the existence of large raindrops whose D0 is greater than 2.3 mm; thus, many large raindrops were formed in the precipitation system and should contribute to the heavy rainfall.
Global Spectral Model (GSM) data provided by the Japan Meteorological Agency (JMA) are used as the initial and boundary conditions of a simulation with a horizontal grid resolution of 2 km using the CReSS. The simulation reproduces well the generation location, propagation direction, and speed of the precipitation system; and the maximum hourly precipitation amount around Aichi Prefecture (Fig. 2). The simulation also reproduces the low-level convergence between the southeasterly and northerly winds below the precipitation system (Fig. 3). The southeasterly is a high equivalent potential temperature (warm and moist) airmass induced by a synoptic-scale situation. On the other hand, the northerly is a relatively low equivalent potential temperature airmass, whose relative humidity is greater than 75%, induced by the downdraft in the precipitation system. Since evaporation cooling is quite limited in a moist environment, the temperature decrease at the surface is only about 2 or 3 °C. Although the propagation of the precipitation system shows a gravity-current-like structure by the relatively cool and dry airmass located on the northwestern side of the system, the propagation speed to the southeastward is slow, because the difference in density of the airmasses between both sides is quite less. The minute effect of evaporation cooling in the precipitation system is the predominant structure under the high humid environment.
Fig. 1: Vertical cross section of distributions of reflectivity (contour) and Doppler velocity (shades) in the Range Height Indicator (RHI) along the direction of 316.08° from the polarimetric radar situated at Nagoya University at 22:23 JST on August 28, 2008. Light and dark shades show southeasterly wind (away from the radar) and northwesterly wind (approaching the radar), respectively.
Fig. 2: Simulated horizontal distribution of hourly precipitation amount (mm: shade) from 2:00 to 23:00. Arrows are horizontal wind vectors at the surface at 23:00 JST. Cross and solid circle show the location of the polarimetric radar and observation range (64 km), respectively.
Fig. 3: Simulated vertical cross section of distributions of equivalent potential temperature (shade), estimated reflectivity from mixing ratio of precipitation (rain, snow, and graupel: contour), and wind vectors shown by horizontal wind along the plane and vertical wind (arrows) along the southeast-northwest direction at 23:00 JST. Nagoya University is located at the origin (O).
A heavy rainfall event was observed around northern Fukui Prefecture on July 18, 2004. The heavy rainfall event is called the Fukui Heavy Rainfall. The event is divided into two stages: the first and second stages are from 01 to 03 and from 05 to 12 JST based on data from the Radar-AMeDAS operated by the JMA. A line-shaped precipitation system is observed initially and changes to an oval-shaped one around 07 JST when the maximum precipitation amount is recorded around Fukui Prefecture.
To reproduce the precipitation system, a simulation with a horizontal grid resolution of 1 km is conducted using the CReSS. The simulation starts at 03 JST on July 18. The simulation reproduces well the distribution and amount of 3-hourly precipitation compared with the Radar-AMeDAS observations (Fig. 4). The maximum precipitation amount, exceeding 50 mm hr-1, around Fukui Prefecture and its time series are also reproduced (Fig. 5). Using the simulation result, the formation and maintenance processes of the precipitation system are examined. Sensitivity tests with modified topography and without evaporation of rain are also conducted to clarify the maintenance process.
The precipitation system is generated around the Oki Islands where high convective unstable stratification and a weak low-level convergence appear using JMA regional objective analysis (RANAL) data and the simulation result; thus, a convective precipitation system would be easily generated. The system propagates eastward over the Sea of Japan after its generation around there. A westerly wind in the lower troposphere supplies a high equivalent potential temperature airmass that maintains the precipitation system. The system only flows eastward due to westerly winds in the middle troposphere and does not show the back-building (BB) structure. When the precipitation system reaches Fukui Prefecture, the convergence zone in the lower troposphere is maintained for about one hour. The convergence zone would be maintained by the effects of mountains located along the coastal region of Fukui Prefecture, based on the result of the sensitivity experiment. As a result, new precipitation cells develop over the western tip of the lower convergence zone, i.e., in the upstream (western) side of the pre-existing cells that propagate eastward; thus, this is the BB structure. As the structure of the precipitation system changes to the BB type, precipitation cells intrude continuously over Fukui Prefecture, producing a large amount of rainfall there.
Gig. 4: Horizontal distributions of 3-hourly accumulated rainfall amounts around Fukui Prefecture between 07 and 10 JST of Radar AMeDAS (a) and the simulation (b).
Fig. 5: (a) AMeDAS observation sites in northern Fukui Prefecture (▲) and the region in which the maximum hourly rainfall amount is calculated in the simulation result (rectangle). (b) Time series of maximum hourly rainfall amount observed by AMeDAS (bar graph) and simulated by the CReSS (solid line) enclosed in the rectangle in (a).
A warm core and an eyewall cloud are characteristic structures of a typhoon. Numerical simulations using the CReSS with a high horizontal grid resolution are conducted to clarify the structure of the inner region of typhoon T0712. The purpose of this study is to clarify the contribution of the warm core to the development of the typhoon, the origin of the airmass of the warm core, and the structure of the forces that act on the airmass in the inner region of the typhoon.
The simulation reproduces well the track and pressure drop of the typhoon. The typical structures of the typhoon, such as the warm core and eyewall clouds, and dynamical characteristics such as low-level inflow, updraft in the eyewall clouds, and upper-level outflow in the typhoon, are also simulated successfully. A warm core with a large positive anomaly of potential temperature (greater than 14 K) is generated at a height of about 15 km after 24 h from the initial time (Fig. 6). The positive anomaly of potential temperature contributes to an increase in thickness between 200 and 100 hPa. Although the height of 50 hPa is almost constant with time, the increase in thickness around these levels contributes to the height of 900 hPa; thus, the pressure drops in the lower troposphere around the center of the typhoon.
A back-trajectory analysis is conducted to show the origin of the airmass in the warm core (Fig. 6). The origin of this airmass is divided into two regions. One moves inward in the lower troposphere and ascends in the eyewall clouds, and the other descends from the lower stratosphere. Figure 7 shows time series of potential temperature and equivalent potential temperature of the airmass coming from the lower troposphere. Potential temperature increases greatly between 18 and 20 h after the simulation starts, when the airmass ascends in the eyewall cloud; thus, the airmass should experience diabatic heating. Consequently, one of the origins of the warm core should be diabatic heating in the eyewall clouds.
A forward trajectory analysis is also performed to investigate the dynamical process from the lower troposphere. The analysis shows that most airmass flows outward in the upper troposphere, and only a small amount moves into the warm core. Although the pressure-gradient force mainly acts on the low-level inflow, the centrifugal force is greater than the pressure-gradient force around the inner side of the eyewall in the lower troposphere, where the radial velocity changes from inward to outward. Thus, it produces the low-level convergence below the eyewall clouds and tilts the updrafts in them in the outward direction. This mechanism is common to both outward flow in the upper troposphere and motion into the warm core. Most of the airmass is accelerated outward by the centrifugal force in the upper troposphere; on the other hand, only a small amount of airmass in which the pressure-gradient force is greater than the centrifugal force drifts into the warm core.
Fig. 6: Radius-height cross section of azimuthally averaged potential temperature anomaly at the center of the simulated typhoon (dashed line) after 24 h from the initial time. Solid lines show parts of the back trajectories. ▲ show the starting points of the back trajectory analysis.
Fig. 7: (a) Radius-height cross section of one back trajectory analysis from the lower troposphere. (b) Time series of potential temperature (solid line) and equivalent potential temperature (dotted line) of the airmass along the trajectory in (a).
The computational domain of most regional models is generally rectangular. However, the region of interest is not always rectangular but is often irregularly shaped. An arbitrary-shaped domain is more suitable for efficient computation of simulations. In the present study, we have developed a new technique to perform parallel computation of a cloud-resolving model in an arbitrary-shaped region, which is named the Tiling Domain Technique (TDT), for the CReSS model.
The arbitrary-shaped domain of parallel computation using the TDT is composed of many small rectangular domains called “domain tiles.” If a side of a domain tile is connected to a neighboring domain tile, a halo region is set along the side and data exchange is performed with the neighboring tile by Message-Passing Interface (MPI). If no neighboring domain tile is present, the side is considered a boundary of the computational domain. Each domain tile is further divided into subdomains if necessary. Another parallel computation between subdomains within the domain tile is performed using MPI. The TDT performs these two types of data exchange at different levels. We refer to this type of parallel computation as “hierarchical parallel computations.” Multiple computational domains consisting of different numbers of domain tiles in each domain are possible, as well as an isolated domain.
Using the TDT, we perform a simulation experiment of typhoon 0418 (T0418), which caused severe damage all over Japan owing to strong winds. In the experiment, 64 domain tiles are used along the track of T0418, and a one-week simulation is performed. The regional objective analysis (RANAL) data of the Japan Meteorological Agency (JMA) are used for initial and boundary conditions. The horizontal grid size is 2000 m, and the experiment is started from the initial conditions of 00UTC September 1, 2004. The central pressure of the simulated typhoon decreases rapidly and reaches that of the best-track data after about 3 days from the initial time. The simulated typhoon almost follows the JMA best track for the 7 days of the simulation period. Figure 8 shows the comparison of simulations of the T0418 using the TDT and all domain at the 6 days from the initial time. The TDT result of the distribution of precipitation amount corresponds to that using the all domain. The corners of the TDT result do not affect the simulation result compared with that using the all domain. When the simulated typhoon approaches western Japan, it causes heavy rainfall over the land. The distribution and intensity of the simulated rainfall corresponds well to the observed rainfall.
Because a typhoon moves along a long curved track, the TDT is an efficient method for high-resolution simulation of the typhoon. Domain tiles are set along the typhoon track, and total computational cost is reduced. The TDT increases the flexibility and applicability of the cloud-resolving model for many different types of weather systems. For example, a computational domain is set along the Japanese Islands, and isolated domain titles are set in the region of individual islands. Another possible application is convective activity in the tropical regions, where the nonhydrostatic effect is crucial. A nested simulation in global circulation model (GCM) using the TDT will be also useful for dynamical downscaling of typhoons.
Fig. 8: Comparison of simulations of the typhoon 0418 in September 2004 using the tiling domain technique (a) and all domains (b) at 6 days (518400 seconds) from the initial time. The gray levels are precipitation intensity (mm/h) and contour lines are sea-level pressure. The arrows are horizontal wind vector at the surface. The solid line is the best track of the typhoon 0418 provided by the Japan Meteorological Agency.
A precipitation cell is a fundamental element of precipitation systems. To clarify the general features of precipitation cells, statistical analyses are needed. In this study, a three-dimensional detection algorithm for precipitation cells has been developed. Figure 9 shows the definition of precipitation cells using reflectivity data. The central parts of the precipitation cells are recognized by connecting negative regions of the second derivative for three-dimensional reflectivity data. After connecting the central parts of precipitation cells vertically, the three-dimensional region of each cell is fixed.
The 811 precipitation cells obtained by Doppler radar in three observation projects conducted in East Asia during the Meiyu/Baiu period are detected by human eyes to validate the algorithm. More than 98% of precipitation cells are detected using the algorithm on the same observation data. Thus, the algorithm would have sufficient accuracy and would be reliable for statistical analyses of precipitation cells.
The algorithm can also detect “updraft cells” and “downdraft cells” to apply three-dimensional wind field data calculated from dual-Doppler analyses. Matching up the precipitation cells with updraft and downdraft cells marks the developing, mature, and decaying stages in their life-cycle. Figure 10 shows a schematic illustration of discriminating the stages of precipitation cells. To verify the discrimination, two precipitation cells that were observed in the dual-Doppler region through almost all of their life-cycle are chased, and stages in their life-cycle are discriminated based on wind fields by human eyes. The algorithm’s results are consistent with those from human eyes.
The discrimination of stages is conducted using dual-Doppler analyses around Miyako Island (the Southwest Islands of Japan) during the Baiu period in 2006. Precipitation cells in the developing, mature, and decaying stages are detected (183, 154, and 124, respectively). The volume of precipitation cells in the mature stage is greater than at other stages. Although the echo top height of precipitation cells in each stage is almost constant through their life-cycle, the area of precipitation cells in the mature stage is almost twice that at other stages. As a result, the volume of precipitation cells in each stage should depend on their area.
Fig. 9: Definition of precipitation cells using reflectivity data.
Fig. 10: Schematic illustration of discriminating the stages of precipitation cells.
Development of a cumulus boundary layer over land is investigated using idealized high-resolution numerical simulations with the focus on the onset of “active cumulus” and the effect of external factors: initial amount of water vapor, static stability, and surface energy partition.
A simulation with a horizontal grid resolution of 100 m is conducted using the CReSS. The parameters of idealized initial profile and land surface are set using observational data obtained in the Huaihe River Basin in China on June 20, 2004. Land surface is set to water paddy fields, which are characterized by large latent and small sensible heat fluxes. In the control experiment, the cumulus boundary layer experiences transition from a condition in which only forced cumulus exists (forced cumulus boundary layer) to a condition in which active cumulus also exist (active cumulus boundary layer) at about 1220 local standard time (LST). The active cumulus onset is almost coincident with an abrupt drop in the level of free convection (LFC: Fig. 11). The LFC drop is attributed to a decrease in the local minimum value of saturated equivalent potential temperature at the bottom height of the inversion layer, in addition to an increase of equivalent potential temperature near the land surface (Fig. 12). After the active cumulus onset, the inversion height maintains a large rate of increase because evaporation of cloud water cools and moistens the inversion layer.
Systematic sensitivity experiments show that the onset time arrives earlier with greater initial water vapor, smaller static stability, and greater evaporative efficiency through changes in the time tendency of saturated equivalent potential temperature at the bottom height of the inversion layer and equivalent potential temperature near the land surface. Since the development of the active cumulus boundary layer causes a large rate of increase of the inversion height, an earlier active cumulus onset leads to a higher inversion height. As a result, water vapor does not accumulate in the lower troposphere and is transported up to the inversion layer; thus, it contributes to the development of a deep moist layer. Therefore, this study demonstrates the interaction between features of the atmospheric environment, such as initial water vapor profile and static stability, and shallow cumulus clouds.
Fig. 11: (a) Time-height cross section of cloud fraction (shade) and time series of lifting condensation level (LCL: solid line), LFC (broken line), and level of neutral buoyancy (LNB: dotted line) calculated from mean values at a height of 45 m (lowest level of model domain). (b) Same as (a) but for mean relative humidity (shade) with mean virtual potential temperature (contours, every 1 K).
Fig. 12: Fig. 12. Profiles of mean equivalent potential temperature (solid lines) and saturated equivalent potential temperature (broken lines) at 11 LST (a) and 13 LST (b). Dotted lines indicate the mean equivalent potential temperature of air near the land surface (at a height of 45 m).