The Tokyo Metropolitan Area Convection Study for Extreme Weather Resilient Cities (TOMACS) began as a Japanese domestic research project in 2010 and aimed to elucidate the mechanisms behind local high-impact weather (LHIW) in urban areas, to improve forecasting techniques for LHIW, and to provide high-resolution weather information to end-users (local governments, private companies, and the general public) through social experiments. Since 2013, the project has been expanded as an international Research and Development Project (RDP) of the World Weather Research Programme (WWRP) of the World Meteorological Organization (WMO). Through this project, the following results were obtained: 1) observation data for LHIW around Tokyo were recorded using a dense network of X-band radars, a C-band polarimetric radar, a Ku-band fast-scanning radar, coherent Doppler lidars, and the Global Navigation Satellite System; 2) quantitative precipitation estimation algorithms for X-band polarimetric radars have been developed as part of an international collaboration; 3) convection initiation by the interaction of sea breezes and urban impacts on the occurrence of heavy precipitation around Tokyo were elucidated by a dense observation network, high-resolution numerical simulations, and different urban surface models; 4) an “imminent” nowcast system based on the vertically integrated liquid water derived from the X-band polarimetric radar network has been developed; 5) assimilation methods for data from advanced observation instruments such as coherent Doppler lidars and polarimetric radars were developed; and 6) public use of high-resolution radar data were promoted through the social experiments.
A dense network of X-band polarimetric radars and other instruments was deployed to study and forecast local high-impact weather around Tokyo.
Urban areas are potentially vulnerable to natural disasters due to their higher concentrations of population and properties. Moreover, it is believed natural disasters in urban areas will be further intensified due to global warming. The Intergovernmental Panel on Climate Change warns that “many global risks of climate change are concentrated in urban areas (medium confidence)” (IPCC 2014). In addition, the distribution of buildings and obstacles, extensive use of impervious materials, reduction of vegetation, release of anthropogenic heat and pollutants by human activities, and street geometry also affect the local climate (Baklanov et al. 2018). Therefore, the occurrence of unprecedented severe weather is anticipated in urban areas.
Under these conditions, the 17th World Meteorological Congress made a resolution to “establish a cross-cutting urban focus” (WMO 2015), and several international research projects studying high-impact weather in urban areas have been conducted, such as the Study of Urban Impacts on Rainfall and Fog/Haze (SURF) in Beijing (Liang et al. 2018), Pan Am in Toronto (Joe et al. 2018), and the Weather Information Service Engine (WISE) in Seoul (Choi et al. 2013). The Tokyo Metropolitan Area Convection Study for Extreme Weather Resilient Cities (TOMACS) is one of these international research projects aimed at studying local high-impact weather (LHIW) in urban areas.
The flood risk in the Tokyo metropolitan area is extremely high, since more than a million people live in areas lower than sea level, and the frequency of extreme precipitation around Tokyo has been increasing in recent years (Fujibe 2015). From the viewpoint of mesoscale meteorology, on the other hand, the Tokyo metropolitan area is an attractive location for research, since convective clouds are frequently initiated by the interaction of sea breezes and synoptic disturbances, and by the heated topography.
TOMACS originally began as a domestic research project in Japan as part of a collaboration between 25 organizations including universities, research institutes, local governments, and private companies. The original goals of TOMACS were as follows:
elucidation of the mechanisms behind LHIW in urban areas (e.g., torrential rain, flash floods, strong winds, and lightning),
improving nowcasting and forecasting techniques for LHIW, and
provision of high-resolution weather information to end users through social experiments.
Because the scientific outcomes obtained in TOMACS would be useful to predict and understand LHIW not just in Japan but also in urban areas around the world, TOMACS was proposed as a Research and Development Project (RDP) of the World Weather Research Programme (WWRP) of the World Meteorological Organization (WMO) and was endorsed as an RDP in July 2013 (Nakatani et al. 2015). In the present report, we briefly summarize the results of the observations, numerical studies, forecasts and social experiments conducted in TOMACS.
OBSERVATIONS IN TOMACS.
Tokyo has a humid subtropical climate according to the Köppen climate classification. It experiences frequent convective storms which are initiated by the interactions between easterly and southerly sea breezes (Saito et al. 2018), and by the heated mountains and basins (Misumi et al. 2018; Sano and Oishi 2018). Such convective storms sometimes cause severe disasters. For example, the torrential rain caused by the meso-γ-scale disturbance on 5 August 2008 killed five people who were repairing a sewage tunnel (Kato and Maki 2009; Kim et al. 2012; Ishihara 2013; Saito et al. 2017). A dense meteorological observation network is necessary for observing the development of such convective storms.
Figure 1a shows the observation network used in TOMACS. The operational instruments of the Automated Meteorological Data Acquisition System (AMeDAS) of the Japan Meteorological Agency (JMA), the Global Navigation Satellite System (GNSS) Earth Observation Network (GEONET) of the Geographical Survey Institute, three C-band Doppler radars and two Doppler lidars of JMA, and five X-band polarimetric radars belonging to the Ministry of Land, Infrastructure and Transport (Fig. 1a) were used. The JMA provides a 1-km mesh rainfall map at 5-min intervals using their operational C-band radars. In addition to the above operational instruments, radars owned by various institutions were utilized to obtain high-resolution data in and around convective storms. Because X-band radars have shorter wavelengths and smaller antenna apertures than C-band or S-band radars, a dense radar network can be easily constructed at a low cost. In TOMACS, seven X-band radars were networked to obtain continuous volume-scan data over Tokyo with a horizontal resolution of 500 m in Cartesian grid at 5-min intervals. A C-band polarimetric radar was also deployed to obtain polarimetric parameters over a wider area. Doppler lidars, which are capable of determining the Doppler velocity in clear air, and a fast-scanning Ku-band Doppler radar, which can obtain volume scan data at 1-min intervals, were also deployed. The specifications of these instruments are as follows.
The Meteorological Research Institute (MRI) Advanced C-Band Solid-State Polarimetric Radar (MACS-POL) was installed at Tsukuba City (36.0550°N, 149.1253°E), located approximately 50 km northeast of Tokyo. The MACS-POL covers an area with a range of 230 km and 150-m resolution in the radial direction and can obtain the radar reflectivity (Z), Doppler velocity (VR), differential radar reflectivity (ZDR), differential propagation phase (ϕDP), and copolar correlation coefficient (ρHV). The minimum detectable reflectivity by the MACS-POL is about −5 dBZ at a range of 50 km.
The X-NET (Maki et al. 2012) is a scientific-purpose radar network consisting of seven X-band Doppler radars belonging to the National Research Institute for Disaster Resilience (NIED), National Defense Academy, Yamanashi University, Chuo University, Central Research Institute of Electric Power Industry, and the Japan Weather Association. Each radar conducts volumetric scans at 5-min intervals. Five of the X-NET radars are polarimetric ones, and can perform quantitative precipitation estimation (QPE) with a higher accuracy than the operational C-band radars. Though an X-band radar tends to suffer from strong rainfall attenuation, the attenuated areas are covered by other radars in the network. All the radar data are collected by NIED in real time to estimate the three-dimensional wind field using a multiple Doppler radar analysis (Maesaka et al. 2007), to create a three-dimensional composite of the radar products (Kim and Maki 2012), and to synthesize the rainfall intensity with that from the operational C-band radars to obtain a rainfall map over a wider area (Kato et al. 2012). The sensitivity of the X-band radars is about 10 dBZ at a range of 60 km.
MRI installed a high-resolution, fast-scanning Ku-band Doppler radar to observe the fine structure and evolution of convective clouds developing in Tokyo (the pink circles in Figs. 1a and 1b). The Ku-band radar consists of two Luneburg lenses and has the ability to obtain the three-dimensional radar reflectivity and Doppler velocity with a resolution of 2.38 m in the radial direction at 1-min intervals (Sato et al. 2013). It is designed to detect radar reflectivity of 20 dBZ at its maximum observation range (about 20 km).
Doppler lidars are a type of laser radars for transmitting infrared rays to detect the Doppler velocity in clear air. In TOMACS, four Doppler lidars were used (Fig. 1b); a three-dimensional scanning coherent Doppler lidar (3D-CDL) owned by Hokkaido University (Fujiyoshi et al. 2006; Fujiwara et al. 2011; Yagi et al. 2017), a scanning Doppler lidar owned by the National Institute of Communication Technology (NICT; Iwai et al. 2013), and two operational Doppler lidars owned by JMA at Tokyo International Airport. The maximum observation ranges of these lidars are 4.4 km for the 3D-CDL, 6.0 km for the NICT lidar, and 10 km for the JMA lidars, although these ranges can vary with the aerosol concentration on the observation days (Iwai et al. 2018). In addition, a dense network of automated weather stations (AWS) at 3-km intervals set up in TOMACS (Fig. 1b) was used to observe the convergence of low-level air before and after the formation of severe storms (Seto et al. 2018) and the development of misocyclones (Saito et al. 2013). AWS were installed in open areas such as the rooftops of school buildings at a height between 20 and 25 m above the ground level to avoid interference from adjacent structures.
Operational upper air soundings are conducted twice a day at Tateno, which is located about 50 km northeast of Tokyo but are not sufficient to detect the diurnal variation of environmental conditions around Tokyo. Therefore, four additional radiosonde stations were set up around Tokyo (green dots in Fig. 1a) and soundings were conducted at 3-h intervals from 0830 to 2030 JST (Japan standard time; UTC + 9 h) for 34 days during the intensive observation period from 2011 to 2013 (Seino et al. 2018b; Sugawara et al. 2018). Because radiosondes falling in urban areas can cause problems on roads or railroads, the radiosondes were launched mainly in autumn when strong westerlies carried the radiosondes away from Tokyo. In 2013, the Ryofu-maru, a research vessel of JMA, was located off the coast near Tokyo and performed special upper-air soundings. In addition, the delay of signals from GNSS was analyzed to estimate the water vapor amount. In TOMACS, five GNSS receivers were set up in addition to the GEONET to observe the water vapor field in detail (Fig. 1b).
Various algorithms were developed to analyze data from these instruments. For MACS-POL, a technique to detect hazardous precipitation cells (Adachi et al. 2013) and an algorithm to estimate raindrop size distributions for accurate QPE (Adachi et al. 2015) were developed. In addition, an estimation method for wind using multiple C-band Doppler radars was studied by MRI (Yamada 2013). To estimate the distribution of precipitable water at the meso-γ scale, methods to analyze the slant path delay of signals from GNSS were investigated (Shoji 2013; Shoji et al. 2014, 2015). It was shown that the “water vapor concentration index” derived from the slant path delay was effective for forecasting of local heavy rainfall (Shoji 2013; Seto et al. 2018).
The dense observation network observed several tornadic storms during TOMACS. Figure 2 shows the evolution of the radar reflectivity and Doppler velocity of a tornadic storm in Tokyo observed by the fast-scanning Ku-band Doppler radar on 23 July 2013. Over about two minutes from 1555:26 to 1557:34 JST, a protruding radar echo appeared accompanied by a pair of positive and negative Doppler velocities. The tornado was generated at the cross mark in Fig. 2 around 1600 JST. Such quick evolution of convective echoes cannot be detected by operational C-band radars which conducts volume scans at 10-min intervals. Other tornadoes that struck Tsukuba City on 6 May 2012 were observed by MACS-POL and X-NET, and characteristics of the polarimetric parameters were revealed, such as the existence of a ZDR arc, ZDR column, and KDP (the specific differential phase) column in the tornadic storm, and the decrease of rhv and ZDR near the ground after the tornado generation (Yamauchi et al. 2013; Suzuki et al. 2018).
Elucidation of the initiation process of convective clouds was also an important topic in TOMACS since it is essential for improving forecasts of LHIW. Around Tokyo, sea-breeze fronts can trigger convective storms. Such sea-breeze fronts were observed as non-precipitation echoes by operational C-band radars whose sensitivity is about 0 dBZ (Iwai et al. 2018), while detection of such sea-breeze fronts was difficult using the X-band radars probably due to their low power. Using the Doppler velocity patterns of Doppler lidars, Iwai et al. (2018) tried to detect sea-breeze fronts using the “perturbation Doppler velocity,” defined as deviations from the mean horizontal wind velocity calculated by the velocity–azimuth display method. However, a forecast method for convective storm initiation using such information has not been established and remains a subject for future work.
INTERNATIONAL COOPERATION FOR THE ANALYSIS OF RADAR DATA.
The use of the radar observations obtained in TOMACS was promoted to encourage international cooperation. This international cooperation was intended to gather advice from experts who had already operated advanced radar systems and to provide information to the countries that are considering introducing new radars.
The Collaborative Adaptive Sensing of the Atmosphere (CASA; Chandrasekar et al. 2018) developed in Dallas–Fort Worth (DFW) provided a useful reference for the construction of X-NET around Tokyo. The CASA-DFW demonstration network is composed of the high-resolution X-band radar network and the S-band radar system of the National Weather Service. Based on these radars, CASA has developed an end-to-end warning system that includes sensors, software architecture, products, data dissemination and visualization, and user decision-making modules. The QPE algorithm of X-NET was developed under joint research between Colorado State University and NIED as academic partners of CASA (Park et al. 2005). The QPE algorithm for mountain regions, where radar beams are partially blocked by topography, was developed collaboratively by Pukyong National University (PKNU) and NIED (P. C. et al. 2013). Collaborative research between PKNU and NIED was also conducted for the S-band polarimetric radar observations on Jeju Island (Lee at al. 2012; Kang et al. 2018).
To obtain rainfall data at 500-m intervals, it is necessary to arrange X-band radars with a beamwidth of 1° at intervals at least 30-km intervals since the radar beam spreads with distance. Because X-band radars have a smaller antenna aperture than C-band or S-band radars, they are suitable for constructing such a dense radar network at a relatively lower cost. Traditional X-band radars had a problem with attenuation of radar signals, but recent polarimetric radars can correct for this attenuation using KDP (Kim et al. 2012). The advantages of X-band radars have been widely recognized and tested in the San Francisco Bay Area (Cifelli et al. 2018), and operationally in São Paulo (Pereira Filho 2012) and in Paris (Schertzer et al. 2014).
CASE STUDIES ON THE IMPACT OF THE URBAN ENVIRONMENT ON THE OCCURRENCE OF HEAVY PRECIPITATION.
According to the recent studies, precipitation is affected by the urban environment such as the low-level air convergence due to the large roughness, the vertical destabilization due to the heat island effect, the change of the movement and structure of storms upwind of cities, enhanced local circulations, and the emission of aerosols (Bélair et al. 2018). Around Tokyo, Fujibe et al. (2009) showed that “no preceding precipitation,” defined as 6-h precipitation not preceded by ≥1 mm precipitation for the preceding 6 h (this criterion was introduced to capture summer afternoon showers), showed an increasing trend in the warm season. Seino et al. (2018a) performed numerical simulations with and without the urban canopy scheme over 8 years and showed that the precipitation in Tokyo increased by 10% when urban effects were included. However, it is not clear how the urban environment around Tokyo affects the individual heavy rainfall events. To address this problem, several numerical simulations incorporating different urban surface schemes were conducted in TOMACS. It should be noted that the experiments described below refer to a single case studied with multiple atmospheric models and surface descriptions at various resolution, with the aim of identifying the most relevant physical processes (urban effects, local atmospheric flows, or otherwise) influencing the precipitation in the Tokyo area.
During the downpour around Tokyo on 26 August 2011 more than 100 mm of precipitation was recorded (Fig. 3) and resulted in flooding. This case was selected as a target to study the urban effects of precipitation. Figure 4 shows schematics of the rainfall event based on the numerical simulation by Saito et al. (2018). In the morning (Fig. 4a), two sea breezes from the east and the south were formed by the heat contrast between the land and the ocean. The easterly sea breeze intruded into the land, while the southerly sea breeze was blocked by the northerly synoptic-scale wind. In the afternoon (Fig. 4b), the easterly sea breeze reached Tokyo and collided with the southerly sea breeze, and strong convective clouds developed. The initiation of convective clouds in Tokyo could not be forecast by the mesoscale model of the JMA.
Bélair et al. (2018), Pereira Filho et al. (2018), Seino et al. (2018b), and Saito et al. (2018) performed numerical simulations of this event using different cloud-resolving models incorporating various urban surface schemes (Table 1). Bélair et al. (2018) used the Global Environmental Multiscale (GEM) model (Côté et al. 1998) incorporating the Town Energy Balance (TEB) scheme (Masson 2000). When the surface conditions were changed from TEB to tall grass, the area with a precipitation rate larger than 10 mm h–1 decreased (Fig. 5). The intense precipitation in the urban environment seemed more related to the enhancement of lateral inflow of low-level moist static energy from Tokyo Bay than to the augmented surface fluxes of heat and humidity from the city itself.
Pereira Filho et al. (2018) simulated the same event using the Advanced Regional Prediction System (ARPS; Xue et al. 2000) incorporating the Tropical Town Energy Budget (T-TEB; Karam et al. 2010) model as the urban surface scheme, and compared the results when the T-TEB was changed to a semi-desert. In the results, the precipitation significantly increased in the urban area when T-TEB was used (Fig. 6a). Seino et al. (2018b) simulated the same event with the JMA nonhydrostatic model (NHM; Saito et al. 2006) with the square prism urban canopy (SPUC) scheme (Aoyagi and Seino 2011), and compared the results with the case where the urban surface is treated as a concrete slab. As a result, they found that the precipitation increased with the inclusion of the urban morphology (Fig. 6b).
Although these simulations were performed with different cloud-resolving models incorporating various urban surface schemes, the conclusions were similar; the precipitation increased under the urban surface schemes, and this was not due to the increase of sensible and latent heat flux from the surface but due to the intensification of moist sea breeze driven by an enhanced temperature contrast between the urbanized land area and the ocean. The same event was also simulated by Saito et al. (2018) from the viewpoint of a forecast experiment, and they showed that the low-level cloudiness, which modulated the solar radiation, strongly affected the location and amount of precipitation. This result also suggested that the temperature contrast between land and ocean, which was increased by solar radiation, influenced the precipitation amount around Tokyo.
NOWCAST AND DATA ASSIMILATION.
The observation data obtained in TOMACS were applied to various forecast experiments including nowcast and numerical prediction. Here, we describe some of the results.
It was expected that the accuracy of the precipitation nowcast would be greatly improved by using the dense X-band radar network around Tokyo. Hirano and Maki (2018) developed an “imminent nowcast” based on the vertically integrated liquid-water content (VIL) derived from the X-band radar network. This nowcast system is an application of “RadVIL” proposed by Boudevillain et al. (2006), which forecasts the VIL using the following equation:
where d/dt is the Lagrangian time derivative, S(t) is the source term for the VIL calculated from the time variation during a period Δt as
and P(t) is the precipitation intensity obtained from radar observations. In this model a linear relationship between VIL(t) and P(t) is assumed as
where the parameters τ and w were estimated by the least squares method using VIL(t) and P(t) derived from the radar observations. The VIL forecast by (1) was converted to P every 5 min using (3). Figure 7 compares the rainfall accumulated from a given time to 10-min ahead using the VIL-based nowcast (VIL-NC), a simple extrapolation of radar echoes at the surface (RR-NC), and the true values observed by X-band radars (Reference). In all three cases, the estimates produced by VIL-NC are superior to those for RR-NC, especially at the onset of precipitation. This is because the VIL includes the information on radar echoes occurring in the upper levels, which is not included in RR-NC. This superior accuracy of VIL-NC is useful for nowcasts at lead times shorter than 10 min. Though the lead time is very short, this system may improve the accuracy of imminent heavy rainfall alerts provided for public activities and emergency alarms.
P. C. et al. (2015) applied the Short Term Ensemble Prediction System (STEPS; Seed et al. 2013) to the rainfall events that occurred during TOMACS to compare the forecast accuracy with the JMA nowcast. The essential difference between STEPS and the JMA nowcast is that the former is an ensemble nowcast that takes into account the errors in observations and advection schemes, while the latter is a deterministic nowcast based on one initial condition. They suggested that STEPS reduced the large errors sometimes found in the JMA nowcast by averaging ensembles.
The tornadic storm that struck Tsukuba city on 6 May 2012 during TOMACS (Kobayashi and Yamaji 2013; Yamauchi et al. 2013; Suzuki et al. 2018) provided a good target for forecast experiments. Observation data obtained from TOMACS were provided to the High Performance Computing Infrastructure (HPCI) Strategic Program for Innovative Research Field3 “Ultra-high Precision Meso-Scale Weather Prediction” (www.jamstec.go.jp/hpci-sp/en/strategy/mswp.html) and the Flagship 2020 Project “social and scientific priority issues (Theme 4) to be tackled by using post K computer.” Seko et al. (2015) suggested the possibility of probabilistic forecasting of mesocyclones through numerical simulations of the tornadic storm using a local ensemble transform Kalman filter (LETKF). They stated that an ensemble forecast of tornadic storms had the advantage that it can provide several possible scenarios and reduce the miss rate compared to a deterministic forecast. Yokota et al. (2016) conducted ensemble forecast experiments with a horizontal resolution of 350 m for the tornadic storm by assimilating VR, KDP, and Z from MACS-POL and the data from the surface weather stations. In addition, Yokota et al. (2018) used NHM with a 50-m horizontal grid and performed 33-member ensemble forecasts for the tornadic storm. They successfully simulated tornadoes and identified factors that were important for forecasting tornadoes.
Assimilating Doppler lidar data has great potential to improve forecast accuracy since such data can provide the wind velocity in clear air. The inflow to the convective storm that caused severe flooding in Tokyo on 5 July 2010 was observed by one of the Doppler lidars deployed in TOMACS. Kawabata et al. (2014) showed that the forecast accuracy for heavy rainfall was improved by assimilating data from the Doppler lidar in order to correct the inflow into the precipitation system (Fig. 8).
Several fundamental studies were conducted with the aim of developing new nowcast models and a data assimilation method. A detection method for the precipitation core was developed for three-dimensional tracking of radar echoes (Shusse et al. 2015). Seko et al. (2017) obtained preliminary results on the assimilation of temporal variations of the refractivity of radar beams, which reflect small-scale water vapor variations, obtained from phase data for radar beams from the JMA operational C-band Doppler radar. By assimilating the temporal variations of refractivity, the forecast accuracy for precipitation was improved through correction of the water-vapor field, although it was only assessed for a case study. The assimilation of polarimetric parameters obtained from C-band dual polarization radars was studied as part of a collaboration between the University of Hohenheim and MRI. After testing several possible relationships between the mixing ratio of rainwater and polarimetric parameters, they found that the relationship using only KDP was the most suitable for the data assimilation (Kawabata et al. 2018).
In TOMACS, social experiments were performed to promote the use of the high-resolution radar data and forecast information for disaster-mitigation activities. Several local governments and private companies, and many citizens participated in the social experiments. Usually, the Tokyo Fire Department prepares for rescue operations when meteorological warnings are issued by the JMA. As a social experiment in TOMACS, they used the precipitation information from X-NET, the 500-m mesh rainfall map updated every 5 min overlaid with the locations of the local fire departments, in their preparation for rescuing people from flood disasters. In the heavy rainfall on 5 July 2010, they ordered firefighters to prepare based on the rainfall information from X-NET. According to the questionnaire conducted after the rainfall event, firefighters highly valued the information from X-NET since they could get almost real-time information on the precipitation intensity at any location (Yoshii et al. 2012).
In Edogawa Ward in Tokyo, the local government provides information to citizens about their facilities using a web-based geographical information system (Web-GIS). To accustom staff to using high-resolution radar data, rainfall derived from the X-NET was provided to the local government using the web Map Service so that they can easily overlay the rainfall distribution on the map. This information was then used not only by local government staff, but also by teachers in public schools and kindergartens, and voluntary disaster prevention organizations in Edogawa Ward. We conducted a questionnaire survey among the users several times to improve the information system. In the survey, some users asked us to improve the system to clearly show whether floods would occur or not. Therefore, we changed the system to display not only rainfall intensity but also 1-h cumulative rainfall to easily understand whether it exceeded 50 mm, which is the criteria for a local flood in Tokyo. Through these processes, the users became proficient at interpreting this information.
In many railroad companies in Japan, train operations have been guided by the information from rain gauges installed at each station. However, rain gauges sometimes cannot capture heavy rainfall occurring between stations. Therefore, participants from railroad companies examined the possibility of using the X-band radar information for their operational management of trains. A construction company tried to use wind information derived from multiple Doppler radar analysis for safety management at construction sites. When strong winds were anticipated, we asked the staff to watch for radar echoes accompanied by a large Doppler velocity and to issue warnings to the construction sites when the echoes approached. They confirmed that it was useful for providing an early warning at their construction sites.
To investigate the perception of rainfall information from X-NET among the general public, questionnaires and interviews were conducted with many participants after providing them with rainfall information from X-NET using mobile phones, tablet personal computers, and digital photo frames. As a result, it was confirmed that a need for high-resolution rainfall information exists in the general public, for use in their daily lives, such as deciding the time to go shopping, and washing and drying their clothes.
As a unique social experiment, Sekiya et al. (2014) attempted to display the rainfall distribution determined by X-NET on a giant screen built at a train station in Tokyo (Fig. 9) and interviewed people walking near the screen. According to the results of the interview, 17.6% of people could not understand what was being shown on the screen. However, 89.5% of people answered that they want to use the rainfall information displayed on the screen in future. It is likely that real-time rainfall distribution would be useful for people traveling to other places by train.
In summary, the social experiments suggested that 1) there is a need for high-resolution rainfall data for the general public, 2) rainfall information is used not only for disaster-mitigation activities but also for the improvement of their daily lives, and 3) by overlaying radar information and web-based geographical information, users can easily interpret radar information.
Intensive observations were conducted in TOMACS using an X-band polarimetric radar network, a C-band polarimetric radar, Doppler lidars, a fast-scanning Ku-band radar, the GNSS observation network, upper-air soundings, and automated weather stations. Through the observations, a large amount of data concerning LHIW were collected, including for the tornado that occurred in Tsukuba on 6 May 2012 and the downpour in Tokyo on 26 August 2011. Moreover, studies of analytical methods for polarimetric radar networks, simulations of convection initiation by the interaction of sea breezes and urban impacts on the occurrence of heavy rainfall, the development of nowcasting and data assimilation methods, and social experiments were conducted as part of an international collaboration. The results indicated that a high-resolution radar network and fast-scanning radars were very useful for monitoring LHIW in urban areas where small-scale phenomena have significant impacts on people’s lives. Construction of a fast-scanning radar network and its application to nowcast and data assimilation are subjects for future work.
Some difficulties were also faced in the project. The launch of radiosondes was greatly restricted when the ambient wind was weak since they might fall and disturb the dense railroad or highway network. Due to this restriction, we could not obtain enough data on convective storms developing under hot and calm conditions. In the social experiment, real-time and nowcast information concerning rainfall was favorably accepted, but it was unknown whether probabilistic information from ensemble forecasts could also be understood by the general public or not. These are also subjects for future work.
Extreme weather in urban areas is a common problem around the world. We believe that the achievements in TOMACS will be useful to many researches around the world for studying high-impact weather in cities. The latest results from TOMACS have been published as a special issue of the Journal of the Meteorological Society of Japan (volume 96A, in English), and the Meteorological Research Note “Extreme Weather in Cities” (in Japanese).
We are grateful to Masahito Ishihara, Takahisa Kobayashi, and Yoshinori Yamada of the Meteorological Research Institute; Isao Nakamura of Toyo University; Naoya Sekiya of the University of Tokyo; Alan Seed of the Bureau of Meteorology in Australia; Volker Wulfmeyer of the University of Hohenheim; Daniel Schertzer of the Ecole des Ponts ParisTech; and all other participants of TOMACS for their great contributions. We would like to extend our thanks to Paul Joe of the Environment and Climate Change Canada and Jeanette Onvlee of the Royal Netherlands Meteorological Institute for their support as chairs of the Nowcasting Mesoscale Research Working Group of the World Weather Research Programme. We greatly appreciate helpful comments from three anonymous reviewers. TOMACS was funded by the Japan Science and Technology Agency (JST) as part of the “Social System Reformation Program for Adaptation to Climate Change.”