Relationships among Rainfall Distribution, Surface Wind, 2 and Precipitable Water Vapor during Heavy Rainfall 3 in Central Tokyo in Summer 4

30 The relationships between the occurrence of intense rainfall and the convergence of surface 31 winds and water vapor concentration for typical heavy-rainfall cases were examined using data 32 from July to August in 2011–2013 obtained from high-density meteorological observations in 33 Tokyo, Japan. Additionally, the temporal variations in wind convergence and water vapor between 34 days with and without heavy rainfall events were compared. Corresponding to the heavy-rainfall 35 area, the convergence of surface winds tended to increase for several tens of minutes prior to the 36 heavy rainfall. The peak of convergence was observed 10–30 min before the heavy rainfall 37 occurrence, and convergence continued to increase for approximately 30 min until the convergence 38 peak time. Around the heavy-rainfall area, the increase in the water vapor concentration index 39 coincided with the increase in convergence. From these results, by monitoring the temporal 40 variations and distributions of these parameters using a high-density observation network, it should 41 be possible to predict the occurrence of heavy rainfall rapidly and accurately.


46
In recent years, short-term heavy-rainfall events that have caused various damages such as flooding and power outages due to lightning strikes have frequently occurred in the Tokyo observation data at Tateno, provided by the University of Wyoming website 140 (http://weather.uwyo.edu/upperair/), were used. The location of Tateno is shown in Fig. 1 where T850, T700, and T500 are the temperature at 850, 700, and 500 hPa, respectively; D850 and 146 D700 are the dew point temperature at 850 and 700 hPa, respectively; TL and TP are the temperature 147 of a parcel lifted from 850 to 500 hPa and a parcel lifted from lowest of the atmosphere, Table 2 shows the percentile values of SSI, K-index, and CAPE at 0900 JST on the heavy-rainfall days and the no-heavy-rainfall days. The SSI values are arranged in descending order because 158 negative SSI values indicate unstable condition. The atmospheric stability indices were unstable on 159 the heavy-rainfall days compared to that on the no-heavy-rainfall days. For the SSI and K-index 160 values, the 25th percentiles on the heavy-rainfall days and the 75th percentiles on the 161 no-heavy-rainfall days were almost equal in value. In contrast, the value of the 25th percentile for 162 CAPE on the heavy-rainfall days was almost equal to that of the median on the no-heavy-rainfall 163 days. The differences in the atmospheric stability indices between the heavy-rainfall days and the 164 no-heavy-rainfall days were large for the SSI and K-index values and small for CAPE. Therefore, 165 the SSI and K-index values, which show a good association with the occurrence of heavy rainfall, 166 were used in this study to evaluate the atmospheric stability on heavy-rainfall days; in addition, the 167 25th percentiles on the heavy-rainfall days (SSI = 0.98; K-index = 31.8) were determined as the 168 thresholds of heavy-rainfall occurrence.
169 Figure 2 shows the relationship between SSI and K-index for July to August in 2011-2013. On 170 heavy-rainfall days, SSI and K-index were both distributed on the unstable side at both 0900 JST 171 (Fig. 2a) and 2100 JST (Fig. 2b). Values of SSI < 0.98 and K-index ≥ 31.8 were generally observed 172 on heavy-rainfall days; however, 13 and 12 of the heavy-rainfall days were outside of these 173 thresholds at 0900 JST and 2100 JST, respectively. It was thought that there are many cases where 174 the time difference between the occurrence of heavy rainfall and the upper-air observation is large 175 (Table 1). In addition, the upper-air data after heavy rainfall were not available when we actually performed the prediction. Therefore, to predict the occurrence of heavy rainfall, it is necessary to 177 use atmospheric stability at a time close to that of the rainfall event (within a few hours) obtained 178 from upper-air data with high temporal resolution (e.g., regional objective analysis data). However,

179
to evaluate the predictability of heavy rainfall occurrence with these atmospheric stability indices, 180 the results obtained using the daily minimum SSI and the daily maximum K-index are illustrated in in no-heavy-rainfall days that satisfied the criteria.

Relationships between heavy rainfall and distributions of wind convergence and water vapor 196
Heavy rainfall, which caused flood damage and power outages, occurred in the western part of 197 the Tokyo Metropolitan area on 12 August 2013. According to the weather chart on this day (Fig. 198 3a), a summer-type pressure pattern was present and disturbances such as typhoons and fronts did than 10 mm of hourly rainfall increases rapidly as PWV exceeds 60 mm. Therefore, it was thought that the amount of water vapor in the atmosphere was comparatively large. However, there was no 234 area in which PWV was remarkably increased relative to 40 min previously, and the same tendency 235 could also be observed at 1700 JST ( Fig. 6b). At 1720 JST (Fig. 6c), PWV was decreased in the  index was not apparent. In addition, the absolute values of the divergence change rate were small 307 and increasing convergence that continued for several tens of minutes was not observed.

308
As mentioned above, the increase in surface wind convergence and WVC index preceding a heavy rainfall was observed only on a heavy-rainfall day. For evaluating the frequency of 310 occurrence of these features, the temporal variations in wind convergence and WVC index on 311 heavy-rainfall days were examined. Figure 10 shows the percentile values from 60 min before to 30 312 min after the occurrence of rainfall on the heavy-rainfall days and the no-heavy-rainfall days. The rainfall was observed when more than 5 mm of 10-min rainfall was observed at 1010-2200 JST.

317
The occurrence of rainfall was defined as the time when more than 1 mm of 10-min rainfall was 318 observed for the first time, and the points wherein the maximum rainfall was observed after 60 min 319 from the occurrence or wherein there was an interruption of rainfall (less than 1 mm of 10-min 320 rainfall) from the occurrence to the maximum time were excluded. From these selected points, we For divergence (Fig. 10a), convergence was observed in approximately three quarters of cases 327 before the occurrence of rainfall, and the decrease in the 25th and the 10th percentiles was observed than that of wind convergence. This result suggests that the temporal and spatial scales of the WVC 366 index are larger than that of the surface wind convergence obtained from the high-density data in 367 this study. However, it is thought that the WVC index and convergence of surface winds represent 368 the concentration of water vapor at each spatial scale.

369
In addition, the increase in surface wind convergence and WVC index preceding the heavy 370 rainfall that was rarely observed on the no-heavy-rainfall days was observed in more than a quarter 371 of cases on the heavy-rainfall days. Although it is necessary to examine the appropriate combination 372 of thresholds to predict occurrences of heavy rainfall, the increase in the WVC index and wind 373 convergence as well as atmospheric stability can be used for prediction of heavy-rainfall occurrence.

374
Therefore, by monitoring the temporal variations and distributions of these parameters using a 375 high-density observation network, we consider that it is possible to predict occurrences of heavy 376 rainfall rapidly and accurately.       (b) 20-min divergence change rate (× 10 −5 s −1 min −1 ) Elapsed time from the occurrence of rainfall (min) (c) WVC index (10 3 × mm km −2 ) (d) Deviation of the WVC index relative to 40 min previously (10 3 × mm km −2 ) Elapsed time from the occurrence of rainfall (min) Elapsed time from the occurrence of rainfall (min) Elapsed time from the occurrence of rainfall (min) No-heavyrainfall days No-heavyrainfall days No-heavyrainfall days No-heavyrainfall days