Flash Flood due to Local and Intensive Rainfall in an Alpine Catchment
Shusuke MIYATA,1,* Masaharu FUJITA,1 Takuji TERATANI2 and Hirofumi TSUJIMOTO21 Disaster Prevention Research Institute, Kyoto University, Japan 2 Japan Weather Association, Japan
*Corresponding author. E-mail: miyata.shusuke.2e@kyoto-u.ac.jp
INTRODUCTION
In mountainous areas, human casualty is caused by flash flood, which is defined as rapid increases in stream water level. Previous studies have found several potential factors of flash floods. Although local and intensive rainstorm can be one of the factors causing flash flood, few studies have addressed the relationship between local rainfall and flash flood in mountainous catchments based on field monitoring. In this study, we examined effects of spatial distributions of rainfall on occurrences of flash flood in an alpine catchment by combining field observations, application of radar rainfall data, and numerical simulations.
STUDY SITE AND FIELD OBSERVATION
Field observations were conducted in Sugoroku Catchment, which is an alpine catchment in central Japan (Fig. 1). The catchment area is 60.9 km2 and the elevation ranged from 1207 to 2897 m. We monitored stream water levels at four gauging stations along the main stream (Fig. 1). Stream water from a major sub-catchment, Ogura-dani, flows into the main stream between the two gauging stations S1 and S2. Precipitation was observed at three locations using rain-gauges. Spatial distributions of precipitation were obtained from a C-band radar rain-gauge. Spatial resolution of the radar rainfall data was approximately 1 x 1 km and time resolution was 5 min.
RESULTS OF FIELD OBSERVATION
Hourly precipitations of the C-band radar generally agreed well with those of the rain- gauges, suggesting that the C-band radar is
applicable to the study site with very large relief geomorphology. During a heavy storm, water levels at the four gauging stations increased and diminished rapidly following changes in rainfall intensity. Increase in stream discharge at S1 was also rapid during a short- duration event; whereas increase in discharge at S2 was relatively moderate (Fig. 2). Intensive rainfall in the Ogura-dani sub-catchment, which is located between S1 and S2, enhanced the
sudden increase of stream discharge at S1. Fig. 1 Location and map of study catchment and monitoring stations
HYDROLOGICAL MODEL AND SIMULATION CASE
We applied numerical simulations to examine hydrological processes of the sudden increase in stream discharge during the intensive rainfall. The study catchment was divided into hillslope and stream sections. The hillslope section consisted of 10 m grids, in which balances of water was calculated for ground surface and A- and B- layers. Discharge from grids of the downslope end was used as input data in the simulation of stream section. Water level and discharge in the stream channels were calculated using one dimensional unsteady flow analysis. Hydrological simulations were conducted for the short and intensive storm event (36.8 mm of total precipitation at S1) under two cases. In Case 1, precipitation at a mesh of downstream area was applied to the entire catchment. Radar rainfall data was applied in Case 2, that is, spatial distribution of precipitation was involved.
SIMULATION RESULTS AND DISCUSSION
In Case 2, simulated timing and amount of sudden increase of discharge at S1 agreed with the observed results (Fig. 2). Simulation results of Case 2 also reproduced runoff characteristics, which was retained stream discharge was followed by the sudden increase of discharge. In contrast, simulated stream discharge increased moderately and longer in Case 1, comparing with the observed discharge. In Case 2, longitudinal profiles of simulated stream flow in the main channel and the Ogura-dani sub-catchment showed that floods from the main channel and the sub-catchment contributed the rapid increases of stream discharge at S1.
Because of thin soil layers in the upstream areas (elevation > 2500 m), another rapid increase of stream water delivered from upstream was found in the main channel. These simulation results indicate that the rapid increase of stream flow at S1 was associated with the simultaneous arrival of the two concentrated floods from the Ogura-dani and the main channel. Comparison of the simulated discharge between Cases 1 and 2 suggests the importance of spatial distribution of rainfall on
prediction of flash flood disasters.
CONCLUSIONS
To examine effects of spatial distribution of rainfall on flash flood, we conducted field observations of stream water level and rainfall in an incised alpine catchment and applied a distributed hydrological simulation model. The combination of field observations and numerical simulation suggested that hydrological processes of a sudden increase in stream water level during a short and intensive storm event were associated with local precipitation and thin soil layers in the upstream areas of the catchment.
Keywords: radar rain gauge, field observation, numerical simulation, intensive rainfall
Fig. 2 Observed and simulated water levels during a short and intensive event
Stream discharge [m3/s]
0 10 20
30 Observed
Case 1 Case 2 5 min rainfall [mm/h]
0 10 20 30
2012/08/18
10:00 14:00 18:00 22:00 Water height [m]
0.3 0.4 0.5 0.6 0.7 0.8
Case 1 Case 2