C3-3428

thC

onfe

renc

e on

Hyd

rolo

gy -

94th

AMS

Annu

al M

eetin

g (F

ebru

ary

2014

, Atla

nta,

GA)

Studies on Mean Areal Rain Rate using Dual-Polarization X-Band Radar over a Small River Basin, Japan

Kohin Hirano(hirano@bosai.go.jp, National Research Institute for Earth Science and Disaster Prevention, JAPAN), Masayuki Maki , Takeshi Maesaka, Koyuru Iwanami

Background areal rain-rate estimators

139 2. ° 139 4. ° 139 6. ° 139 8. °35°

35.2°

35.4°

35.6°

35.8°

36°

LONGITUDE

LA

TIT

UD

E

Tokyo Bay

Tokyo

Kanagawa

Prefecture

Ebina Radar

37°

50 km50 km

Hayabuchi

-55°

-500

0

500

1000

1500

2000

Kisarazu Radar

Elevation [m]

139 53. ° 139 55. ° 139 57. ° RR [mm/hr]35.54°

35.55°

35.56°

35.57°

35.58°

35.59°

35.60°

0

10

20

30

40

50

60

LONGITUDE

LA

TIT

UD

E

37°35°

39°

33°31°

41°

Ebina Radar

Kisarazu Radar

r)( 12 r

1°

)( 22 r

)( 32 r )( 42 r

)( 41 r

)( 31 r)( 21 r

)( 11 rこの範囲でφDP が一

定であると仮定

r1、r2 はθに応じて

変化する r1,r2 changes

with

dp is constant

→ 3.96 10 , 0.551 → 4.26 10 , 0.644

(1) Z-R method

(2) KDP-R method

(3) DP-AR-Ryzhkov method

(4) DP-AR-Bringi method

0 1 2 3 4 5 60102030405060708090

100

R=23.4･KDP (0.1～0.5)R=20.4･KDP (0.5～1.0)

R=18.1･KDP (1.0～2.0)

R=14.7･KDP (2.0～7.0)

KDP [deg/km]

Rai

n ra

te [m

m/h

] R=18.9･KDP0.856

R=c･KDP

(5) Composite DP-AR method ( DP-AR-NIED)

DP­AR­Ryzhkov DP­AR­Bringi

(6) MKDP–R method ( )

Since National Research Institute for Earth Science andDisaster Prevention (NIED) deployed the first dual-polarization X-band radar around the Tokyometropolitan area in 2000, a widespread use of dual-polarization X-band radars has gained significantmomentum in Japan.Rainfall estimators using polarimetric parameters fromX-band radar have been proved to be in the bestharmony with rain gauge measurements without anycorrections from surface observations.Most of the common used hydrological models still relyon rain gauges and are usually based on theassumption of uniform rainfall over the catchment.What is the best rainfall estimating method in thecatchment areas for X-band radars? Can radar derivedrainfall take the place of traditional gauges be used inhydrological models?

XRAIN 38 Radars up to 2013

Objectiveresults verification

139.52 139.54 139.56LONGITUDE

35.54

35.55

35.56

35.57

35.58

35.59

0 5 10 20 30 40 50 60

LATI

TUD

E

[mm/hr]

139.52 139.54 139.56LONGITUDE

35.54

35.55

35.56

35.57

35.58

35.59

0 5 10 20 30 40 50 60

LATI

TUD

E

[mm/hr]

139.52 139.54 139.56LONGITUDE

35.54

35.55

35.56

35.57

35.58

35.59

0 5 10 20 30 40 50 60

LATI

TUD

E

[mm/hr]

139.52 139.54 139.56LONGITUDE

35.54

35.55

35.56

35.57

35.58

35.59

0 5 10 20 30 40 50 60

LATI

TUD

E

[mm/hr]

Estimate the areal rainfallrate using dual-polarization X-band radar and compare thederived results against a highdensity rain gauge networkof 30 rain gauges within thearea of 20 km2 around a verysmall river basin in Japan.

experiment areaHayabuchi river basin

Flow along Yokoama city Merge into Tsurumi river

and flow into Tokyo Bay Length : about 13.7 km Max. width: about 20 m Many water parks along

the Hayabuchi river Typical city river prone to

flash flood

Rain gauges 12 rain gauges in the

Hayabuchi river basin, 18outside

11 rain gauges between36 – 38 azimuth anglesof Ebina Radar

Figure1. Location of experiment area , rain gauges, and radars used in this study

Figure2.PPI images ofEbina Radar* Volume scan covers 12 elevations within 5 minutes.*KDP (NIEDstable) isestimated usingthe iterative filterand local linearregression.

conclusion and future workMean areal rain-rate estimators using differential propagation phase shift return the best harmony to Thiessen (gauge) – derived rainfall rates.Mean areal rain-rate estimators using reflectivity tends to underestimate rainfall rate, while classic KDP gives negative values because thedifferential propagation phase is contaminated by noise sometimes .Although MKDP (estimated using variational method) avoids the negative KDP values, the computation cost is heavy.Introduce mean areal rain-rate estimators into hydrological models to forecast real time flash flood is our future work.

:05 00 :07 00 :09 00

AVER

AGERRPER5MIN

0

10

20

30

40

50RainGaugeRefKdpPdp Ryzhkov_Pdp Bringi_Pdp NIED_MKdp

:15 00 :16 00 :07 00 :08 00 :18 00 :19 00

AVERAGERRPE

R5MIN

0

10

20

30

40

50[mm/hr]

LOCAL TIME (UTC+9)LOCAL TIME (UTC+9)LOCAL TIME (UTC+9)LOCAL TIME (UTC+9)

RAINFA

LLAM

OUNT

0

5

10

15

20

25

30RainGaugeRefKdpPdp Ryzhkov_Pdp Bringi_Pdp NIED_MKdp

RAINFA

LLAMOUNT

0

10

20

30

40

50

60

[mm]

CASE04 - 2011.07.30CASE03 - 2011.07.19CASE02 - 2011.06.30CASE01 - 2011.06.11:05 00 :07 00 :09 00 :15 00 :16 00 :07 00 :08 00 :18 00 :19 00

Case studies were carried out forthe verification of 6 areal rainfallrate estimators4 rainfall events occurred in 2011summer season, which included 1stratiform rainfall (CASE01) and 3convective rainfalls (CASE02 - 04)Radar ray-based verification, andbasin-based verification areconducted respectivelyBasin-based verificationresult is shown in this posterAreal rainfall rate for raingauges was obtained byapplying Thiessen algorismResults are noise in the caseof stratiform case comparedwith convective casesOnly DP estimators producedtime-continuous results,missing data occurs whileusing other parameters

Figures on the right Upper panels : distributions of rain rate at a

sample time for each case Middle panels : basin averaged areal rainfall

rates with respect to time for each case Bottom panels : basin-based rainfall amounts

with respect to time for each case

Rainfall amount

Rainfall rate

DPestimators

One no data periodWhen using other estimators

DP gives negative values

DP do over estimate

do under estimate