3rd Workshop of theInternational Precipitation Working Group

Melbourne Australia 23-27 October 2006Melbourne, Australia, 23 27 October 2006

Modeling storm hydrographs in a mid-size basin using satellite rainfall estimatesbasin using satellite rainfall estimates

Daniel Barrera Viviana Zucarelli andDaniel Barrera, Viviana Zucarelli and María V. Morresi

Universidad Nacional del LitoralUniversidad Nacional del LitoralCONICET / Universidad de Buenos Aires

ARGENTINAARGENTINA

Rainfall-runoff models of any type need rainfall areal estimates as input.

Characteristics of operational rain-gauge networks:

1. Low spatial density of stations

l l d1. Irregular spatial distribution

Intrinsic characteristic of rainfall:Hi h ti l i bilitHigh spatial variability

Consequence: Important errors in areal rainfall Consequence: Important errors in areal rainfall estimates over sub-basins or other model input

areasareas.

i f h li i i iLocation of the Feliciano River Basinin the Province of Entre Ríos (Argentina)

26

28

30

(deg

ree s

)

34

32

uth

latit

ude

36

34

So

64 62 60 58 56 5438

West longitude (degrees)

Feliciano river basin. Sub-basins and drainage system.

Location of Feliciano river basin30.0

31

30

30

)

31 0

31

31

TUD

S UR

( GR

A DO

S

31.0

31

31

31

LATI

T

32.060 60 59 59 59 59 59 59 59 59 59 59 58

A division in 17 sub basins and basin A division in 17 sub-basins and basin segments was made for hydrologic

modelling purposes33.0

L ti f thi id i b i

60 0 59 0 58 0

34.0

Lag time for this mid-size basin(5500 km2): ~ 4-5 days

60.0 59.0 58.0

WEST LONGITUDE (degrees)

On the other hand: Satellite-based methods of rainfall ti ti t i l it ff t d b i ifi t t testimation at a pixel unit are affected by significant errors at present,

which depend on the type of rain among other factors.

Ad t i t b tiAdvantage over point observations:

They provide full spatial coverage (precipitation estimations can beperformed at all pixels) and account for spatial variability of rainfallperformed at all pixels) and account for spatial variability of rainfall.

Therefore, they are more adequate to get the area-average rainfallTherefore, they are more adequate to get the area average rainfallover a sub-basin containing several to many pixels.

Location of pixel centers andLocation of pixel centers and Paso Medina station (outlet) over the watershed

30.4

30.6

(de

gre

es)

30.8

thla

t i tu d

e

31.0

So

ut

59.6 59.4 59.2 59.0 58.8 58.6 58.4

West longitude (degrees)

31.0

g ( g )

Techniques to obtain accumulated Techniques to obtain accumulated precipitation estimates

at a given pixel consist on two steps:at a given pixel consist on two steps:

1. Estimation of mean spatial rain rate at every pixel

2. Time integration over a lapse assigned to the analyzed imagethe analyzed image

Th A t ti t t hi (Vi t S fi ld The Auto-estimator techique (Vicente, Scofield and Menzel, 1998) with further modifications is called Hydro-estimator (HE). A version of this technique was developed at the University of t n qu wa p at t n r ty f Buenos Aires and is run operationally at the

National Meteorological Service of ArgentinaNational Meteorological Service of Argentina.

Auto-estimator technique: Designed to estimate Auto estimator technique: Designed to estimate convective rainfall in an operational way,

accumulated on lapses lesser than 1 day with a accumulated on lapses lesser than 1 day, with a spatial resolution of 1 pixel (4km x 4km).

A power-law empirical function relates the p pbrightness temperature at the cloud top with the

rain rate at the cloud base

Rc = ko exp (- Tb1.2) The relationship applies to unicellular ppand multicell convective clouds (cumuloninmbus) at mature stage.

Vicente, Scofield and Menzel, 1998

A further modification introduced the atmospheric precipitable water as a parameter family of curvesp p p y

Rc = ko exp (- Tb1.2)o p ( )

ko , functions of PW

(Scofield and Kuligowski, 2003)

Satellite GOES East: 4 sectors of images generation

Scanning frequencyScanning frequency

CONUS: 1 every 15 minutes

SOUTHERN HEMISPHERE:

1 every 30 minutes over 70%1 every 30 minutes over 70% of the time during the day.

1 every 90 minutes over the1 every 90 minutes over the lasting 30% of the time.

Source: NOAA / NESDISSource: NOAA / NESDIS

To better describe the displacement of cloud systems and minimize the p ydeformation in the obtained rainfall fields, it was proposed the generation of

synthetic brightness temperature images interpolated between two consecutive observed images.g

The image was sectorized in boxes of 4x4 degreesThe image was sectorized in boxes of 4x4 degrees.

In order to determine the mean vector displacement In order to determine the mean vector displacement of clouds on each box, two criteria were proposed, applied and tested The vector difference (in pixel applied and tested. The vector difference (in pixel units) between pixels with cloud tops at the same tropospheric layer in both the analyzed and the tropospheric layer in both the analyzed and the precedent images was searched, which complies with one of the following conditions:g

• The maximum cross-correlation coefficient of temperature values at pixels in two consecutive images; and

• The highest joint frequency of the same interval class of temperature (or same atmospheric layer) at pixels in two consecutive images.

Storm of April 10, 2002

28S

Isoyetal map obtained from

30S

GOES images (procedure 1) (6h i d)

32S

30S(6hr period)

62W 60W 58W 56W 54W32S

020406080100

28S

Isoyetal map obtained from

30S

GOES plus synthetic images( d )

32S

30S(procedure 2) (same period)

62W 60W 58W 56W 54W32S

Estimated area-averaged rainfallover sub-basin 13

Left bars: Procedure 1 Right bars: Procedure 260

50

60

40

mm

)

20

30

Rai

nfal

l (m

10

20

01 2 3 4 5 6 7 8 9 10 11 12 13 14

3-hour lapses starting on Apr/10/02 at 0 UTC3 hour lapses starting on Apr/10/02 at 0 UTC

6-hour lapses evolution of area-averaged rainfall. Procedure 1 and Procedure 2

90

70

80

50

60

mm

)

30

40PP (m

10

20

01 2 3 4 5 6 7 8 9

Time in 6hr lapses

Procedure 1 Procedure 2

OCINE2 Rainfall-Runoff Model

It is a conceptual, pseudo-distributedp , pparameters model developed by the project team which computes bothproject team, which computes both overland flows and stream flows by applying the kinematic wave theory.

B Basic equations

fpqy

qAQ

fp

xt

q

tx

Continuity for stream segment Continuity for catchment segment

cmyq *smAQ *

Continuity for stream segment Continuity for catchment segment

cc yq *s

s AQ *Momentum for stream segment Momentum for catchment segment

x: distance along the stream segment (m); Q: runoff in the stream segment (m3/sec) ; A: area of the wetted stream cross-section (m2) ; q: discharge per unit width in the catchment segment (m2/sec); y: average depth of flow (m); p: rainfall rate (mm/hr);catchment segment (m2/sec); y: average depth of flow (m); p: rainfall rate (mm/hr);

f: infiltration rate (mm/h);

c, mc, s, ms, are calibration parameters for the catchment and streamc, mc, s, ms, are calibration parameters for the catchment and stream segments, respectively.

APPLICATION TO THE FELICIANO RIVER BASIN WITH OUTLET IN PASO MEDINA STATION

As part of the calibration process a As part of the calibration process, a segmentation in 17 sub-basins was performed after an analysis of the performed after an analysis of the

dynamics of hydrological responses to f ll rainfall inputs.

A Topologic framework with 51 segments (34 A Topologic framework with 51 segments (34 for overland flow and 17 for stream flow)

was utilizedwas utilized.

Storm hydrographs modelled.Due to a lack of information about soil

characteristics and soil moisture conditions previous to the storm to be modelled, it is not

p ssibl t this st t pr dict th run ff possible at this stage to predict the runoff component of the rainfall. Therefore, for each of the cases modelled the volume of the simulated the cases modelled, the volume of the simulated

hydrograph has been adjusted to match the volume of the observed hydrograph in order to y g p

evaluate the model skill in predicting the distribution of that runoff in time, that is, the

h f th t h d hshape of the storm hydrograph.

Observed and simulated hydrographsStorm starting on April 9, 2002g p ,

250

200

150

m3/

s)

100Q (m

0

50

01 2 3 4 5 6 7 8 9 10 11 12 13 14 15

observed simulatedobserved simulated

Observed and simulated hydrographsStorm starting on April 22, 2002 g p ,

800

600

700

400

500

600

/sec

)

200

300

400

Q(m

3/

100

200

01 3 5 7 9 11 13 15 17 19

observed simulated

Observed and simulated hydrographsStorm starting on December 28, 2002 g ,

300

250

150

200

m3/

sec)

100

Q (m

0

50

0

27/12

/0229

/12/02

31/12

/0202

/01/03

04/01

/0306

/01/03

08/01

/0310

/01/03

12/01

/0314

/01/03

16/01

/0318

/01/03

20/01

/0322

/01/03

24/01

/03

observed modeled

Observed and simulated hydrographsStorm starting on March 9, 2003g ,

250

200

150

/sec

)

100Q (m

3/

50

01 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

observed simulated

Ob d d i l d h d hObserved and simulated hydrographsStorm starting on April 23, 2003

2500

2000

1500

500

1000

0

500

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 241 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

observed sim. f rom HE estimates sim. from raingauges

CONCLUSIONS

HE estimates lead to compute area-averaged rainfall over large sub-basins which are as good as the ones obtained g g

from conventional raingauge networks to be used as input in rainfall-runoff models.

The present state-of-the-art permits to implement p p poperational hydrological forecast systems in ungauged

basins of developing countries, based on satellite rainfall p gestimates as model inputs.

Thank youy