Sie sind auf Seite 1von 21

IRAD 2019

Applications of radar data in Meteorology, Hydrology and Urban planning.

Comparison between radar-, rain gauge- and


satellite- based rainfall for the Hyderabad region

Vimal Chandra Sharma1*, Satish Kumar Regonda2,Y.V. Rama Rao3,YK Reddy4, K Nagaratna4
1* PhD Scholar, Department of Civil Engineering, IIT Hyderabad, Kandi,Telangana, India
2 Assistant Professor, Department of Civil Engineering, IIT Hyderabad, Kandi, Telangana, India
3Consultant (Weather forecast/analysis), Telangana State Development Planning Society (TSDPS), Hyderabad, India
4Scientist, Meteorological Centre, IMD, Hyderabad,Telangana, India

Corresponding author email: ce16resch01005@iith.ac.in


Presented by,
V.Vimal Chandra Sharma
PhD scholar, IIT Hyderabad
1
Total rainfall
S.No Date Remarks
(mm)

Background 1 28-9-1908 153.2 The great Musi floods

2 01-8-1954 190.5 Sever flood

3 1970 140 Sever flood

 Hyderabad is one of the fastest growing metropolitan 4 24-8-2000 240 10 to 15 feet submerged
August,2001 230.4 Sever flood
cities, increase in population ~ from 7.7 millions to 10 5
August,2002 179.4 Sever flood
millions over the decade 6
7 8 -10August,2008 220.7 36hr rainfall
 Urban flooding is an issue, and a few facts 20-9-2016 120 to 160
Repeat of 2000 event
13 due to La Nina
 Encroachment, old drainage system and poor maintenance 14 15-7-2017 20 to 70 Cyclonic circulation
of storm water system (Vemula et al., 2018) Table 1: Floods due to heavy rainfall (source: NIUA report, September 2016)

 Both, extreme rainfall events covering large areas as well as S.No Daily average rainfall(mm) Count
1 0 to10 21515
localized high intense rainfall events of short duration result
2 10 to 20 964
floods (Ahmed et al., 2013) 3 20 to 30 304

 Reliable rainfall information in real-time as well as future 4 30 to 40 110


5 40 to 50 58
times assists decision makers including general public 6 50 to 60 27
7 60 to 70 10
 The objective of this study is to compare rainfall for the 8 70 to 80 7
Hyderabad region from different sources 9 80 to 90 6
10 90 to 100 2
11 ≥ 100 8
Table 2: Classification of rainfall events from the time series 1951 to 2013
(IMD gridded rainfall) 2
Study area: Greater Hyderabad Municipal Corporation (GHMC) region

Aspect Value
Area (Sq.Km) 650

Population (2018) ~ 10 millions

Admin. divisions 36

Annual average
~ 800mm
rainfall(mm) (1985-2005)

Sources of the rainfall


• Gauge measured, “28” stations
• Satellite estimates
• Doppler weather radar estimates

3
Information of different sources of rainfall

Rainfall Spatial Temporal


Product Wavelength Period of the record Coverage
source resolution resolution
Gauge - - - Daily 2004 – 2016 -
TIR-1 (10.3 to 11.3 20º N-S x 20º E-W
KALPANA-1 IMSRA 10 km um ) TIR-2 (11.5 to Daily 2010-2014 (Full scan) at 35,786km
12.5) in 26 min
18° E-W x 18°N-
TIR-1 (10.3 to 11.3
Daily* S(Full scan) at
INSAT-3D IMSRA 10 km um ) TIR-2 (11.5 to 2014 - present
35,786km in 26 min
12.5)

Doppler Maximum extend of


S band (7.5 to 15
Weather PAC 53 m Daily** 2005 to present reflectivity at 500km
cm)
Radar (DWR) radius.

Period from 2010 - 2016 is considered as the common Period of the Record
*Satellite based daily rainfall products/information obtained specific to a gauge location as well as for the entire GHMC
region
**1-day accumulated rainfall products (PAC) generated for 25 km radius with beam width of 0.5 and 530 above MSL
4
Doppler Weather Radar (DWR)
INSAT – IMSRA Algorithm Rainfall estimation

IMSRA algorithm gives the half-hourly rainfall using the below  Volumetric scan of the cloud and drop size distribution.
relationship. [Prakash et al., 2009, 2010]
N(D) = ධ 𝑁0 𝑒 −λ𝐷 ⅆ𝐷 where;
R = 8.61309 × exp(−(Tb − 197.97)/15.7061)
D = diameter of a rain drop in the volume
Where;
𝑁0 = Number of rain drops in the volume
R = rain rate mm/hr, Tb=Cloud top brightness temperature in
Kelvin.  Generation of SRI (surface rainfall intensity) products
from the scan (dBz) for 10min interval.
Tb = C2 / (cwl * log(C1 / (1.0E6 * rad_w_m2 * pow (cwn,
5.0)) + 1)) Z = aRb
cwl(metres) = central wavelength(um)/1000000.0(Field central where; a = 200 and b = 1.6 [Marshall and
wavelength in product) Palmer., 1948]
rad_w_m2 = radiance*10.0 (For converting from mw cm-2 sr  Generation of 1Hr accumulated rainfall from SRI
-1 um-1 to w m-2 sr-1 um-1) product (.RN1).
h = 6.6260755e-34 kg m2 s-1; c = 2.9979246e+8 ms-1;  Generation of one day accumulated rainfall (PAC) from
hourly accumulated rainfall(.RNN).
k = 1.380658e-23 kg m2 s-2 k-1; C1 = 2.0 * h *c * c;
C2 = (h * c) / k;

5
Rainfall products

Rainfall Gauge INSAT-IMSRA DWR - daily PAC


Point data Manually measured data Station- specific (point data) Station- specific (point data)
(station-specific) collected extracted from the IMSRA extracted from the PAC
from TSDPS for 28 locations daily rainfall products using daily rainfall products using
in the GHMC region. ArcGIS tools. ArcGIS tools.

Spatial average Average values of 27 Used GHMC boundary file Used GHMC boundary file
locations for selected and extracted rainfall from and extracted rainfall from
events. daily TIFF/HDF5 files using daily PAC (.RNN) files using
weighted average method. weighted average method.

6
Rainfall time series
Rainfall

Date [monsoon 2010 – 2016]


7
Selection of rainfall events

• Rainfall amounts sorted in the


decreasing magnitude for each
location, separately.
• Top ‘50’ highest rainfall events selected
for each location, separately.
• Date of the highest rainfall is not
necessarily same across all locations.
Rainfall

• Therefore, unique ‘dates’ are much


more than ‘50’
• Sort the above events as per the date
of spatial coverage of the event.

Date (may not same for all locations)


8
Selected rainfall events

• Sorted top ’50’ events as per the date of spatial coverage of the event. In this study top 26 events that have large
spatial coverage are considered.

9
Spatial-average daily rainfall estimates

10
Comparison of spatial-average estimates

Both satellite- and radar- estimated


INSAT-IMSRA
values appear to have similar 250.00 DWR-PAC
behaviour, i.e., no clear relation, Linear (INSAT-
however, for a few specific dates 200.00 IMSRA)

Estimated rain (mm)


radar estimated values exhibited 150.00
variations with respect to the
magnitude of observed values. 100.00

Accuracy Assessment Statistic Satellite Radar 50.00

Mean Error (ME) 17.58 20.75


0.00
Mean Absolute Error (MAE) 31.9 24.28 0.00 50.00 100.00 150.00
Root Mean Square Error
47.52 31.06 Gauge rain (mm)
(RMSE)
Normalized ME 0.33 0.39
Normalized MAE 0.61 0.46
NormalizedRMSE 111.5 124.8
Correlation 0.21 0.58
11
Station-specific rainfall estimates

12
Station-specific rainfall estimates across all stations

 Both satellite- and radar- estimated values are


underestimated (below the 1:1 line)
 No variation among the estimated rainfall values
(values along a horizontal line) INSAT
DWR
 A few instances where both gauge- and estimated
rainfall are in agreement.

Estimated rain (mm)


 The local fit suggests a relation between gauge
measured and radar estimates [red line] beyond
gauge rainfall values of certain threshold (50 mm)
Accuracy Assessment Statistic Satellite Radar
Mean Error (ME) -28.62 -35.71
Mean Absolute Error (MAE) 31.98 36.32
Root Mean Square Error
44.51 47.03
(RMSE)
Normalized ME -0.59 -0.74
Gauge rain (mm)
Normalized MAE 0.66 0.75 Point wise comparison between rain gauge, INSAT(IMSRA) and DWR- PAC rainfall data
NormalizedRMSE 0.92 0.97 of 26 events for 27 stations
Correlation 0.22 0.49
13
Comparison between rain gauge, INSAT(IMSRA) and DWR (PAC)
rainfall data

 The relation varies with


location
 Relatively DWR estimates
exhibit variation and in
agreement with the gauge data.

INSAT
DWR

14
Comparison between rain gauge, INSAT(IMSRA) and DWR (PAC)
rainfall data

15
Comparison between rain gauge, INSAT(IMSRA) and DWR (PAC)
rainfall data (Cont’d)

16
Comparison between rain gauge, INSAT(IMSRA) and DWR (PAC)
rainfall data (Cont’d)

 

17
Comparison between rain gauge, INSAT(IMSRA) and DWR (PAC)
rainfall data (cont’d)

 Apparently, no relation for


many dates
 Radar estimated values do not
exhibit any variation (values
along a horizontal line)

18
Conclusions

• No clear relation found between gauge, radar and satellite based spatial average rainfall
estimates
• The relation between gauge, radar and satellite based station-specific rainfall estimates vary with
location and date
• In general, radar-based estimates appear to be in better agreement with gauge rainfall estimates
for a few locations and dates
• The verification metrics appear to be influenced by small sample size and presence of a few
large rainfall amounts
• Both satellite and radar based estimates need to be improved for decision making

19
Future work

 Increase in sample size and pursue detailed analysis


 Analysis in the context of hourly values as well as integrating climate information
 Enhancement of calibration of weather radar, this includes analysis of values of parameter in
Z-R relationship for different events
 Use of rainfall estimates for urban flood modelling as well as riverine flood modelling

20
Thank you

Acknowledgments:
I express my special thanks of gratitude to Shri. Sridhar, IMD, Hyderabad, who provided guidance and assisted in the
technical aspects of radar data processing.
Also, I thank Mr.Azharuddin, PhD scholar, IIT Hyderabad and Mr. Prasad, B.Tech graduate for the provided help.

Questions ?

21

Das könnte Ihnen auch gefallen