Sagar Ratna Bajracharya , Mandira Shrestha and Pradeep Mool sagbajracharya@icimod

International Centre for Integrated Mountain Development

Validation of satellite rainfall estimation in the summer monsoon dominated area

of the Hindu Kush Himalayan Region

Sagar Ratna Bajracharya, Mandira Shrestha and Pradeep [email protected]

Integrated Water and Hazard ManagementInternational Centre for Integrated Mountain Development (ICIMOD)

www.icimod.org

4th Workshop of theInternational Precipitation Working Group

13-17 October, 2008, Beijing, CHINA


Outline General description and climatic condition of HKH region

What is NOAA CPC-RFE 2.0 Methodology and Analysis

Results

Recommendations and Road Ahead



The Himalayan Region Extends over 3500 km from Afghanistan,

Pakistan, India, China, Nepal, Bhutan to Bangladesh and Myanmar

Geologically youngest mountain range in the world, giving rise to the high degree of slope instability and landslide hazards

High mountains, Plane and Tibetan Plateau

Variable background – snow cover etc

High spatial variations with widely varying physical and climatic conditions


The Hindu Kush-Himalayan Context

Meteorologically diverse

•Orography and continental influences•Convective precipitation, Cloud burst, Monsoon Influence•Seasonal variations – Extremely cold vs hot and humid temperatures•Variety and variability of climate due to complex topography•Plenty of intense rain intensities…


Track of Monsoon Depression


L, 990 mbar

Active Monsoon Trough


Kitini Khola 1993 flood


Orography and Rain Shadow Orographic lift occurs when an air mass is forceda low elevation to higher elevation as it moves over rising terrain. As the air mass gains altitude it expands and cools adiabatically. This cooler aircannot hold the moisture as well as warm air and this effectively raises the relative humidity to 100%,creating clouds and frequently precipitation.


PrecipitationSouthern part of the Himalayas receive higher rainfall whereas northern receive less rainfall

Higher in the east and gradually decreases towards west

More than 80% rainfall during monsoon (June-September)

High seasonal and spatial variation

Note: -ve value indicates Ocean

Source: World Water and Climate Atlas, IWMI


Ocean0 - 1010 - 5050 - 100100 - 250250 - 500> 500

Pre Monsoon (Mar-May)

Ocean0 - 1010 - 5050 - 100100 - 250250 - 500> 500

Post Monsoon (Oct-Nov)

Ocean0 - 1010 - 5050 - 100100 - 250250 - 500> 500

Post Monsoon (Oct-Nov)

Winter (Oct-Nov)

Ocean0 - 1010 - 5050 - 100100 - 250250 - 500> 500

Winter (Oct-Nov)

Ocean0 - 1010 - 5050 - 100100 - 250250 - 500> 500

Seasonal Variation of PrecipitationPre Monsoon Monsoon

Post Monsoon Winter

Monsoon (Jun-Sep)

> 25002000 - 25001500 - 20001000 - 1500500 - 1000100 - 5000 - 100Ocean


NOAA CPC RFE2.0 Initial version became operational in May 2001

Originally run over the African continent then expanded to southern Asia and western Asia / eastern Europe

Product is a combination of surface and satellite precipitation information

Spatial resolution: 0.1 degree

Temporal resolution: daily

Availability: 5°-35°N; 70°-110°E


Final Product

Source: Tim Love


NOAA CPC RFE Domain


1515

Source: Tim Love

GTS Inputs

International Centre for Integrated Mountain DevelopmentData Preparation

Daily independent rain gauge data in word file convert into appropriate format provided

by individual country from 2002-2004

Data Quality Control

Data Conversion- RFE2 Data downloaded by NOAA ftp server

- Observed rain gauge data in GIS format

Change the projection parameter of GIS dataset

Interpolationa) Kriging

- 0.1˚ spatial resolution for individual country

-0.25 to 2.5˚for regional level

Working Area- ICIMOD Whole HKH (Regional)

- Partner institutes their individual country

Considered ScalesIndividual country

- 0.1 to 0.25˚ spatial resolution- 24 hours, 10 and 30 days temporal

ICIMOD-0.25 to 2.5˚spatial resolution

- 24 hours, 10 and 30 days temporal

Estimated Data-NOAA CPC_RFE Product-Whole HKH Daily product (24 hours)-0.1 degree spatial resolution-In Lambert Azimuthal Area

Comparison or Overlay

Validationa) Visual analysisb) Descriptive statistics - through contingency tables - POD, FAR (e.g. with zero and 1mm/day rain/no rain threshold) c) Statistical analysis-Bias-RMSE-linear correlation coefficient-Skill score index-% error-etc

Methodology for validation


International Centre for Integrated Mountain DevelopmentComparison


Visual Analysis

Scatter plot of Observed V Estimated rainfall

Descriptive statistics

Contingency tables use of POD and FAR

Statistical analysis

Bias, RMSE, Correlation, Skill, % error etc

Validation of RFE









Sumarized statistical summary of regional validation

Continuous verification statistics

Categorical verification statistics

Days Pixel No

Bias(mm)

Corr RMSE(mm)

% error

POD FAR Skill

2002203(Max) 9466 -5 0.72 23.42 -24.27 0.85 0.8 0.02

2002243(Min) 9763 -0.74 0.1 5.62 -28.46 0.38 0.49 0.3

2003190(Max) 9763 -7 0.54 22.2 -40.65 0.85 0.01 0.98

2003153(Min) 9763 -0.62 0.01 5.96 -31.2 0.48 0.42 0.41

2004190(Max) 9763 -9 0.55 26.36 -35.47 0.88 0 0.5

2004162(Min) 9763 -0.14 0.34 5.44 -5.5 0.5 0.3 0.3


The CPC-RFE technique overestimates rainfall particularly over a region where there is persistence of cirrus cloud, snow and ice.

Underestimates rainfall in a region where there is orographic precipitation and precipitation by warm cloud.

Rainfall occurrence is underestimated by about half and more than half in monsoon during heavy rainfall and overestimates in pre monsoon

Limitation of SRE is that it cannot produce more than certain amount of rainfall in 24 hours

Results


How lag time of the data can be reduced? Improving Orographic effects in rainfall estimation . RFE- the shape of precipitation is given by the combination of

satellite estimates, magnitude is inferred from GTS station data, need the maximum availability of the rain gauge stations

Including radar data for validation where available in HKH and Incorporate more gauge data for validation

Validation considering different rainfall regimes. Validation considering temporal variable like decadal, monthly,

yearly, rainy season etc using different spatial resolution (0.25˚, 0.5˚, 1˚, etc)

Improve Satellite estimates over the ice and snow cover estimates over the Himalayas

Application of improved RFE in flood early warning and flood monitoring activities in flood season.

Next Steps in SRE application


Sagar Ratna Bajracharya , Mandira Shrestha and Pradeep Mool sagbajracharya@icimod

Documents