Global Flood and Drought Prediction
GEWEX 2005 Meeting, June 20-24
Role of Modeling in Predictability and Prediction Studies
Nathalie Voisin, Dennis P. Lettenmaier, Eric F. Wood
Global Floods and Droughts
• Floods– $50-60 billion USD /year, worldwide– 520+ million people impacted per year worldwide – Estimates of up to 25,000 annual deaths
Mostly in developing countries; Mozambique in 2000 and 2001, Vietnam and others (Mekong) in 2000.
• Droughts– 1988 US Drought: $40 billion– Famine in many countries: 200,000 people killed in
Ethiopia in 1973-74
Source: United Nations University, http://update.unu.edu/archive/issue32_2.htmhttp://www.unu.edu/env/govern/ElNIno/CountryReports/inside/ethopia/Executive%20Summary/Executive%20Summary-txt.html1988 drought: NCDC : http://lwf.ncdc.noaa.gov/oa/reports/billionz.html
New technologies
In the last 25 years:• Climate models• Hydrological models • Land surface schemes• Remote sensing devices • Archives, storage
Despite all these advances, no capability for performing global hydrological prediction
But discontinuity on a global scale…
– Uneven observations– local hydrological models
Hydrologic warnings tend to be localized
Objectives
Develop a global flood and drought nowcast and prediction system
Using
• climate ensemble forecasts• Distributed hydrologic model VIC ( U. of
Washington, Princeton University)• Satellite remote sensing information• NCEP / ECMWF data sets
Forecast System Schematic *
Satellite precipitation estimates
local scale (1/2 degree) weather inputs
soil moisturesnowpack
Hydrologic model spin up
SNOTEL
Update
streamflow, soil moisture, snow water equivalent, runoff
Month 0
1-2 years back
G-LDAS /other real-time met.
forcings for spin-up
Hydrologic forecast simulation
NOWCASTS
INITIAL STATE
AMSR-E MODISUpdate
ensemble forecasts NCEP GSM ensemble
* Similar experimental procedure as used by Wood et al (2005) West-wide seasonal hydrologic forecast system
SEASONAL FORECASTS (drought)
SHORT TERM FORECASTS (flood)
Hydrologic Model Spin Up
Preliminary studies• Compare Hydrological variables as
simulated by the distributed model VIC using
– Climatology: Adam and Lettenmaier (2003) and Adam et al (2005) precipitation data sets
• gauge undercatch and orography correction • 1979-1999• Refer to A & al. later on
– Satellite precipitation estimates
Satellite datasets
Choosing satellite data sets
– Availability ( near real time later on)– Time resolution (daily and less)– Spatial resolution ( ½ lat/lon degree maximum)– Spatial coverage (global)
Satellite based precipitation estimates
Combined IR and PMW data sets
Spatial Domain
Spatial res.
Time res. Period available
Avail.
CMORPH (Joyce et al 2004)
Global 0.25 hourly Dec 2002- daily
PERSIANN (Sorooshian et al 2000)
50oS-50oN
0.25 6 hourly Mar 2000- 2 days
CMAP (Xie and Arkin 1996, 1997)
Global 2.5 monthly Jan 1979- 1 week
GPCP 1DD (Huffman et al 2001)
Global 1 daily Oct 1996- 3 mths
3B42RT (Huffman et al 2002)
50oS-50oN
0.25 3 hourly Feb 2002 - 6 hrs
Satellite precipitation estimates
Surrogate for future near real time satellite estimates:
GPCP 1DD daily precipitation
– Huffman et al 2001– Infra-Red (TMPI) over 40oS-40oN– Recalibrated TIROS Operational Vertical Sounder (TOVS)
beyond 40oS and 40oN– Scaled to match monthly GPCP Version 2 Satellite-
Gauge precipitation estimates– 1997-present
Major Basins to be simulated first
World Basins
Study basins
6 simulated basins
Mackenzie
Mississippi Mekong
Danube
CongoAmazon
1997-99 Water Balance (mm)
We compare hydrologic variables as simulated by VIC driven by 1997-99:
• A & al.precipitation estimates• GPCP 1DD (Huffman et al 2001) precipitation estimates
Annual (mm) Amazon Congo Danube Mack. Mekong Missip.Precip A & al. 2378 1687 888 406 1405 947
GPCP - A & al. -526 -439 -34 -10 184 -147Runoff Adam et al 1582 532 356 157 685 397
GPCP - A & al. -550 -325 -44 -26 71 -150Evap Adam et al 792 1095 527 237 720 568
GPCP - A & al. 25 -63 15 28 107 8
Monthly Water Balances 1997-99
Mackenzie
0
10
20
30
40
50
60
70
80
90
1 2 3 4 5 6 7 8 9 10 11 12Months
mm
Amazon
0
50
100
150
200
250
300
350
1 2 3 4 5 6 7 8 9 10 11 12Months
mm
Precip Clim
Precip GPCP
Runoff Clim
Runoff GPCP
Evap Clim
Evap GPCP
Danube
0
20
40
60
80
100
120
140
1 2 3 4 5 6 7 8 9 10 11 12Months
mm
Precip Clim
Precip GPCP
Runoff Clim
Runoff GPCP
Evap Clim
Evap GPCP
Monthly Water Balances 1997-99
Mississippi
0
20
40
60
80
100
120
140
1 2 3 4 5 6 7 8 9 10 11 12Months
mm
Precip Clim
Precip GPCP
Runoff Clim
Runoff GPCP
Evap Clim
Evap GPCP
Monthly Water Balances 1997-99
Mekong
0
50
100
150
200
250
300
1 2 3 4 5 6 7 8 9 10 11 12Months
mm
Precip Clim
Precip GPCP
Runoff Clim
Runoff GPCP
Evap Clim
Evap GPCP
Monthly Water Balances 1997-99
Congo
0
50
100
150
200
250
1 2 3 4 5 6 7 8 9 10 11 12Months
mm
Precip Clim
Precip GPCP
Runoff Clim
Runoff GPCP
Evap Clim
Evap GPCP
Future work
• Model Spin up – Further analysis : assess bias in simulated hydrologic
variables when using satellite precipitation estimates• Extend A & al data sets to 2004
• Use CMORPH (Joyce et al 2004)
– Use other precipitation estimates: • NCEP
• ECMWF ERA 40
– Bias adjustment of forcing data set : need 10 years of observations at least
Future Work
• Data Assimilation: – use satellite soil moisture
• still experimental, • need further validation and assess the additional skill in
forecast
– Use MODIS: experimental as well
• Forecasts: – Validation with retrospective forecasts, near real time
forecasts / nowcasts– Assess predictability skills ;
• initial conditions• precipitation forecast
Thank You!
Credit: Philip Wijmans/ACT-LWF Trevo, Mozambique, February 2000 , http://gbgm-umc.org/umcor/00/mozphotos.stm
Amazon
Mekong
Congo
Mackenzie
Danube
Mississippi