1. Upstream Satellite-derived Flow Signals for River Discharge Prediction
Tom Hopson1
F. A. Hirpa2; T. De Groeve3; G. R. Brakenridge4;
M. Gebremichael2; P. J. Restrepo5 1
2Uconn; 3Joint Res Counc; 4U. of CO; 5Office Hydro Devel

2. Outline
Monitoring river flows using Satellite-based Passive-Microwave data
Case study: Bangladesh’s vulnerability to flooding
Estimating flood wave celerity
River discharge estimation “nowcasting”
Discharge forecasting
Future Work

3. Objective Monitoring of River Stage and Flow:
Satellite-based Passive Microwave Radiometer
-- Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E) & NASA TRMM
(Future: Global Precipitation Measurement System)
- Utilizing 36-37Ghz (unaffected by cloud)
- pixel size ~20km
- ~2day complete global coverage (night-time brightness temperatures)
- data range: 1997 to present
Here is the satellite we are currently using. Similar data are also provided today by the NASA TRRM satellite, with an extensive follow-on constellation of satellites being planned (GPM, or Global Precipitation Measurement System). There is also a long series of Defense Department satellites whose data are publically available: the DMSP SSM/I sensor. Thus, we may be able to extend many river records back to These satellites were not designed to measure rivers, but their data stream can in fact provide such measurements. We use the 36.5 GHz AMSR-E band that is not affected by cloud cover.
Other Approaches: satellite altimeter-derived water level (and discharge derived through rating curve):
e.g. Birkett, 1998; Alsdorf et al. 2000; Jung et al. 2010, Papa et al. 2010, Alsdorf et al. 2011, Biancamaria et al. 2011

4. One day of data collection (high latitudes revisited most frequently)
AMSR-E covers the whole globe once every 2.5 days, but higher latitudes more frequently.
=> On average, global coverage 1-2 days

5. Example: Wabash River, Indiana, USA
GDACS, JRC
Gridded Approach
Black square shows measurement pixel. White square is calibration pixel.
De Groeve, et al, 2009, 2010
=> Use “M/C” ratio
Brakenridge, et al, 1998, 2005, 2007

6. Case Study: Bangladesh River Flooding
Damaging Floods:
large peak or extended duration
Affect agriculture: early floods in May, late floods in September
Recent severe flooding: 1974, 1987, 1988, 1997, 1998, 2000, 2004, and 2007
1998: 60% of country inundated for 3 months, 1000 killed, 40 million homeless, 10-20% total food production
2004: Brahmaputra floods killed 500 people, displaced 30 million, 40% of capitol city Dhaka under water
2007: Brahmaputra floods displaced over 20 million
(World Food Program)

7. Bahadurabad gauge
(Brahmaputra)
Hardinge Bridge gauge
(Ganges)
Gauging data reference: Hopson 2005; Hopson and Webster 2010

8. Correlation vs. lag time and distance between Gage Q & satellite signals
Ganges
Brahmaputra

9. Estimating the kinematic wave celerity
Ganges
Brahmaputra
Assumptions (large): hydraulic parameters homogenous; rainfall predominantly upper-catchment; wave speed variation with depth lower-order; dynamic wave effects lower-order (confirmed through rating curve analysis – not shown), etc.
Independent analysis: Kleitz-Seddon Law for kinematic wave celerity c:
W channel top-width, Q discharge, y the river stage.
Brahmaputra at Bahadurabad: 4m/s < c < 8m/s; Ganges at Hardinge Bridge: 2m/s < c < 6m/s.

10. Generation of Nowcasts and Forecasts
Regressor and model selection procedure:
Correlate all variables with gauged observations
Sort in decreasing level
Select using step-wise forward-selection using K-fold cross-validation and RMSE cost function
Repeat ii) – iii) for all lead-times

11. Discharge estimation “Nowcasting”
Ganges 2003
Brahmaputra 2007

12. 1-, 5-, 15-day Forecasts

13. Skill Scores Nash-Sutcliffe efficiency
(fraction of variance explained):
-- 80% (1-day) to 55% (15-day)
explained
Now, generate a forecast including “persistence”:
-- satellite information improves forecasts by 10-20% at all lead-times

14. Future Work Blend with ECMWF-derived forecasts
(Hopson and Webster 2010; Operational in 2003)
2007 Brahmaputra Ensemble Forecasts and Danger Level Probabilities
7-10 day Ensemble Forecasts
7-10 day Danger Levels
7 day
8 day
7 day
8 day
9 day
10 day
9 day
10 day

15. Summary Useful for estimating: -- flood wave celerity
Satellite-based passive microwave imagery useful tool for monitoring river conditions for limited-gauged basins
Useful for estimating:
-- flood wave celerity
-- discharge estimates (nowcasts) including filling in missing data
Combined with other methodologies, a potentially useful tool for flood forecasting of large river basins
Further questions:

17. MODIS sequence of 2006 Winter Flooding
2/24/2006 C/M: /15/2006 C/M: /22/2006 C/M:
Some measurement sites are so sensitive we can see changes even from, for example, low flow up to bank full flow. But along thousands of river measurement sites, we can most easily see the changes that occur from normal flow into various flood flow states. The C/M ratio above is the AMSR-E remote sensing data, with the two pixels included to provide the ratio. The images shown are from MODIS.
Note the dates. We captured MODIS image data on these dates, but AMSR-E provided daily updates of the discharge estimator: the C/M number. Higher ratios = higher discharge, more land flooded. The two remote sensing data systems provide synergy, and are together more accurate and powerful than using just one alone.

18. Site 98, Wabash River at New Harmony, Indiana, USA
We compare here the discharge obtained by the River Watch approach, so C/M numbers transformed to discharge using a rating equation, with that measured by a ground gaging station. You see that, as is commonly the case, smaller flood events are not always seen by River Watch, but the timing, duration, and relative magnitudes of the larger events are. We can also look at the flow series, and set thresholds: automatically, when AMSR-E sees a rise in the C/M values indicating a flood exceeding a certan size, the site can be flagged.

19. So we compare the microwave signal coming from an image pixel centered over the river, and one nearby, not affected itself by the rising and falling of river discharge, stage, and surface area.