Hydrologic forecasting for flood risk management Predicting flows and inundation in data-limited catchments Dr. Kelly Kibler Assistant professor Civil, Environmental, & Construction Engineering University of Central Florida
Roadmap 1 Hydrologic forecasting for early warning Lead time Accuracy Challenge of the data-limited basin 2 Remotely sensed precipitation data Opportunities and limitations Designing observation networks to integrate remotely sensed data 3 Inundation Calibrating/validating hydrologic models 2
Flood Forecasting for Early Warning Two critical components : 1. Forecast accuracy 2. Forecast lead time 3
Flood Forecasting for Early Warning Two critical components : 1. Forecast accuracy 2. Forecast lead time HOURS 4
Flood Forecasting for Early Warning Two critical components : 1. Forecast accuracy 2. Forecast lead time HOURS DAYS 5
Flood Forecasting for Early Warning Two critical components : 1. Forecast accuracy 2. Forecast lead time HOURS DAYS WEEKS 6
Flood Forecasting for Early Warning Two critical components : 1. Forecast accuracy 2. Forecast lead time HOURS DAYS WEEKS Increasing forecast lead time, increasing potential response and loss reduction 7
Flood Forecasting for Early Warning Economic benefit- damage/loss reduction In other words, for every USD 1 invested in this EWS, there is a return of USD 558.87 in benefits. Subbiah and Bilden, 2008, World Bank Report 8
Flood Forecasting for Early Warning Economic benefit- damage/loss reduction: risk assessment in Surma River basin, Bangladesh Benefit-cost ratio of early warning system for a 5-year event was 2.71, 12.44 and 33.21 for 24 hours, 48 hours and 7 days lead time, respectively. Hyder, 2013 9
Flood Forecasting for Early Warning Important to accurately predict: 1. Onset of flooding time Discharge Precipitation, flood stage Peak of water level Flooding Peak of rainfall It started to rain. Event 10
Flood Forecasting for Early Warning Important to accurately predict: 1. Onset of flooding 2. Peak discharge time Discharge Precipitation, flood stage Peak of water level Flooding Peak of rainfall It started to rain. Event 11
Flood Forecasting for Early Warning Important to accurately predict: 1. Onset of flooding 2. Peak discharge 3. Duration of flood time Discharge Precipitation, flood stage Peak of water level Flooding Peak of rainfall It started to rain. Event 12
Flood Forecasting for Early Warning Important to accurately predict: 1. Onset of flooding 2. Peak discharge 3. Duration of flood 4. Spatial extent of inundation, through time 13
Flood Forecasting for Early Warning Two critical components : 1. Forecast accuracy 2. Forecast lead time Accuracy of forecast with 10 days lead time Webster et al., 2010
Flood forecasting in poorly gauged basins Challenges: Insufficient implementation and maintenance of ground-based, real-time hydrologic observation. Lag time between data observation and availability for flood forecasters. Lead time compromised, poor forecasting skill. Management across boundaries. Administrative barriers to data availability. Manual data observation Basins cross boundaries Lack of transmission 15
Geopolitically ungauged catchment area Hydrologic prediction in ungauged basins: a wicked problem for hydrologists Geopolitically ungauged catchment area: Areas where comprehensive data observation networks may exist, but due to geopolitical constraints, catchments are effectively ungauged. 16
Water-related disaster in transboundary river basins Flood disasters in transboundary river basins: Historically more severe, affect larger areas and result in higher costs of human life and economic damages. Suggests that international river basins may be uniquely vulnerable to flood hazards. Bakker, 2009 Limitations in capacity for preparedness a potential source of vulnerability in transboundary basins? 17
Integrated Flood Analysis System Global data: topography, land use, etc. Ground-gauged and/or satellite rainfall Courtesy of JAXA Model creation Run-off analysis Surfacemodel River discharge, Water level, Rainfall distribution Aquifermodel River course model 17
IFAS Satellite rainfall data Several remotely-sensed precipitation products have global coverage. Resolution (time and space) and observation accuracy are low compared with ground observation rainfall. Product name 3B42RT CMORPH QMORPH GSMaP Builder NASA/GSFC NOAA/CPC NOAA/CPC JAXA/EORC Coverage 50N~50S 60N~60S 60N~60S 60N~60S Spatial resolution 0.25 0.073 0.073 0.1 Time resolution 3 hours 30 minutes 30 minutes 1 hour Delay of delivery 6 hours 18 hours 3 hours 4 hours Coordinate system WGS Data archive Dec. 1997~ Recent 1week Recent 1week Dec.2007~ Data source (sensor) Aqua/AMSR-E, AMSU-B, DMSP/SSM/I and TRMM/TMI and IR TRMM/TMI, Aqua/AMSR-E, AMSU-B, DMSP/SSM/I and IR TRMM/TMI, Aqua/AMSR-E, DMSP-F13-15/SSM/I, DMSP- F16-17/SSMIS, IR data SATELLITE RAINFALL IS NOT A SUBSTITUTE FOR AN OBSERVATION NETWORK! 19
IFAS Satellite rainfall data calibration Rain event with slow movement of rainy area present 1hour later 2 hours later 3hour total Rain event with quick movement of a rainy area present 1hour later 2 hours later 3hour total > Satellite rainfall (GSMaP) (mm/3h) correction method to describe the difference of rainfall events in terms of wind speed Ground-based / Satellite rainfall 2 5 8 7 6 1 3 9 4 quick Slow Movement of rain area Yoshino river after 雨域の移動速度 ( 停滞度 ) 指標とピーク雨量比率の関係 1 9 3 5 48 2 9 1 3 8 2 6 7 before 5 6 Ground rainfall (mm/3h) 7 20
IFAS Satellite rainfall data calibration Corrected GSMaP is more accurate than raw data. User may specify any calibration equation. Ground gauged rainfall GSMaP(original) GSMaP(corrected) 3 hourly rainfall in Taiwan on typhoon Morakot, 2009 Aug.08 20-23(UTC) 21
IFAS Satellite rainfall data calibration Sendaigawa River length =137km Basin Area =1,600km 2 Correction of satellite data is successful Discharge (m 3 /sec) 4,000 3,500 3,000 2,500 2,000 1,500 1,000 500 Ground-gauged rainfall Satellite rainfall(3b42rt) (original GSMaP) (corrected GSMaP) Calculated discharge(ground) (3B42RT) (original GSMaP) (corrected GSMaP) Measured discharge 0 20 40 60 80 Rainfall (mm/hour) 0 9/6 9/7 9/8 9/9 9/10 Date (2004,GMT) 100 22
IFAS Satellite rainfall data calibration Kikuchigawa River Length =71 km Basin Area =996 km 2 Why was correction of satellite data unsuccessful? 2,500 0 Discharge (m 3 /sec) 2,000 1,500 1,000 500 Ground-gauged rainfall Satellite rainfall(3b42rt) (original GSMaP) (corrected GSMaP) Calculated discharge(ground) (3B42RT) (original GSMaP) (corrected GSMaP) Measured discharge 20 40 60 80 Rainfall (mm/hour) 0 7/4 7/5 7/6 7/7 Date (2006,GMT) 100 23
IFAS Satellite rainfall data calibration Discharge [m3/s] 3500 3000 2500 2000 1500 1000 500 successful case : Sendai river MWR observation Ground gauged rainfall GSMaP Corrected GSMaP Calculated discharge (GSMaP) Calculated discharge (Corrected GSMaP) Observed discharge 0 100 9/6 0:00 9/6 6:00 9/6 12:009/6 18:00 9/7 0:00 9/7 6:00 observed 9/7 12:009/7 18:00 9/8 0:00 Date (UTC) MWR frequently 0 10 20 30 40 50 60 70 80 90 Rainfall [mm/h] Discharge [m 3 /s] 1600 1400 1200 1000 800 600 400 200 No MWR during peak rainfall unsuccessful case : Kikuchi river 0 100 7/4 0:00 7/4 6:00 7/4 12:00 7/4 18:00 7/5 0:00 7/5 6:00 7/5 12:00 7/5 18:00 7/6 0:00 Date (UTC) MWR observation Ground gauged rainfall GSMaP Corrected GSMaP Calculated discharge (GSMaP) Calculated discharge (Corrected GSMaP) Observed discharge 0 10 20 30 40 50 60 70 80 90 Rainfall [mm/h] Accuracy of rainfall area distribution depends on frequency of MWR observation (& accuracy of IR motion vectors) Image Source : JAXA Ozawa et al. (2010) 24
Forecasting Inundation for Early Warning a) 2 year b) 5 year c) 10 year d) 25 year e) 50 year f) 100 year Bicol River basin, Philippines: simulated inundation extent and depth for a) 2-year, b) 5-year, c) 10-year, d) 25-year, e) 50-year and f) 100-year return periods. 25
Forecasting Inundation for Early Warning The Rainfall-Runoff Inundation (RRI) model Rainfall-Runoff Model: simulating streamflow discharge with rainfall input. Rainfall-Runoff Model River Routing Model: tracking flood wave movement along an open channel with upstream hydrograph. River Routing Model Flood Inundation Model Flood Inundation Model: simulating flooded water spreading on floodplains with inflow discharge. 26
Traditional calibration with discharge NSE=0.58 Discharge (m 3 /h) Discharge (m 3 /h) Rainfall (mm/h) Inundation map (2009-2010), Pampanga river basin. Observed and simulated discharge (2009-2010) at Arayat station. 18
Calibrating by spatial extent of inundation: RRI vs MODIS 0.3 NDVI 0.24 0.29 NDFI1 0.25 22-30/09/2009 Ondoy 8-16/10/2009 Pepeng 8/10/2010-1/11/2010 Juan 22-30/09/2009 Ondoy 8-16/10/2009 Pepeng 8/10/2010-1/11/2010 Juan 0.30 LSWI 0.25 0.14 MLSWI 0.10 22-30/09/2009 Ondoy RRI MODIS 8-16/10/2009 Pepeng 8/10/2010-1/11/2010 Juan 22-30/09/2009 Ondoy 8-16/10/2009 Pepeng Comparison of inundation extent from RRI model and MODIS images using different NDSIs during selected flood events in Pampanga river basin. 8/10/2010-1/11/2010 Juan 19
Sumary 1 Hydrologic forecasting for early warning Objective to achieve long(er) lead time with confidence (accuracy, or at least understanding of uncertainty) in datalimited basins. 2 Remotely sensed precipitation data We still need ground-gauged precipitation, but do we need the same information as before? 3 Spatial inundation patterns Calibrating/validating hydrologic models using discharge or inundation through space and time? 29