Global Flood Partnership Annual Conference, Guanzhou, China 11-13 - - PowerPoint PPT Presentation

Risk financing against floods: an application leveraging on global models and EO data

Global Flood Partnership Annual Conference, Guanzhou, China

11-13 June 2019

Roberto Rudari, CIMA Research Foundation Joost Beckers, Deltares Patrick Matgen, LIST

The Application

• Immediate support and relief to population in

SEA

• Risk financing options: Parametric insurance

The Application

Risk Profile

• Hydrologic/hy

draulic scenarios modelling

• EO based

Validation

• Impact

scenarios modelling NRT Trigger

• Model Based

estimates

• EO based

estimates

• Merged

estimates Access to Information

• Platform

implementatio n

• Operations

maintenance

The Application

The Interface

The Interface

Risk Profile

Hazard maps and scenarios

GLOFFIS for riverine flood hazard GLOSSIS for coastal flood hazard

Unosat image 2015 50 yr flood map 100 yr flood map

Risk Profile

Hazard maps Validation

Risk Profile

Event maps validation

Risk Profile

Impact Scenarios validation

NRT Flood Analyzer

Telemetry Satellite imagery Global models

1 2 …

BM

Best Match Algorithm Best Flood Map

• Gauge level similarity:

S2GA = (hfloodmap - hgauge - hbias)2/stdev(hfloodmap)2

• GLOFFIS level similarity: S2GL = (hfloodmap - hGLOFFIS - hbias)2/stdev(hfloodmap)2
• EO map similarity:

SEO = (Npospos+Nnegneg) / Ntotal

• Ntotal = Npospos+Nnegneg+Nposneg+Nnegpos
• Total similarity = S2GA + S2GL + (1-SEO)2
• Weights can be added (but not used so far):
• Total similarity = WGAS2GA + WGLS2GL + WEO(1-SEO)2

Water level (masl) Flood extent (km2) T=1.5 T=2 T=5 T=10 … Model result Gauge reading Flood extent from EO image

Best Match: Map selection

NRT Flood Analyzer

BM

GUF GHS-JRC WPOP Best Flood Map Population Layers NRT Affected People Estimates

The need

After 1 year of pre-operation:

• Strengthen the EO component
• Concentrating on SAR data processing

Challenges:

• Operation in NRT
• Geographical scope (full country coverage)
• Reliability (compared with the 7-day window)
• Urban areas

e-DRIFT

Virtual Platform:

• Directly based on the

DIAS

• Accessible through

Machine-to-machine protocols

• Providing added value

EO services

Back Office

The flood Archive Algorithm

WASDI Processing Output User

Change Detection (per tile) 1. Clip Input Raster based on 100 km x 100 km reference grid 2. Compute difference image S1i-S1i-1 3. Mask blind spots 4. Parameterize distribution functions 5. Thresholding & region growing to generate maps of positive/negative changes

∆WB-

i-1:i

Water bodies mapping (per tile) 1. Compute WBi 2. Subtract permanent water layer

∆WB+

i-1:i

S1i S1i-1

Permanent water bodies Reference grid (i.e. Tiles) Blind spots layer

Flood record Sentinel-1 data hub WBi WBi-1 Archiving water bodies maps / floodwater maps

Reference image selection Systematic query to identify and retrieve Sentinel-1 imagery

• ver tiles

Mosaicking

The flood Archive Algorithm

Methodology: flood mapping in Myanmar

S1i S1i-1

∆WB+

i-1:i

∆WB-

i-1:i

WBi-1 WBi

«Blind spots» Permanent Water Floodwater

Methodology: flood mapping in Myanmar

The SEA DRIF eDRIFT Integration

EO Augmented Historical Scenario

External data sources

Hydrological data Operational Models EO Flood Maps Authomatic Flood Scenarios Matching Selected flood scenarios Enhanced Population Density Layers Impact assessment Population affected/ Economic impact estimates Historic Flood Scenarios Selection Historical Scenarios Selection Man Made Scenarios Synthetic Flood Scenarios External Trigger / Man made scenario detection Manual Flood Scenarios Matching EXTERNAL SOURCES SEA DRIF COMPONENTS eDrift SERVICES

Combined use of Model and EO data

Sentinel-1, August 11, 2015

Rapid assessment by matching of EO and model simulation E.g. 2015 monsoon floods, Myanmar

T=20 modelled flood map

Flood impact NRT estimation

Built-up area: World Settlement Footprint based on Landsat and Sentinel1 Critical infrastructure from various sources:

• Airports
• Seaports
• Railways
• Roads
• Hospitals
• Fire stations
• Police stations
• Schools
• Government buildings

police station railway station flooded road village flooded railroad

GRAZIE