flood mapping exploiting hr and vhr satellite images physical … · 2019-09-10 ·...

theia-land.frtheia-land.fr Dr H. YESOU, 2018

Flood mapping exploiting HR and VHR satellite images

Physical basis and contextual elements

Dr Hervé YESOU

French-Vietnam Remote Sensing Summer School

Remote Sensing of inundation, flooding and wetlands

Quy Nhom : 9 to 14 September 2019


Presentation outline

Introduction: Why water bodies and flood mapping and monitoring

Flood and lakes in the landscape

Short cut of Physical basis for Water bodies mapping• Interest for SWIR bands

Elements for water bodies extraction based on optical imagery

Optical sensors for water bodies and/or flood mapping• Medium• High resolution• VHR sensors

Rapid mapping examples


What are we are looking for: Flood events definition


• Floods: 34% world natural hazards between 1974-2003

• Half of affected people related to flood events

• Near 200 millions of affected people each year (more than half of affected people by a natural hazards)

• More than 170 000 deceases from 1980 to 2000

• Average: 4700 persons /year (2006-2015)

Source:EM-DAT - International Disaster Database

Why it is relevant to map and monitor flood events?


• Floods: worldwide

• Important mortality in Asia, Central-South America, Eastern Africa

• Important economic losses in Europe, Northern America as well as Asia

• Most dramatic are not the most costly ones (Nargis: 140 000 , none insurance prime, whereas 2008 spring floods in US and Germany 1,1billion $ each

Global Distribution of Flood Risk

(Source: Columbia University, UNDP, CRED)



• Closer period• 2016: (Nino Year) China, one flood

affected 60 million (134 million affected in 2010, 105 million in 2007 and 68 million in 2011).

200 000 homes damages > 5 Billion $ of damages

• 2018: • Kerala (India) flood event in summer

2018, near 500 casualties, 1 million of displaced people, 3 B US$

• With climate change it would become worse worldwide

• Fitting floods is one of the most important environmental challenge

Source:EM-DAT - International Disaster Database

2016



(SERTIT 2008)

• Floods: Europe

• Central Europe

• British Islands

• South France

(Europe 2002-2007: SERTIT 2008)


2002

2003

2004

2005

2006

2007inondation

feux de forêt

Incident technologique

Pollution

Tremblement de terre

Tempête

Glissement de terrain

Eruption volcanique

0

5

10

15

20

25

30


Flood patterns recognition

Event signature

Not only flood extent

But also associated features: mud, erosion process etc ..

Haïti Charter action

May 2004


Event signatures


Ireland SAFER 027, 2010: Water within bogs



Angola Charter action

Water within inter dune depression

May 2009



Okavongo delta and river

Niger Inner delta and river

Lakes and water bodies: Landscape variability


Lac du BourgetTonle Sap

LofotenSchiessrothried, Vosges, Fr



Boreal CanadaNorthern territories

Newfoundland



EO sensors for hydrology








Flood mapping examples


29 August 2016

29 August 201625 August 2016

Clear skySunny weatherSentinel 2 SPOT 6-7 VNREDSat-1Pléiades HR

Pasive Remote SensingOptical sensors


Short cut of Physical basis for Water bodies mapping

• Water absorbs the longer wavelengths of visible and NIR and SWIR domains Reflects the shorter wavelengths of the visible domain (blue, green)

More precisely water color depends on:

• Depth (ground influence sand/rocks)

• Materials in suspension

• Vegetation or algae


Short cut of Physical basis for Water bodies mapping


Water bodies: optically complex system


15 km

© Copernicus, 2015; Credits: ESA 2015, Processing SERTIT 2015

Sentinel2

2015-10-20

10 m

Water bodies: spatially complex system


High variability of spectral answer and contrast land/waterShenjian Lake, Anhui Province

Water bodies: temporally complex system


Classical colour composite

PIR, R, YG en RGB

XS3, XS2, XS1 in RGB

SPOT 4 10th September

2002

Optical Flood mapping : channel selection

©CNES [2002], distribution SPOT Image,

traitement SERTIT


SPOT 4, 10th September

2002

Better identification of the

flood affected area with:

NIR, SWIR, RED in RGB

Xi3, Xi4, Xi2 in RGB

Applicable with SPOT4&5,

landsat TM ETM,

VEGETATION

Future Sentinel 2

©CNES [2002], distribution SPOT Image,

traitement SERTIT

Optical Flood mapping : Contribution of the SWIR channel


T0

SPOT 4

Xi 3,4, 2 RVB 20m

10th September 2002

Flood imprint analysis

Soil drying

Multitemporal approach:

contribution of the SWIR channel for flood imprint mapping


T0 + 1 day

SPOT 4

Xi 3,4, 2 RGB 20m

11th September

2002

Flood extent monitoring

Flood imprint analysis

Soil drying

Multitemporal approach:

contribution of the SWIR channel for flood imprint mapping


Spectral basis for water bodies mapping: VIS & SWIRActual and future optical sensors more or less suitable for water surface mapping

SPOT5

400nm 500nm 600nm 700nm 800nm 900nm 1000nm 1100nm 1200nm 1300nm 1400nm 1500nm 1600nm 1700nm 1800nm 1900nm 2000nm 2100nm 2200nm 2300nm

VHR

HR

SWIR domain


Extraction of water bodies on:•raw channels, particularly the NIR, SWIR ones•Indices, NDWI, , AWEI, INH, SBI•Saturation or Hue indices of a HIS transformation

Methods of thresholding • Simple/double • Otsu • Kalman, Hysteresis filters• ...

Methods of classification •Supervised •None supervised •Oriented object methods•SVM•Random Forest• Deep leaning•…

Water bodies mapping based on Optical data

Ta

blea

u 1

: Va

riatio

ns d

es cou

leurs d

e différen

ts corp

s en ea

u en

pla

ine d

’Alsa

ce (g

ravières, co

urs d

u R

hin

, ba

ssins

po

rtua

ires) extraits d

e l'ima

ge SP

OT

-7 d

u 3

0 sep

temb

re 20

15

Variations in space of water bodies spectral signatures


Yearly time of submersion

Sentinel-2

PreProcessing

Validation

Images Features

characterisation

ESA SciHub

DEM

Band selectionMeta data

MasksMetadata+ métas

ExtractionDatabase

Water products

Data ingest

PrimitivesImage

features/patterns

Methods

Validation Database

Processing chain for Sentinel 2


Simple or multiple

thresholds--

Water mask+ metadata

Binary mask+ métas(GeoTiff)

SVM, classification

Indices:NDVI, NDWI, , AWEI, INH, SBI

Spectral space

IHS, ….

Primitives, features

S2 primitives’ image extraction

Thematic extraction



Simple or multiple

thresholds--

Water mask+ metadata


SVM, classification

Indices:NDVI, NDWI, , AWEI, INH, SBI

Spectral space

IHS, ….

Primitives, features

S2 primitives’ image extraction

Thematic extraction

False friendof SWIR signatures



Flood mapping based on classification from test areas

Performance analysis: a jungle

High water level


Performance analysis: a jungle

Low water level

Flood mapping based on classification from test areas


Sensors properties Lakes Characteristics Data processing

Spatial Spatial resolution Size and ShapeSize of the transition zones between dry and flooded areas.. Ice..

Spatial sampling

Radiometric Radiometric resolutionSpectral coverage (bands)

Spectral response of water/land, floating or submerged vegetation

Radiometric samplingSensitivity of the descriptors

Temporal Temporal resolutionAcquisition date

Seasonal water surfaces fluctuationsPeriodic/recurrent inundations

Co registrationNb exploitable images

Factors, spatial, radiometric and temporal, factors contributing to errors in water surface mapping

Lyons and Sheng, RS, 2016Lyons and Sheng, RS, 2016

Errors and sources of errors in water surfaces mapping

Errors distribution in function of water bodies size


Errors and sources of errors in water surfaces mapping

Azolla filiculoides

Nymphoides Pelatum

Shape /resolution

Floating vegetation

Definition of limits of flooded areas

Presence of vegetation/alga bloomPresence of Ice


La Coste pound, Saint-Julien, Côtes-d'Armor, FR, 04-2016

Azolla filiculoides

Potential limitation on water surface regognitionwater flooded vegetation and floating vegetation

41H. YESOU 2017

theia-land.fr Dr H. YESOU, 2018

Potential limitation on water surface regognitionwater flooded vegetation and floating vegetation

Wuchan lake, Anhui Province China

42H. YESOU 2017


Potential limitation on water surface regognition water: alga bloom


S2 8 November 2017 S2 13 November 2017

43H. YESOU 2017


Potential limitation on water surface regognition water: alga bloom


S2 8 November 2017 S2 13 November 2017








Rapid mapping example


Near 30 years of exploitation of EO data for water bodies mapping and monitoring

Improvement from one generation to another one • SPOT1-3 to SPOT4-5=> SPOT 6-7• SPOT => Pleaides VHR •S POT /Landsat => Sentinel2• MODIS => MERIS=> OCLI• ERS =>ENVISAT=> Sentinel 1A/B• HJ 1C => Chang Zheng 4C• Radarsat 1 => Radarsat 2• VHR SAR TerraSar X and CSK•Improvement in term of • Swath• Resolution• Radiometric quality• Revisiting time• Access to images• Derived products


Influence of spatial resolution


DMC Familly: DEIMOS 1 – Beijing 1 DMC 2G:2005 October 27: Beijing 12009 July 29 : Deismos 12009 July 29: UK DMC 22011 August 17: Nigeria Sat 2

22m 650 km swath DMC 1G:

Beijing 1 : 32 m, 600 km swath

3 bandsGreen: 0.52-0.60 µmRed: 0.63-0,69 µmNIR: 0.77-0,90 µm

Daily revisit for constellation

Bj 1: September 2007


Bj 1: March 2008

Poyang Lake, RP

China

DMC Family: Beijing 1


Bj 1: March 2008

Poopo Lake, Bolivia

2013-2016DMC Family: DEIMOS


Systematic acquisition16 days revisitHuge archive

Since Landsat 4-5 . SWIR band 30 m

Landsat Family


SPOT 1-5

SPOT1-3 3XS + 1 PAN 20m -10 m

SPOT4-5 4XS + PAN20 m + 10m (red band)10 m + 2,5m

Green: 0.53-0.60 µmRed: 0.61-0,68 µmNIR: 0.78-0,89 µmSWIR: 1.58-1,75 µm

PAN: 0.48-0,71 µm

Swath 60 km

•SPOT 1: 22 February 1986 / 17 November 2003•SPOT 2: 22 January 1990 / July 2009•SPOT 3: 26 September 1993 / November 1996•SPOT4: 26 March 1998 / •SPOT5: 4 May 2002 /31 March 2015

•=> Very rich archive, no so well known and exploited


SPOT 6-7

Following of SPOT 5 with improved spatial resolution 1,5-6m at nadir

2 satellites in constellation with Pleaides

Launch September 2012 and June 2014

4 bands + PanBlue: 0.45-0.52 µmGreen: 0.53-0.60 µmRed: 0.76-0,69 µmNIR: 0.76-0,89 µmPAN: 0.45-0,75 µm

Swath 60 km (agil)

=> Daily coverage capacity


SPOT 6-7

Flood in Bridgewater, England, Before event


SPOT 6-7

Flood in Bridgewater, England, 11 January 2014


SPOT 6-7

Lake Poopo, Bolivia24 March 2013- Spot 616 January 2016 – SPOT 7


SPOT5

400nm

500nm

600nm

700nm 800nm 900nm 1000nm

1100nm

1200nm

1300nm

1400nm

1500nm

1600nm

1700nm

1800nm

1900nm

2000nm

2100nm

2200nm

2300nm

Sentinel 2 like: Applicable to others optical sensors

SWIR


67

Sentinel2 mission

Launch date: Sentinel 2A: June 2,3 2015 , Sentinel 2B: March 7, 2017Nominal end of the mission: 2028 + S2 1D and S2 1C satellites (2035?)

Sentinel 2 main innovations:

• Nominal 10 days revisit , 5 days for constellation• Multiresolution: 10, 20, 60m • Continuity with SPOT and Landsat series• Systematic acquisition and open & free access

Mission objectives: part of Copernicus constellation:• provide systematic global acquisitions of high-resolution,

multispectral images, enhanced revisit frequency, for operational services and applications requiring long time series.

• Inland and ocean water; water quality monitoring • Monitoring polar environment: ice shelves, glaciers• Changing lands: agriculture/forestry/uran• Emergency response: floods, forest fires, landslides


Sentinel 2

Sentinel 2

• Highest Resolution same as SPOT5 (10m)

• Presence of two SWIR bands (heritage of landsat)

• Large swath (MERIS heritage)

• Revisiting time 10 – 5 days

• Free access

Sentinel-2A : on 23 June 2015Sentinel-2B : on 7 march 2017


Water Surface

Water Height

Copernicus missions (ESA) exploitable for hydrology


Sentinel2

Sentinel 2

• Resolution depending of the spectral coverage

Sentinel-2A : on 23 June 2015Sentinel-2B : on 7 march 2017


Sentinel2 images acquired over Poyang Lake area

155 images S2 acquired since October 2015 (at the 11-11-2017)Increase of acquisitions since the rump up phase of Sentinel 2B


http://flightlessboyds.blogspot.fr/2011/07/spies-in-sky.html

High resolution: military heritage

Spy dove escadron: 1903

Deutsches Museum Munich


PLEIADES2 satellites in constellationLaunch December 2011 and 2012

0,70 cm in PAN2,8 in XS

4 bands + PanBlue: 0.43-0.55 µmGreen: 0.50-0.62 µmRed: 0.59-0,71 µmNIR: 0.70-0,94 µmPAN: 0.47-0,83 µm

Swath 20 km (agil)

=> Daily coverage capacity


Water bodies: temporally complex system

POYANG

LAKE

PLEIADES

IMAGES


PLEIADES

Pléiades HR acquired on the 12 April 2013Over Poyang lake


PLEIADES

Pléiades HR acquired on the 5th of August 2013Over Poyang lake


Worldview 2 Safer action 42 Xynthia storm: coastal flooding

Optical VHR and flood mapping:very fine description of the flood field


AVANTAPRES

Lourdes

Mud traces recognition: within cities


Extraction of narrow water bodies

Identification of mud deposit

Impact on river pathway

Impact on agricultural field

Reference image, IGN BD topo

22 of June 2013

Hautes-Pyrénées: flash flood of

Gave de Pau (June 2013)

Pléiades days, Toulouse, 1srt of April 2014

Pléiades © CNES 2013, Distribution Astrium Services /Spot Image S.A., France

tous droits réservés. Usage commercial interdit. »

Optical VHR : parcelling and flood

VHR &flood mapping: Impact on agricultural parcels


Optical VHR : Dike break

Agly 2013 flood event

Based on Pleiades HR

Pre event image

Post event image acquired on the 9 March 2013

Optical VHR and flood mapping:very fine description of the flood impact on hydraulic elements


Optical VHR : Dike over flooded and partially erased

Xynthia 2010 – Western coast of France


Xynthia 2010

Western coast of France

Optical VHR : Dike over flooded and partially erased


Image pré-crise (Google)

Image post-crise (Pléiades)

Damaged infrastructure: Bridge

theia-land.frtheia-land.fr Dr H. YESOU, 2018Beijing December 2011

Flash flood in Algeria, October 2011


Image pré-crise (Google)

Image post-crise (Pléiades)

Détection des traces d’inondation : eau résiduelle, traces de boue

Krymsk , Rusia, july 2012

Mud traces recognition: within cities


Morombe, affected individual housingViewed with Pleiades HR

The Haruna Cyclone that hit Madagascar the 22nd of February 2013 caused much damage due to violent winds plus rising river and ocean waters.

More than 20 people were killed,Affected population: > 20 000

•Toleria•Morombe•Sakaraha

National Office of Risk and Disaster Management (BNGRC) supported by French Civil Security(DGSCGC), requested the International Charter to assist rescue organization

Madagascar Haruna Cyclone- February 2013- Charter call 434


1er Produit Crise12H UTC

22/02 23/02 24/02 25/02 26/02 27/02 28/02

CHARTER CALL 434 – MADAGASCAR

91

Evènement

Demande d’activation15H30 UTC

Activation19H34

Suite cartes crise18H23 UTC

En parallèle des acquisition SPOT et Pléiades HR, Tasking de Radarsat 2, TerraSAR X, et des THR USAcquisitions du ont eu lieu de 26 février au 2 mars.

MAD SPOT 57H00 UTC

DP PLD HR7H15 UTC

DP SPOT 510H00 UTC

MAD PLD HR10H00 UTC

Acquisition criseSPOT 5 (7H) et

PLEIADES (7H23UTC)

Transfert Sat – Terre Soirée

Essai MADSPOT Réunion18H UTC - ERR

Gestion de crise

Catastrophe

Désignation PM9H10 UTC

ProgrammationPléiades 10H00 UTC

Moyens satellitaires

Call fermé le 01/03 à 14H UTC

theia-land.frtheia-land.fr Dr H. YESOU, 201893Pléaides days, 04-01-2014, Toulouse

Large scale mapping thanks to SPOT5 image acquired by SEAS- OI

Levees breaksInundated villages

Madagascar Haruna Cyclone, February 2013 -Charter call 434


Latest rapid mapping action Madagascar

Haruna Cyclone, February 2013

Tolaria, affectedindividual housingViewed withPleiades HR


Morombe, affected individual housingMapped using Pléiades HR

Focus on affected towns based on Pléiades HR

Pléiades © CNES 2013, Distribution Astrium Services /Spot Image S.A., France tous droits réservés. Usage commercial interdit. »

Madagascar Haruna Cyclone, February 2013 - Charter call 434


Case of Gave de Pau flash flood

18-19 of June 2013: flash flood on the Gave du Pau (Pyrenees, South France)

Acquisition RushProduction None Rush

18-06-2013 14h

• Wednesday 19: Request of Caisse Centrale de Réassurance

(Ministry of Finance) & Acceptation by CNES RTU team

• Thursday 20: Definition of AOIs and satellite programming

• Saturday 22: First attempt of acquisition

• Sunday 23rd: Second attempt of acquisition

• Monday 24th: Image selection & production and reception at SERTTIT (18h30)

• Mardi 25: Thematic exploitation, validation, export shape (format define by CCR)

• Wednesday 26: export kmz, maps products and WEB site (restricted access)


Definition of AOIs: demonstration scheme

97

Initial request20-06-2013 à 11h30


18-06-2013 14h


First attempt of acquisition: Saturday 22 of June 2013

More or less cloudy




Second attempt of acquisition: Sunday 23rd of June 2013

Very cloudy




Definition of the four segments to be produced

•Orthez area : 22 km of river bed

•Peyrefitte – Nay area: 48 km of river bed

=> 3/5 of Gave de Pau within the AOIs are covered




AVANTAPRES

Confluence gaves de Pau et de Cauterets




AVANTAPRES

Lourdes




Affected areas such as observed on Pleiades HR imagesKMZ format as requested by CCR





Thematic information also provided as space maps 1:15 000 at

theia-land.frtheia-land.fr Dr H. YESOU, 2018 105

Zoom on space map


Flood plain mapping exploiting HR and VHR satellite images

Physical basis and contextual elements

Dr Herve YESOUJuly 2018


Yangtze river’s monsoons lakes monitoring

Health of Yangtze is a major concern for 400 000 000 of inhabitants as a fresh water resource.

The river basin gives - 70% rice production- 40% cereal production- 40% industry

- Biodiversity stakes

Climate fluctuation and man activities (ie Three Gorges dam) could have significant impact.


January 2004: 1000 km2 August 2004: 4000 km2

Poyang Lake, RP China

Bj 1: 2 July 2006Bj 1: 2 March 2008

Monsoon lake: important annual variations of water surface

700 Km2 3500 Km2


Example of water body monitoring: Poyang

+550 images

A mixed resource

Initially: Mix of EO dataNow two major resource Sentinel 1 et 2


Water extent monitoring: Poyang

1 2 2 3 3 3 4 5 5 6 7 5 2 1 2 2 2 3 3 3 1 6 2 2 1 2 2 2 2 5 5 2 3 2 20

5

10

15

J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D

Date

Am

ou

nt o

f D

ata

500

1000

1500

2000

2500

3000

3500

2004 2005 2006

Su

rface (km

²)

2002 2003 2004 2005 2006 2007 2008 2009

?

?

2010

Dragon 2 objectives: Continue and complete water surfaces’ monitoring

3GD event


Water extent monitoring: Poyang

1 2 2 3 3 3 4 5 5 6 7 5 2 1 2 2 2 3 3 3 1 6 2 2 1 2 2 2 2 5 5 2 3 2 20

5

10

15

J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D

Date

Am

ou

nt

of

Data

500

1000

1500

2000

2500

3000

3500

2004 2005 2006

Su

rface (

km

²)

2002 2003 2004 2005 2006 2007 2008 2009

??

2010

Dragon3objectives: Continue and complet water surface monitoring


SENTINEL 1

The Sentinel-1 series : part of the GMES programme

Sentinel1A, 2014 Sentinel1B, 2016

Priority : ensure continuity for C-band data

Improvement of SAR signal (30% better than ENVISAT)

Multi mode

• Strip map: 80 km swath , 5m

• Interferometric Wide swath mode IW, 250km

• Extra wide EW Swath , 400 km , 25x100 m

• Wave mode, WV, low data rate, 5x20m

• Swath 250 km

Polarisation modes:

• VV or HHi n wave mode

• Selectable dual pol for all other mode HH+HV; VV+VH

113H. YESOU 2017


20140921 20140926

20141003 20141008

20141015

20141015

2014110120141108

20141020

2014111320141120

2014112520141202

2014120720141214 20141219

20141226

Sentinel 1: High temporal revisit T0 , +5, +7


First Sentinel 2 Time series over Poyang

© Copernicus, 2015; Credits: ESA 2015, Processing SERTIT 2015

Water draw off over Nanjinshan natural reserve

15 km

23-10-201520-10-201503-10-201513-09-2015


Request to a secured resource allowing to monitoring large areas with a reduced revisiting time (10 – 15 days)

Moving from MR to HR SPOT 4&5 TakeFive, HJ1A, preparing Sentinel 2 venue Archive TerraSAR, New modes TerraSAR TandemX Cosmo Skymed from ASI (supporting Envisat Gap) Sentinel 1A First Sentinel2 First Sentinel1B

Dragon 3


Poyang lake water surface monitoring:

Regional analysis and global interactions


-0,80

-0,60

-0,40

-0,20

0,00

0,20

0,40

0,60

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Poyang water surface

Centrée réduite

Water surface Statistical analysis

Centred reduced

2000-2005 : positive

2006-2009 : negative

2010_2015 : variations from one extreme to another

WP7 : Regional and global interactions


In literature draw off; Mid September, mid OctoberDraw off becomes very early over the years with a shortness of the inundation

period First time observed in mid August 2016 In 2011 very short flooding period, max in 25-06 In 2013, redraw in mid-July In 2015 same behaviors, max flood extent in end of June

25-06 28-0615-08 09-07

WP7 : Regional and global interactions


Water extent monitoring: Submersion time: residual analysis


Thank you for your attentionMerci pour votre attention

http://sertit.unistra.fr/

[email protected]

http://sertit.unistra.fr/

flood mapping exploiting hr and vhr satellite images physical … · 2019-09-10 ·...

Documents