Application of Satellite Remote Sensing for Natural Disasters Observation

Transcription

1 Application of Satellite Remote Sensing for Natural Disasters Observation Prof. Krištof Oštir, Ph.D. University of Ljubljana Faculty of Civil and Geodetic Engineering

2 Outline Earth observation current state and trends Increase of resolution temporal, spatial, spectral Imaging capabilities and data access Copernicus Earth observation and natural disasters Landslides Floods Drought

3 Earth observation is more than Google Earth Google Earth Aerial photography Multispectral images High resolution satellite images Radar Hyperspectral images Lidar

4 Where are the satellites?

5 Active Earth observation satellites Belward, Alan S. in Jon O. Skøien Who launched what, when and why; trends in global land-cover observation capacity from civilian earth observation satellites. ISPRS Journal of Photogrammetry and Remote Sensing 103: doi: /j.isprsjprs

6 Satellites by country 6% 32% 62% Civil Military Commercial

7 Increase of resolution Spatial, spectral, temporal

8 Image spatial resolution

9 Spectral, radiometric and temporal resolution

10 Spatial resolution

11 WorldView-3 31 cm panchromatic 1.24 m multispectral 3.7 m short wave infrared resolution 8 Multispectral bands 8 SWIR bands

12 Increase of data quantity Cloud computing and big data

13 Data amount Sentinel 2 Europe ~ 180 TB per year

14 Data access policy is changing Open access

15 Opening of the Landsat archive

16 Satellite data is more and more accessible Copernicus Free, full and open data All data is available for free, regardless of use Freely available satellite imagery will improve science and environmental-monitoring products Wulder, Michael A and Nicholas C Coops Make Earth Observations Open Access. Nature 513:

17 Copernicus European Union Programme based on satellite Earth Observation and in situ data

18 Copernicus Global Monitoring for Environment and Security GMES European information services based on satellite Earth Observation and in situ data Overall funding by the EU and ESA has reached over 3 billion Large part dedicated to the development of satellites (Sentinels) Emergency Management Service natural or man-made disasters

19 Sentinel satellites Sentinel 1: all-weather, day-and-night radar for land and ocean services Sentinel 2: multispectral high-resolution imaging mission for land monitoring Sentinel 3: multi-instrument mission for sea and land surface observation Sentinel 4: atmospheric monitoring from geostationary orbit Sentinel 5: atmosphere from polar orbit

20 Small satellites Complementing big systems

21 Planet Planet successfully launched 149 Dove satellites Largest satellite constellation ever PlanetScope covers 150 mio km 2 per day 3.7 m resolution Blue, green, red, near infrared Image all of Earth s landmass every day

22 Disaster observation examples 22

23 Disasters Natural and technological disasters are causing huge damage and loss of lives They are more and more frequent

24

25

26 Current state There is a huge and diverse number of end users that need rapid mapping data public authorities civil protection fire fighters public

27 Space and Major Disasters Charter operational November 2000 founding agencies: Canadian Space Agency (CSA) Centre National d Etudes Spatiales (CNES) European Space Agency (ESA) more than 15 space agencies and partners organizes data capture and delivery: solidarity almost 500 activations in 15 years

28 Space and Major Disasters Charter

29 Copernicus Emergency Management Service Copernicus EMS - Mapping timely and accurate geospatial information derived from satellite remote sensing completed by available in situ or open data sources Digital or printed map outputs Support geospatial analysis and decision making processes of emergency managers It can be activated only by authorised users Rapid Mapping consists of the on-demand and fast provision within hours or days products are standardized Risk and Recovery Mapping On-demand provision of geospatial information

30 Copernicus Emergency Management Service

31 Copernicus Emergency Management Service

32 Mangart landslide First Charter activation

33 Mangart landslide two landslides (15 and 17 November 2000) liquefaction of material 8 15 m/s 7 dead huge damage: destroyed village two bridges vital for the valley disappeared hydroelectric power plant under meters of material landscape in the valley completely changed

34

35 Data capture 13 images used (8 acquired through Charter) 5 ERS (both ERS-1 and 2) 2 RADARSAT 4 SPOT (two PAN and two MS) 2 Landsat images

36

37

38 Land use open agricultural water bushes mixed forest impact area landslide area deciduous forest coniferous forest individual houses built-up %

39 Usefulness of the Charter The study has proved the usefulness of the Space and Major Disasters Charter. It has shown that remote sensing can be used to estimate the damage and, under suitable conditions, also in rescue operations. Mail delivery, used in this first study in the frame of the Charter, was far too slow.

40 Flash floods of 2007 torrential rains on 18 September 2007 rainfall reached 300 l/m 2, 80 l/m 2 in 50 minutes 6 people died estimated damage over 233 million houses, roads, bridges, cars factories flooded health care centre destroyed electricity supply affected, telephone connections disturbed

41

42

43 Satellite imaging

44 Flooded area classification Attributes Flooded {yes, no} Land cover Satellite images: SPOT MS, SPOT PAN, vegetation index NDVI Relief: elevation, slope, curvature Distance from water Decision tree Accuracy more than 95%

45 Mapping 41 maps were produced 7 overview (1: :150,000) 34 detailed (1:25:000) format A4 for easier handling on the field

46 Space and Major Disasters Charter findings Space and Major Disasters Charter is operational proved to be very useful response times are extremely short mostly based on the voluntary basis end user has limited influence on acquisition plan delivery speeds and quality of imagery were sufficient for damage assessment activities the delay in imaging/mapping/delivery is still an issue for rescue operation

47 Floods September 2010 heavy rain 1/5 to 1/4 of yearly precipitation in three days almost in the whole country half of the country problems with flooded areas Southern part of Ljubljana flooded estimated damage 240 million

48

49 Satellite data more then 20 images received (11 archive) mostly used: 4 Radarsat-2 (12.5 m) 3 Envisat (12.5 m) 2 Landsat (30 m) RapidEye (6.5 m) also used: DOF IR (0.2 m) DEM 12.5 m

50 Received and used satellite data Radar ENVISAT (Lj) RADARSAT-2 (Lj) RADARSAT-2 (Lj) RADARSAT-2 (Lj) ENVISAT (Krka)RADARSAT-2 (Zasavje) Optical Landsat (Lj) RapidEye (Lj) DOF IR (Lj)

51 Landsat and

52

53 RapidEye Bands 543

54 RapidEye Bands 542

55 Mapping comparison Radarsat-2 (12.5 m) RapidEye (6.5 m)

56 Radarsat-2 RapidEye

57 Validation of water detection problems Radar over-detection in agitated relief (noise, shadows) under-detections in heterogenous environments (urban, agricultural) no detection discontinuous floods pattern high plants (corn, trees, shrubs) fields are typically long and narrow Optical water has different spectral signature (dependent on soluables and depth) discontinuous floods pattern high plants (corn, trees, shrubs) fields are typically long and narrow

58 Drought mapping

59 Drought mapping Photo: Mary Sheft

60 Damage structure in Slovenia 19% drought hail 12% 8% flood storm (strong wind) 15% landslide and snowslide other 26% 20% Source: SURS, 2009

61 Data processing workflow

62 Data Ground data April October Satellite data Ground characteristics 21 meteorological stations MERIS and MODIS DEM landuse crop type forest type available water capacity

63 Satellite data MERIS RR 1P MERIS FR 2P MERIS, NDVI and FAPAR, 16 days (Vito) MODIS, bands 1-7 MODIS 250 m, NDVI and EVI, 16 days MODIS 1000 m, LAI and FAPAR, 4 days

64 Selection of sample points regular, 250 m grid more than 56,000 sample points all data must be available all data must be relevant visually inspected by human expert reduction more than 2000 points

65 Sample points for labelling

66 Drought labelling selected 2008 points manually labelled expert knowledge: field observations, archived records of drought impacts, meteorological measurements, agrometeorological modelling, and satellite data, including vegetation indices drought severity levels: severe drought drought no drought no data

67 Location Nova Gorica MODIS signal MODIS data top LAI bottom NDVI Nova Gorica location (SW Slovenia)

68 Data modelling for drought detection General architecture for data modelling General architecture for drought detection

69 Drought detection performance evaluation data: 500 evaluation points seasons 2006 to drought events non-drought events high accuracy: over 90% for non-drought over 92% for drought Land use type nondrought detection drought detection fields 82.4 % 94.1 % plantations 88.2 % 88.6 % olive 88.6 % 91.3 % vineyard 87.3 % 95.1 % coniferous forest 96.3 % 88.5 % deciduous forest 95.6 % 98.2 % mixed land use 92.1 % 92.8 %

70 Drought 2012 development Build up, rocks, water Drought Non-drought

71 Drought

72 Satellite-data-based drought detection system Satellite data processed and aggregated at a 250 m grid Study period extended to 2014 Human expert labelled drought on a subset of 2008 grid-points Over 90% for non-drought and over 92% for drought overall accuracy Drought maps produced for every five days for the period Cumulative drought maps per year Temporal evolution and spatial distribution of droughts

73 Conclusions

74 Natural disasters mapping with satellite data situation maps of larger areas within hours immediate publishing on internet time series (daily monitoring during/after the event) radar data is weather independent optical data gives better identification of geographical phenomena ancillary data is important land use, digital elevation model, hydrography fast mapping is standard accurate classification takes time data validation is necessary: local knowledge, local data (DOF, lidar)

75 What can be expected? monitoring every day or every few hours integration of high resolution radar and optical satellite data (resolutions from 0.5 to 10 m: more accurate location and impact) mapping with data from airplanes, helicopters and/or UAVs (drones) almost automatic procedure from the activation to publishing the results in one place

76