Remote Sensing and GIS Integration Workshop

1
Remote Sensing and GIS Integration Workshop

• Barry Haack, George Mason University
• October 19, 2004
• GeoTech2004
• Potomac Region American Society for
  Photogrammetry and Remote
• Sensing
• Silver Spring, MD

2
Welcome and Logistics

• Introduction
• Bound handout, note appendices
• Time 930 to 1230, break 1030 - 1050
• Workshop evaluation please

3
Assumptions, Caveats, Disclaimers

• Assume GIS experience
• Directed towards GIS analyst interested in
  remote sensing
• Will not include aerial photography,
  photogrammetry, base map creation
• Emphasis on spaceborne remote sensing for
  thematic information

4
Objective and Outline

• Explore the relationship between spaceborne
  remote sensing and Geographic Information Systems
  (GIS)
• Definitions/background of remote sensing and GIS
• Current status of spaceborne remote sensing
• Remote sensing input of thematic data to a GIS
• GIS data layers assist in remote sensing
  information extraction

5
Geographic Information Sciences (GISci)

• Geographic Information Systems
• Remote Sensing/Photogrammetry
• Global Positioning Systems/surveying
• Cartography
• Spatial Statistics
• Mapping Sciences, GeoInformatics, Geospatial
  Sciences?

6
Remote Sensing (RS)

• Collection of information without direct contact
• Remote sensing is a primary source of data for a
  GIS
• maintains a historical record of the
  Earths surface
• provides current information
• allows for change detection and
  predictive models

7
RS Information Extraction Procedures

• Visual/human/interpretation from hard or soft
  copy product
• Computer/automated from digital data
• Many hybrid or combination techniques

8
Geographic Information Systems (GIS or GISys)

• Multiple definitions
• System to input, store, manage, analyze and
  output spatial data
• Includes hardware, software, data,
  infrastructure, staff
• Raster and/or vector
• Error - accuracy

9
Remote Sensing Roles for GIS

• Base maps
• photogrammetric considerations
• generally air photo based (perhaps
  hyperspatial spaceborne)
• great spatial detail
• contours, transportation, buildings,
  infrastructure, utilities
• Thematic information
• single or multiple classes
• often generalized
• focus of this workshop
• spaceborne platforms

10
Air Photo George Mason University
11
Air Photo Derived Base Map
12
Color Infrared Film
April 5, 1988 NAAP VG B VR G NIR R
13
Two Directional Interaction of Remote Sensing and
GIS

• Remote sensing creates data layers for a GIS
• GIS data layers assist in remote sensing analysis
  and classification
• Remote sensing image/photos may be a backdrop
  layer in a GIS

14
Major Issues RS Integration to GIS

• Geometric rectification to coordinate system
• Cartographic generalization - scale compatibility
• Data structure (raster - vector)
• Error - accuracy

15
Resolution in Remote Sensing

• Spatial, degree of spatial detail, often in
  meters, pixel size
• Spectral, number and types of energy
  bands/wavelengths
• Temporal, frequency of data acquisition, days or
  hours
• Radiometric, level of discrimination in energy
  recorded
• Concept of resolutions useful for
• remote sensing data evaluation
• data specifications for
  informational needs

16
Spatial Resolution
17
Remote Sensing Classification Parameters

• Platform
• Energy type (Electromagnetic Spectrum- EMS)

18
Remote Sensing Platform

• Height above surface
• Airborne or spaceborne
• Historically tradeoff between footprint and
  spatial resolution
• low altitude, small footprint and high
  spatial detail
• high altitude, synoptic view and low
  spatial resolution
• Exceptions in national assets data and recent
  high spatial resolution
• spaceborne platforms

19
Remote Sensing Platform Tradeoff Spatial
Resolution vs Footprint/Synoptic Coverage
20
Early Platforms 1903
21
Electromagnetic Spectrum

• Classified by wavelength and/or frequency
• Inverse relationship between wavelength and
  frequency
• Units in micrometers (one one-millionth of a
  meter)
• Reflected or emitted energy
• 04 .4 .5 .6 .7 1.5 4.5
  300 1m
• ultraviolet visible
  infrared microwave (radar)
  
• B G R near mid
  thermal

22
Electromagnetic Spectrum
23
Energy Flow Profile

• Energy source
• Source to surface
• Interaction at surface
• Surface to sensor
• Sensor to user
• All wavelength dependent and site/time specific
• Bi-directional Reflectance Distribution Function
  (BRDF)
• Atmospheric corrections necessary or possible?
• Signature extension problem (over space and time)

24
Components Of EFP Wavelength, Time and Location
Dependent
25
Spaceborne Remote Sensing

• Experimental and research
• Operational
• National assets (dual purpose)

26
Operational Spaceborne Remote Sensing

• Medium spatial resolution multispectral (10 to
  100 m)
• Radar
• High spatial resolution (lt10 m)
• Low spatial resolution multispectral (gt100m)
  includes meteorological
• Hyperspectral

27
Multispectral Sensors

• Match feature to a spectral signature
• Cameras
• Scanners or wisk broom
• Linear array or push broom
• Expanding to hyper and ultraspectral sensors

28
Selected Spectral Signatures-Reflectance Curves
29
Medium Spatial Resolution MSS (10-100 m)

• Landsat
• SPOT
• IRS
• JERS

30
Landsat

• US system
• Seven platforms since 1972 (six successful)
• Primary sensors MSS, TM, ETM

31
Landsat Orbit Parameters

• 570 mile or 920 km height
• 16 to 18 day repeat coverage
• Near polar NE to SW orbit
• 81 north to 81 south
• Sun synchronous 930 am
• Archived by global path/row location

32
Landsat Multispectral Scanner MSS

• On first five platforms
• Four band scanner - VG, VR, NIR, NIR
• 79 by 56 m pixel, 1.1 acre, 0.4 hectare
• Standard frame, 185 by 185 kms, 115 by 115 miles
• 7.5 million pixels per frame
• 18.5 cm print at scale of 11,000,000, enlarge to
  170,000
• Digital data, varied radiometric resolutions

33
Landsat Thematic Mapper TM

• Since 1982, Landsats 4 and 5
• Seven spectral bands
• VB,VG,VR,NIR, MIR, TIR,MIR
• 30 meter pixel, 120 m TIR
• 256, 8 bit radiometric resolution
• Enhanced Thematic Mapper ETM
• Landsat 7 1999
• Seven bands
• Panchromatic band at 15m
• About 600 digital data
• Serious data problems since May, 2003

34
Landsat TM 30 m Atlanta
35
SPOT

• French
• Five platforms since 1986
• Linear array or push broom system
• SPOTs 1 to 3
• 10 m panchromatic, 20 m three band multispectral
• 60 by 60 km format
• Pointable sensor, stereo and greater temporal
  resolution
• SPOT 4 1998
• Added fourth MSS band (Mid IR 1.5 to 1.75)
• SPOT 5, 2002
• 2.5 and 5 m panchromatic at 60 km swath
• Vegetation mapper on 4 and 5 at 1km, daily
  coverage

36
SPOT 20 m MSS
37
Attributes of Spaceborne Data

• Synoptic view
• Global
• Repetitive
• Uniform over time
• Uniform over space
• Often multispectral
• Digital
• Planimetric
• Available, open sky
• Inexpensive?

38
Visual Interpretation of Spaceborne Data

• Do not expect perfect understanding
• Use ancillary information
• Minimize joy of recognition
• Spatial extendibility
• Convergence of evidence
• All changes in tone/texture are real
• Importance of seasonality
• Field work
• Accuracy assessment

39
Radar

• ERS
• JERS
• RADARSAT
• Shuttle Imaging Radar SIR
• Almaz

40
Advantages of Radar

• Day and night
• Weather independent /cloud penetration
• Vegetation and surface penetration
• Determine distance

41
Radar Backscatter

• Aspect/geometry
• Composition/dielectric constant
• Texture/roughness
• Radar is good for shape/form
• Optical is good for composition

42
Change Radar Return

• Wavelength (X,L,C, K)
• Polarization (quad-pole, HH,VV, HV,VH)
• Look angle
• Look direction
• Strategies with single band radar
• Multidate
• Multi incidence angle
• Derived measures such as texture
• Fuse with optical

43
RADARSAT

• Canadian sensor
• 4 November 1995 launch
• C-band, 5.6 cm, HH polarization
• Programmable incident angle, spatial resolution,
  and swath/footprint
• Spatial resolution from 8 to 100 m
• Footprint from 50 x 50 km to 500 x 500 km

44
Radarsat Kathmandu Nepal
45
Radar Applications

• Areas of cloud cover
• Geology/geomorphology/structure
• Ice and snow cover - iceberg monitoring
• Oceanography - oil spills
• Cultural features
• Deforestation
• Land use/land cover
• Agriculture

46
Large Spatial Resolution MSS (gt 100 m)

• Meteorological Sensors (AVHRR)
• NOAA Platform
• 5 spectral bands (visible to thermal)
• 1.1 km spatial resolution at nadir
• 2800 km swath
• 6 hour temporal resolution with two
  platforms
• SPOT Vegetation Mapper
• Global daily coverage
• Four bands
• 1 km pixel, 2200 km swath

47
Non-Meteorological Applications of Met Satellites

• Oceanographic temperature and color
• Oil spills
• Geologic thermal inertia, volcanoes
• Forest and grassland burning
• Monitor dust storms (volcanic)
• Vegetation assessment (NGVI), desertification
• Crop Monitoring

48
AVHRR Nile Delta Temperature Difference
49
Fine Spatial Resolution (lt 10 m) Hyperspatial

• Space Imaging IKONOS, 1999
• 1 m panchromatic, 4 m three band MSS
• 11 x 11 km footprint
• 3-5 day temporal resolution
• ImageSat EROS 1-A, 2000
• 1.8 m pan, 13.5 x 13.5 km footprint
• Digital Globe QuickBird, 2001
• 0.6 m pan and 2.6 m MSS,1-3.5 days,
  16.5 km swath
• SPOT 5, 2002
• 2.5 and 5 m panchromatic, 60 km swath
• Orbital Imaging OrbView-3, 2003
• 1.0 pan, 4 m four band MSS, 8 x 8 km
  footprint

50
IKONOS August 5, 2002 1 and 4 m merged
51
QuickBird Washington DC .6 m
52
Hyperspectral

• Various definitions, gt 10 wavelengths, often
  hundreds
• Developed for geologic/earth materials scientists
• Several airborne systems exist AVIRIS
• Spaceborne
• MODIS Hyperion

53
Hyperspectral Image Cube
54
MODIS

• NASA, November 1999
• 36 bands visible green to thermal
• Varied spatial resolutions
• bands 1 and 2 250m
• bands 3 - 7 500m
• bands 8 - 36 1000m
• Swath of 2330 km
• One to two day temporal repeat
• Many applications and derived information

55
(No Transcript)
56
NASA Hyperion

• November 2000
• 30 m
• 220 bands
• 7.5 km swath
• Orbits with Landsat 7

57
Two Directional Interaction of Remote Sensing and
GIS

• Remote sensing creates data layers for a GIS
• GIS data layers assist in remote sensing analysis
  and classification

58
Major Issues RS Integration to GIS

• Geometric rectification to coordinate system
• Cartographic generalization - scale compatibility
• Data structure (raster - vector)
• Error - accuracy

59
Geometric Rectification

• Image to image (registration) or image to grid
  (rectification)
• Coordinate system (UTM, Lat/Long, State Plane)
• Ground Control Points GCPs
• from image/map or field GPS
• Transformation level (first, second)
• Error (RMS in pixels, under 1 pixel optimal)
• Resampling (nearest neighbor, bilinear, cubic
  convolution)

60
Cartographic Generalization

• All maps are generalizations of the real world
  (RF gt 11)
• Maps should indicate smallest feature (minimum
  mapping unit - mmu)
• GIS only valid at smallest scale or largest mmu
  of any single data layer
• Often remote sensing digital data are too
  detailed
• Spatial filtering, clumping or spatial
  aggregation necessary

61
Data Structure

• Raster
• traditional remote sensing
• ease for modeling
• Vector
• traditional cartographic
• Vector and raster conversions
• Issues
• display
• storage

62
Accuracy Assessment

• Extremely important, data without accuracy of
  questionable value
• Accuracy should be a component of metadata
• Very difficult and often avoided, embarrassing
• Expensive
• Several types of accuracy, locational and
  thematic
• Thematic the most difficult to evaluate
• Primary difficulty is identification of truth,
  temporal differences
• Extensive literature on statistics of accuracy
• Accuracy often summarized in error or confusion
  matrix
• Overall accuracy, users and producers per class

63
Contingency Table Sudan

• Urban Veg Other Totals
  Users
• Urban 15,248 335
  1,502 17,085 89.2
• Agriculture 2,012 3,015 1,159 6,186 48.7
• Other 934 200 21,961 23,095 95.0
• Totals 18,194 3,551 24,622 46,367
• Producers 83.8 84.9 89.2
• Correctly Identified Pixels 40,225/46,367
  86.8

64
Example GIS Data from Remote Sensing

• Biophysical variables
• DEM, biomass, temperature, soil
  moisture, snow and ice
• Point and Line features buildings,
  transportation, hydrology
• (high spatial resolution may
  be required)
• Thematic features
• single layers wetlands, urban,
  water, forest, agriculture
• multiple classes land use/land cover
• Change detection
• Remote sensing images may be a GIS backdrop layer

65
Remote Sensing as Thematic Input to a GIS

• Visual mapping from photos/images and input to
  GIS
• by digitizing or scanning (may be hard or
  soft copy)
• Use digital RS data to visually update existing
  GIS layers
• Automated classification of digital RS data to a
  GIS layer

66
Visual Image Interpretation

• Geometric rectificatoin
• before or after interpretation
• creation of mosaics/image maps
• Classification system (single or multiple land
  use/cover)
• USGS Anderson, Hardy, Roach and Witmer
• Class definitions
• Minimum mapping unit
• Hardcopy or softcopy data sources
• Conversion to GIS
• direct softcopy, digitizing, scanning
• Accuracy assessment

67
USGS Profession Paper 964, 1976, Anderson, Hardy,
Roach and Witmer
68
USGS Land Use/Cover Map
69
LU-LC Afghanistan
Source Image Landsat TM at 1250,000
70
LULC Kabul Province
71
Land Use/Land Cover Kathmandu, Nepal
72
Digitally Update Existing GISLayer

• Geometric registration
• Appropriate dates of sources
• Compatible classification system
• Viable minimum mapping units
• Accuracy assessment/quality control
• Example use RS to update USGS DLGs

73
Update DLG from Air Photo
74
Digital Image Processing

• Preprocessing
• Analysis

75
Preprocessing

• Data selection, subset, merge
• Radiometric
• missing data
• striping
• atmospheric correction
• Geometric
• registration to grid or image
• scale, resampling

76
Analysis

• Image enhancement
• Automated classification

77
Image Enhancement

• Provide improved image for visual information
  extraction
• Spatial
• texture
• filters
• edge detection
• Spectral
• band selection
• ratio
• principle components
• green vegetation index

78
Automated Classification

• Signature matching process
• Classification system selection and definition
• Difficulties
• Signature not unique or too unique
• EFP atmospheric considerations
• Mixed pixels
• Signature extension issue (over space
  and time)
• Signature extraction (most important, GIGO)
• calibration-training sites or
  supervised (from GIS layers possible)
• clustering or unsupervised
• Signature evaluation (visual or statistical)
• multiple signatures per class
• Application of a decision rule
• Accuracy assessment
• Spatial filtering for GIS compatibility, product
  delivery

79
(No Transcript)
80
Automated and Filtered Classification Kathmandu,
Nepal
81
Classification Improvement Strategies

• Data
• Multisensor
• Multitemporal
• Ancillary data/GIS
• Texture/context
• Nested sampling/varied spatial
  resolution
• Procedure
• Decision rules
• Processing strategy
• Hierarchical
• Image segmentation

82
Radar and Optical Classification Kericho, Kenya
Forest-Red Tea-Yellow Bare Soil-Green Urban-purple
83
GIS Assists Remote Sensing

• Pre-classification image segmentation
• Calibration-training site selection
• Direct integration of GIS and RS data
• Post classification sorting

84
Pre-classification Image Segmentation

• Advantages exist to segmenting study area/data
  prior to analysis
• GIS layers may provide segmentation not possible
  via remote sensing
• Examples elevation, slope, soil type, historical
  land use/cover
• Each segment examined independently and then
  combined

85
Calibration-Training Site Selection

• Signature extraction for automated classification
• Existing GIS layers may be source or helpful
• Elevation, land use, vegetation, soil layers for
  example

86
Direct Integration of GIS Data with Remote
Sensing Data

• Ancillary information intrinsic in RS
  information extraction
• DEM, historic land use/cover,
  vegetation maps
• Add GIS layers as bands of data in traditional
  classification
• Logical channel addition
• Continuous and not categorical GIS
  data
• GIS layers in other procedures
• Classification and Regression Tree
  Analysis (CART)
• Neural networks
• Hierarchical strategies
• Can use categorical data

87
Post Classification Sorting

• Use GIS to finalize classes
• Forest class above 350 m is forest 13, below 350
  m is forest 16

88
Conclusions

• Integration of RS and GIS extremely useful
• RS and GIS integration is two directional
• Multiple existing uses and also multiple research
  topics
• Multiple new RS platforms and sensors in near
  future
• THANK YOU!
• Please complete workshop evaluation