Michael Ramsey

1
Advances in Urban Ecosystem Science using ASTER

• Michael Ramsey
• University of Pittsburgh
• Department of Geology Planetary Science
• IVIS Laboratory

2
Outline

• Urban science
• motivation?
• integration into NASA Research Science Strategy
• The role of ASTER
• Urban Environmental Monitoring STAR
• acquiring ASTER urban data
• Methodology
• data classification
• target cities, hazards applications
• Results Conclusions

3
Why Cities??

• The role of humans
• integral components of ecosystems, both driving
  biogeophysical change and effected by these same
  changes
• logical starting point to gain understanding of
  ecosystem processes in human-dominated systems
• The role of remote sensing
• synoptic characterization and monitoring urban
  land cover change, degree of landscape
  fragmentation, heat islands, air/water pollution
• land cover data and spatial metric
  characterization important
• analysis of urban climate, energy and mass
  fluxes, hazard assessment, and ecosystem change

4
ALM STAR (cont)
5
Relevance of Urban Science

• Present justification
• ever-increasing interest in the science/policies
  of the urban environment (NASA, NSF, EPA, others)
• efforts will directly impact the largest
  percentage of a countrys population / dollars
• Future urgency (2025)
• estimated that 2/3 of the global population will
  be urbanized
• gt 5 billion people WRI, 1996 UN, 2001
• majority of the fastest growing urban centers are
  located in semi-arid, coastal and/or fragile
  environments
• vulnerable to natural hazards, ecological/economic
  degradation
• forcing changes in land use surface
  transformations

6
Research Strategy Science Questions
Variability
Forcing
Response
Consequence
Prediction
Precipitation, evaporation cycling of water
changing?
Atmospheric constituents solar radiation on
climate?
Clouds surface hydrological processes on
climate?
Weather variation related to climate variation?
Weather forecasting improvement?
Global ocean circulation varying?
Changes in land cover land use?
Ecosystem responses affects on global carbon
cycle?
Consequences in land cover land use?
Transient climate variations?
Global ecosystems changing?
Surface transformation?
Changes in global ocean circulation?
Trends in long-term climate?
Coastal region change?
Stratospheric ozone changing?
Stratospheric trace constituent responses?
Future atmospheric chemical impacts?
Ice cover mass changing?
Sea level affected by climate change?
Future concentrations of carbon dioxide and
methane?
Motions of Earth interior processes?
Pollution effects?
7
ASTER Data

• ASTER is scheduled (unlike Landsat TM)
• due to the large data volume
• 8 duty cycle during the lifetime of the Terra
  spacecraft
• 50 of resource time is allocated to the one-time
  global map
• Science Team Acquisition Request (STAR)
• dedicated to large global science objectives
• demand larger resources from the instrument
• 25 allocation of the total instrument time over
  6 years
• example STARs include
• volcano monitoring, land ice, global deserts,
  coral reefs, deforestation observations,
• urban science Urban Environmental Monitoring
  (UEM) STAR

8
UEM Program

• Implementation strategy
• 100 cities targeted globally
• population increasing from 1 million
• current or expected sprawl
• most located in arid regions ( 75)
• both high low priority targets
• Designed as a collaborative effort
• global partnerships, data dissemination,
  education
• assures that data are collected, calibrated and
  archived
• provides a point of contact to the ASTER team
• allows for feedback
• local scientists studying local issues

9
UEM Target Cities (U.S.)

• City State Priority City State Priority
• Albuquerque NM High Los Angeles CA High
• Anchorage AK Low Miami FL Low
• Atlanta GA High New York NY Low
• Baltimore MD High Phoenix AZ High
• Chicago IL Low Salt Lake City UT High
• Dallas TX High San Diego CA High
• Denver CO Low San Francisco CA Low
• Detroit MI Low Seattle WA Low
• El Paso TX High St. Louis MO Low
• Houston TX High Tucson AZ High
• Las Vegas NV High Washington DC High

10
UEM Example Las Vegas, NV
ETM 30m/pixel
11
UEM Progress to Date

• Collaborations
• 29 institutions, 45 urban centers, 23 countries,
  more than 60 independent research projects
• Publications
• 25 journals/conference abstracts/proceedings
  papers
• UEM program description in Earth Science in the
  Cities - AGU Special Monograph, Ramsey 2003
  (in press)
• After 30 months of ASTER data
• 91,761 level 1B (L1B) scenes collected
• 2,908 scenes within 60 km (1 ASTER scene) of a
  UEM city
• 2,434 are below 25 cloud cover
• 656 centered directly on a target city

12
All ASTER L1B Scenes
13
Project Web Sites
14
Project Web Sites
15
Project Web Sites
16
Project Web Sites
17
Project Web Sites
18
Project Web Sites
19
Project Web Sites
20
Project Web Sites
21
Project Web Sites
22
Project Web Sites
23
Project Web Sites
24
ASTER Scene Locator (U. Pittsburgh) http//aster.e
ps.pitt.edu/
UEM total scenes (MODIS background)
25
Urban Growth

• Land cover/Land use
• land transformation cycle f (time, wealth,
  environment, )
• sprawl vs. land-grabs
• response?
• population, environment, local/regional climate
• monitoring provides critical data for GIS-derived
  models
• infrastructure modifications
• utility needs
• economic development
• vulnerability of the population to natural
  hazards and environmental damage
• several case study examples

26
Landsat-based Classification

• Landsat coverage
• MSS (1974 -1980) TM (1985 -1998) ETM (1999
  -2000)
• Maximum likelihood classification performed
• using calibrated reflectance and vegetation index
  (SAVI)
• land cover classes used
• water, undisturbed, vegetation, and disturbed
  (with subclasses commercial/industrial, compacted
  soil, mesic/xeric residential)
• Additional data sets
• texture image is derived from the TM data
• land use, water rights, city boundaries, etc.
• Scene reclassified using boolean logic Stefanov,
  Ramsey and Christensen, Rem. Sens. Environ., 2001

27
Land Cover Classification
Cluster NDVI texture data into classes (ISODATA)
Level 1B VNIR Data
28
Hypothesis Testing

• Improving accuracy through ancillary data
• decision-based rule classification can be used to
  refine remote sensing classification

Final Classification
Hypothesis Testing
Land Cover
Initial Classification
60 active vegetation 20
concrete/ asphalt 10 tar roofing 10
water
Mesic residential class
29
Ancillary Data Land Use
30
(No Transcript)
31
Land Cover Classification Matrix
Low Vegetation Moderate Vegetation High Vegetation
Low Texture Bare Soil/Low Vegetation, Roadways Moderate Vegetation High Vegetation
Moderate Texture Low Density Urban, Roadways, Dry Washes Moderate Vegetation High Vegetation
High Texture High Density Urban Moderate Density Urban Moderate Density Urban

• Vegetation and texture value groupings obtained
  from ISODATA classification
• Water classified using VNIR, SWIR, or TIR
  spectral radiance values
• Boolean logic rules used in expert classifier to
  obtain final pixel classifications
• Model can be refined using additional spectral
  and ancillary information

32
San Francisco, CA June 14, 2000

• Patch diversity (f) (number of classes per unit
  area)/ ASTER pixels
• Unit area (250 m x 250 m) ASTER pixels
  1089
• Percent area (relative to ASTER scene) for each
  index value calculated to allow comparison
  between urban centers

33
Fragmentation Analysis Results

• High landscape fragmentation of the urban regions
  is present in over half of the metropolitan areas
  examined
• Highest regions
• African (A), Asian/Indian (B), and European (D)
• reflects the high population density
• megacities
• The results presented here are based solely on
  ASTER data
• classification accuracy of 75 to 80
  overall
• however spatial and temporal coverage of the data
  is not uniform

34
UEM Case Study Sites

• Phoenix, AZ
• presence of LTER project resources
• excellent setting for remote sensing
• one of the fastest growth rates in the US for the
  past decade
• existing air- and space- borne data sets
• Pittsburgh, PA
• comparative opposite end-member to Phoenix
• climate, land cover, growth rates, environmental
  issues
• São Paulo - Rio Claro, Brazil
• comparative opposite end-member to Phoenix
• mega-city growth rates, climate,
  geo/environmental hazards

35
Calibrated Data Landsat TM
Calibrated Reflectance (Landsat TM) Phoenix, AZ
(1993)
36
(No Transcript)
37
(No Transcript)
38
Phoenix Urban Heat Islands
39
UEM Case Study Sites

• Phoenix, AZ
• presence of LTER project resources
• excellent setting for remote sensing
• one of the fastest growth rates in the US for the
  past decade
• existing air- and space- borne data sets
• Pittsburgh, PA
• comparative opposite end-member to Phoenix
• climate, land cover, growth rates, environmental
  issues
• São Paulo - Rio Claro, Brazil
• comparative opposite end-member to Phoenix
• mega-city growth rates, climate,
  geo/environmental hazards

40
UEM Case Study Sites

• Pittsburgh, PA
• population decline for the past 3 decades (within
  the city)
• evacuation of industry (large brown-field sites)
• environmentally-sensitive locations
• urban renewal project sites
• rapid sprawl in suburbs
• relatively constant population in the metro
  region
• initial land cover classification
• focus of ASTER/MTI study using seasonal change
  detection
• geo-hazard implications
• landslides, flooding and waterway pollution

41
(No Transcript)
42
Image Sharpening Other Data

• Multispectral Thermal Imager (MTI)
• DOD instrument
• 15 spectral bands (0.4 - 11.5 microns)
• 5 - 20 meter spatial resolution
• restricted data use
• Approved urban targets
• Pittsburgh, Phoenix, Rome, Calcutta São Paulo

MTI (5m VNIR) false color composite Downtown
Pittsburgh, PA
43
Cross-Sensor Calibration
44
UEM Study Pittsburgh, PA
VNIR Color Composite
Vegetation Index (NDVI)
45
ASTER Pittsburgh, PA
ASTER L1B VNIR (8/19/02)
46
Land Cover vs. Slope Landslide Flooding
Hazards
slope stability?
valley flooding?
47
UEM Case Study Sites

• Phoenix, AZ
• presence of LTER project resources
• excellent setting for remote sensing
• one of the fastest growth rates in the US for the
  past decade
• existing air- and space- borne data sets
• Pittsburgh, PA
• comparative opposite end-member to Phoenix
• climate, land cover, growth rates, environmental
  issues
• São Paulo - Rio Claro, Brazil
• comparative opposite end-member to Phoenix
• mega-city growth rates, climate,
  geo/environmental hazards

48
UEM Study São Paulo, Brazil

• Population vulnerability in mega-cities
• Brazilian-US, Sustainable Urban Environment
  Project (SUEP)
• consortium composed of
• Universidade de Campinas (UNICAMP)
• Universidade de Sao Paulo (USP)
• Universidade Estadual Paulista (UNESP)
• Carnegie Mellon University (CMU)
• University of Pittsburgh (UP)
• Over-arching goals
• conduct research in the São Paulo region and
  Pittsburgh that will influence policy and
  practice in sustainable urban environmental
  development
• enhance research activities and outcomes for
  researchers in both countries

49
UEM Study São Paulo, Brazil

• SUEP Objectives
• Urban Sprawl
• using remote sensing (ASTER and MTI), GIS,
  survey, and social research methods to examine
  the planning options for managing sprawl in the
  region
• comparative targets Pittsburgh, Pennsylvania and
  the São Paulo to Campinas corridor
• Socio-economic
• developing remote sensing techniques that will
  aid
• understanding environmental equality
• food vulnerability
• population health
• other topics?

50
UEM Study São Paulo, Brazil

• Urbanized Brazil
• land cover classifications
• expert system of ASTER VNIR/SWIR
• NDVI
• GIS data layers
• initial issues/problems
• cloud/shadow mask
• variations in urban patch dynamics
• very poor ancillary data sets

ASTER VNIR (Rio Claro to Campinas Corridor)
51
Landsat ETM VIS (São Paulo, Brazil)
52
Application to Urban Hazards

• Vulnerability Fire
• dramatic increase of forest fires over the past 5
  years in the western US
• cost of over 1 billion dollars
• expansion of urban centers into previously
  unpopulated areas
• fire/flood hazards using remote sensing
• underway in Phoenix
• expand to other UEM cities
• using ASTER MTI data

53
Application to Urban Hazards

• Methodology
• field study of burned and non-burned regions
  surface properties
• soil type distribution
• vegetation cover re-growth
• changes in sediment transport patterns
• Purpose
• examine semi-arid brush fire scars
• assess the potential for future fire flash
  flooding
• integrate analysis into other ASTER urban data

54
Application to Urban Hazards
55
Application to Urban Hazards

• Initial results
• classification accuracy improved from 50 to 62
  using multi-wavelength data
• collecting sediment during rain events
• monitoring changes
• expansion
• Albuquerque, San Diego, El Paso

56
Application to Urban Hazards

• Vulnerability Air Quality
• particulate matter/pollution
• National Ambient Air Quality Standards (NAAQS 40
  CFR 50)
• PM10 (less than 10 ?m)
• allowable annual average concentration is 50
  ?g/m3 (24-hours)
• PM2.5 (less than 2.5 ?m)
• allowable annual average concentration is 15
  ?g/m3 (24-hours)
• health implications??
• pesticides, endotoxins, allergens, heavy metals,
  particulates

57
Application to Urban Hazards

• Vulnerability Air Quality
• completed project in Nogales, AZ region Stefanov
  et al., 2003
• examine the ability of remote sensing to identify
  
• dust generation, dust transport, dust
  depositional sites

58
Application to Urban Hazards

• Dust Pathways

Urban Canyons Paved and Unpaved Roadways Parking
Lots Traffic Industrial Sources
Agriculture Grazing Development Recreation Eolian
Processes
Natural Vegetation Golf Courses Agriculture Hillsl
opes
59
Application to Urban Hazards
winter
60
Application to Urban Hazards

• Air quality results
• accuracy of land cover classification is 74
  overall
• dust generation and deposition sites (81 95 )
• transport sites (44 61 )
• major causes of urban misclassification
• sub-pixel mixing
• spectral similarity with natural surficial
  materials
• significant change detected in land cover classes
  due to seasonal variation in vegetation cover
  (grasses)
• classified data useful for first-order
  assessments
• higher resolution data necessary for better
  accuracy

61
Urban Remote Sensing

• Conclusions
• land cover/use change is an important indicator
  for urban health
• high spatial/moderate spectral resolution data
  are critical
• tools are being developed to take advantage of
  the new data
• ASTER urban science has direct linkage into the
  NASA strategies
• variability, forcing, response, consequence,
  prediction
• community growth, public heath, disaster prep.,
  air quality
• where do we go from here?
• new initiatives in urban monitoring, mapping and
  science
• improvements in existing models
• hierarchical classifications
• spectral vs. spatial resolution
• time series analyses

62
Key Model Improvements

• Spectral vs. Spatial
• clear improvement with sub 5m/pixel urban data
• increased accuracy in urban land cover model (6
  land cover classes)

63
Conclusions

• ASTER data is available being used!
• providing a valuable resource for urban science
• in combination with historical records/ancillary
  data
• excellent spatial and spectral quality
• Quantitative land cover / land use analysis
• is possible with ASTER and other data sets
• improved accuracy with the inclusion of ancillary
  data sets
• individual classes 72 - 98
• overall 85
• involves a more sustained effort
• Enormous potential
• hazard assessment / growth issues
• local and regional climate change ??