A Proposal for Caribbean-Wide Land Cover/Use Mapping

1
A Proposal for Caribbean-Wide Land Cover/Use
Mapping

U.S. Geological Survey National Mapping
Division EROS Data Center
2
Developing a Regional Planning Framework for
Biodiversity Conservation, Disaster Preparedness,
and Economic and Agricultural Development In the
Insular Caribbean
A Proposal funded by the USAID Caribbean
Regional Office supported by FAO, NASA
The Nature Conservancy
U.S. Forest Service
U.S. Geological Survey International
Conservation Program International
Institute of EROS Data Center
Conservation Science Department
Tropical Forestry
3
WHY make Consistent Land Cover/Use Products for
the Caribbean?

• To respond to Strategic Objective 2 of the
  Caribbean Regional Program Strategy (USAID,
  2000) Environmental Management Improved
  environmental management by public and private
  entities
•  
• To enable integrated environmental and economic
  planning by
• providing improved access to information for
  regional and in-region planners and other
  stakeholders,
• bringing the spatial dimension to the development
  of best practices, and
• developing national capacity and management
  resources in an integrated and cooperative
  regional framework.

4
IMS Implementation
5
IMS Implementation
6
Integrated Land Cover/Use System for the
Caribbean Islands
Static Products Digital (CD, Internet) Hard
Copy
Documented and Adaptable Land Classification
Dynamic Products Digital (CD, Internet Map
Server)
Landslide Risk Flood Risk Basin
Delineation Vulnerable Populations Ridge to Reef
Analyses Terrestrial/Marine DEMs Commodity/Parcel
Certification Historical Land Change
Land Cover Land Use Urban Expansion Natural
Vegetation Gap Analysis Agricultural
Class. Forest Density Carbon Stocks
7
Final Products will Consist of

• High resolution land cover/use database maps
  with best georectification for each island
  (MRLC-like)
• Ancillary databases including where possible
• prior land cover maps, GIS data layers, digital
  elevation models, forest inventory data, soils
  maps, and others
• Consistently labeled and interpreted products for
  the Caribbean
• A flexible mapping framework for specific client
  refinement and modification
• Vector-based land cover units designed for
  conservation planning
• A network of knowledgeable users among the
  Caribbean Islands
• (An ArcIMS Data Querying Distribution System
  for each country and cooperating regional
  organizations)

8
WHO will do the work?

• Collaborating Institutions

• USGS, EROS Data Center
• USFS, International Institute for Tropical
  Forestry
• The Nature Conservancy,
• International Conservation Program
• National Cooperators
• Regional Organizations

9
Anticipated Work Flow An Iterative Process

• The Lead Institution Primary Analyst is
  responsible for coordinating all efforts for an
  island.
• EDC will have specific roles in
• training,
• image acquisition,
• processing all data inputs, and
• implementing a distributed data dissemination
  system

Secure, Process TM Images
Calibrate Subset Mask
Create Field Map
Plan/Perform Fieldwork
Spectral Data Stack
Field Map
Combine Create Metadata
Review Edit Extract
Digitize Rasterize Georegister Resample
Field Data
Training Data
Ancillary Data
Ancillary Data Stack
Metadata
Submit to EDC
Explain Rejection
All Data Stack
Review/Edit/Identify Corrections Create Improved
Training Data
Review
Reject Return
Accept
Generate Additional Data
Submit To Primary Analyst
Run Preliminary Decision Tree
Preliminary Classification Image
New Training Data
Submit to EDC
Review
Submit To Primary Analyst
Accept
Run Final Decision Tree
Final Classification Image
Review/Edit
Final Land Cover Maps Databases
EROS/USGS SCIENTIST
Final Land Cover Maps Databases
Final Land Cover Maps Databases
ArcIMS-based Distribution by Country
10
Acquire Landsat 7 ETM Imagery
Multiple date acquisitions are anticipated to
negate clouds and seasonal impacts.
11
Finished Land Cover/Use Maps and Databases

• After edits and reviews are incorporated by all
  interested local agencies, the final Land
  Cover/Use Maps and ancillary databases are
  released for use and distribution.

12
Distribution of Products

• Complete data set to collaborators and each
  island.
• Digital versions on CDs with convenient
  interface.
• A Caribbean web site for information and down
  load.
• Data for a Caribbean Digital Atlas.
• Source data available from web sites,
  collaborators and others.

13
Multiple Product Benefits
-National Planning/Management -Natural Resource
Monitoring -Regional Assessments
and Planning -Smallholder Development -Public
Relations, Advertising, and Tourism -Personal
Networking
Premium Value for Banana Certification
14
What do we gain with new and Consistent Land
Cover/Use Mapping?

• A standardized classification schema, high
  resolution land cover/use maps, and associated
  databases will
• Permit inter-island comparisons of the status and
  ecological variability of similar landscape
  units,
• Enable regional conservation strategies,
• Provide a baseline for regional biodiversity
  monitoring programs.
• Protect coastal environments,
• Enable regional land use planning
• Facilitate environmental management
• Support agricultural certification, and
• Enable change comparisons

15
Dont Land Cover/Use Maps Already Exist?

• Although maps of vegetation exist for many of
  the individual Caribbean islands, these are
  often
• out of date,
• erroneous,
• lacking information,
• frequently too coarse, both spatially and in
  classification, for many practical applications,
• unsuitable for diverse client needs, i.e. too
  categoric
• non-standardized in both classification and
  methodology,
• not integrated with ancillary databases,
• not incorporated in integrated applications,
• not in a framework that promotes upgrading and
  refinement
• not available to a general group of users via
  browsers, and
• not integrated into an Internet Mapping System
•  

16
Arent the Global Regional Maps Adequate?

• Global and regional land cover maps that already
  exist are not suitable for the Insular Caribbean
  because of
• Spatial Scales
• Classification Details
• Lack of Ancillary Databases
• Lack of Internet Map Serving

• Therefore, most existing maps are unsuitable for
  local terrestrial land use and coastal planning,
  monitoring, management or for comparisons of
  vegetation across multiple islands.

17
HOW do we propose to do this?

• Draw on existing methods developed by USGS for
  the National Land Cover project (MRLC)
• Landsat 7 ETM Source
• 30 Meter Pixel
• Standardized Projection
• Consistent Classification
• Large-area Applications
• Associated Databases

18
Landsat 7 ETM Instrument Characteristics
Band Spectral
Ground Number Range (micrometers)
Resolution (m) 1 .450 to
.515 30 2 .525 to .605 30 3 .630 to
.690 30 4 .750 to .900 30 5
1.55 to 1.75 30 6 10.40 to
12.5 60 7 2.09 to 2.35 30 Pan .520 to
.900 15 Swath Width 185 km Repeat Coverage
Interval 16 days (233 orbits) Altitude 705
km Quantization Best 8 of 9 bits On-board data
storage 375Gb (solid state) Inclination Sun-sy
nchronous, 98.2o Equatorial Crossing Descending,
1000am
/- 15 min. Launch
vehicle Delta II Launch Date April 15, 1998

19
Use On-Island Expertise
Identify and contact island specific agencies and
experts to enlist their interest, expertise, and
input.

• Those islands able to contribute to the process
  are encouraged to do so.
• Field validation
• Image interpretation
• Image processing
• Data distribution

Appropriate Land Use at Forest Margins
20
Obtain Existing Datasets
Acquire any available ancillary data (prior land
cover mapping efforts, GIS data layers, digital
elevation models, forest inventory data, soils
maps, etc.). Where such data are only available
in hardcopy format, they will be transformed into
digital formats.
21
Responsibilities of The Primary Analyst

• All ancillary datasets will be digitized,
  rasterized, and georegistered to the Landsat 7
  ETM imagery.
• Metadata describing each dataset will be
  generated.
• Metadata and Data will be Placed in Distributed
  Clearinghouses.

Support Parcel and Commodity Identity or
Certification
22
Responsibilities of The Primary Analyst (cont)

• An unsupervised classification will be used to
  create cloud, cloud shadow, and water masks.

Protect reef to ridge Systems and Environmental
Protection
23
Responsibilities of The Primary Analyst (cont)

• Secure Local Knowledge of the island to help
  identify spectral groups which correspond to the
  proposed land cover schema.
• This will provide a realistic check on spectral
  separability of land cover classes.
• This will guide productive field data collection.

Identify, Conserve and Protect Rare Communities
and Species
24
Responsibilities of The Primary Analyst (cont)
Prior Classification Proposed Classification
Agriculture Mixed Commercial and Subsistence Plantations
Elfin and Alpine Meadow Montane Cloud Forest and Montane Grasslands
Littoral Woodland Lowland Semi-deciduous Forest
Montane Rain Forest Montane Rain Forest
Montane Thicket Evergreen Montane Shrubland
Secondary Rain Forest Disturbed Submontane Rain Forest
Mature Rain Forest Submontane Rainforest
Semi-Evergreen Forest Lowland/Submontane Seasonal Evergreen Forest
Scrub Woodland Lowland Drought Deciduous Shrubland
Grasslands Short/Medium/Tall Grasslands
Swamp and Wetlands Seasonally Flooded Rainforest/Woodland/Grassland

• Islands with existing land cover maps will be
  crosswalked into the proposed regional land
  cover classification schema.
• Unchanged portions of the island may be used to
  provide training data for an intermediate
  classification on this new schema.

25
Responsibilities of The Primary Analyst (cont)

• Field data, collected via GPS, recent aerial
  photography, or other means, will be used to
  verify and correct both of these preliminary
  efforts at generating training data.
• A final training dataset will be assembled.

Assist Management of Natural Resources
26
Responsibilities of The Primary Analyst (cont)

• Provide all imagery, ancillary data, metadata,
  and training data for initial EDC processing.
• EDC prepares a preliminary classification using a
  decision tree linear regression technique.
• Identify local experts who can take advantage of
  sub-regional training as processing is done.

27
EROS Data Center will Process all Data
Input Database
Training Data
Land Cover Map
Slope
Multiple Dates and Spectral Bands of LANDSAT TM
ETM Imagery
Elevation
Position Index
Textual Expression
Decision Tree
If tm2b1 gt 30 and tm1b5 lt 45 and tm1b3 gt 30 and
tm2b4 lt 25 and tm1b1 lt 127 and slope lt 5 and
elevation gt 4500 and position lt 3 and aspect eq
north and soil eq 2 Then SHRUB
Aspect
Spatial Expression
Soil
28
Responsibilities of the Primary Analyst (cont)

• The land cover map resulting from this
  preliminary classification will be returned to
  primary analyst and each island for review,
  edits, and corrections.
• Corrections will be used to provide improved
  training data to EDC.
• EDC will use the new training data to run a final
  classification with the same decision tree
  technique. This land cover map will be the last
  submission EDC plans to make in the process.
• Any further classification efforts will be done
  by more traditional, supervised classification
  methods, by the Primary Analyst and Island
  entities.

29
Contributors to Caribbean Mapping

• IITF, USDA 225k
• NASA Data Buy 75
• IITF, USDA 180
• UN FAO Caribbean Training 15
• USAID/W, TNC (Dominica and Bahamas) 150
• USAID/Jamaica 70
• USAID/Dominican Republic 200
• EDC and TNC 27
• EDC/MRLC (Puerto Rico and U.S.V.I.) 100
• TOTAL 1,022k

30
Landsat TM Band/color Combinations for Image
Interpretation

• Bands 3,2,1
• True color composite
• No infrared band

31
Landsat TM Band/color Combinations

• Bands 4, 3, 2
• Color Infrared
• (false color composite)
• Similar to CIR Photos

32
Landsat TM Band/color Combinations

• Bands 7,4,2
• Arid regions maximum separation of rock types