USDA Forest Service Forest Inventory and Analysis (FIA)

1
USDA Forest ServiceForest Inventory and
Analysis(FIA)

• MRLC Land Characterization Partners Meeting
• Nov. 7-8, 2000

2
OUTLINE

• Federal mandates that FIA more effectively use
  remote sensing
• FIA Business needs from satellite data
• Classification detail
• Classification accuracy
• Geographic priorities
• Information needed by FIA Management Team

3
Federal mandates that FIA more effectively use
remote sensing

• 1998 Farm Bill
• White House Office of Science and Technology,
  Committee on the Environment and Natural
  Resources
• RAND Corporation review of forest monitoring
  conducted by federal agencies
• FIA Staff Director Rich Guldin http//fia.fs.fed.u
  s/library.htm - Papers

4
Improve consistency of data and process using a
top down approach

• Consistent data is like a common language
• Centralized data collection, documentation and
  dissemination
• Decentralized analyses and decision making
• Economies of scale

5
FIA Business needs from satellite data

• Stable, dependable and economical production of
  accurate and consistent forest cover and land use
  maps
• Cover entire USA every 3 to 10 years
• Adherence to Federal Geographic Data Committee
  (FGDC) standards

6
FIA Business needs from satellite data

• Automated image processing algorithms that
  require little human intervention
• Product consistency and accuracy
• Cost reduction
• Timeliness
• Diversity of geospatial products
• Henry Ford analogy

7
FIA Business needs from satellite data

• Improve accuracy of FIA statistics
• Improve statistical efficiency through
  stratification on forest v. nonforest cover
• Improve statistical estimates for small
  geographic areas (e.g., counties) using remotely
  sensed ancillary data

8
FIA Business needs from satellite data

• Improve timeliness of statistics in annualized
  FIA
• 10 - 15 of field plots re-measured each year
• Remotely sensed data refreshed every 3 to 5
  years
• This is a goal, not an absolute design
  requirement
• Could use change detection to update
  forest/nonforest in a 10-year MRLC product

9
FIA Business needs from satellite data

• Change detection
• Keep forest/nonforest map current to maintain FIA
  statistical efficiency through stratification
• 2005 update to 2000 landcover map
• Better identify spatial patterns of change in
  broad landscapes

10
FIA Business needs from satellite data

• Change detection
• Improve accuracy of FIA statistical estimates for
• Timber removals
• Reforestation
• Afforestation

11
FIA Business needs from satellite data

• Help provide 30-m/124,000 products to FIA
  customers
• User-friendly data base for GIS analyses
• Attractive maps for distribution
• Spatial analysis tool box (internal and external
  users)

12
FIA Business needs from satellite data

• Characterize context surrounding each FIA field
  plot that are not easily measured in field
• Landscape fragmentation
• Size and shape of forest stand
• Distance to roads, surface waters, other land
  uses (important components of wildlife habitat)

13
FIA Business needs from satellite data

• Substitute satellite data for 140,000 NAPP
• Reduce cost of FIA stratification with Phase 1
  plots (1-km grid)
• Continue to provide imagery for navigation by
  field crews
• 15-m pan-sharpened Landsat 7
• 10-m pan-sharpened SPOT
• Superimpose ancillary geospatial data (DLG, DEM,
  topos., etc.)
• Downloadable to field crews (federal, state,
  contractors)

14
FIA Business needs from satellite data

• Implementation schedule
• Prototype products available for 10 -20 of USA
  by September 2002
• Production system functional by September, 2003

15
FIA Business needs from satellite data

• New remotely sensed products in the future
• Net primary productivity or photosynthesis rates
• Tree mortality
• Indicators of drought, acidic deposition, or pest
  attack
• Boundaries between different forest stands
• Indicators of human infrastructure (e.g.,
  individual buildings)

16
FIA Business needs from satellite data

• Developers tools to implement a variety of
  spatial models with centralized database
• Linkages to other geospatial databases (e.g.,
  Census Bureau)
• Sharing geomatic models
• Facilitate local improvements to national map
  products
• Accuracy
• Classification detail

17
Minimum spatial resolution

• 1-km pixel for global/national assessments
• 250-m to 30-m pixel for regional assessments
• FIA definition of forest requires 30-m scale
• Special assessment needs require 30-m scale
  (e.g., riparian management zones)
• Functionality request
• change spatial scale of data to balance
  assessment needs with technology

18
Classification detail

• Might need separate MRLC products for forest
  cover and timberland use
• Forest v. nonforest (most valuable for
  statistical efficiency through stratification)

19
Classification detail

• FIA definition for forest uses
• 10 stocking, which can be applied with field
  data but not directly with remotely sensed data
• At least 1-acre and 120-foot wide
• Includes non-stocked clearcuts and
  seedling/sapling stands
• Accuracy of remotely sensed classifications need
  to be high, but not necessarily 100

20
Classification detail

• FIA definition for nonforested land use includes
• Urban and suburban areas with tree cover
• tree stocking less than 10
• Pasture with tree cover
• Rangeland

21
Classification detail

• Broad forest types (global/national assessments)
• Softwoods
• Bottomland hardwoods
• Upland hardwoods
• Mixed hardwoods and softwoods

22
Classification detail More specific cover types

• Softwood forest
• White-red-jack pine
• Spruce-fir
• Longleaf-slash pine
• Loblolly-shortleaf pine
• Douglas-fir
• Hemlock-Sitka spruce
• Ponderosa pine
• Western white pine
• Lodgepole pine
• Larch
• Fir-spruce
• Redwood

• Upland hardwood forest
• Oak-hickory
• Maple-beech-birch
• Aspen-birch
• Western hardwoods
• Bottomland hardwoods
• Oak-gum-cypress
• Elm-ash-cottonwood
• Oak-pine
• Woodland
• Chaparral
• Pinyon-juniper

23
Classification detail

• Open v. closed stands
• Non-timber land use (e.g., urban with forest
  cover)
• Special categories
• Forested wetlands
• Mesquite
• Krummholtz

24
Classification detail

• National Forest System needs for Map Product 2
  (Forest Planning)
• Cover Type
• 30-35 categories of forest
• 6-10 categories of grass/forb/shrub types
• 6 non-vegetated categories (rock, snow/ice, etc.)
• Stand Size Class (5 categories)
• Stand Crown Closure Class (4 categories)

25
Classification detail

• National Forest System needs for Map Product 2
  (less detailed )
• Cover Type
• 9 categories of forest
• 4 categories of grass/forb/shrub types
• 5 non-vegetated categories (rock, snow/ice, etc.)
• Stand Size Class (2 categories)
• Stand Crown Closure Class (3 categories)

26
Classification detail

• Need to agree on detailed description
• Classification rules for each category
• Devil is in the details

27
Classification Accuracy

• Forest v. nonforest 90 to 99 accuracy
• Needed for stratification efficiency
• Inaccuracies caused by FIA field-definition of
  forest included with usual classification error
• No formal FIA accuracy standards for more
  detailed categorizations
• Known accuracy relative to FIA field data

28
Classification Accuracy

• National Forest System (Montana, Idaho) Map
  Product 2 (most detailed)
• 60-65 overall for cover types
• at least 40 for any individual class
• 40 overall for stand size class
• 60-70 for stand density classes

29
Classification Accuracy

• National Forest System (Montana, Idaho) Map
  Product 3 (less detailed)
• 75 overall for cover types
• at least 65 for any individual class
• 75 overall for stand size class
• 75 for stand density classes

30
Timeliness

• Less than 5 net change in forest cover since
  date of imagery
• stratification efficiency
• Less than 5 years old is desirable

31
Registration Accuracy

• Sufficient to link 1-acre FIA field plots to 30-m
  pixels

32
Geographic priorities Forest/non-forest mask
September 2002
33
Geographic priorities Forest/non-forest mask
September 2002

• Maine
• Iowa
• Indiana
• Minnesota
• Missouri
• Wisconsin
• Utah
• Arizona
• Colorado
• Oregon

• Alabama
• Virginia
• Georgia
• Kentucky
• South Carolina
• Tennessee

34
Geographic priorities Forest/non-forest mask
September 2003

• Arkansas
• Louisiana
• Tennessee
• Texas

• Pennsylvania
• Michigan
• Puerto Rico
• Hawaii

35
Information needed by FIA

• Cost to FIA for Part II of MRLC

36
Information needed by FIA

• Timing of coverage
• Will MRLC land characterizations always be 5 to
  15 years out of date?
• Can MRLC incorporate re-characterization or
  change detection in between 10-year MRLC cycle?

37
Information needed by FIA

• Classification detail
• Potential role of FIA in determining detail of
  classification system
• What decisions have already been made
• What is on the table?
• Need a thorough review of detailed classification
  descriptions and rules
• Can MRLC produce map of forest cover optimized to
  FIA definitions of forest land use?
• Consistency of MRLC and FGDC standards?