AN OPERATIONAL SYSTEM FOR DEFORESTATION ESTIMATION FOR CANADA

1
AN OPERATIONAL SYSTEM FOR DEFORESTATION
ESTIMATION FOR CANADA
Donald G. Leckie Canadian Forest Service,
Natural Resources Canada Victoria, British
Columbia
2
Forest/Deforestation Definition

• Forest
• Area 1 ha minimum area of forest
• Height 5 m at maturity
• Crown Closure 25 at maturity
• Deforestation
• Minimum Width 20 m tree base to tree base
• Clear permanent land use change
• Harvest not deforestation

3
Requirements

• UNFCCC - greenhouse gas (GHG) emissions due to
  deforestation annually 1990-present
• thus need 1970-present deforestation area
• Kyoto Protocol - greenhouse gas emitted annually
  2008-2012 due to deforestation from 1990 onward
• thus need 1990 - 2012 deforestation
• Convert area and type of forest to carbon other
  GHG emissions
• Track carbon GHG on deforested land after
  deforestation under the new land use (e.g.,
  agriculture)

4
Urban and Oil Gas Development
Oil and Gas
Suburban Development
5
Forestry and Extraction
Road Building
Gravel Pit Expansion
6
Agricultural Expansion
7
Hydro Reservoirs
Large reservoirs flooded in the 1980s in north
Quebec
The Eastmain-1 reservoir flooded 2006 is 600km2
8
Deforestation Analysis Flowchart
9
Deforestation Monitoring Strata
10
Area Sampled
1975-90 16.8 Mha 1990-2000 43.4
Mha 2000-2006 5 Mha Sampled from 264
Mha Plus 55.6 Mha N. Quebec

11
Scaling for National Level Reporting

• Deforestation data are scaled from samples to
  deforestation strata units, to 60 reconciliation
  units then aggregated to 18 reporting zones

12
Areas of Full Cover Mapping 1990-2000

13
Full EOSD/CTI Earthsat 2000 Coverage 1975
and 1990 coverages as well
14
Circa 1990 Winter Landsat Imagery
Blue winter image frames 110
images Deforestation strata in background
15
Circa 1990 Digitized Aerial Photography
Red - Federal Yellow -
provincial 70,000 photos Deforestation strata
in background
16
Deforestation Sample Plot Design
17
Steps

• Training, field calibration
• Interpretation mostly on contract with internal
  QC
• Real time QC by CFS
• Correction of interpretation biases, error types
• Full QC by CFS includes field checks
• Revision
• Final vetting
• Format checks before GIS processing

18
Deforestation Interpretation Guide

• The Guide is our bible for mapping methods,
  specifications and database structures
• Standardized methods are used during mapping
• Revised based on new realities, interpreter
  comments and suggestions

19
Rules for mapping deforestation for different
scenarios (combinations of areas of unit "C",
lengths and width "g" and "w", and cases of treed
or non treed for areas A, B, C and D).
20
Base Manual Interpretation
21
Ancillary Data

• Roads
• Pipelines and well pads (complete coverage)
• Wooded vs non-wooded from provincial national
  base maps
• Pits and quarry license areas
• Forest management zones
• Parks
• Forest inventory (provinces)
• Potential future (population census block data,
  911 emergency GIS database)

22
(No Transcript)
23
(No Transcript)
24
Event Interpretation

• Interpreter records pre-deforestation forest
  types, post-D classes, confidence in
  interpretation, request for validation/help
• Omissions, commissions, attribution and boundary
  errors, plus required field checks are noted in QC

25
Quality Control
26
Planned Flight Line and Targets
27
Actual Flight Path and Photos
28
Area of Field Validation (Saskatchewan Prairie
Fringe)

29
Saskatchewan Validation Mission Fights (July 2004)
30
Records Data

• Records data were used in part for
• - forest roads
• - hydro lines
• - hydro reservoirs
• - oil and gas pipelines
• - well pads
• Data are from published and internal reports and
  individuals within responsible agencies
• Records data analyzed to determine completeness
  and to extract the portion being Kyoto Protocol
  deforestation

31
Uncertainty Assessment

• Expert judgment to determine omissions and
  commissions (20), boundary errors (20) and
  sampling errors (25) supported by pilots and
  Monte Carlo sample design study
• These errors are combined to solve uncertainty
• U (O/C2 D2 SE2)0.5
• or (.22 .22 .252)1/2 0.38 as a
  preliminary estimate
• Where O/C Combined Omission/Commission Error
• D Delineation Error
• SE Sampling Error

32
Time Trend Interpolation
33
Deforestation Estimates for Canada

34
Deforestation by Industrial Class

• In normal years, D is dominated by agricultural
  clearing with lower amounts of urban development
  and forest roads

• When large hydro reservoirs are built, they
  dominate the national total (for area)

35
Number of Deforestation Events by Size Class

36
Deforestation Area by Event Size Class

37
Carbon Estimation

• NRCAN Canadian Forest Service
• D mapping, records extraction and compilation
• D event carbon estimation (CAT CBM-CFS3)
• D portion of NIR Report production

• Agriculture and AgriFoods Canada
• Post-D agricultural GHG calculation

• Environment Canada
• Assembly of estimates from different sectors
• Non-Agricultural Post-D GHG calculation
• Compilation of full D carbon estimates
• NIR Report production

38
Deforestation GHG Estimates

39
Principles

• Not practical or cost effective in Canada to map
  T1 and T2 forest and compare
• 4.3 million km2, ½ forested
• fine delineations needed
• small amount of deforestation
• Deforestation is both land cover and use
• many factors integrated, automated not
  appropriate as primary identifying tool
• Land cover classification comparison too
  imprecise due to vagaries of classification no
  or difficult land use attribution
• Low resolution (e.g., MODIS, SPOT VGT) can not be
  used successfully to identify deforestation or
  zones with deforestation
• Annual mapping not viable for Canada
  interpolation over time acceptable

40
Canadas National Deforestation Estimates Mapping
Scope

• 48 million ha mapped (1975-1990 1990-2000 or
  both)
• 25 interpreters from companies, provincial
  mapping agencies and RD organizations
• 3-5 main QC personnel in CFS
• 5-20 minutes to map reasonably complex 3.5 x3.5
  km sample cells contract costs 30-40/plot 10
  QC 40-50/plot
• In total 258,864 ha of deforestation events
  delineated
• 54,827 events mapped

41
Lessons Learned

• Interpreter training and experience key QC and
  vetting important
• Efficiencies and results suffer if steps are
  bypassed or diminished
• Circa 1990 winter imagery and digitized air
  photos very useful
• Too much reliance (faith in) ancillary data
  yields problems
• Imagery 5-10 years after deforestation useful to
  confirm land use change (e.g. 2000 imagery for
  1975-1990 deforestation)
• Roads, power lines, pipelines near 20 m width
  limit difficult to definitively determine width
  criteria (i.e. gt 20 m)
• Image quality important phenology, crop stage,
  haze

42
Conclusions

• Full methodology and approach is sound and a
  keeper
• Infrastructure of expertise, tools and methods
  has been developed
• For first time there is a national accounting of
  deforestation
• Improvements can be made
• Procedures framework established, will form the
  foundation for deforestation estimation in Canada
  throughout the KP commitment period and beyond,
  ever evolving and adapting to new technology,
  resource levels, institutional circumstances and
  the lessons learned

43