Regional presentations by the FRA staff members and consultants responsible for Asia, Anglophone Africa, Latin America and the Caribbean illustrated the parameters for estimating the utility of existing data sources for change assessment. The presentations highlighted the following key parameters: type of change information or data; coverage; periodicity of data; compatibility of time series; reliability of inventory field work; validation of remote sensing data; compatibility with FAO classifications; overall quality.
Table 1. Metadata about country-level forest inventory data available to FRA 2000.
(1) Reference with recent data Reference contains the most recent data or the reference provides change data. In case of two independent inventories/surveys see option # 3. |
|
(2) Reference with old data Select the reference containing the latest data. |
|
(3) Type of change information or data a) Change information is derived from a series or cycle of (national) forest inventories, for example the National Forest Survey of India. b) Change information is based on two independent inventories. c) Basic information is available that could be used to predict change. |
|
(4) Coverage
|
|
(5) Year(s) of data Reference years of the two data sets in the time series. |
|
(6) Classification
|
|
(7) Methodology
|
(8) Interdependent analysis
|
(9) Type of remote sensing data Here the type of remote sensing data (aerial photos, Landsat TM, etc.) is listed. In general, the higher the resolution, the higher the utility/quality for change assessment. |
(10) Type of field work
|
(11) Comparability with FAO classification Judgement on compatibility of the national definitions of forest and deforestation (applied in the time series or change study) with FAO’s definitions (see FRA 2000 Working Paper 1, Terms and Definitions).
|
(12) Overall quality Based on all criteria a judgement can be made on the overall quality of the forest change information for estimating national forest change in FRA 2000. Three classes are distinguished:
|
The larger group was split into three sub-regional working groups: Asia, Africa and Latin America. An attempt was made to have a distribution of technical skills (modelling, remote sensing, statistics) in each group. FRA Country Information Coordinators (CICs) joined each group.
The FRA Country Information Coordinators made a first attempt to categorize the available information and the utility of the data for change evaluation. The regional groups revised the information sets and modified or confirmed the ranking of data availability scores within their data sets. Participants decided to categorize data availability in terms of high, medium and low. This ranking was established according the reliability of the data, information dates, periodicity and source of information. These rankings were later referred to as traffic lights, and were used to produce a map showing the availability of FRA data (see Appendixes 1 and 2).
Using the revised information, the regional groups constructed scenarios for generating estimates of forest state and change, including options, criteria, resources required and the rationale for their region. The groups presented their scenarios and modified the "traffic lights" in plenary session. The summary conclusion was that the Asia working group works with high-utility information from its countries, the Africa working group works with low-utility information while the Latin America working group works with medium-utility information.
The group found no grounds to change from its assessment of low utility data. Information is poor in most countries.
Conclusions
Ø AVHRR will not be useful, mainly because of poor performance in dry and semidry areas.
Ø Intensifying the remote sensing survey seems to be a feasible approach. This is further elaborated.
Ø Narratives on forest change processes are useful to provide an understanding and to justify the approximations.
Ø Use of expert opinion when weighting the various data sets is a promising approach.
The Africa group developed the following table to show how all available data for each country could be summarized for purposes of review and eventually weighting to arrive at a final estimate.
Utility criteria include: time period, coverage and precision/accuracy. It is important to define how deal with partial/regional observations, how to weight experts opinions to get a FAO estimate, how to weight observations outside the time period, how to establish guidelines on weighting, and how to disseminate the information and procedures (should the whole table be published).
Table 2. Assessment table.
Source |
Utility info |
precision/accuracy |
Weights by "experts" | ||||||
FRAPanel |
Expert opinion |
Original sources of information | |||||||
COUNTRY DATA |
|||||||||
State A-State B |
C |
2 |
0 |
0.5 |
0 | ||||
State A- State C |
C |
1.5 |
1 |
0 |
0.5 | ||||
FRA RSS |
|||||||||
Samples |
P |
NA |
NA |
NA | |||||
Country |
C |
3 |
0 |
0 |
0.5 | ||||
Regional |
R |
2.11 |
NA |
NA |
NA | ||||
MODELS |
|||||||||
Model 1990 |
C |
4.7 |
0 |
0 |
0 | ||||
"OTHER" |
|||||||||
Province X |
P |
11 |
NA |
NA |
NA | ||||
EZ Y |
R |
-0.3 |
NA |
NA |
NA | ||||
Expert NN |
C |
0 |
0 |
0.5 |
0 | ||||
Combination ABC |
C |
2 |
0 |
0 |
0 | ||||
TOTAL |
1.5 |
1 |
2.25 | ||||||
"NEW INFO" |
|||||||||
State A-D |
NCY |
NCY |
NCY | ||||||
-> FRA Total = 1.5 + 2.25 : 2 = 1.875
It was considered relevant the information provided by the FRA panel and the Original sources of information. The weigth gathered by the Expert opinion was considered with low precision/accuracy.
NA=not applicable; NCY=not considered yet
Africa scenario for intensified RSS to determine cover state and change
Ø Restratify using another source of information from a particular province (ecological zone) and/or groups of countries (say 3 ecological zones by 4 sub-regions) and make a new sample design;
Ø Poststratify existing (ca. 40) samples in the new strata;
Ø Add new samples (2 dates) to top up the new sample design;
Ø Interpret new samples;
Ø Derive sub-regional estimates (by ecological zone);
Ø Apply ecological zone statistics to countries, using known distribution of ecological zones in each country.
Resources needed:
Scenarios should be country-specific. A scenario is defined as a combination of available information for change estimates (high, medium, low) and timeliness of information (<5 years or >5 years). This defines six possible combinations, divided into four scenarios:
Table 3. Possible scenarios for Asia.
Available Information |
<5 yrs old |
>5 yrs old |
High |
A |
B |
Medium |
B |
C |
Low |
C |
D |
Scenario A: Formal data extrapolation
Ø Countries with high quality data; and
Ø Data source not dated earlier than 1995.
Scenario B: Country’s own estimations for 2000, as requested by FAO*
Ø Countries with medium quality data;
Ø Data source dated earlier than 1995.
*FAO needs to develop a decision rule to accept these estimates or to go to Scenario III.
Scenario C: Combination of estimates
Ø The remainder of the countries that do not adhere to the first two scenarios will be assessed based on a combination of the following “independent” estimates:
Country estimates.
Model-based estimates.
New remote sensing estimates.
Disaggregated data from regional estimates to country level expert estimates.
Ø An expert system must be used to define weights appropriate for each of the above estimates based on measured or expected precision of the data.
Scenario D: Combination of estimates
Ø With source data earlier than 1995.
The group revised the existing information available and alternative options for the establishment of possible methods to evaluate change. The following options were highlighted:
Ø Estimating change based on some weighting of existing country data;
Ø Deriving country level estimates by stratifying regions into areas of high, medium, and low rates of change, then using regional thematic mapper samples to obtain weighted averages.
This option may call for more field work, either in the office or through field visits to the country and would need maps by region.
Some concern was expressed about including the Brazil information in the low precision regional estimates because the country has good data available. The FRA classifications should be crosschecked with Brasil’s classification of same image.
The criteria for selecting the method to carry out in-depth work in priority countries would be related to: forest area, quality of existing information, rate of change, type of forest and opportunity to work with established technical networks.
Possible process for country estimates in Latin America
Develop different estimation scenarios (expert opinion, regional thematic mapper basis):
Ø Identify information reporting units.
Ø List possible existing data inputs:
National or sub-national inventory reports.
Satellite samples (FRA and others).
Local sub-national inventories.
Old FRA modelled estimates.
Population growth rates by regions.
Policies on land use and forestry.
Ownership structure.
Land use and land productivity information by regions.
Forest map(s).
Expert opinions.
AVHRR map (FRA and others).
Thematic mapper sample unit change information for units.
Rutgers study.
Local forest change studies.
Ø Sort and classify data sources.
Ø Identify best complement of information by information unit.
Ø Rank information.
Ø Discard unreliable or unnecessary information for information unit.
Ø Weigh information inputs according to importance in deriving estimate.
Ø Processing:
Experts draw map of state and change and compare.
Experts identify where their knowledge is best.
Validate results from AVHRR to get broad idea of change, consult old AVHRR maps and images if available.
FRA expert synthesize and present conclusions.
FRA present cover map to validate and identify change areas receiving comments from national experts.
Expert delineates high change areas on small scale map.
Ratio estimates.
Build a regional state and change map (inputs are expert opinion, gradients built from remote sensing sample units).
Ø Remote Sensing Studies
FRA to request or provide new studies
Post stratification based on expert opinion based on estimate of change
Ø Tests
Test old model in selected areas.
Review the sampling procedure before you know how to apply properly.
Ø Experts
Key government agency officials.
International institutions and projects.
NGOs
Ø Presentation
Reference year estimates.
Estimates of change.
Confidence interval.
