Section 7 Annexes

Annexes include:

1. Classification of NWFPs - examples of approaches used
2. Understanding plots and subplots
3. Example of NWFP inventory outputs
4. Some currently used and emerging sampling methods
5. Useful institutions and Web sites

7.1 Annex 1. Classification of NWFPs - examples of approaches used

Approaches

International trade reporting: for example Customs and Excise, tend to group resources according to:

· product type (e.g. `live plants', `prepared beverages', `animal fats', `prepared bark products'); or

· end use (e.g. `chewing sponge or stick', `cloth', `edible leaves', `wine', `resin').

Biodiversity inventories usually group animals and plants according to scientific names of family and genera.

Ethnobotanic studies classify according to local end uses (e.g. construction, edible, fuel, medicinal, poisons).

Foresters and forest-based assessments use groupings according plant form and parts used (e.g. non-wood tree parts, tree fruit, herbs, climbers, shrubs, etc.)

Wildlife ecologists usually group according to the scientific family and size (e.g. insectivores, primates, reptiles, rodents, ungulates).

Land/resource managers sometimes group according to management characteristics (e.g. ease of propagation or cultivation, accessibility, who collects it, for regular household consumption, occasional use, for sale in local markets).

Examples

A. Typology for national NWFP accounting (after Chandrasekharan, 1995)

A. Live plants and parts of plants

Live plants

Parts of plants (fresh, cut, dried or crushed), collected for specific uses
Specific parts of plants with multiple uses, not included under the previous group
Vegetable materials not elsewhere classified
Raw exudates and similar natural products

B. Animal and animal products

Live animals
Animal products

C. Prepared/manufactured products

Prepared (provisionally preserved) edible products
Prepared beverages
Prepared animal feed/fodder
Vegetable oils/fats
Animal fats/oils
Prepared waxes of animal or vegetable origin
Dying and colouring extracts of plant or animal origin
Phytopharmaceutical/medical extracts, galenicals, medicaments
Essential oils and their concentrates
Rosin and rosin derivatives
Processed gums and latex
Fuels and alcohols
Other basic organic/phytochemicals
Prepared bark products
Plaited products
Products of natural fibre
Tanned leather, fur and products of taxidermy
Miscellaneous products, manufactured from non-wood forest raw materials
Other non-wood plant and animal products n.e.c.

D. Services

Forest-based services

B. End use classification (after Wyatt, 1991)

References

Boom, B.M. 1989. Use of plant resources by the Chacobo. pp. 78-96. In: Posey, D.A. & Balee, W. (eds). Resource management in Amazonia: Indigenous and folk strategies. Advances in Economic Botany 7. 287 pp.

Chandrasekharan, C. 1995. Terminology, definition and classification of forest products other than wood. pp. 345-380. In: Report of the International expert consultation on non-wood forest products. Yogyakarta, Indonesia. 17-27 January 1995. Non-wood forest products no. 3. FAO, Rome. 465 pp.

Dunn, R.M., Out, D.O. & Wong, J.L.G. 1994. Report of the reconnaissance inventory of the high forest and swamp forest areas in Cross River State, Nigeria. Cross River State Forestry Project (ODA Assisted), Calabar, Nigeria. 7 pp.

Edwards, I. 1991. Quantitative ethnobotanical survey of a hectare of tropical forest near Toraut, Dumogo Bone National Park, Northern Sulawesi, Indonesia. Sulawesi Ethnobotanical Project. Preliminary Report. 8 pp.

FitzGibbon, C.D., Mogaka, H. & Fanshawe, J.H. 1995. Subsistence hunting in Arabuko-Sokoke Forest, Kenya, and its effects on mammal populations. Conservation Biology 9 (5): 1116-1126.

Gadsby, E.L. & Jenkins, P.D. 1992. Report on wildlife and hunting in the proposed Etinde Forest Reserve Consultants report to Limbe Botanic Garden and rainforest genetic conservation project (ODA). Unpublished. 43 pp.

Kleinn, C., Laamanen, R. & Malla, S.B. 1996. Integrating the assessment of non-wood forest products into the forest inventory of a large area: Experiences from Nepal. pp. 23-31. In: Domestication and commercialization of non-timber forest products in agroforestry systems. Proceedings of an International Conference held in Nairobi. FAO.

Lahm, S.A. 1993. Utilization of forest resources and local variation of wildlife populations in northeastern Gabon. pp. 213-226. In: Tropical forests, people and food. MAB Series Vol. 13. Hladik C.M., Hladik A., Linares O.F., Pagezy H., Semple A. & Hadley M. (eds). UNESCO 852 pp.

Malhotra, K.C., Poffenberger, M., Bhattacharya, A. & Dev, D. 1991. Rapid appraisal methodology trials in Southwest Bengal: assessing natural forest regeneration patterns and non-wood forest product harvesting practice. Forest, Trees and People Newsletter 15/16: 18-25.

McCormack, A. 1998. Guidelines for inventorying non-timber forest products. M.Sc. thesis, Oxford. 127 pp.

Prance, G.T., Balée, W., Boom, B.M. & Carbeuri, R.L. 1987. Quantitative ethnobotany and the case for conservation in Amazonia. Conservation Biology 1 (4): 296-310.

Salick, J., Mejia, A. & Anderson, T. 1995. Non-timber forest products integrated with natural forest management, Rio San Juan, Nicaragua. Ecological Applications 5 (4): 878-895.

van V

Valkenburg, J.L.C.H. 1997. Non-timber forest products of East Kalimantan. Potentials for sustainable forest use. Tropenbos Series 16. Tropenbos Foundation. 202 pp.

van Wieren, S. 1999. Towards the sustainable use of wildlife in tropical forests. pp. 175-178. In: Seminar proceedings 'NTFP research in the Tropenbos Programme: Results and perspectives', 28 January 1999. Ros-Tonen, M.A.F. (ed.). Tropenbos Foundation, the Netherlands. 203 pp.

Wong, J.L.G. 1998. Non-timber forest products from the reserved forests of Ghana. Consultancy report 11. Forest Sector Development Project, Accra, Ghana. Unpublished. 31 pp.

Wyatt, N.L. 1991. A methodology for the evaluation of non-timber forest resources. Case study: the forest reserves of southern Ghana. M.Sc. thesis, Silsoe College, Cranfield Institute of Technology. 102 pp.

7.2 Annex 2. Understanding plots and subplots

7.3 Annex 3. Example of NWFP inventory outputs

From Case study 3 - Ghana national inventory. Climbers - Hunhun - Fruit - Manniophyton fulvum (L5)

1. Vegetation zone preferences

TUKEY HSD MULTIPLE COMPARISONS.

MATRIX OF PAIRWISE COMPARISON PROBABILITIES:

WE ME MSSE MSNW DS

WE 1.000

ME 0.000 1.000

MSSE 0.545 0.000 1.000

MSNW 0.000 0.043 0.000 1.000

DS 0.000 0.071 0.000 0.992 1.000

WE & MSSE not different

MSNW & DS not different

2. Tree basal area 3. Tree pioneer index

Mean BA = 23.218 SE = 0.260 Mean PI = 59.787 SE = 22.284

4. Economic index for trees > 30 cm d 5. Management zones

6. Relative abundance

Zone	WE & MSSE	ME	MSNW & DS
Occupancy (%)	78.9	58.9	27.4
Mean density (stems ha^-1) *	26.531	18.692	9.271
Standard error	20.939	16.142	13.131
Maximum density (stems ha^-1)	118	95	59
Area to search to find 10 (ha)	0.5	0.9	3.9

^*Density in area occupied by species

7.4 Annex 4. Some currently used and emerging sampling methods

a) Sampling designs possible for NWFP inventory.

Subjective sampling

Generally not statistically acceptable but often used in 'cruise' or rapid assessments to ensure complete range of environments are sampled. Also used in PSP location to ensure that all types of forest are represented. Take care that a design does not inadvertently become subjective. Watch for bias e.g. leaving out areas where access is difficult.

Gradsects - used in ecological surveys to ensure that all vegetation types along major environmental gradients are sampled.

Sampling systems

Objective sampling

The most commonly used types of designs for natural resource inventory.

Complete census - Measuring and recording every individual. Only practical for small areas. Generally used for stock survey of forest compartments due to be logged.

Simple random - Samples drawn using random numbers from a pre-determined sampling frame. E.g. set up a grid of 1x1 km numbered squares, select squares for sampling using random number tables.

Systematic - Samples selected according to pre-determined rules, i.e. plots placed at the intersections of a 1x1 km grid, every fifth tree measured, etc. There has been some argument about whether this is statistically acceptable. However, it is generally considered that such designs are acceptable as long as care has been taken to reduce the risk of the sampling grid coinciding with some regular feature of the landscape. Note: the sampling error can be calculated using the formula for a simple random sample with the assumption that the underlying population is random (i.e. that the placement of trees is itself random). If it is not safe to assume this, then calculation of the sampling error can be problematic. Note that the systematic grid can be considered a single plot, replication of the grid could therefore be used to estimate errors.

Probability sampling

Samples where the probability of selecting an individual is proportional to its size. Note: all other methods discussed sample with constant probability of selection which can mean that rarer, large individuals are undersampled given that they contribute disproportionately to the total quantities present.

List sampling - make a list of all individuals and their size. Calculate cumulative size i.e. the sum of sizes of all smaller individuals should be tabulated for all individuals. Assign numbers for selecting individual according to cumulative size (see example). Probability of selection given by cumulative size/sum of sizes.

Individual	Size	Cumulative size	Numbers
1	2	2	1-2
2	5	7	3-7
3	10	17	8-17
4	15	32	18-32

If random number drawn is 5 then individual 2 is selected, if it is 20 then individual 4 is selected. The larger individuals have a higher probability of being chosen because they have more number assigned to them.

3P sampling - developed for estimating volume of timber in a timber sale. Do a visual assessment of tree, select sample with probability proportional to the predicted size of the tree. Use of selection rules to determine which trees to be sampled. It requires that every tree in the tract is visited. Estimate the maximum tree volume in the stand.

At each tree:

If the tree is bigger than the maximum estimated size then estimate its volume and measure it.
Otherwise use a random number table to determine if tree is sampled.

If random number is less than estimated size measure the tree

Or move onto next tree.

Use data to estimate total volume on the stand.

Line intersect sampling - sample individuals that touch or intersect a line - the bigger they are the higher the chance that they will touch the line. Originally developed to estimate the amount of material e.g. slash or fuelwood lying on the ground. Has also been suggested for sampling lianas and used for wildlife tracks and signs.

Besides these basic designs it is also possible to use more or less any of them within a larger plan which can be used to achieve sampling efficiencies or to ensure that all subpopulations are adequately sampled. These plans are:

Stratified sampling - Dividing the population into sub-populations.

· Pre-stratification - Dividing the population into sections which are generally less variable and therefore can lead to savings in terms of the overall number of plots required. Also help to ensure that small subpopulations are adequately sampled. Generally stratification is beneficial and can reduce errors by 5 to 20 percent compared with an independent measurement of the total stand.

· Post-stratification - uses characters of the plots to group similar plots to improve precision of overall estimates (not strictly statistically correct unless sampling is random).

Note that more or less any design can be stratified hence: stratified random, stratified systematic etc. Strata may be decided by mapping or be systematic, e.g. dividing an area into 10x10 km blocks.

Multi-stage sampling

Sampling a series of nested plots, generally smaller plots located within larger ones. For example 1x1 km areas may be selected for land use mapping, within this a 1 ha plot may be randomly selected, every fifth tree in the 1 ha plot may have 10 percent of its branches sampled for fruit.

· Often used in extensive inventory as a simple layout would give too many plots.

· Sampling design at each level can be different and the highest level often uses remote sensing.

· If the subplots are selected systematically then these designs effectively become cluster plots.

· Better to use a multi-stage design than to undertake a low intensity sample of whole area as you at least have good data within the largest sampling units.

Double sampling

Independent selection of two different samples selected from the same population of individuals with the objective of measuring different characteristics in each sample. Often there is at least one character in common which can be used in regression-type models to predict a character that is more difficult to measure from a simpler one. E.g. using an independent, small sample of trees for which fruit yield is measured, this information used to interpolate fruit yields from a larger sample of trees for which only diameter is measured. Choose designs most efficient for each type/scale of sampling. The two inventories are related using ratio or regression estimators.

b) Emerging sampling designs

Adaptive sampling

General class of methods in which the number of plots sampled responds to the occurrence and number of individuals encountered during sampling.

Features:

+ Efficient (precise and cost-effective) and unbiased sampling strategy for rare, clustered or spatially uneven populations;

+ Increases numbers of observations for a given sampling effort than SRS²;

+ Locates and incorporates local hot spots;

- Cannot know number/cost of sampling at start of exercise;

- Special calculations of mean and variance required.

Adaptive cluster sampling - Method for locating and recording the size and composition of clumps in heterogeneous populations. Start with a low intensity sample and when the item of interest is located, add additional samples until you run out of individuals to sample. This forms a cluster of plots.

It is especially useful where density is clumped across large areas, allowing maximum number of individuals to be sampled for minimum sampling effort. A drawback is that additional plots may get disturbed through the sampling of the first. The principle is that the plot data are aggregated so the whole clump becomes the sample unit, so it does not matter if the plots touch. A problem is that you do not know how long or how expensive the inventory will be until you have finished.

Variants which may suit different situations are:

Initial simple random sample. Add plots (usually adjacent to `filled' plot in a fixed pattern, recommended in a cross configuration) whenever a plot contains more than a threshold number of individuals (adding rule) stop adding when all new plots do not satisfy the adding rule.	+Forms clusters of sample plots that grow towards local maxima and completely include aggregations of individuals
Initial strip samples (plot clusters grow sideways from strip once species is discovered)	+Good for covering large areas
Initial systematic sample (with random starting point)	+Very efficient for rare, clustered populations
Order statistics adding rule (adding rule uses rank order of samples i.e. if new plot has density > 4^th highest then add new plots)	+Population of unknown density for which an a priori adding rule cannot be determined - Computations more involved
Stratified: (1) Clusters not allowed to cross strata boundaries (2) Clusters allowed to cross strata boundaries	+Permits use of prior information (1) Strata independence maintained (2) More efficient but requires special calculation of mean
Adjusted for imperfect detectability: (1) Constant detectability (2) Variable detectability	+Good for motile or cryptic organisms -Uses non standard mean and variance calculations

Adaptive allocation - Two stage adaptive designs. Take initial sample in a conventional manner. Allocate next set of plots according to the density of target trees in first set of plots. This permits the final number of plots to be known in advance. Approaches include:

Sample sizes based on initial observations in each stratum:

Stage 1: Area divided into strata and SRS (or other allocation system) used in each strata

Stage 2: More plots added using SRS (in proportion to number of plots per strata that qualify on adding rule or to minimize estimated final variance)

+Maximizes the value of pilot studies

+Permits collection of additional data in areas discovered to have high population density without compromising design

+Costs easier to control

-Requires two passes over study area

-Small negative bias in estimates for total pooled sample

Sample size based on observations from previous strata (sequential SRS of strata; allocation of plots to subsequent strata based on adding rule and observations in previous strata)

+Only requires one pass

+Good for sampling across large scale environmental gradients, e.g. mountain slopes where target species may be confined to certain altitude zones

+Traditional stratified SRS calculations apply

Subjective - Rank set sampling. Novel technique that is actually unbiased and efficient. It ranks plots laid out in groups at different locations according to average value (e.g. size) of the measured characteristic. For example, three plots could be laid out in each of three sample locations. At each location the three plots are ranked (high, medium, low) according to density of the resource species. At the first location, the high density plot is measured, at the second the medium, and at the third the low density plot is measured. The mean of the three measured plots is used for calculating overall estimates about the population. It is useful where there is a lot of local variation, avoids bias and can potentially use local knowledge. This needs further development for use on NWFPs.

Features:

+ Gives unbiased estimates and better precision than SRS of same sample size

+ Works best for populations with high local variability and can be tailored to match the level of local variability

+ Permits the incorporation of subjective knowledge

- Requires visual comparison of plot sets for ranking so they must be close together

- Cost of locating plots for ranking needs to be small compared to cost of enumeration

Guided transect sampling. A two-stage unbiased design for transect survey utilizing high resolution prior information.

Stage 1: Wide strips laid out as primary units and divided into grid-cells of suitable dimensions for which prior information, i.e. remote sensing is available.

Stage 2: One survey line for subsampling per strip is randomly selected. The grid-cells that will form the route of the survey line are selected with probabilities proportional to their covariate values. The strategy for selecting cells can be varied.

Along selected lines the inventory is performed using some transect based method such as line transect sampling, strip surveying, line intersect sampling, etc.

Features:

+ Can use high resolution a priori data, i.e. classified pixels from remote sensing interpretation

+ Better alternative to gradsect sampling, etc., as line selection is based on probability rather than subjectivity

+ Good for sparse populations

- Requires large amounts of detailed prior information

Further reading

Brown, J.A. Unknown. The application of adaptive cluster sampling to ecological studies. pp. 86-97. In: Statistics in Ecology and Environmental Monitoring. Otago Conference Series No. 2. Fletcher, D.J. & Manly, B.F.J. (eds). University of Otago Press, Dunedin, New Zealand.

Cochran, W.G. 1977. Sampling techniques. Third edition. Wiley. 428 pp.

Halls, L.K. & Dell, T.R. 1966. Trial of ranked-set sampling for forage yields. Forest Science 12 (1): 22-26.

McIntyre, G.A. 1952. A method for unbiased selective sampling, using ranked sets. Australian Journal of Agricultural Research 3: 385-390.

Muttlak, H.A. & McDonald, L.L. 1990. Ranked set sampling with size-biased probability of selection. Biometrics 46: 435-445.

Patil, G.P., Sinha, A.K. & Taillie, C. 1994. Ranked set sampling. pp. 167-200. In: Environmental Statistics. Handbook of Statistics Vol. 12. Patil, G.P. & Rao, C.R. (eds). Elsevier Science. 927 pp.

Seber, G.A.F. & Thompson, S.K. 1994. Environmental adaptive sampling. pp. 201-220. In: Environmental Statistics. Handbook of Statistics Vol. 12. Patil, G.P. & Rao, C.R. (eds). Elsevier Science. 927 pp.

Shiver, B.D. & Borders, B.E. 1996. Sampling techniques for forest resource inventory. Wiley. 356 pp.

Ståhl, G., Ringvall, A. & Lämås, T. 2000. Guided transect sampling for assessing sparse populations. Forest Science 46: 108-115.

Thompson, S.K. 1991. Stratified adaptive cluster sampling. Biometrika 78 (2): 389-397.

Thompson, S.K. 1992. Sampling. John Wiley & Sons. 343 pp.

Thompson, S.K. 1997. Spatial sampling. pp. 161-172. In: Precision agriculture: spatial and temporal variability of environmental quality. Ciba Foundation Symposium 210. Wiley. 251 pp.

Thompson, S.K. & Seber, G.A.F. 1994. Detectability in conventional and adaptive sampling. Biometrics 50: 712-724.

7.5 Annex 5. Useful institutions and Web sites

Material from a range of resource institutions informed the original review. These are available to further your knowledge in this area, some of which have useful Web sites.

*Institution*	*Web site*
AERDD, University of Reading	http://www.rdg.ac.uk/AERDD/
Afrirattan	www.africanrattanresearch.fsnet.co.uk
Birdlife International	http://www.wing-wbsj.or.jp/birdlife
Bushmeat Crisis Taskforce	www.africanrattanresearch.fsnet.co.uk
CABI Online Publishing	http://www.cabi.org/Bookshop/Readingroom/
CARPE - Central African Regional Programme	http://carpe.umd.edu/
Centro Agronomic Tropical de Investigación y Enseñanza (CATIE)	http://www.catie.ac.cr/research/research.asp
Centre for International Forestry Research (CIFOR) -Criteria and Indicators	www.cifor.cgiar.org/ www.cifor.cgiar.org/acm/methods/candi.html
Conservation International	www.conservation.org/
Department for International Development (DFID), United Kingdom	http://www.dfid.gov.uk/
Department of Forestry, University of Aberdeen	http://www.abdn.ac.uk/~for257/forestry.hti
European Forest Institute - Certification information service	www.efi.fi www.efi.fi/cis
European Tropical Forest Research Network (ETFRN) - NWFP workshop report	www.tropenbos.nl www.etfrn.org/workshops/index.html
Falls Brook Centre, Canada (Certification of NTFPs)	http://www.fallsbrookcentre.ca/programs/International/certmark/certmark.html#ntfp
FAO	www.fao.org
Institute for Culture and Ecology - NTFP programme	www.ifcae.org/ntfp
Institute of Ecology and Resource Management, University of Edinburgh	http://www.ierm.ed.ac.uk/ierm/research/index.htm
Instituto Nacional de Biodiversidad (INBio) - inventory	www.inbio.ac.cr/en www.inbio.ac.cr/en/inv/invent.html
International Institute for Environment and Development (IIED), London	www.iied.org
International Union of Forest Research Organisations (IUFRO)	http://iufro.boku.ac.at/
IUCN, Sustainable Use Initiative	www.iucn.org/themes/sui/
Natural Resources Institute (NRI)	http://www.nri.org/Themes/forest.htm
New York Botanical Garden - herbaria information	www.nybg.org www.nybg.org/bsci/ih/
Overseas Development Institute (ODI)	www.odi.org.uk/fpeg/rdfn
Oxford Forestry Institute (OFI)	www.plants.ox.ac.uk/ofi
ProFound: Advisers in Development - NTFP information	www.thisisprofound.com/ www.ntfp.org
Royal Botanic Gardens, Kew Royal Botanic Gardens, Edinburgh	www.rbgkew.org.uk/ www.rbge.org.uk/
School of Agriculture and Forest Sciences, University of Wales	http://www.safs.bangor.ac.uk
Statistical Advisory Centre, University of Reading	www.rdg.ac.uk/ssc
Tropenbos, NTFP Programme, University of Wageningen, the Netherlands	www.tropenbos.nl/tropenbos/thementfp.html
Tropical Forest Forum (United Kingdom)	www.nri.org/TFF/forumfra.htm
University of St. Andrews, RUWPA Downloadable DISTANCE software	www.ruwpa.st-and.ac.uk www.ruwpa.st-and.ac.uk/distance
USGS Biodiversity monitoring program Downloadable MONITOR software	http://www.mp1pwrc.usgs.gov/powcase/index.html
UNEP World Conservation Monitoring Centre	www.unep-wcmc.org
UNESCO, People and Plants Initiative	www.rbgkew.org.uk/peopleplants
Wildlife Conservation Society	www.wcs.org

Prance et al., 1987	Edwards, 1991	Boom, 1989	Valkenberg, 1997	Salick et al., 1995
Edible	No use	Food	Timber	Aesthetic
Construction material	General purpose	Fuel	Special purpose wood	Construction
Technology	Timber	Construction	Bark/leaves	Edible
Miscellaneous	NTFPs not in trade	Medicinal	Edible fat	Firewood
Remedies	NTFPs in trade	Poisonous	Fruit	Hunting
Religion		Commercial	Exudate	Animal habitat
		Miscellaneous	Medicinal	Intoxicant
			No use (including firewood)	Medicinal
				Oils Poison Resins, etc. Shade Timber Utility Non-timber wood Other
Malhotra et al., 1991
Raw materials for commercial sale or processing
Subsistence food or drinks
Animal fodder
Fuel
Timber and fibres for tools and construction purposes
Medicinals

NWFP group	Group description	Examples	Comments
1	Non-wood tree parts	Fruits, leaves, twigs	Can be related to tree dimensions
2	Products from `tree like' plants	Bamboo, rattan	Relatively easy measurable dimensions
3	Herbs and other plants	Medicinal and aromatic herbs	Some specific properties to be taken into consideration when incorporating into standard forest inventories

Animals	No sub-division
Plants	Perennial species and products	Trees	Wood
			Bark
		Non-trees	Climbers	Lianas
				Rattans
			Non-climbers	Palms
				Bamboo
				Epiphytes
				Shrubs
	Ephemeral products from perennial species	E.g. Fruit, fluff from seed cases, nuts/seeds, oil seeds, apical buds, leaves
	Ephemeral species	E.g. Herbs, mushrooms, wild honey

Wong, 1998	., 1994	*FitzGibbon et al.,* 1995**	Lahm, 1993	Gadsby & Jenkins, 1992
Non-timber trees	Climbers	Primates	Reptiles	Insectivores
Herbs	Shrubs	Duikers	Pangolin	Bats
Climbers	Palms/bamboo	Elephant shrews	Rodents	Primates
Rattans	Marantacae	Squirrels	Primates	Rodents
	Non-timber trees		Carnivores	Carnivores
	Rattan		Ungulates