Section 3 Quantitative studies on NWFPs

This section describes current experience in NWFP resource assessment. It covers inventory, yield measurement, growth studies, harvest determination and monitoring.

3.1 Finding out how much of a resource is present

This is often called `quantitative inventory'.

Quantification of resources can mean something different to an ecologist than it does to a forester, and so on. But definitions boil down to it being: A biometrically rigorous enumeration of the abundance and distribution of resource populations.

Many designs, single structure

There are very many different designs, partly because there are very many different types of NWFP - plant and animal. Whilst methods to inventory these appear to be very different, they all contain four basic elements, as Figure 3 shows. How to select one of the wide range of different methodologies that have been used at each stage in the design of quantitative NWFP inventory will be further discussed in Section 5.

Methodologies used for NWFP inventory are adapted from the wide range of experience available across the botanical and zoological sciences. NWFP inventory has used relatively few of the available methods because of:

· NWFP assessment being a relatively recent issue;

· nature of the NWFP resources that have been studied - e.g. the disproportionate focus on fruit; and

· context of NWFP assessments - whether the NWFP is the main focus of the inventory.

Figure 3: Basic structure of a quantitative inventory design

Methodologies can be adapted according to the species being studied and the availability of time, money and human resources. The level of adaptation also depends on the importance of NWFPs in the inventory. Three clear contexts can be distinguished:

· single resource inventory: where the inventory seeks to quantify the abundance and distribution of a single NWFP species;

· single purpose, multiple resource inventory: where the inventory looks at more than one resource for the same reason - i.e. a strategic inventory, for several different NWFPs; and

· multi-purpose resource inventory: where NWFP inventory takes place during inventories for other purposes, such as timber management or watershed protection.

Inventory for a single resource species

There are few inventories for a single species - it is a costly process so the product has to be very valuable (usually for export) or subject to specific legislation. Even then, few studies actually aim to quantify the species in situ, and methodologies are rarely tailored specifically to the characteristics of the species.

Six main reasons for carrying out single species inventory have been identified:

· provide new/initial knowledge about the consequences of harvesting a species (e.g. tapirs in Belize - Fragoso, 1991);

· assess the potential of specific species to support increased product demand (e.g. palm products in Namibia - Sullivan et al., 1995);

· assess the potential of an area for viable harvesting of a commercial product (e.g. rattans on Barateng Island in India - Sharma & Bhatt, 1982);

· investigate where a commercial product can be found (e.g. savanna fruit trees in Benin - Schreckenberg, 1996);

· provide information for determining harvesting quota levels for species under national or international regulation (e.g. CITES listed species, such as caiman skins from Venezuela - Velasco et al., 1996); and

· academic inquiry to better understand individual species, for ecological, historical or cultural reasons (e.g. understanding of the role of wild yams in historical human diets in central Africa - Hladick & Dounias, 1993).

Examples of the designs used in these studies are noted in Table 7. In summary, whilst single resource inventories are a good opportunity for developing reliable NWFP inventory protocols, there has been insufficient work done to make much progress.

 

Table 7: Inventory designs used by single resource studies

Product type	Sampling design	Plot configuration	Enumeration	Author
Tree bark	Systematic (1%)	50x50 m square	Diameter of trees >10 cm d	Acworth et al., 1998
Tree exudate	Aerial cruise, 2 flights	11 possible locations	Visual estimates	Zieck, 1968
Tree fruit	Subjective transects	10 m wide transects up to 1 km long	Diameter of trees >10 cm d	Shankar et al., 1996
	Six systematic radial line-plot transects (plots every 100 m on transect 3 km long)	Point centred quarter	Diameter for trees >3 cm and stumps >50 cm	Schreckenberg, 1996
Palm fibre	Stratified: Oxisols and podsols: unreported plot layout Gleys: line-plot 600 m long, 20 m between plots	Oxisols and podsols: 100x50 m rectangular Gleys: point quarter method	Height measured for all stems in plot	Lescure et al., 1992
Rattans	Subjective site selection	Single 3 ha (300x100 m) plot divided into 10x10 m subplots	Tally of clumps and stems	Stockdale, 1994
	Multi-stage sampling Random selection of 32 from 123 primary blocks	Three secondary 1 ha blocks selected from each selected primary block	Tally of commercial and non-commercial culms per plot	Sharma & Bhatt, 1982
Herb fibre	Line-plot transects, unreported distribution	Circular 50 m² plots every 10 m.	Tally and % cover of plants in plot	Cevallos undated
Tubers	Four sites - unreported plot location of 4-9 transects in each	Transects 4 m x up to 2.5 km long	Tally of yam stems	Hladik & Dounias, 1993
Large birds	Available trails and tracks (haphazard - biased)	180 variable width transects	Tallies of individuals	Silva & Strahl, 1991
Tapir	Randomly located line transects River transects	Line-intercept transects	Indirect (tracks)	Fragoso, 1991

Inventory for more than one resource species

In the studies reviewed, the multiple species are a range of NWFPs, and the single purpose for inventory is usually to provide quantitative information to assist management planning.

Using census methods for small areas

Some NWFP inventories have used methods developed for stock survey (the census of exploitable timber trees in a harvesting area). These methods can be used for measuring relative abundance of NWFPs in different land use types (e.g. Gronow & Safo, 1996). Although census methods provide truly accurate data, using them has drawbacks:

· does not take account of difficulties in finding species such as animals, small herbs or epiphytes;

· errors cannot be quantified, as it is a single sample; and

· such census type surveys are expensive unless over very small areas.

Using participatory rural appraisal (PRA) techniques (e.g. Poffenberger et al., 1992).

PRA based methods were iused in Poffenberger's manual to guide inventory and monitoring in the Indian joint forest management (JFM) process. This advocates a mix of methods for scoring relative abundances, and quadrats and plotless sampling for looking at vegetation change. Whilst the recommendations made are sound, they are rather broad and do not include detailed protocols.

Using basic forestry sample plot techniques (e.g. Cunningham, 1996a).

This method was designed to allow quantification of key plant resources - trees and bamboo - to support management planning in a National Park in Uganda. In this particular study, only three or four plots were used at each of the three sites. The lack of replication means that data could be imprecise, inaccurate and biased. The results are more suited to strategic planning than detailed management planning.

Sampling across forest and non-forest land (e.g. Dijk, 1999a).

Sampling strata in southern Cameroon were identified using air photos. Data collected from plots in each habitat type were used to prepare NWFP stand density tables grouped by product types and marketability, and to map their distribution. Looking at NWFPs outside forest land is important as many NWFPs are actually sourced from `farmbush' (farmland in cleared/regenerating forest areas). These areas are more usually assessed using a wide range of participatory approaches, such as described in Section 4.

Doing stock survey for timber management:

Immediately before logging operations, every timber tree in a forest management compartment is located, identified, numbered and measured. This usually gives a 100 percent enumeration in consecutive strips of the compartment. The stock survey data allows calculation of the volume of timber that can be harvested sustainably, and trees for felling are selected accordingly, to ensure future harvests.

Including NWFPs in inventories for other purposes

The mounting interest in NWFPs is driving a trend towards including NWFPs in inventories for other purposes.

For example, NWFPs can be incorporated into routine stock survey of commercial logging areas (e.g. Smith, 1995). Typically, routine forest inventory is becoming `multi-purpose resource inventory' (MRI). This is to improve economic efficiency, as well as in recognition of the increasingly wide range of products and services for which forest is managed. Experience in State forests is important here - the forest authority generally has responsibility for maintaining up to date information on important forest resources, which may include important NWFPs.

NWFPs are commonly included in routine forest inventory (typically every 10-20 years) across northern and eastern Europe, where berries, mushrooms, medicinal herbs and resins are traditionally important (Lund et al., 1998). However, little literature is available outside these countries. Also, whilst stock survey provides useful and biometrically sound information on the distribution, abundance and potential of NWFPs in areas to be logged, it does not take account of difficulties in finding species such as animals, small herbs or epiphytes; or of seasonality.

Further MRI reading: Lund, 1998

In the tropics NWFP componenent of MRI focuses onTraditional export products, such as bamboo and rattan - national assessment of resources by government forest agencies is often carried out, especially in South East Asia and India where exploitation is intense, with increasing interest in West Africa, as Case studies 2 and 3 show.

Case study 2: NTFPs in the Philippine national forest inventory An example of the inclusion of NTFP is the Philippines MRI which included rattans, palms and bamboo. Here a national-scale stratified inventory used quite different plot configurations in different regions (presumably, partially as a result of different donor/advisory support). The basic design was cluster sampling on an 8x8 km systematic grid. In regions 10 and 11, the clusters were four 20x250 m strips arranged on the arms of a swastika 1 km across. In these plots rattans were sampled in a 10x10 m subplot centred on the plot mid-line at the start and every 100 m along the strips. In all other regions the cluster was a triangular arrangement of six Bitterlich point samples at 50 m intervals. In this design rattans were sampled from a 5-m radius plot at the corner points of the triangle. Given the large quantities of data for each configuration, it would be interesting to compare their performance in terms of the precision and accuracy of rattan densities. (Serna, 1990)

Quantitative timber inventory. At its most basic this is a count of all individuals of interest within plot or transect. Plot totals allow estimation of average density in a given area. If tree diameter is measured, then basal area per tree can be calculated and volume estimated per unit area through yield tables. Methods are common, well understood and usually included in professional forestry training. Case study 3: NTFPs in the Ghana national forest inventory In Ghana NTFPs were enumerated from 1 ha plots (20x500 m) and included rattans, climbers and herbs. In this case, as plants became less like trees the enumeration methods, due to the absence of botanical survey advice, became increasingly crude. For example, rattan stems and clumps were counted and tallied into juvenile, mature and cut categories while abundance of herbs was represented as a simple tally of clumps. The herb data are unreliable as it is difficult to determine clump boundaries in dense stands and the size of clumps varies widely between species. As a consequence the herb data had to be reduced to presence/absence of the species. With hindsight % cover or even relative abundance would have been a better measure of quantities. The information derived from this inventory was intended for use by policy-makers and for national and international reporting requirements who required management-orientated interpretations. This was provided as tables and graphs illustrating the distribution and abundances of NTFPs across the country (see Annex 3). This type of analysis and presentation of inventory data are relatively rare. (Wong, 1998)

Getting beyond a timber focus

Typically, MRI inventories are carried out by forestry staff, and the primary aim remains timber management. This often restricts the range of products and interests, including recreation or agriculture, that can be included in MRIs. The focus on forestry can also limit the quality of the NWFP assessment, as it constrains:

· number of NWFPs included - generally only around 20 unfamiliar species can be identified without botanical expertise (Kleinn et al., 1996). Species included are usually restricted to the most familiar and more heavily traded ones;

· including difficult to find species, which may require specific observational methods - for example, animals may be nocturnal or avoid humans, fungi are seasonal. Some small herbs are difficult to see - detectability is recognized as a major problem;

· skill and effort with which they are assessed - forestry staff may be more skilled in tree enumeration, and plot enumeration often is limited to what can be accomplished in one day. Enumeration of NWFPs is often limited to a few subplots; and

· design of the inventory - this is determined by timber information needs, which may not be ideal for NWFPs. Protocols for including NWFP are not well developed, and yield tables to convert easily field data into resource estimates are not generally available. One of the biggest failings in MRI in tropical countries is that they do not include animals, despite bushmeat, for example, often being the most important product for local people.

Whilst separate inventories for different NWFP groups is usually impossible, due to cost and logistical constraints, to improve the value of the information gathered there should be some balance and co-ordination between different elements of MRI. For example, survey lines cut through tropical forest for timber inventory could also be used for animal inventory¹. Making the most of such opportunities can reduce costs and improve information for forest use planning.

Picking NWFPs out of existing timber inventory datasets

In many cases, NWFP tree species are included in formal timber inventory. Sometimes it is possible to extract and analyse NWFP data from records of historical timber inventories.

Some single-purpose timber inventories have been re-interpreted to provide information on NWFPs, and demonstrate that there are some useful NWFP data in older tree inventories (see Case study 4). With a picture of what trees are in an area, predictions can be made of what other plants and animals are likely to be present.

Case study 4: Use of existing inventories Southern Ghana. An excellent example of the use of a timber inventory is the analysis of the national forest inventory in southern Ghana. The ecological profiles of nearly 300 tree species were presented, based on information from the timber inventory. This thorough work provided information on NWFP trees. (Hawthorne, 1995b). Uttar Pradesh, India. Re-analysis of Forestry Department stand table data provided an estimate of total, state-wide quantities of edible oil from 25 important oil-seed species. (Rai, 1983)

Developing methods specifically for NWFPs - some examples

From the preceding sections it becomes apparent that NWFPs often fall between divisional responsibilities and professional expertise. This has contributed to:

· limited success in using recognized inventory methods reliably for NWFPs; and

· a lack of established methodology specifically for NWFPs.

Some pilot studies on methodologies for plant NWFPs have been done, mainly for some of the more economically important ones such as rattan, mushrooms, medicinal plants - Box 2 and Case study 5 below give some examples. Studies have often been very innovative, but there are as yet few comparisons between approaches. What is clear is that different approaches suit different types of NWFPs - further complicating the development of standard methodologies.

Box 2: Developing plot layout and measurement techniques for rattan inventory Rattans are the subject of a relatively large number of studies (12.6 percent of reviewed studies, see Table 6) and there are a number of researchers who have been investigating the relative efficiency of different plot sizes and shapes. Tandug (1978), Siswanto & Soemarna (1988, 1990), Siswanto (1991), Stockdale (1994) and Stockdale & Wright (1996) have all used basically similar techniques to determine optimal plot size and shape for rattan inventory. The technique used is to demarcate areas (ranging from 1 ha to 16 ha) into small (5x5 m or 10x10 m) square quadrats and to enumerate all rattan stems in the quadrats. The quadrat data is then aggregated to represent plots of different sizes and shapes and the relative efficiency of the different plot configurations compared in terms of sampling error and cost efficiency. This is apparently an efficient means of determining optimal plot dimensions but has a few drawbacks and some significant pitfalls. Although the chosen study site is large, it is effectively a single plot so the applicability of the results will depend on the representativeness of the study site which, in the absence of replication, is unknowable. Also, several studies have the test plots touching, which means that the "plots" are not independent (see p. 18). Tandug (1978) measured cost-efficiency of different plot configurations by comparing the sampling error for the number of plots that could be enumerated within three hours in a 1-ha study area. The sample plots were laid out on a systematic grid to cover the area so that more time was spend travelling between plots as the plots became smaller. The optimal plot size emerged as a 10x10 m square plot at a 50 percent sampling intensity. However, later recommendations of Tandug (1988) suggest using two 10x200 m strips arranged in a cross formation at a sampling intensity of between one and three percent. Unfortunately, it was not possible to locate a larger scale application of these recommendations to judge their overall merit. Identical studies were undertaken to determine optimal plot size and configuration in West (Siswanto & Soemarna, 1988), Central (Siswanto & Soemarna, 1990) and South (Siswanto, 1991) Kalimantan. The methodology employed was to subdivide a 16-ha study area into 10x10 m quadrats and to form these into a range of strip and line-plot configurations at a range of sampling intensities from 10 to 25 percent. In each case a 10-m wide continuous strip at sampling intensities of 25 or 20 percent was recommended. Stockdale (1994) and Stockdale & Wright (1996) used a similar technique to that of Tandug (1978) and subdivided a 1.5-ha (300x50 m) study site into 10x10 m quadrats. However, there is an significant difference between Stockdale's work and that of Tandug, Siswanto and Soemarna in that Stockdale's trial plots are contiguous while the others used non-contiguous plots. Consequently, in Stockdale's study the variation in sampling efficiency is more a function of spatial pattern in the rattan clumps and their coincidence with plot shape and size than a true test of the efficiency of different plot configurations. Stockdale & Wright (1996) found that strip plots were more efficient than square plots and recommend that 5-m wide strips forming plots of 0.005 to 0.025 ha for enumerating stems per hectare. Nandakumar & Menon (1992) developed a protocol for rattan inventory in Kerala State, India and recommended that 4x20 m strips be used in strips 100 m apart to give a 4-percent sampling intensity. However, they do not report any field work so it is not possible to judge the efficiency of their recommendations. An investigation of optimal plot sizes and shapes in Lao (Evans submitted) arrived at a 5x50 m plot sizes using six replicates randomly located along a transect line. This overcomes the problem of contiguous plots but the level of replication is still not sufficient for work of this kind. A methodology for estimating the length of rattan stems using a ruler hypsometer was developed (Stockdale, 1994; Stockdale & Power, 1994). This was tested against various other methods for estimating lengths and was found to be cheap, relatively easy to learn and significantly more accurate than visual estimation, internode counts (as used by Nur Supardi [1993]) and clinometer readings.

Case study 5: Developing protocols for mushroom monitoring A team at the Forest Service Pacific Northwest Research Station in Corvallis of the United States Department of Agriculture (USDA) has been developing methodology to inventory, assess yields and monitor production of edible wild mushrooms since 1993. Their experience and process of development of ideas is very well documented in a series of publications and provides an informative case study of the problems of inventorying non-tree forest products. The major problems facing the design team is that the target fungi (Matsutake, chanterelle and morels) occur as scattered colonies with patchiness at a range of spatial scales as well as being cryptic (largely invisible on the surface) and seasonal. It was recognized from the beginning that the patchy distribution would require the development of novel sampling schemes and analyses (Molina et al., 1994). The first attempt at inventory used methods borrowed from fungus diversity surveys in three study forests (Molina et al., 1994; Pilz et al., 1996a; Hosford et al., 1997). In each forest, three locations were selected to represent the three most productive vegetation types within the forest. At each location, three 225x225 m (5 ha) study sites were chosen to represent altitudinal, aspect and accessibility across the vegetation type giving nine study sites in each of the three study forests. Each site was surrounded by restricted access notices and, within each, six permanently marked strip plots 2x50 m were located systematically on a random orientation. The plots were located during the period when the fungi were not fruiting so as to avoid conscious bias. Mushrooms were enumerated by measuring cap and stem diameters, vertical distance from veil to cap, distance to nearest tree and volume predated. Measured caps were marked to avoid repeat enumeration. The plots were re-surveyed weekly through the fruiting period. After a couple of years' experience this method was largely abandoned as it was too expensive and time consuming, the sample area was also found to be far too small to adequately represent any individual species, the plots had been also compromised by illegal picking and vandalism, while legitimate pickers were intimidated and did not harvest the plots normally. In addition, off-site weather records did not correlate with yields. After this experience it was decided to change the sampling methodology. Japanese experience suggested that shiro (individual mycelium body or `castle' in Japanese) monitoring would be a useful for Matsutake mushrooms (Hosford, 1996). However, this methodology is very time-consuming and could only be considered for research work and not for routine monitoring. There have been two developments in methodology based on this early experience. It is proposed that regional monitoring should utilize volunteer enumerators drawn from the local pickers and a proposal to this effect has been circulated (Pilz & Molina, 1998). The plan is to use exclusive harvest agreements as an incentive for volunteers to make detailed harvest records from marked sample plots. Regional systematic stratified sampling is to be used to select local monitoring sites and the data used to investigate the relationship between forest management and mushroom productivity. Control sites are to be monitored by Forest Service staff. The programme is intended to be voluntary and based on flexible, decentralized collaboration encouraging volunteer ownership of the programme. For Matsutake, a mapping approach has been adopted with mushrooms mapped to reference trees which are located using GPS (Pilz & Molina et al., 1996). A cluster of mushrooms is taken to include mushrooms no further than 0.5 m apart with distances between clusters being at least 2 m. The demarcated clusters were assigned to experimental harvest treatments and monitored by Forestry staff with the co-operation of local pickers who kept the surrounding area well picked to discourage opportunistic collectors. The method of selecting reference trees is not given but this method would appear to be an efficient means of sampling for mushrooms. The search is still on for a suitable protocol for monitoring mushrooms (Pilz et al., 1997; Pilz & Molina, in press; Pilz et al., in press). A manual describing present experience and best advice is under preparation (Pilz, pers. comm.).

3.2 Finding out the yield of a resource

What is yield?

`Yield' typically refers to the amount of product available and useful for collection or harvest at a given point in time (i.e. that which can be used commercially). However, it can also mean the total biological potential of a species (i.e. how much actually grows there). The difference can seriously influence the conclusions of an assessment, as the latter is typically much greater than the former.

Measuring yield

Measuring product availability is generally known as yield assessment. It is the quantification of the amount of a product that can be harvested from an area of forest.

How is it done? First, the product quantity is measured for a small sample of the population. This is then related to an easily measurable characteristic of individuals enumerated in the overall inventory, using models such as regression equations.

Measuring the product

Measuring parts. NWFPs can be almost any part of a plant or animal, and each needs to be measured using a different technique. Hence there are a huge number of different enumeration techniques in use - Table 8 illustrates a small range.

Methods for measurement. Specific methods are few, and there is little standardization even for the same type of plant part - for example, Table 8 gives three different ways of measuring fruit yields. Such differences can relate to:

· ecology: e.g. marking and repeat counts may be the only method if fruit do not fall when ripe, or harvesting might stimulate fruit production;

· tree structure: e.g. random branch sampling is only realistic if the branches can be reached, whilst ground level traps may be the only alternative for fruit which are inaccessible or difficult to see from the ground;

· objectives of the assessment influencing the units which must be counted: e.g. for marketing or legislation - rattan in India must be measured in terms of dry weight as that is what is used for permits, whilst in Indonesia it is quantified in lengths; and

· ownership and number of users: e.g. if a tree is owned and harvested by a single individual, they can be asked to count ripening fruit.

Deciding on a measurement technique thus involves consideration of the product type, characteristics, and objectives of the assessment along with pragmatism about what will work. Section 5 gives some guidelines.

Table 8: Examples of techniques used for quantifying product yield

Variable	Methodology	Source
Fruit yield per season	Ground level traps. Four isolated trees selected, 15 1-m² plots randomly located beneath crown. Number of intact, predated, immature and mature fruit recorded every 7-10 days in plot.	Peters, 1996a
Fruit yield per season	Fruit counted in situ on sample trees at frequent (weekly) intervals. Counted fruit marked with paint to avoid repeat counts.	Peters, 1990
Fruit, leaves, etc.	Randomized branch sampling. Branching pattern defined as numbered segments between branch nodes. Path from trunk to branch tip selected using random selection at each node. Fruit/leaf/etc. counts undertaken at distal end of path. Pooled results from several randomly selected branches is a non-destructive, precise and statistically reliable method of estimating fruit yield of tree. There are several refinements of method, e.g. path selection proportional to size of available segments at a node, importance sampling, etc.	Gregoire et al., 1995 Jessen, 1955 Nguvulu, 1997
Leaves	Pipe model. Non-destructive regression technique for estimating leaf biomass and area from branch cross-sectional area. Pipe model based on observation that transpiration rate of canopy is proportional to leaf area, sapwood cross-sectional area and conductivity of water transporting tissue. Therefore size of stem is proportional to leaf mass and area. So can estimate leaf mass and area from measurement of stem cross-sectional area (NB: needs to be very accurate ~mm). Sample branches selected systematically to represent different branch heights. Regression analysis without constant.	Nygren et al., 1993
Palm leaves	All leaves measured. Partially open leaves counted as fraction of open leaf. Leaf length measured monthly to track growth.	Cunningham,  1988
Palm stem increment	Leaf scars counted at monthly intervals. Stem growth quantified as height increment (cm) per leaf scar.	Olmstead & Alvarez-Buylla, 1995
Palm age	Count of leaf scars, assume constant rate of leaf production to give estimates of age and numbers of years to reach critical heights.	Pinard, 1993
Bulb size	Measurement of maximum width of largest leaf on each plant. Regression analysis performed on a random sample of 50 plants at each site indicated that the largest leaf's maximum width is strongly correlated to total leaf area. Total leaf area already shown to be an indicator of bulb size.	Rock, 1996
Bamboo biomass	Measure clump dimensions on orthogonal axes at ground level, 1 m and full canopy extent. Map these as concentric ellipses. Determine biomass as volume of cone projected upwards from the base of the clump. Site index = _ clump volume/clump density in plot. Site clump area = _ clump area.	Widmer, 1998
Bushmeat weight	Opportunistic records of weights of captured animals in three villages used to supplement animal census.	Lahm, 1993

Choice of sampling scheme

Whilst there are many ways of measuring yield, the choice of sampling scheme for selecting individuals to measure is more restricted. Subsampling of a small number of individuals within the overall inventory sample is usual - as making detailed yield measurements on every individual in the inventory is hard work. There are two main ways of sampling individuals on which to measure yield.

Double sampling: this is done separately from the main inventory, and does not need to use the main inventory's plots. Using a smaller, independent sample, detailed yield measurements are taken on individuals in the population. The data from these measurements are then used to form models of yield against an easily measured indicator of overall size. Yields for each unit area can then be extrapolated from the measurements taken in the main inventory. Case study 6 is a useful example, others include: a survey of bark thickness independent of the main inventory of Prunus trees; berry yields from research plots applied to MRI inventory in Finland; inventory to quantify animal biomass based on bodyweights from market records. Double sampling has some constraints:

· sampling of measured individuals must be statistically sound;

· the sample measured must have some predictive variable in common with the main inventory, i.e. at least one of the measurements taken should also be taken in the main inventory (e.g. tree diameter or leaf length);

· ideally, the yield sample should cover the full range of sites of the main inventory, i.e. should be representative of the whole inventory area.

Case study 6: Developing a biomass table for shrub bark in Nepal Height and diameter at 30 cm above ground were measured for all samples of a shrub used in paper production. A subsample of each utilizable component was taken to a laboratory for oven-dry weight determination. The regression of utilizable dry bark weight against diameter at 30 cm above ground was selected as the best model. The results gave a basis for determining the rotation length for bark removal and allowed management recommendations to be made. (Jeanrenaud & Thompson, 1986).

Multi-stage sampling: This takes place alongside the main inventory but uses 'nested' subsamples in each inventory plot, creating a hierarchy of plots within plots. The advantage of this approach is that the sample is evenly distributed across the area, making estimation of yields in certain parts of the area possible. Such site-specific estimation might be useful where conditions that influence yield vary across the assessment area, making a single subsample unrepresentative. Case studies 7 and 8 give examples.

Case study 7: Assessing the potential of cane products on Barateng Island, India In an area of 123 primary blocks, 32 were randomly selected. In each of those 32, three secondary blocks were demarcated. Within the secondary blocks the number of commercial and non-commercial canes was counted. In a quarter of the secondary block, canes were cut then weighed every day until the weight was constant (i.e. until it was dry). The data was used to prepare cane stock tables by stem density, length and weight per hectare. (Sharma & Bhatt, 1982)

Case study 8: Multi-stage bamboo enumeration protocol in India Clumping species: three levels of sampling Whole plot - all clumps counted N-W quadrant (one-quarter of plot) clump diameters measured One in eight clumps - culm number, age, soundness, size, condition, average culm height and quality recorded Non-clumping species One-eighth plot - condition, age, average height, total number of culms, etc., recorded Utilizable green to dry weight relationship One mature culm from each culm diameter class cut from the first clump in each plot. Length is measured 25 cm from ground to 1 cm diameter. Whole culm weighted in field and 30 cm section taken from bottom, middle and top of cut culm for determination of dry weight. (Rai & Chauhan, 1998)

Disadvantages and difficulties include:

· detailed measurements may not be possible, for example if equipment is not mobile; and

· there may be insufficient samples of rarer but important individuals, such as infrequent large trees that contribute disproportionately to yield, for statistical analysis. If this is the case then additional subsamples might be needed or double sampling might be preferable.

Sampling decisions might be influenced by measurement practicalities. An example is if dry weights are needed, it might be preferred to sample in areas of high density to collect enough samples to efficiently use drying facilities.

Conversion factors? Usually the average amount of product per individual. Depending on the type of product, this can be the average size of an individual (e.g. for fruit or bushmeat), or the ratio of dry weight to green (undried) weight (e.g. for bark used in the dried form).

The key factor to the adequacy of a sampling scheme is often in the number of subsample replicates. Often NWFP assessments use too few. For example, one using only 8-15 fruit fall traps per tree for eight trees can at best only give a rough guide to the total crop yield. For management purposes over wider areas, replications need to be much higher, for example, sampling at least 30 and preferably hundreds of trees in a number of several sites.

Working out overall yield estimates

Measuring the yield in the subsample area gives results that can be applied to the data on overall population densities from the main inventory to estimate a total product yield for the area. There are several methods, using conversion factors to relate individual yields to total product quantity. Table 9 below briefly describes some of them.

Table 9: Summary of alternative methods for calculating an overall yield

Method	Description	Use	Example
Single conversion factor	The simplest method is to multiply the average yield per individual by the total number of individuals estimated by the inventory. Adjustments can use only the accessible or commercially sized individuals from the inventory.	This method is most appropriate when size of individuals does not vary much or is not related to product quantity.	Bushmeat. Single average bodyweights can be applied to the whole population to estimate total bushmeat biomass (Lahm, 1993).
Yield as a function of size	Simplest methods involve dividing individuals into size classes and calculating a conversion factor for each class. Refinements involve relating yield to another measurable characteristic of the individuals, such as length or bark thickness, through regression equations.	Most appropriate where yield is strongly related to size - i.e. the bigger the individual the bigger the overall yield.	Traditional methods: Use of diameter at breast height as a predictor of tree volume yields. Bark yield of yew trees: assume the tree is conical and bark area is the cone's area. Bark volume = area x thickness. Bark weight is then estimated, using a dry/green weight conversion factor, as volume x 0.4 (Jong & Bonner, 1995).

Yield estimation models can become more sophisticated with more data and longer studies. For example, in northern and eastern Europe models for berry yields are highly developed from long-term studies using regular inventory and permanent sample plots. Case study 9 describes systems devised in Finland.

Case study 9: Inventory and forecast system of wild berry yields for Finland Saastamoinen et al. (1998) and Salo (1999) describe what is termed the Marsi Enquiry which is a volunteer-based national berry production monitoring system. Berries = cowberry (Vaccinium vitis-idaea), bilberry (Vaccinium myrtillus) and cloudberry (Rubus chamaemorus) Yields of target species are observed in 1 110 permanent experimental plots in 57 municipalities across the country. In each compartment five plots of 1 m³ for berry yield were located in suitable vegetation types or where shrubs occurred and permanently marked. Enumeration Berries - species, count of flowers, raw and ripe berries. Site characters - timing of flowering and ripening, type of growing site, proportions of tree species, development class of trees. Sites visited several times throughout the flowering and fruiting seasons. Modelling Models are yet to be developed as there are insufficient years of observation to characterize annual yield variability. Previous models for Finland (Raatikainen et al., 1984) are: · V. vitis-idaea _ vegetation type, shrub cover, shrub height, stand age class, crown density and weather conditions; and · V. myrtillus _ vegetation type, stand age class, tree crown density, method and degree of coppice control and weather conditions. Dissemination Information sent in electronic form to Joensuu Research Station where it was processed using MASI (= berry and mushroom system) and presented as thematic maps. These maps are intended to inform the berry pickers about the time of flowering and the development of raw and ripe berries across Finland. The maps and notes describe the kind of sites on which the main yields were to occur and the level of yield for each species during 1997. Maps were distributed to media as five situation reports and were cited widely by press, radio and appeared as part of the evening news and morning television broadcasts.

Such systems may not be entirely appropriate for use in the tropics, but there are relevant features. Berry yields vary from year to year. To account for this, in their models the Finns considered:

· weather conditions for each year. Rainfall can be a particularly useful yield predictor;

· site quality;

· production data for a range of (good, medium and bad) years; and

· data from several seasons, to take annual variation into account.

Few NWFP yield models in the tropics use these factors. Instead they are typically very simple. The use of more complex methods has the potential to bring great improvements to predicting yields of NWFPs, particularly for fruit.

3.3 Measuring growth and production rates

A little about growth and productivity studies.

Inventory data provides snapshot information about the distribution and abundance of a resource species. Implementation of sustainable management plans also requires reliable data on the dynamics of the species, including information collected over long time periods on:

· population dynamics - recruitment, death rates, migration;

· growth rates and patterns;

· productivity; and

· impacts of harvesting on the species being harvested.

Methods used to determine growth and productivity for NWFP management are few, varied and mostly short-term studies. Most are based on forestry methods, and are only really suitable for trees and perennial plants - very few are designed specifically for NWFPs. Notable exceptions are for berry yields in Scandinavia (e.g. Case study 9). This section briefly describes the main types of method used for NWFPs.

Using permanent sample plots

Using PSPs for timber.

All individual trees are tagged, mapped, identified and measured at periodic intervals (2-5 years) over long periods of time (+15 years). The aim is to quantify growth throughout the lifespan of the trees. In practice, the focus is on the growth of established trees, and observations of early growth and fruiting are often omitted.

For trees

Permanent sample plots (PSPs) are the most common way of monitoring tree growth and yield. Data from such plots are used for predicting and modelling timber yields. The PSPs are essential for use with long-lived trees where timber accumulates slowly. In several cases NWFPs have been enumerated in PSPs, and in some studies the PSP protocols have been adapted for NWFPs by:

· including phenological observations - i.e. of seed and fruit yields;

· focusing on early establishment rather than long-term growth; and

· using shorter time periods and seasonal observation to observe fruiting.

Further PSP reading: Adlard, 1990; Alder & Synott, 1992

The use of PSPs has been adapted for fruit producing neotropical trees, palms and shrubs. Most have basically similar protocols (see Box 3). The protocols tend to focus on a single species, either because it is the only one in the stand, or the only one of interest. This is different from most timber PSPs, which include all species to provide a wider understanding of the dynamics of the forest as a ecological community, and to pre-empt changing demands for species.

Box 3: Permanent sampling plots (PSPs) protocols used for fruit production Studies on Amazonian fruit from trees, shrubs and palms. Sampling design: single plot subjectively located in dense stands of target species. Plot configuration: arrangement of 20x20 m contiguous subplots into a 1 ha square or rectangle. Enumeration: Adult trees tagged, mapped and measured for diameter and height. Seedlings counted in a number of random 1 m² quadrats within main plot and measured for height. Phenology - direct observations of small numbers of tagged trees spread across size range at frequent (weekly) intervals. Fruit - several protocols including fruit marking in situ and ground level traps observations repeated at frequent intervals (weekly). Study period: one-to-two year study (though the intention is that several will be longer-term studies). (Peters et al., 1989; Peters, 1990; Peters & Hammond, 1990; Peters, 1996b)

For non-tree products

PSPs have also been used for non-tree NWFP species (e.g palms as in box 4), usually using plots already in place -use of subplots for sampling NWFPs inside the PSP is common for focusing on seedlings and saplings, as well as for looking at fruiting and seeding of adults.

Box 4: Using permanent sampling plots (PSPs) for palms in Mexico A study of a single palm species in Mexico used data from biannual observations over seven years. It recorded mortality at all stages, and used large numbers of replicates (50 samples) on three sites (Piñero, 1984). Similarly, a small-scale study of growth and reproduction of a Mexican palm recorded data from biannual observations over four years on three sites (Oyama, 1990).

New developments

There is as yet few PSP systems being developed specifically for NWFPs. Users of PSPs for NWFPs should note the long experience of plant biologists, ecologists and foresters in developing effective, new systems for NWFPs. An important consideration is use of an appropriate interval for re-measurement - this should reflect the seasonality and longevity of the resource species.

Comparing harvested and non-harvested sites

Further reading:

Cunningham, 2001

This method uses observations at paired sites, often using PSP style plots. The paired sites should be of similar conditions (vegetation, topography, etc.), with one site having been harvested while the other has not. Ideally, there should be several matched pairs of sites in the study. This allows statistical testing of differences between harvest and non-harvest regimes, but may not allow analysis of different levels of harvest or different types of management practice. Table 10 below gives some examples, but note that most of these have weaknesses - including insufficient replication, short-term studies and subjectively located plots. Alternative designs pair plants rather than plots and consider harvesting impacts on the resource populations.


Table 10: Productivity studies undertaken on paired study sites

Author	Product, location	Sites	Sampling design	Replicates	Duration
Waters et al., 1997	Truffles, California	Two sites representing old growth & mature plantation	Systematic	Four plots at each site	2 years
Olmstead & Alvarez-Buylla, 1995	Palm leaves, Mexico	Four sites representing harvested and unharvested secondary forest	Subjective	No replication, one plot per site	2 years
Runk, 1998	Palm leaves and seeds, Ecuador	Three management regimes stratified by degree of inundation	Subjective	No replication, one plot per site	1 year
Konstant et al., 1995 & Sullivan et al., 1995	Palm, Namibia	Two sites representing high and low human and livestock densities	Systematic	Ten plots at each site	Not reported

Experimental harvests

This method allows the researcher to test different harvesting levels and thus assess the impact of different management practices.

Different harvest rates are applied to different plots. The results are compared to each other and to a control, where there has been no harvesting. This is the most direct way to assess the impact of NWFP harvesting on the population and has been used for a range of heavily exploited species. An example is shown in Case study 10.

Case study 10: A harvesting impact study A study in Great Smoky Mountains National Park in the United States of America responded to concerns about the casual harvesting of 'ramps' (Allium tricoccum). Three sites were selected according to the ease of access, abundance (at least 15 m² with 20 plants/m² and not regularly harvested). Fifteen 1x1 m plots (three replicates of five treatments) in a non-linear arrangement at each site. Maximum leaf width of largest leaf of each plant in plot measured (leaf width = bulb size). Five harvesting treatments: control, 25, 50, 75 and 100 percent harvesting. Plants harvested without bias using traditional methods. Leaf widths and flower/fruit production of remaining plants and recruits measured for 4 years post-harvest without further harvesting. Harvest technique: Three replicates of three 0.5x0.5 m plots established at one site. Plants harvested using three methods; control, complete removal and partial removal of plants. All plants too small to have been harvested removed from plots to avoid counting as regeneration in subsequent years. (Rock, 1996)

For all harvest comparisons, standard experimental procedures should be followed and attention paid to ensuring sufficient replication and use of a control. A control is especially important as there is likely to be a high level of uncontrolled variance between plots in the first place. Protocols are typically purpose made for each resource species.

Measuring individual plants on several occasions

This method uses repeated measurements of individuals which are not inside permanent plots. Individuals are marked, and productivity measured regularly to determine growth over a relevant period (i.e. a month, season or year - depending on the species or product being measured). Case study 11 is an example. To be able to extrapolate total yield from this, it is important to have an estimate of the total size of the population.

Case study 11: Palm leaves in southern Africa Two areas of dense palms, with a high possibility of commercial harvesting, were selected because they were easy to find and access. A number (16 and 34) of palms were marked in two sites and grouped into three size classes. Leaf lengths of each palm were measured at monthly intervals. Growth was calculated for each size class as the annual production of new leaves, defined on the basis of leaf length. (Cunningham, 1998)

Key point: It is critical that reliable methods of selecting sample individuals are used. Ideally, random selection should be used, using either stratification (for example, by size) or using nearest-neighbour methods (selecting the nearest individual to random points).

This has advantages over using actual PSPs:

· adequate numbers of trees can be rapidly identified, whilst there is no guarantee of getting adequate numbers of samples in plots;

· there is no need to maintain permanently marked plots, making it low cost and straightforward; and

· can be used for animals through marking and re-capture, but is rarely used as capture is often difficult.

Disadvantages include:

· a lack of detailed information about the environs of the samples (e.g. about density of competing vegetation); and

· sampling is at risk of being opportunistic and subjective, which is poor from a biometric point of view.

This method can be useful where the resource species is widely scattered and easily observed, and where interactions with other vegetation are not important. It is often used in farm-bush and savanna situations.

3.4 Determining sustainable harvest levels

The complexity of `sustainability'

Sustainability is a complex concept with many interpretations, ranging from idealist definitions to practical guidelines.

Defining `sustainability'

A rather idealist definition for a sustainable NWFP harvest is:

"If the harvest has no long-term deleterious effect on the reproduction and regeneration of populations being harvested in comparison to equivalent non-harvested natural populations. Furthermore, sustainable harvest should have no discernible adverse effect on other species in the community, or on ecosystem structure and function." (Hall & Bawa, 1993)

However, it is virtually impossible to remove anything from natural forests without creating some noticeable change. A more pragmatic approach to sustainable harvesting might require there to be "no loss in species and no irreversible changes in ecosystem processes" (Boot & Gullison, 1995), but even this is difficult to demonstrate.

Practical interpretations of sustainability for timber management include:

· allowable harvest levels should not exceed a level that can be harvested from the population in perpetuity without damaging its vitality; and

· annual harvest should be constant and available in perpetuity.

Sustainability and NWFPs

Further reading: Cunningham, 2001

Even for timber where growth rates are slow and there is considerable experience of sustained yield management the attainment of a relatively constant level of production is difficult. The search for sustainability for NWFPs is even more complex:

· there is often strong variability between production from year to year (e.g. good fruit crops one year, poor the next); and

· extensive, regulated management is unusual.

Working out what is a `sustainable' harvest for many NWFPs remains problematic. Thorough understanding of their productivity must be interpreted from ecological and harvesting studies. These involve determining the rates and patterns of variation in recruitment, growth, mortality, and reproduction, and how these patterns relate to environmental and management changes.

There have been few methodological developments for determining sustainability, for several reasons:

· a common assumption is that traditional management practices are sustainable;

· available resources are often limited and seldom directed towards doing biological research on NWFPs; and

· implementing sustainable management is perceived as costly and infeasible, so developing such systems are not prioritized.

Population studies and assessing harvests

Matrix models predict future populations using probabilities to calculate the likelihood of individuals surviving, growing, dying or reproducing.

The studies that have been done to try to work out sustainable yields have used a range of approaches, including:

A matrix model is...

Looking at population dynamics - biological approaches use simple matrix models or `rules of thumb' based on population dynamics. If sufficient data on birth, death and growth rates are available, then these approaches can identify the theoretical upper limits of sustainable extraction, i.e. the total productivity.

Establishing appropriate harvests - look at the impacts of harvesting on the ecosystem and economic returns from the forest. Combinations of resource-based assessments and socio-economic surveys are not uncommon, with a focus on harvesting and revenues, rather than biologically sustainable harvesting levels.

A step-by-step methodology incorporating all these approaches might be a helpful start for determining sustainable harvests. One proposal (Gould et al., 1998) is to:

· delineate the current supply area;

· determine current supply;

· estimate growth and yield or target species;

· determine current demand;

· compare short-term supply and demand, and evaluate management options;

· assess secondary ecological effects;

· repeat the process for future time periods; and

· summarize results.

Some of the best examples of determinations of sustainable yield for NWFPs are again in north and eastern Europe for berries - despite harvesting rates being as yet far below available yields (Rutakauskas, 1998; Saastomoinen et al., 1998).

Further RVA reading: Cunningham 1994, 1996a, 2001

Assessing how close a species is to overexploitation

Identifying species at risk

Indigenous and scientific information must be integrated, with matched local and scientific names making the link.

The `rapid vulnerability assessment' (RVA) method collects information to identify species, resources or sites that may be at risk of overexploitation. It was developed as a quick way of collating both scientific and indigenous information about a resource species, and has been used to recommend whether or not that resource species is suitable for harvest.

Information: life-form, habit specificity, abundance and distribution, growth rate, response to harvesting, parts used, patterns of selection and use, demand, seasonal harvesting, traditional conservation practices, commercialization.


The assessment is made in several stages:

· standardized field sheets are used to collect information for each species;

· information is evaluated according to sustainability criteria drawn from ecology, socio-economics, and economics - some examples are shown in Table 11; and

· each species is then assigned a management category from a selection of eight, each with a set of management recommendations.

Table 11: Criteria used in rapid vulnerability assessment

Criteria	Potential for sustainable use
	Low	High
Ecology	Low abundance	High abundance
	Slow growth	Fast growth
	Slow reproduction	Fast reproduction
	Sexual reproduction only	Vegetative reproduction
	Habitat-specific	Habitat non-specific
	High habitat diversity	Low habitat diversity
	High life form diversity	Low life form diversity
Life form	E.g use of grasses is likely to be more sustainable than trees
Parts used	E.g the use of leaves/fruit/stem is more sustainable than of the roots (if damaging) or the whole plant.
Method of harvesting	Potential for sustainable harvesting is higher if size/age classes are not selected - harvesting of only one particular age or size class can place pressure on the whole population.

Table taken from Watts et al., 1996

Is it biometrically reliable?

The comprehensive requirements of RVA provide a useful checklist for collating information from a wide range of sources, and provide information that is available for later updating and modification, for example during inventory. However, it does not:

· guarantee good information - much is often lacking;

· include inventory activities - it is a rapid `first look' at a species; and

· quantify information - it is subject to interpretation.

Whilst it may seem complicated at first, experience suggests that new users of the method can usefully develop scoring systems for the criteria and simple and transparent ways of translating these into management categories. Such evaluation systems are a very informative first look at the problem and for selecting candidate species for commercialization.

Adjusting harvest levels if they appear unsustainable

A simple and attractive method has been proposed for periodically adjusting harvest levels to ensure sustainability, using minimum regeneration levels as baseline (e.g. Peters, 1994, 1996a). Box 5 gives more details of how this works, and Figure 4 illustrates basic principles of the cycle.

Box 5: Harvest adjustment method for assessing sustainable yield from trees Regeneration survey. Network of permanent, small (5x5 or 10x10 m) regeneration subplots in PSPs for larger trees. Total number of desired seedlings and saplings of the required species less than minimum diameter for inventory tallied into four size classes and recorded. The initial data represent the threshold values by which sustainability is evaluated. These plots are enumerated at five-year intervals. If at a subsequent enumeration seedling or sapling density drops below the threshold value, the harvest intensity is reduced. If levels rise then harvesting levels can be increased. Successive approximations are used to try and stabilize seedling and sapling densities preferably at the original threshold level. Harvest assessments. Visual appraisals of the behaviour and condition of adult trees conducted along with harvesting activities. During routine harvests the health, flower and seed abundance and harvesting impacts are recorded for marked trees in yield plots. Information collated and tracked for the individual trees. If specific problems are identified, e.g. loss in vigour, increased seed predation, drop in productivity, etc. this should also initiate harvest adjustments. (Peters, 1994, 1996a)

Care is needed with this method, due some of the assumptions made in it:

· It assumes that constant annual yields are possible and desired.

· Threshold regeneration density is determined from only one season's sampling, with the assumption that this can be the constant. Given that reproduction notoriously varies from year to year, this may be highly inaccurate.

· It gives no guidance on how to determine initial harvesting level, and seems to assume that current levels are an approximation of the sustainable yield.

· It assumes that the damaging effects of overharvesting can be observed over short time periods. However, for long-lived species dependent on natural regeneration, effects may only be apparent after years of close observation.

This type of method could be improved by introducing the scale and pattern of variability in productivity. This would require observations over a number of years, and could be complemented by climatic data, especially rainfall. This would provide a basis for developing yield prediction models such as those mentioned earlier for berries in northern Europe. Appropriate harvest levels could then be decided relative to:

· long-term yields, future population level, or a forecast of annual yields;

or

· current demands, with an understanding of impacts on future yields.

Figure 4: Flow chart of basic strategy for establishing sustainable harvest of NWFP plant resources



Monitoring cycle

(after Peters, 1994)

Using models to predict future yields of plants

Long-term sustainability of an exploited species depends on the impacts of harvesting on its whole life cycle. Using models that represent life-cycle dynamics can contribute to developing reliable estimates of sustainable harvest levels.

Using life cycle dynamics

What are life cycle dynamics?

Every individual goes through several life-stages from birth or germination to maturity, old age and death. For each stage, models can apply different rates of: growth; fecundity/new births; and mortality/deaths.

Models use these data to predict the state of future populations under different conditions.

Life cycle models are used in forestry and wildlife management, but have only recently been used for working out sustainable harvesting levels for NWFPs. The most commonly used NWFP population model is the `matrix method' (Peters, 1996a). It predicts the number of individuals in different age or size groups (the `size-class structure') in future populations under different harvesting regimes, based on the present population and estimates of birth, death and growth rates in each age or size class.

How does a matrix model work?

It works by jumping forward in fixed time intervals, usually one year. First, the population is divided into size (or age) classes. The structure of the current population is the number of plants in each class, determined from the field survey data. The model uses life-cycle dynamics to estimate the probabilities of each individual's chance of surviving from one life-stage to the next over given time periods. This is done for several years to predict the future population structure. Box 6 provides an example to explain such a model. The sustainability of different harvesting regimes is tested by changing the reproduction and death rates (because of harvesting) used in specific size classes in the model.

 

Box 6: An example matrix model to work through Looking at a non-annual plant, the size classes might be: 1=seeds; 2=small, non-reproductive plants; 3=larger plants producing some seed; 4=fully reproductive plant. Time steps are in intervals of one year, and year 1 is 'now'. Example calculations: The number of plants in size class 3 in year 2 is the proportion of those now in class 2 which survive and grow fast enough to enter class 3 in one year. The number of seeds produced by each size class during year 1 gives the number of seeds in class 1 in year 2. Germination rate predicts the number of seeds (size class 1) in year 1 which will be in size class 2 in year 2. The transition matrix represents this data and calculations. The probabilities used can be altered to represent different harvesting levels and regimes. The state matrix represents the population structure at a given time.

 

Further reading on sampling issues for population dynamics: Bowden et al., 2000); Gagnon, 1999. Also see USGS. Biodiversity monitoring programme Web site: www.mp1-pwrc.usgs.gov/powcase/index.html

Biometric adequacy at risk

There are several areas which reduce the usefulness of these methods.

· The models require information from PSPs, including the level of flowering, seed production and dispersal, germination and predation. This kind of information requires frequent observations, ideally over several seasons to allow for variation. As noted earlier, few PSPs for NWFPs have been studied for more than one or two years, and rarely provide biometrically reliable data, but rather a general indication of rates. This means that often the basis of life cycle models may not be reliable.

· Most of the impacts of different management methods are hypothetical, and have not been demonstrated by observations or experiments.

· They assume that mortality, growth, and regeneration rates remain constant, unless changed by management. In reality they will also vary due to other factors, such as forest fires or climate variations. An alternative approach is population viability modelling, which attempts to predict whether the species can withstand chance events over long periods. This has not yet been used in NWFP assessments.

Further reading on MSY: Caughley and Sinclair, 1994

Models for assessing the sustainability of hunting

Managing for a sustainable yield of animals is clearly different from doing so for trees, and different methods have been developed by wildlife managers. Population dynamics are different as survival is generally density dependent and there is a maximum carrying capacity on any area of land. Carrying capacity is the maximum number of animals that can populate an area. At this point, births are equal to deaths and the population is stable. Above it, deaths increase and the population decreases. Below it, the population is growing. If the population falls too low, breeding will not occur. There is an optimal population level which ensures maximum birth rates and survival of new individuals. This is known as the maximum sustained yield (MSY).

Direct estimation of MSY is difficult - again because of the lack of detailed data on population dynamics, and how they respond to changing animal density. With this in mind, a simplified approach has been developed and has become popular (Robinson and Redford, 1991, see Box 7 for more details):

· determining population growth rate and establishing population density, in order to

· estimate overall annual production, then

¬ setting sustainable harvest at 20-40 percent of the annual production, depending on the species' longevity.

This method has limitations:

· it gives only an initial assessment of hunting impacts, and should not be used for setting quotas; and

· it assumes animals do not die before they reproduce, thus overestimating population growth. Including mortality in young animals highlights a greater level of overexploitation than previously estimated.

What is optimal foraging theory? This describes hunting choices in terms of the costs and benefits in energy of a particular hunting strategy to the hunter.

Models using optimal foraging theory to simulate the impacts of hunting strategies on wildlife populations may also be useful. One example (Winterhalder & Lu, 1997) simulates resources species, ranking them according to which ones hunters would prefer. This approach takes account of: differences in local preferences; availability of different resource species; and different types of hunting practices. In this way it calibrates for real, multi-resource situations, but may be expensive and complex to set up with real data.

Box 7: Method for assessing sustainability Calculate maximum production in animals km-2 Model variables (actual density measured; the other parameter values estimated or taken from literature). Actual density (D) - numbers per square kilometre. Predicted density (D2) - predicted from linear regression of log10 population density against log10 body mass by dietary categories. Intrinsic rate of natural increase (rmax) - highest possible without any limitations estimated using Cole's (1954) equation: 1 = e-rmax + be-rmax(a) - be-rmax W+1 where: a = age of first reproduction, w = age of last reproduction, b = annual rate of female births. Cole's formula assumes no mortality (error small if mortality is not significant before age of first reproduction). Maximum finite rate of increase (_max) - exponential of the intrinsic rate of natural increase (ermax) and is the increase in population size from time t to time t+1. Production (P) - addition to population = (birth + immigration) - (death + outmigration) + survival (to end of specified time period i.e. 1 year). Assume: · Predicted densities are higher than observed and that the predicted densities are close to, or at carrying capacity for the species. · Maximum production will be achieved when the population density is at 60 percent of carrying capacity. Subtracting 0.6 D2 maintains the population at the same density. Production Pmax = (0.6 D2 x _max)- 0.6 D2 Estimate of potential harvest It is assumed that the average life span of a species is a good index of the extent to which the harvest takes animals that would have died anyway. Thus harvest levels are set according to the longevity = age of last reproduction of the species: for very short-lived spp, < 5 years, harvest levels = 0.6 P; for short-lived species, between 5-10 years, harvest = 0.4 P; long lived spp, > 10 years, harvest = 0.2 P. Most significant assumptions are: · Model assumes density dependence in production, i.e. production increase with decreasing density such that it is at a maximum at 0.6 of the carrying capacity. · The proportion of production that can be harvested without depleting the standing population. Indirect confirmation could be to compare the weight distribution of harvested and non-hunted populations, if they are the same then can argue that hunting is taking animals that would have died anyway and hunting can take a higher proportion of total production. NEED to have the following to use this method: average body mass, food preferences (if using predictive equation), age of first reproduction, age of last reproduction, and annual birth rate of female offspring (average population density for study site - if not using predictive equation). Model should NOT be used to generate single species harvesting schedules. It CAN provide a first assessment of the impact of hunting on wildlife populations. (Robinson & Redford, 1991)

3.5 Monitoring the success of management actions

Why monitor?

Monitoring allows you to assess whether the interventions made have been successful and how they could be improved. It should always be part of the management process. Ideally monitoring should begin before any changes are made in order to provide a baseline against which to assess success of management actions.

Measuring change over time

Reliable assessments of changes due to management actions require quantitative and biometrically rigorous inventory. However, less direct approaches - for example, market surveys, harvesting records, forestry indicators - are useful for bringing attention to potential problem areas and informing management decisions.

There is no specific methodology for resource monitoring - all of the preceding methods can be used. The difference is that the inventory is repeated at intervals, often every year or five years. What is important is the ability of the inventory design to separate real trends from random errors in estimates. This is only possible if sampling errors are low each time the resource is measured. This requires large numbers of plots, which could be too expensive. The expense of rigorous monitoring means that it has not often, if ever, been used for NWFPs in the tropics. Instead, efforts have focused on developing simple and cheap indicators.

Further reading on indicators: Lindenmayer et al., 2000

Using indicators

Whilst they are attractive, care must be taken with the choice and measurement of indicators. In particular there is a risk that they might not adequately reflect the state of the resource. Notable problems include:

· they can be difficult to measure - e.g. regeneration assessment used to monitor the state of the whole population (Peters, 1994);

· ensuring that the indicators are representative of the whole resource - e.g. use of market records, without knowledge of where the product is sourced from, may reveal nothing about local resource depletions (Milner-Gulland and Mace, 1998); and

· failure to use the indicators to refine or adjust management practices - there is little evidence of this happening, but it may be due to the lack of long-term monitoring programmes in place for NWFPs.

The two main approaches to monitoring the impact of NWFP extraction are:

· forest-based: monitoring the health of the forest and/or its resources remaining after harvesting; and

· market-based: monitoring the size and quality of what is harvested (e.g. Watts et al., 1996).

Ideally, both should be used. Participatory monitoring at the local level is also useful to improve the sharing of understanding about the resource between regulators and harvesters.

Further reading on harvesting impact studies: Cunningham, 2001

Looking at what is left after harvesting

In forestry, PSPs have typically been used to monitor the impacts of harvesting on future yields and on other elements of the forest environment. The NWFP monitoring schemes can be similarly established, but it is important that they are done according to strict biometric principles. This will increase the likelihood of the data being representative and useful for extrapolation over larger areas. As yet there are few established protocols.

In order to allow the impact of harvesting to be properly measured, there should be a non-harvested `baseline' that the indicators can be compared against. This is often neglected, but should take the form of a pre-exploitation survey or use comparisons between harvested and non-harvested sites. Important considerations include:

· ensuring that baseline surveys are done as early as possible and before management changes are made;

· verifying that they efficiently measure the relevant indicators of forest condition and productivity;

· linking social (PRA methods such as the `walk-in-the-woods') to the more technical (quantitative, mapping) approaches; and

· Benefits of local monitoring.

Data collected far from the forest is the less likely to reflect what is happening in the forest. Resource based information is therefore likely to be closer to the truth than Customs and Excise records.

In addition, records measured after the product has left the village cannot link the product to its harvest location.

limiting enumeration to resource species or indicator species rather than trying to measure everything - this is the preferred option for species harvested from outside forests i.e. in farmlands.

The PSPs for NWFPs have often been seen as high-technology, high-input methods only useful for relatively small, intensively managed areas, such as National Parks (e.g. Sheil, 1997).

Measuring what has been harvested

Records of harvests are a popular, quick and straightforward way of monitoring NWFPs. They can be quantitative (absolute weights and measures) or qualitative (relative measures - lots, few, etc.) measures of the collected product. Case study 12 gives an example of using harvest records.

Case study 12: Monitoring harvests in a Ugandan National Park In the past, National Parks staff in Uganda have monitored harvests using records official quotas. Forest-level monitoring plots are now anticipated in areas of heavy harvests. It is planned that information for monitoring of participatory management agreements for NWFP use should be collected from three sources. · resource users - involved with collecting information on resource harvests, illegal activities and the state of the resource; · patrol rangers - as part of their routine activities; and · formal ecological monitoring - PSP plots established for species indicated by the communities. Monitoring Ensuri (Smilax kraussiana) in Rutugunda Parish, Bwindi National Park. Park staff went with collectors to traditional collection sites in the forest. Each collection site was marked with a red flag and the date and product harvested was noted. Ensuri is a liana from which ground runners are harvested. Harvesting followed traditional practices which are apparently sustainable - the collectors have reported no change in extent of any of the patches. Each bundle of collected product was uncoiled and the park staff measured the length, number of nodes and middle diameter of each vine. Records are kept against the names of individual collectors. (Watts et al., 1996 )

Recording can take place at various points in the supply chain:

· at source in the forest;

· in village, local or national markets; and

· at the international trade level, as Customs and Excise or CITES import/export statistics.

Harvest records can be used for a variety of purposes:

· market-based records can put a monetary value on the product, which makes them useful to market researchers and socio-economists;

· detailed harvesting records are often used to determine hunting/collection quotas for the next season, generally supported by direct population assessments; and

· as an indicator of change in resources they can highlight the need for more detailed studies.

Harvest records are the most widespread form of NWFP resource data. However, there are important points for caution when using them:

· It can be difficult to distinguish whether the product in the market is from a wild or a domestic population. If the proportion of domestic product is overestimated, there will be an underestimate of amounts harvested from the wild population, which can make management or quota decisions inappropriate.

· Harvest records do not indicate changes in the way products are harvested. Two different harvesting methods may produce the same amount of product in the market, but one may be destructive to the resource and not be sustainable, whilst the other harvests sustainably. This is not uncommon, particularly where outsiders have been drawn into an area for valuable commercial products.

· Their use tends to assume that changes in harvest levels do actually reflect changes in the resource. This is not always the case - many factors may influence harvesting levels.

· They often do not account for changing social, market or price factors. For example, a product in high demand will remain in the market even if the resource is in decline - but the price will increase to reflect the scarcity. Thus, measuring the amount in the market can prolong overharvesting until the supply collapses. Case study 13 provides a further example. If market measures are not used in conjunction with resource assessments, it is necessary to ensure that they really are sensitive to and calibrated against actual resource availability.

Case study 13: The influence of socio-economic factors American ginseng harvesters in Virginia, United States, must report sales every 30 days during the harvest season. Changing levels in sales are assumed to represent changing levels in the population. In reality, changing levels in the trade of ginseng is more closely related to changing patterns of employment among the harvesters - when unemployment is high, more is harvested; when people have jobs, harvests go down as they have less time and don't need the extra money. With a predominantly rural and poor population this is not unusual. (Bailey, pers. comm.)

Local participation in monitoring

It is important for the local people who are using the NWFP resource to understand the basis for quotas and other management prescriptions in order to make the management credible to them. Their participation in resource monitoring can be an important strategy in gaining support for management prescriptions amongst harvesters.

Local people can also adapt and improve methods through their knowledge of the resource. Using indicators of resource condition chosen by local people can increase their commitment to both the monitoring, and to consequent management adjustments.

Appropriate methods can be critical in achieving the objective of continued monitoring by local people. This accounts for the widespread interest in developing participatory monitoring techniques which are appropriate to local capabilities.

3.6 Participatory approaches

Many NWFP studies have focused strongly on local involvement for improved and sustainable livelihoods.

Involving local people

Further reading on participation:

Ingles et al, 1999;

Carter, 1996

Many practitioners argue that if an NWFP inventory is to contribute to improved sustainability of local livelihoods then local people should participate actively at all stages of decision-making - deciding whether do the inventory, its objectives and design, fieldwork and data analysis. The reasoning is that participation can:

· be an opportunity for a two-way learning process;

· help to generate a sense of responsibility for the environment;

· help people to understand how and why management decisions are made, making decisions more acceptable locally in the long term, and making the whole process more sustainable;

· help people see the potential economic benefit of management changes and thus ensure those management practices are adhered to;

· help to resolve conflicts between managers and harvesters of the resource by building trust and securing access (e.g. Watts et al., 1996; Pilz & Molina, 1998); and

· ensure that the data collected will actually be useful for management.

Levels of involvement vary, as noted in Table 12. Establishing who will own and have the 'intellectual property rights' to the information collected is becoming increasingly important and should be decided early in the process.

Table 12: Degrees of participation - from co-option to collective action

Mode of local people's participation	Type of participation	Outsider control	Potential for sustaining local action and ownership	Role of local people in research and action
Co-option	Tokenism - representatives are chosen but have no real input of power	***********		Subjects
Co-operation	Tasks are assigned, with incentives; outsiders decide agenda and direct the process	*********		Employees/ subordinates
Consultation	Opinions asked; outsiders analyse information and decide on a course of action	*******		Clients
Collaboration	Local people work together with outsiders to determine priorities; outsiders have responsibility for directing the process	*****	***	Collaborators
Co-learning	Local people and outsiders share their knowledge to create new understanding and work together to form action plans; outsiders facilitate	***	******	Partners
Collective action	Local people set and implement their own agenda; outsiders absent		*********	Directors

(adapted from Cornwall, 1995, in Carter, 1996) ** indicate the relative strengths

Key message: Local knowledge can provide useful information about the resource, but verification of the data is usually required.

Use and value of local knowledge

Local knowledge can be useful in designing and implementing a NWFP inventory for various reasons, as highlighted in Case study 14 and noted in Table 13. Participation can improve the efficiency of the inventory and optimize local benefits. Enhanced efficiency can come from:

· local knowledge providing basic information about a resource (for example, through RVA methods described in 3.4); and this then

· helping to decide whether an inventory is actually necessary, and if so what kind of sampling design and enumeration methodology are appropriate.

Case study 14: Using local knowledge to design and implement an inventory for Pacific yew in British Columbia An interesting North American study that contains some lessons for tropical NWFP assessment is the work that has been done in Canada on inventory for Pacific Yew (Taxus brevifolia, bark harvested to manufacture a breast cancer drug). As a preliminary to large-scale inventory a decision needed to be made about which of two available forest maps (an ecosystem map and a forest cover map) would provide the best stratification and whether local knowledge could be used to select which strata should be sampled for yew. A questionnaire was sent to local foresters and ecologists to elicit their knowledge about the occurrence and distribution of yew among the mapped units on each map. The questionnaire data were compiled and used to identify high and low probability strata for yew on each map. Field sampling was weighted so that 80 percent and 20 percent of samples were placed in high and low probability strata respectively. Analysis of the data showed that overall estimates of yew populations produced using either map as a basis for stratification were not statistically different, but the standard errors for the ecosystem map were much smaller indicating that it is more precise and hence a more efficient stratification. The validity of the high and low yew occurrences strata as determined by the questionnaire was not challenged or tested in the analysis of results. Presumably this is because they were confirmed as accurate. If this is the case then the local knowledge was reliable even though a diverse range of individual opinions on yew distribution were expressed by the questionnaire respondents. This study demonstrates a means of using local knowledge in the context of a biometrically sound sampling scheme that does not compromise its integrity and may offer useful lessons for the use of indigenous knowledge in tropical NWFP inventory. (Jong & Bonner, 1995)

Table 13: Examples of areas of local knowledge and their possible uses in NWFP inventory

Local knowledge	Use in inventory
Species identification	Local tree spotters can be useful in the field (but see section below on taxonomy)
Important economic species	Can highlight species to include in inventory (e.g. RVA)
Vegetation classification/description	Can be used for stratification
Micro-climate types and distribution	Can be used for stratification
Soil types and distribution	Can be used for stratification
Harvesting techniques and frequency	Can improve enumeration methods and frequency
History of availability	Helps to prioritize species to include according to the level of threat or change
Current estimation of availability	Helps to prioritize species to include - and influences the decision on whether inventory is necessary
Ecology and distribution of species	Helps to decide on the most appropriate sampling method
Human interaction with environment (e.g. existing management)	Influences inventory objectives and design
Forest and resource value	Influences management objectives and hence inventory objectives
Socio-economic factors affecting NWFP management	Influences decision on whether to have an inventory and its objectives and influences interpretation of inventory results

Note that this table is not comprehensive and that the above uses of local knowledge are case specific, i.e. the types of local knowledge listed cannot always be used in inventory in the manner described.

To ensure that local knowledge is helpful in NWFP inventory, care should be taken to:

· ensure the local knowledge is collected and analysed using suitable methods and that it reflects reality; and

· design the inventory to investigate areas of doubt, and to allow the flexibility for review after the data collection has started.

Combining local and scientific knowledge

Getting the name right.

Consultation with herbaria experts from national herbaria or leading botanical gardens is often needed to identify samples and apply the correct botanical name. Specimens must be prepared properly and identification can take months, even years. Whilst this can be costly and time consuming, it is the most reliable way of naming. Developing links with herbaria helps ensure competence in collecting samples.

The gap between local knowledge and scientific knowledge cannot be bridged unless local and scientific names can be matched. The NWFP surveys tend to use local names in data collection - rather than codes or Latin names - because it is easier for local staff and collaborators to record the names with which they are familiar. However there are some important problems in using local names this way.

Incomplete and inconsistent use of names by local informants - There is considerable variability in local names. For example, in central Kalimantan (Wilkie, 1998), less than a quarter of local names are applied consistently, and these are typically those which are of specific use to the informant or are especially distinctive. Switching between product name and individual species name is common.

Mis-match between local and scientific names - Analysis of experience in Uganda shows single local names being matched to several botanical names (Cunningham, 1996a). Local names often refer to scientific genera rather than to species (see Table 14).

Further reading: Cunningham, 2001

Taxonomic difficulties - The taxonomic description and naming of species in tropical forests are notably incomplete, even for trees. This means that it is often not possible to give a taxonomic identity to a locally named NWFP species, and species identification can often only be taken as far as genera. This is the case even for rattans in most south Asian countries, where local names describe genera rather than species (Stockdale, 1995a).

However, it is essential to determine scientific names if information is to be compared between different areas - where different languages and local names may be used. This is difficult as few field guides or botanical keys are available for NWFPs and they require flowers or fruit (often inaccessible and seasonal) for identification. The most reliable way of verifying identifications is to collect samples during the inventory and have them identified by experts. Attempts to match names to samples later have been found to be unreliable. However, vouchering specimens in an inventory can be costly, time consuming and require skills which are not available.

Table 14: Correspondence of folk and scientific names

Correspondence	Explanation	Correspondence*
One-to-one correspondence	Folk generic = scientific species	61%
Under differentiated (I)	Folk generic = two or more scientific species of the same genus	21%
Underdifferentiated (II)	Folk species = more than one scientific genus	14%
Overdifferentiated	More than one folk generic = scientific species	4%

* Correspondence of Tzeltal Maya names (n=471) taken from Berlin (1994) quoted in Martin, 1994

Some synergy between local and scientific knowledge develops informally through participatory or informal approaches. However, formal methodologies (e.g. Box 8) for dealing with informally collected local knowledge are important in order to gain a deeper understanding of the knowledge and its relationship with scientific investigations. Formal methodologies also have the potential for more objective verification of qualitative ecological information, and for improved storage and retrieval of knowledge.

Box 8: Formal methodology for linking and analysing formal and informal information `Knowledge-based systems' have played a significant role in agriculture and are developing rapidly. They formally represent qualitative knowledge on computers. The processing power of the computer can then be applied to the representation, analysis and use of large and complex sets of knowledge. They have particular potential to be useful when integrated with numerical approaches to create integrated decision support systems. It keeps the links between 'units of knowledge' in a way that allows retrieval of overall concepts, rather than simple lists of the units of knowledge collected. However, this methodology may be of more use and relevance to researchers and development workers than it is for local people themselves, and may best be used in complement with informal methods. (Sinclair & Walker, 1999)

The role of participation in NWFP inventory

Few would question the value of local people's participation in resource assessment. However, there is great debate about whether participatory inventory can or should be biometrically rigorous.

· Biometric methods typically require sophisticated techniques, which are inappropriate and/or undesirable for use by local, untrained people. Where participation and learning is more important than biometric rigour, it is argued that the latter can be sacrificed.

· However, non-biometric, social science techniques rarely collect information that is reliable enough to guide management decisions regarding sustainable harvesting levels. Sacrificing biometric rigour means denying that local people need reliable information or robust management prescriptions.

The risk of collecting poor data is that local people will be disappointed with the results and lose interest in the process. It is important to ensure that all risks involved regarding the reliability of the information should be clearly understood by local people and outsiders. Box 9 highlights a few examples of how local people can learn and change their methodologies - doing it for themselves will encourage learning and understanding of the process.

The challenge? To make biometric methods accessible to communities.

Increasingly, it is being realized that communities do need biometric data (for example, to provide the basis of a management plan for submission to government for approval), and sometimes there may be an urgent need for reliable information (for example, where a resource is severely threatened). In these circumstances, despite the learning opportunities of doing it their own way, it may be appropriate to suggest that the local people undertake a biometric inventory.

Box 9: Local people and evolving understanding - some examples · Local communities might want to collect information only from areas they currently use - this is what they want to manage now. This might not help the researcher who wants to verify local knowledge about distribution and get figures for the whole species population, but the community might later decide they need wider information. · Local people might decide that indirect measures, such as average walking time to the resource or harvesting statistics, are easier to assess. Whilst this might seem sufficient when the resource is abundant, if it becomes threatened they will find they need more information for management.

It is always important to ensure that data are simple to collect and to analyse.

This does not mean that methods and designs should be unsophisticated, but that they should be simply presented. Methods are available that allow a complicated protocol to be performed, even where literacy is low (see Box 10).

Box 10: Appropriate methods? Some examples of success Illiteracy problems are increasingly being broken down by advances in technology - for example, the use of palm-top computers with symbols rather than words has been successful (Cunningham & Liebenburg, 1998; Cunningham, 2001), as have hand-held GPS (Stockdale & Ambrose, 1996). Use of older technology, such as mainframe computers and surveying equipment, requires much more training and may not be appropriate.

Experience and debate has shown that methods need to be adaptive to local needs whilst providing replicable, standardized data. In Nepal, the qualitative methods that have emerged from the promotion of JFM have not provided the detailed information required, and have now begun to evolve into more formal quantification of resources.

1 Cut lines will remain open only for a limited amount of time and communication between teams about access periods is important.