(Note: We decided to drop the bamboo inventory component of the original terms of reference, since it is a product of lower priority and is not yet a market chain being targeted by WN.)
Recommendation for Inventory and estimation of productivity of the Mbepp Gum Tree (Sterculia setigera)
The tree Sterculia setigera produces karaya or mbepp gum (Figure 12), an important commodity for export that is used in world-wide in cosmetics, food, and pharmaceutical industries. During the period of 1982 through 2002, Senegal’s production has fluctuated from a low of approximately 350,000 kg in 1999 to nearly 2,500,000 kg in 1989 (Johnson et al. 2005). Although much has been written on the production and value of mbepp gum, little has been published on the number of gum trees in the population. At least one study by Mbaye et al. (2005) only covered a 300-hectare area.
Fortunately, all the inventories conducted by PROGEDE included the species Sterculia setigera. A review of the database shed light on whether the existing plot design would be adequate for estimating the number of trees within a given area (Appendix A Tables 1&2). It appears that within the Tamba-Kolda inventory area, the tree is represented reasonably well: 26% of the plots had at least one tree recorded, and there are approximately three stems in plots where the species was present. These numbers indicate that while the stem density is not high, the density is sufficient to consider only using SIEF’s existing plot design. This is a practical decision. A specifically designed inventory would probably result in a larger plot, but the existing inventory already identifies the gum species and records associated attributes such as diameter and height reasonably well. The only attribute that needs more explicit mention is the state of the trees; it would be interesting to see the evaluation of the health by size class, since the future productivity will depend on it. Momar Mbaye performed such an assessment but it has not been used in management planning yet.
This inventory would give the user an average number of stems per hectare. Once the variability of the existing data is calculated, the production per tree (which we were told is on the order of 2.5 kg of gum per tree per year) can be multiplied by the stems present and the annual potential can be inferred. Ideally, the appropriate statistics (confidence interval) indicating the minimum and maximum potential should be worked up at the same time and based on specific target areas.
We recommend that the available inventory data on Sterculia setigera be evaluated, and then the need for further information assessed and prioritized as compared to the need for information on other non-wood products.
How to inventory the Baobab and estimate its productivity
Baobab trees have numerous uses. The leaves are an ingredient in sauces, the branches are cut for forage, the strips of bark are used for short rope cords, and the fruit’s flesh is edible and even exportable as a condiment and also as cream of tartar. Wula Nafaa is primarily concerned with estimating the quantity of marketable fruit produced so that contracts can be signed with processors and exporters.
We presumed that the diameter (age) distribution of trees is linked with the quantifiable supply of fruit.
A review of the frequency of occurrence of the baobab recorded in the SIEF inventory shows that this species is relatively rare on the landscape (Appendix A Tables 3&4). In addition, it is well known that this tree occurs in clumps or ‘parks’ (see Figure 13).
While unbiased estimates of the number of trees within the areas covered by the existing inventories can be calculated, the simple random design, or even a stratified random design with a low correlation of the stratification to the attribute of interest, is an inefficient sampling strategy. Also, it is clear from casual observations of the spatial arrangement of the trees that a plot significantly larger than 20 meters in radius should be used. For areas the size of a classified or community forest, the existing inventories will only provide marginal information for managing this species or for estimating the quantity of fruit. Several sampling strategies will be proposed depending on the availability of aerial photography and the ability to discern baobab trees on the photographs.
If aerial photographs are available
The objective is to test the use of aerial photographs to identify the presence of baobab parks. PROGEDE has 1-meter panchromatic photography for the regions of Tambacounda and Kolda. There are 3- and 9-meter photos also available, but it is important to use the highest resolution photography to assess the validity of this approach. The steps are:
Locate and GPS some baobab parks across the area where there are photos. The baobab parks should include ranges of sizes of the parks, diameters of the trees, and land uses. Not all baobab parks should be in agricultural fields. Also, locate and GPS groups of other large tree species that are found scattered in the area.
Bring the photo interpreter(s) to these areas with a lap top computer with the aerial photographs so that the photo interpreters can see the trees from the ground and see the trees on the photographs. Discuss the characteristics, if any, that can be used to distinguished the differences between the species.
Construct a blind test. GPS and locate on the photos about 6 baobab parks and 6 areas of large non-baobab trees. Give these locations to the photo interpreter(s) and ask them to identify which are baobab parks. Do not identify which of these areas have baobab trees or even say how many baobab parks are in the group. If there is more than one interpreter, be sure that they interpret the photos independently. Assess the results.
If baobab parks can be identified reasonably well on the photography:
It is assumed that the area to be inventoried is within either the Tamba-Kolda or the national inventory sample frame. Henceforth, both inventories will be called ‘extensive surveys’. Now divide the landscape into two strata: baobab parks and everything else. The baobab stratum will be described below. The existing SIEF inventory will be used to sample all areas outside of the baobab parks. The use of the existing inventory allows the estimation of baobab trees that have not been identified on the aerial photos.
Method for identifying the baobab parks: There are over eight hundred 1-meter resolution aerial photographs in the Tambacounda and Kolda regions. Only a sample of them will be interpreted for baobab. The basic steps are:
Divide the regions into square blocks by using UTM coordinates in a GIS layer.
Intersect the blocks with the aerial photo coverage and determine the portion of the photo that is within the block.
Randomly choose a relatively large number of blocks using a simple random sample without replacement.
Delineate all probable baobab parks within the chosen block; place blocks with any such parks into a “high” potential stratum and all remaining blocks into a “low” potential stratum.
Randomly choose a subset of the photointerpreted blocks using a simple random sample without replacement within the high and low potential strata; visit these blocks and record the baobab trees greater than XX cm in diameter at 1.3 meters above the ground. Sampling can be more intense for the high potential stratum and less in the low potential stratum.
For trees less than XX cm, randomly locate points within the parks delineated in step 4, and collect information on smaller trees and regeneration as desired on fixed-radius plots.
Count the number of fruits on each tree.
The size of the block to superimpose on photos in the GIS layer can be adjusted by piloting several possible grid sizes, and choosing one that has a reasonable chance of containing a baobab park but that is not so large that it will take a long time to interpret. Some suggested lengths of the sides are 100, 200, 400, and 800 meters; but any size can be used.
Here is a variation on the preceding design for visiting parks delineated on the photos:
It is simpler to adjust the size of the block and visit all of the parks than to sub-sample the parks, but sub-sampling is a legitimate sampling technique. Only some of the baobab parks are assessed on the ground on each photo. The parks to be assessed can be chosen by a number of methods but two methods will be described here.
Select the baobab park with probability proportionate to the size of the park. Follow the steps described above at each one of the selected parks.
If an approximate number of baobab trees greater than a certain diameter can be estimated from the photo, then a double sample could be constructed where a relatively large number of photographs are selected, the baobab parks identified and the number of trees estimated. A sub-sample of these parks is selected for ground measurement. This method assumes a linear relationship between the estimated number of trees and the actual number of trees. The advantage of this method is that if the relationship is strong between the estimated and the actual, then the estimate of the variance is reduced or fewer plots need to be recorded in the field. The big disadvantage is that if some of the ‘parks’ identified on the photos are not baobab, then there will be some positive estimate of baobab trees but zero recorded when the field crew visits the area. At the minimum, this will reduce the advantage of this method and at worst violate the linear relationship, producing biased estimates.
It is possible to measure diameters and count fruit on a sub-sample of the baobab trees and then model the number of fruit on trees by diameter class; but this procedure is not recommended when working with baobab parks. This would add to the complexity of the inventory, while a good estimate on large trees with many fruits is adequate.
One concern is that large trees vary significantly in the number of fruits they produce depending on whether the branches have been cut for forage or food by villagers. This added variability will decrease the effectiveness of modeling the fruit based on diameter, unless other factors are recorded such as classes for the size and health of the crown. But again, these other factors will further complicate the analysis.
The concern over varying fruit productivity will be relaxed when estimating fruit production outside of baobab parks, where no direct measurement of fruits is possible without return visits to the field.
In summary, it is recommended to use a simple survey strategy that relies upon few assumptions, at least for the first couple of forests to be inventoried. With additional information, more complex designs could be assessed for use later.
If there are no photographs of the area , or baobab parks cannot be identified reasonably well on the photography
Implement the first three steps under the heading for photography being available. The size of the blocks should be smaller than 200m since each block will be visited on the ground and the entire block surveyed for baobab trees. If the ground plot has a baobab tree, then sample the neighborhood of the plot using an adaptive sample design as discussed in Thompson and Sebe, 1996.* Do step 6 if information on smaller trees is desired.
*Thompson, Steven K., George A.F. Seber 1996. Adaptive sampling. John Wiley & Sons, Inc. New York 265p.