The effect of having a biased result is the improper assignment of the contribution of a given sample plot to the calculated average, and there is no way to know if the bias is in the direction of overestimation or underestimation. Thus it may take some years to detect that over- or under-harvesting is taking place. But overall, SIEF-1 works well for its intended use with the above caveat.
An example of likely bias in reported results due to differing sample intensity is shown in Figure 7. The differences in sampling intensity within Kothiary Rural Community (CR) are due to the inclusion of part of the legally-recognized Missirah Zone de Production Contrôlée, the Forêt Classée of Bala-Ouest, and an area with no label in the northeast corner of the rural community, all having gone through different plot selection. Some of the plots are part of the national inventory and are not present in the original SIEF database (SIEF-1). Other plots are part of the Tamba-Kolda inventory, and others may have been added as part of an intensification policy in forests of interest.
Only if all of the 5 key Yangambi forest strata in a rural community are wholly included in its boundary, each with its own uniform sampling intensity, will the results be unbiased. The results for rural communities with a mixture of sampling intensity will be biased towards the conditions found in areas that are heavily sampled. The magnitude of the current bias cannot be ascertained.
The same principle will apply to any forest area’s reported volume per hectare: plots within each given stratum should have their own sampling intensity and their own error (variance and confidence interval) calculations; and the combined stratum polygons that make up the forest would each contribute an appropriate weight to the average and the error.
In sum, the SIEF-2 program ignores how plots are distributed on the map and can produce biased results of unknown magnitude if the user requests average volumes per hectare in areas containing different sampling intensities. The good news is that this problem can be rectified with changes to the program and does not require new field information. There need to be provisions for differences in sampling intensity by using references to hectares represented by each plot. Some of this information is already available within the database; other information can easily be derived using the GIS interface.
Applying these needed changes to the program would be part of a larger issue involving the role of PROGEDE and that project’s relationship with other projects that are involved with writing PAFs for/with the Senegalese Forest Service. Corrections to SIEF-2 should be a high priority for PROGEDE since the program is providing critical information for land management plans.
Using the SIEF in new forests or in plantations
It is possible to enter additional plot data into the SIEF-1 database, but before this is attempted, there are two aspects that should be observed.
1) The database is a series of tables that are accessed by visual basic macros embedded within the database. It is recommended that PROGEDE staff assist, at least initially, someone from the Wula Nafaa project or that PROGEDE assigns someone from their staff to work on Wula Nafaa projects when needed, so that the addition of new plot data and its analysis can be done correctly.
2) Entering additional plots may change the sample design of the target area. The sample design with the additional plots should be fully understood and the ability of the program to accommodate any design changes should be reviewed prior to collecting additional field data. Such a review needs to be carried out by persons with sufficient statistical background to avoid building on faulty bases.
Based on the above information, it is recommended that the Wula Nafaa project continue to use the original SIEF-1 program with the help of someone in PRODEGE that completely understands the database, until members of the Senegalese Forest Service are sufficiently trained in its use. SIEF-2 may be useful in the future for writing PAFs once its mission and the sampling intensity issues are resolved.
There are some plantations containing Khaya, Gmelina,
teak (Tectonis grandifolia
), and other species, especially in the regions of western Kolda and Ziguinchor. Since these plantations are typically more dense and uniform than the Yangambi class-based forests discussed until now
, a different system of inventory may be more efficient when sampling these forests, for example, if they are large enough to become a community forest. The first requirement would be to map these plantations from photography (if the maps don’t already exist), then make decisions on the appropriate sample design. This needs to be accomplished together with Centre de Suivi Ecologique and the regional Senegalese Forest Service offices, as well as with any other forestry projects intervening in the regions.
Potential changes to plot design
In the context of using the SIEF in natural forests in the rainfall zones already covered, the nested plot design appears to be an efficient design with the right balance of subplot size and diameter range. The only subplot size that may need to be reviewed is the regeneration plots. If natural regeneration from seed is a concern for the Wula Nafaa project
, a review of the data should be made to determine if a larger subplot sizes are warranted for management purposes.
If forest types inventoried in new areas differ much from what is managed now, i.e., if they become more sawtimber-oriented or if they are plantations, a smaller, more intense plot design or a prism/relascope sample may be more appropriate. In order to recommend a plot configuration, it will be necessary to see the hectare size and the variability of the composition of the forests in question, with the assistance of aerial or satellite imagery interpreters from CSE.
The SIEF programs report outputs in terms of basal area, number of stems per hectare, and cubic volume per hectare. While these attributes should be retained, these units of measure are not very meaningful to villagers. The program should have the option of displaying the results in terms of sacks of charcoal or quintaux, the units of measure familiar to the villagers, using the appropriate conversion factors (see chapter on factors under charcoal below). The current practice is to convert by hand for use in the PAF
, which is another opportunity for the introduction of human error.
SIEF-1 and -2 both display sample errors for basal area, number of stems, and cubic volume for the total population of trees on the plots selected, but the output should include sample errors in all of the strata’s output tables separately. A common method is to produce the standard deviations within parentheses as a second output line immediately below the estimate of the mean or total population. This would give the user some sense of the variability inherent in the data and the confidence with which one can harvest the estimated product.
Include SIEF nontimber product species outputs in the PAF
The program should expand output tables to include non-wood products. The SIEF-2 already includes some non-wood product options which are inoperable due to the lack of data. These should be expanded, in cooperation with their clients such as Wula Nafaa
, and have the outputs in terms of units that are used in the marketplace. Examples of the SIEF output for selected nontimber product trees are in Appendix A
Include SIEF regeneration outputs in the PAF
The legal requirement for volume information is linked to the status of regeneration in the forest. Therefore, in order to better orient activities in the workplan, the PAF should contain SIEF-generated reports of the amount of regeneration found on plots during the inventories. There are data from hundreds of regeneration subplots available, so it should be possible to provide a thorough analysis to include in the management plans (PAFs). An automated process for generating useful regeneration tables for different Yangambi classes would be simple to build into the SIEF.
Assess and adjust volume equations
Adequacy of current models:
The selection of pre-inventory plots on which trees were cut for volume equation development was clustered in several areas. While it would have been preferred to not cluster these plots, the clustered areas are distributed fairly well across the intended area and probably adequate when using the two parameter models in the SIEF. There are some three parameter models that attempt to adjust for difference in the form of a tree with the same diameter and height, but such refinement, if possible in hardwoods, is not warranted.
Systematic overestimate of volumes:
The current volume equations were constructed to estimate the total biomass above ground and not just the merchantable portion of the stem for charcoal production. While it is valuable to have above-ground biomass equations, using these equations for charcoal production will overestimate the quantity of wood available to harvest. The above-ground biomass equations were derived from two components
, the woody portion of the stem that is used for charcoal and the remaining small branches. Fortunately both pieces of information appear to be available for all of the trees used to construct the biomass equations. A preliminary estimation of the unusable proportion of volume in the overestimate is about one-third for fuel wood and one-fifth for saw timber. Calculations showing the overestimate are in Appendix C
Extension into new areas or species:
If the Wula Nafaa project extends to ecologically different areas, particularly to the west of Kolda, then additional tree data should be collected to validate or revise the existing models. Also, if other species in these areas are important to management, then volume equations should be assigned to these other species by using one of the alternatives listed below. The entire list of species treated in the SIEF is in Appendix E
USE DATA FROM THE AREA OF INTEREST:
If there is already an existing volume equation for this species but the data come from outside of the area of interest, visit active logging operations within the area of interest. Work with sawyers to record the basic dimensions of the tree and record its yield to validate the equation. The visited logging operations should be well distributed across the area of interest and not clumped in one portion of the intended area of use.
If there are no logging operations in the area of interest, fell trees and record measurements for volume calculations within the area of interest.
If there are logging operations in only a portion of the area of interest, collect information from those where it is available and complement with either felling or non-destructive sampling of standing trees.
OBTAIN A NON-DESTRUCTIVE SAMPLE: Non-destructive sampling can be accomplished by measuring a subset of the trees on PROGEDE or WN plots. Use a Barr & Stroud dendrometer or similarly accurate instrument that can take upper stem measurements of standing trees (non-destructive sampling).
BORROW FROM ANOTHER SPECIES: If there is an existing volume equation from another species with a similar growth form, validate it by following the procedure in Step 1 above.
Development of a future Wula Nafaa – PROGEDE Relationship
PARTICIPATION IN THE NATIONAL INVENTORY: From a broad perspective
, Wula Nafaa should support the maintenance of PROGEDE as leader of the national inventory program. PROGEDE has most of the elements to perform this service and is willing to play this role. However, with this role come additional responsibilities that PROGEDE should embrace and that Wula Nafaa, through the Senegalese Forest Service, should promote. PROGEDE should have a charter clearly stating its role in the national inventory program. The charter should include the service-oriented nature of the unit.
The objective of PROGEDE to estimate the cubic volume of the major charcoal species was achieved by a simple but very effective sample design for the Tamba-Kolda area. Now the objective of the unit is changing to a supporting role for the all of the Forest Service and projects such as Wula Nafaa. This changing role requires a different strategy that grapples with the question as to whether previously installed plots are a one time phenomenon or part of a monitoring system where some of these plots will be re-measured.
There were different objectives for each of the three types of inventory plots within the SIEF-2 database: pre-national inventory plots to calculate variability; Tamba-Kolda inventory plots to estimate fuelwood in selected forests; and national inventory plots to estimate ecological characteristics and wood volumes at the same time. Projects such as Wula Nafaa simply need a volume estimate to satisfy legal requirements for their community-based management plans. It is tempting to use the SIEF “off-the shelf” to obtain the volume estimate; complex software and supporting research were used to develop it so its credibility is high. However, its current limitations should be understood by those who want to sustainably manage forests. This is an opportune time to carefully set up the SIEF for future use at the forest level as well as the national level. For this reason, it is worthwhile for WN and other forest management projects to support the continual improvement of PROGEDE’s SIEF system.
SIEF AT THE NATIONAL VERSUS THE FOREST LEVEL: An example of WN assistance to PROGEDE that will be valuable to Senegal is supporting the remeasurement of permanent growth plots established in Senegal’s eco-geographical zones. If, at a minimum, WN supports the remeasurement of Tamba-Kolda permanent growth plots, even though they also fall within a national monitoring framework, then the need for more growth data to support rotation ages at the forest level would be fulfilled, while contributing to the improvement of the SIEF.
An example of the confusion between national- and forest-level data needs is the stratification of the existing 2,900 plots by the Yangambi system (as in SIEF-1) versus their classification by the more generalized satellite system (as in SIEF-2). The satellite classification system was added at a later date to track ecological trends by taking advantage of the repeatability of satellite coverage. There is no doubt that this satellite stratification, which is based on density and size of trees, is an efficient design for estimates of current volume. But potential problems arise when an efficient design such as this, made for current wood volume, transitions into a national ecological monitoring program, where the objective is to track trends over time. By requesting targeted area-based information, projects such as WN can support those aspects of PROGEDE that lead to improvements in the forest-level components of the SIEF.
CONSIDERATIONS FOR FUTURE INVENTORY:
In the coming years, it would be natural to avail a fresh set of photography or satellite imagery. If these were to be stratified into the same classes as before
, these strata will be different from the past due to changes in land uses, growth, photo interpreters, or technological advances. Should the re-measurement be on the original stratification or should the plots be re-stratified? A rule in sampling is that the original sample design must always be taken into account. Estimates using a new stratification scheme and ignoring the original sample design will be biased (Schreuder and Alegria 1995). It is possible to re-stratify, and properly take into account the original sample design, but this is another level of complexity that should be made with full understanding of the maintenance involved.
Finally, PROGEDE developed relationships with many external parties including the Direction des Travaux Géographiques et Cartographiques (DTGC), CSE, CNRF, and ISE, as well as Universities of Freiburg, Munich, and Gembloux. Relationships with all these organizations, as well as with CIRAD the French agronomic research institute and the US Forest Service, should be supported by WN.