Chapter 5 Carbon Assessment and Monitoring by Local Communities
5.4 Steps for carbon assessment in managed forests
5.4.2 Pilot survey to calculate variance
For estimation of variance of carbon stock of the main carbon pool in forests i.e. trees, stand basal area9 was determined from 15 randomly laid out sample plots. The shape of the plots was circular with different sizes depending on the vegetation of the forests as follows:
• For the dry miombo woodlands, all trees greater than 1 cm diameter at breast height (dbh) were measured all over a plot of 5.6 m radius, i.e. with a total area of 100 m2.
• For the montane and lowland forests, concentric plots of varying radius were used depending on the sizes of trees. Measurements of dbh of trees/shrubs were done in a plot of 5.6 m radius for all trees of greater than 5 cm dbh, while trees of less than 5 cm but greater or equal to 1 cm dbh were measured in a plot of 2 m radius.
Concentric plots were used for montane and lowland forest since these were observed to have many small trees and few large size trees. It is a standard procedure to adopt different plot sizes depending on the vegetation type. In this case, this is because woodland forests were expected to have few small diameter trees of less than 5 cm dbh and vice versa for montane and lowland forests.
It was difficult to enter data on each tree immediately into the handheld computer in the field as observed during the training. Therefore a paper record was made by the villagers and data entered to the computer later at home by the staff from the supporting organizations who were among the trainees. This increased the field efficiency and the trainees were able to collect all the pilot data needed after only one day in each forest. Existing checklists for the
9 Stand basal area is the cross-sectional area of all trees at breast height (1.3m) per hectare of a forest. It is a summary of the number and size of trees in a stand and therefore relates much to stand volume and biomass.
areas were used to get the botanical names for species identified by the community teams.
One day was spent for matching local names to botanical names, assigning species identification codes and punching the data to the computer for each forest. This work was done by the staff from the supporting organizations.
With data from the 15 plots, it was possible to calculate standard deviation and average basal area per hectare. Then the number of sampling units (n) required to attain a desired precision at sampling error (E) of 10% is given by:
Where: CV = Coefficient of variation = standard deviation/mean
t = the value of t obtained from the students’ t distribution table at n-1 degree of freedom of the pilot study plots at 10% probability.
For LULUCF projects, it is recommended by IPCC to adopt a sampling error of 5% which is expected to give estimates within the precision of ±10% of the mean with 95% confidence (IPCC, 2003). Given the nature of CFM forests with fragmented degraded and intact patches as shown in Sub-Section 5.4.1, the 10% sampling error rather than 5% was used as it considerably reduces the number of plots required as shown in Table 10, but this means compromising the level of precision as shown in Figure 13 by Brown (2002). However, the level of precision required for carbon inventory has direct effect on inventory cost. With 5%
sampling error four times as many plots would be required (Table 10) and that implies more time needed for the inventory, which of course greatly increases the costs. It is foreseen that the market value for CO2 credits from REDD policy might be less than that of other mitigation options such as CERs from CDM projects (Stern, 2007). A compromise may therefore have to be made between the precision desired and the costs of a REDD project including the inventory cost. Therefore, bearing in mind the cost element and ease of handling the plots by local communities, it was decided to adopt the 10% sampling error.
Having decided on the use of 10% sampling error the calculation of the number of plots is done. For consistency of the computations of pilot data from all the sites, a pre-designed database on Microsoft Access program was prepared. This database does these calculations by replacing the default data file stored on the computer. This was simple for the staff of the supporting organization to operate. At the end of this step, the number of plots, n, needed for each forest was obtained and distributed systematically in the respective geo-referenced base
2 2 2
E t n=CV
maps of the forests as shown in Table 10. This was done in the office by the staff of the supporting organizations following instructions from the field methodological guide.
For the KSUATFR, the highway which formed the southern border of the forest with a ground distance of about 3,300 m was taken as a basis for transects layout. A total of 11 transects were established at an interval of 300 m. In order to space out these transects the first transect was laid out at 150 m from the eastern boundary of the forest. The total transect length was estimated to be about 13,350 m. For the 89 plots to be established, the interval between plots was 150 m. The first plot was laid out at half plot interval for better spacing out of the plots along the whole transect length. Figure 11 illustrates this layout. Forward bearing of the transects was 300° while the back bearing was 120°. Similarly, plots for the rest of the forests were established as detailed in Table 10.
Figure 13. Relationship between the number of plots and precision level
Table 10. Number of permanent sample plots for the CFM forests in the studied villages Number of plots