1St Order Titles: Arial 16Pt Bold, First Letter of All Words Caps (Heading 1)
Report to Climate Air Action Corporation
December 2004
Tim Pearson, Sandra Brown and David Shoch
Submitted by:
Timothy Pearson
Email:
Ecosystem Services Unit
1621 N. Kent St, Suite 1200
Arlington, VA 22209
Background
Methodology Assessment
Sampling
Measurement
Number of trees
Tree height
Tree circumference
Allometric equations
Additionality, Permanence, Leakage and the Baseline
Additionality
Permanence
Leakage
Baseline
Assessment of Milan Options
Assessment of Baseline Carbon
Kyoto Definitions
Minimum Area
Minimum Height
Canopy Cover
tCER / lCER
ISSUES/Recommendations
References
Appendix 1: Creating Biomass Regression Equations
i. Method-1: Developing biomass equation
ii. Method-II; Mean tree biomass estimate
References
Appendix 2: Calculating sample size
Background
Between 2nd and 5th December Tim Pearson with Eduardo Locatelli visited groves and spoke with quantifiers, coordinators and group members. The visited groves spanned a wide range of actual and potential biomass from sites with few to no live trees to sites with trees of between 2 and 4 meters tall. Sites included trees planted in single rows, densely planted seedlings and sparsely planted agroforestry groves. Measurements methods were witnessed and discussed with each of the levels of operators.
In this brief report we will discuss the methods used in the field in Tanzania and the methods for analysis of the collected data. Also discussed will be the baseline approach, leakage from the project, additionality and permanence of the project, appropriate Kyoto forest definitions and certified emissions reductions. Finally recommendations will be given based on our observations in Tanzania.
Methodology Assessment
Sampling
Sampling is necessary because it is not realistic to measure each and every tree. As every tree is not measured any measurements that are taken are merely estimates of what the actual result would be if every tree had been measured. It is necessary to know that the estimate is sufficiently strong that we can have confidence in it. Two concepts are important here. Accuracy is how close to the actual value your sample measurements are. Precision is how well a value is defined.
A popular analogy is a bull’s eye on a target. In this analogy how tightly the arrows are grouped is the precision, how close they are to the center is the accuracy. Below in (A), the points are close to the center and are therefore accurate but they are widely spaced and therefore are imprecise. In (B), the points are closely grouped and therefore are precise but are far from the center and so are inaccurate. Finally, in (C), the points are close to the center and tightly grouped and are both accurate and precise.
Given the shear number of trees that have been planted and will likely continue to be planted over the next 10-15 years or so, a system is needed to sample the plantings to achieve precise and accurate numbers (case c) at the least cost as possible.
Typically we would sample for carbon in forest-based projects using permanent plots containing a number of trees because this method diminishes the variance and facilitates the estimation of changes in carbon stocks throughout the project life. However, for this project we are not convinced that plots would be a useful approach to take. The existence of groves which consist of single rows of trees, or widely spaced agroforest trees means that it is likely that tree-level sampling design is more appropriate.
The methods for sampling for biomass were probably the largest source of error that we determined during our time in the field in Tanzania. Current methods are for quantifiers to randomly select ten representative trees to measure. Select and randomly are two words that should not be used together, unless it also involves a computer doing the selection. Any human selection is subject to bias, which, conscious or unconscious will govern the results and lead to results that are in not representative of the site. I heard quantifiers say that they choose some small trees, some intermediate trees and some large trees roughly taken from all areas of the grove. There is no way that this can record the natural range of sizes that exist.
If a computer is not used to select trees randomly then the alternative is a systematic selection. From a given starting point every Xth tree, in a row for example, would be measured. It would be important to vary either the starting point or the number represented by X so that the identical trees are not being measured every time.
The normal procedure for determining sample size would be to choose the age of plantings at which you first want to report for carbon. You would then take some preliminary measurements in groves of this age to determine the mean carbon stock, standard deviation and variance. From these data using the appropriate equations (see Appendix 2) the correct number of trees would be determined to achieve the desired level of precision.
A problem with the tree level sampling is that variance between trees could be high. If you are measuring plots, each sample point is a combination of several trees, so some of the variation in tree sizes is absorbed. As a consequence of the high variance the required number of trees to achieve high precision is likely to be high.
An important factor at this stage is the reporting level. It is only the quantity of carbon that is sold as certified emissions reductions that has to be known with high precision. If the sampling were designed so that at the project level it was possible to know that in a given five year period there was a sequestration of, for example, 130,000 tons of CO2 equivalent plus or minus 13,000 tons, then this could be sold (minus perhaps 25 or 30 % for insurance purposes) with confidence. Results would be available at the node, group or even grove level but with a great deal less precision, which would not matter as long as sales of carbon offsets were not made at this level.
It is possible to design a sampling scheme that would achieve this purpose with stratification at the group level and/or the age cohort level and/or the species planted level. This would require an immense amount of set-up work to determine the variance, standard deviation and mean carbon stock in each of the strata. It would also require a lot of administration to tell quantifiers that in this particular grove 18 trees must be measured while in this grove 6 trees and in this grove 43 trees. This level of complexity is bound to introduce errors.
Instead we advocate a sampling scheme that would allow the application of simple rules for the quantifiers and that would be measuring more than the required number to achieve the desired precision at the project level.
Below is presented a potential sampling methodology:
To decrease variation we would suggest a sampling scheme with consideration of cohort age and species. If sampling is at an adequately high level it will permit a post-stratification by node or group.
Cohort age (if possible) should be divided into each planting year.
Species, at least at the measurement stage, could be divided into Eucalyptus and non-Eucalyptus species. Note that banana should not be measured for carbon purposes.
As the equations for converting tree measurements into biomass carbon require diameter at breast height, there is no value in measuring circumference until the tree has a recordable dbh. Height can be measured for any age of tree but to save effort while using equations based solely on dbh there is no need to measure height.
Step 1: Count all trees in each of the classes --age and species group--that have a measurable diameter at breast height.
We suggest that in each grove 20 trees be measured for biomass carbon for each species group (Eucalyptus vs non-Eucalyptus) by age cohort class. There is nothing statistical about the number 20, it is just a reasonable compromise between reasonable effort and the need for data. If there are less than 20 trees in a given class (age-species group) then measure as many as there are. Currently this would give a maximum of 4 years by 2 groups = 80 trees. However, only rarely will all combinations of age and species group be present and even more rarely will there be 20 living example of each class. Measuring circumference is relatively simple and the effort involved in taking even 100 measurements is not excessive, especially as height measurements are not needed for carbon estimation.
If a reduction in effort were desired, it would not be necessary to measure circumference every year as reporting would be on a 5-year cycle. In the intermediate years just the number of trees could be counted.
It is very important that the circumference of each measured tree is recorded in the palm along with the species and age, and not the average for a class as this will introduce large errors.
Step 2: The method of tree selection should be to start in one “corner” of the grove and work along the rows. If there are less than 40 trees in a given class, measure the first 20, if there are 40-100, measure every second tree, if there are more than 100, measure every fifth tree. If possible, each year the measurements should begin in a different corner of the grove.
If the trees are in lines, then start at one end of the line and note which end this is in the field data recording devices.
When it comes to analysis we (at Winrock) could help develop a spreadsheet, that should contain the following steps:
1.Determine strata – e.g. – group, species, age
2.Sum the number of trees within each strata
3.List all the dbh measurements in each strata for measured trees and calculate the biomass of each tree
4.Take the mean biomass per tree for each strata and multiply by the sum of trees in each strata (from 2)
5.Sum the strata and multiply by 1/2000 (to convert from kg to tons of carbon) then by 44/12 (to convert to tons of CO2eq).
If data are required at the grove level then (with the understanding that it is less precise, unless all trees in a grove are measured (e.g. in groves with <20 trees per age/species class), in which case carbon stock is calculated from the actual (i.e. true) mean dbh, rather than the estimated mean) the number of trees within each of the relevant strata in the grove can be multiplied by the mean for that strata. Summing the strata gives the grove level carbon stock.
Measurement
Number of trees
Counting the number of trees by definition is an accurate method of determining totals. However, a systematic method is required to retain the accuracy of the count especially when large numbers of trees area present. Working along rows and tallying in groups of five might be a good method.
Tree height
The method used for tree height is simplistic but effective. Errors can arise with non-expert practitioners using clinometers. Comparing dimensions between someone of a known size and a tree using a transparent rule is a method that should minimize errors. It is crucial, however, that the height of the person is accurately determined as all errors in this measurement will be magnified. Care should be taken to make these visual estimations from a vantage point well set back from the tree to minimize errors associated with optical angles. However, as noted above, measurements of height is not needed for biomass carbon estimates (for the proposed equations).
Tree circumference
Tree circumference is measured using a plasticized tailor’s tape measure. This is a good measuring implement, as it will not stretch which is a potential source for errors. Current methodology proposes that trees of less than 2.5 m in height the tree should be measured at half this height. This is incorrect for application to equations based on dbh. The varying measurement height is also a potential source for confusion for quantifiers. We witnessed quantifiers measuring circumference at points of the tree seemingly unconnected with half the tree height or breast height.
As long as breast height is the dimension required in the allometric equation, diameter should be measured at this height and this height only. See exceptions below (few of which apply to trees growing in TIST groves in Tanzania).
Circumference should not be measured unless the tree has attained a height greater than breast height (1.37 m) and also the minimum diameter required for the regression equations at breast height. For the general equation presented below, the minimum diameter is 1.1 cm. Trees that do not have a diameter of more than 1.1 cm diameter (3.46 cm circumference) should not be measured.
If on a given tree there are multiple stems at breast height, then all the stems should be measured as all represent sequestered biomass. Currently quantifiers only measure the largest stem on any tree.
Examples of correct circumference measurement
Allometric equations
In the monitoring plan it is proposed that the moist equation of Brown (1997; updated) be used. This will result in a significant overestimation of biomass. The moist equation is designed for moist tropical forests where trees can attain heights of 40 m or more and a diameter of brest hegith of up to 200 cm and thus have very high biomass. In particular, tree heights and diameters are low in dry forests typified by the TIST sites in the Dodoma Region of Tanzania.
Two alternative solutions exist:
- Create a new equation – creating an equation, though a lot of work, is probably the most reliable way of determining biomass as site-specific considerations are included. Methods for creating biomass equations are included in the appendix of this document.
- Use a general equation for dry regions – analyzing data for dry forests in India, Mexico and Thailand, I created the following equation:
Where x is the dbh in cm. The r2 value is 97 % and the range of dbh that can be applied is 1.1 – 63.4 cm.
While in Tanzania we took measurements of the height and diameter of trees, from this we estimated the volume to which we applied an average wood density for Africa (Brown 1997) to gain biomass. To extrapolate from bole biomass to total biomass we used a factor of 1.2. While this is a method full of approximations, the results fitted fairly well to the general regression curve calculated for dry forests.
An additional complication is provided by Eucalyptus. The biomass allocation of Eucalyptus is atypical and so general equations do not well represent the relationship between diameter at breast height and biomass. No equation could be found in our initial search for the Eucalyptus species that are planted in Tanzania. Here we present an equation calculated from Eucalyptus trees growing in Australia in a semi-arid area (Ashton 1976):
Where M is mass in (kg) and G is the circumference at breast height (cm).
Additionality, Permanence, Leakage and the Baseline
Additionality
Additionality can be reduced to the basic question – would this project have happened in the absence of carbon financing?
After spending time in the field at the sites in Tanzania, I would confidently say that trees would not have been planted without the promises to the group members of payment for the planting. The farmers would instead have continued growing crops regardless of the benefits of shade or soil quality.
I have one caveat. One group was headed by a priest, who was a forester by training. He was planting trees prior to signing up for TIST, primarily it seems to provide firewood. In this case, the baseline comes into consideration and it should be shown that TIST enhanced the area being planted.
Permanence
The group members will retain trees as long as they are receiving the financial benefit of doing so. If the flow of money stopped it is likely that all but a few trees would be cut.
However, the evolving decisions of the CDM afforestation/reforestation working group have eradicated the problem of permanence. See discussion on tCERs and lCERs.
Leakage
For this project leakage should only be positive. Firewood is being grown that otherwise would have been cut from existing non-project trees. One question does remain regarding leakage –
If trees are being grown on farmland that would otherwise be used for growing crops then there is a decrease in area of farmland but no decrease in the demand for food. I can imagine two responses to this, and which response is prevalently chosen will have implications:
1. Farmers will cut areas of native bushes to create additional areas for growing crops – this is a negative leakage and would have to be accounted for.
2. The conservation farming programs encouraged by TIST make farming more efficient so that for a given area of farmland more crops can be grown compensating for the land lost to tree plantations.
Minimal leakage will occur through project transport but this is accounted for and should not exceed the positive leakage provided by growing firewood.
Baseline
Assessment of Milan Options
1. Existing or historical, as applicable, changes in carbon stocks in the carbon pools within the project boundary
2. Changes in carbon stocks in the carbon pools within the project boundary from a land use that represents an economically attractive course of action, taking into account barriers to investment
3. Changes in carbon stocks in the pools within the project boundary from the most likely land use at the time the project starts
As long as the project boundary is defined as the land farmed by the group members in the TIST program, options 1 and 3 are identical. Current, historical and likely future land use is (low biomass) farming. Option 2 is not applicable as there is currently no likely land use that would be more economically attractive to the project participants than continuing farming.