The main purpose of the inventory is to:
(a) estimate wood stocks under the different types or classes of soil use; and
(b) estimate the productivity of these stocks.
It is best to divide the activities into two stages: first estimating stocks, then their productivity.
Stratification of soil cover and land use with potential for fuelwood production is the first task. This can be done with the help of remote sensing (aerial photographs, spectral filters on orbital platforms or aircraft, videographs, etc). The purpose here is to discriminate as many classes of land cover as possible, and thus reduce the internal variability within each class in order to reduce the sampling effort and error. (Cf. stratified sampling in Chapter 3). The strata must be recognizable at all times from direct observation on the ground, otherwise the classification will be unusable by persons not trained in image interpretation – as will be the case with the majority of direct and indirect users.
The interpretation of remote sensing images also provides a tool for the plotting of vegetation maps or land use maps or charts, from which the areas occupied by each stratum can be calculated.
Once the strata have been defined and have been seen to be recognizable on the ground, the pilot sampling can begin, with not fewer than 15 plots per stratum. In strata of limited importance  because of small area or low intensity of use  it is best to carry out the pilot sampling on five plots per stratum. This serves to estimate the variability displayed by the major variables within each stratum when estimating woodfuel stocks. Usually, the variable best correlating with stocks is the cylindrical volume (basal area x height).
The shape and size of the plots do not affect the estimation of mean values but do influence the estimation of variability of basal area and cylindrical volume, depending on the pattern of spatial distribution of trees and shrubs. If the population is heterogeneous or in clumps, small plots will exhibit a high degree of variability. Generally, the most efficient plots by reason of simplicity of determination and measurement are rectangular in shape, with a width of not less than 20 m and a length of between 20 and 500 m (400 to 10 000 m^{2}), depending on the spatial heterogeneity of the vegetation. It is useful to divide these into subplots of 20 m length to analyse the influence of plot size on variance estimation. Where there are low density trees and good visibility, circular plots make for faster measuring.
Plots must be selected at random throughout the area occupied by each stratum. This is achieved by randomly plotting them on a land cover map and transposing them on the ground through satellite geopositioning.
The minimum measurements needed for each plot (using one field record sheet per plot, see Annex V) are:
• species, shape, quality of bole, health^{10};
• diameters at 0.5 and 1.3 m height (beginning with 3 cm diameter or 9.5 cm girth, without decimal);
• total height and height of bole (in metres, without decimal).
The field record data are transferred to an Excel worksheet to input the measurements and calculate (see Annex VII):
• basal area (G) at 0.5 and 1.3 m height (in m^{2} to four figures);
• dominant height (Hd), which is the average of the largest trees measured (the upper decile of each height stratum);
• cylindrical volume (in m^{3} to four figures);
• number of trees (or stems) per hectare.
Distributions per diameter class and summations per species and plot can be computed for all these variables.
A dendrometric table is compiled from the final values (or summations) of these variables per plot, with each plot occupying one line. Excel tabulation makes for easy calculation of the statistical parameters, using the "functions” application. This permits the calculation of the mean, standard deviation, standard error and confidence interval for each variable. If the confidence interval obtained for cylindrical volume is greater than envisaged as acceptable, further samples will need to be taken – i.e. more plots will have to be measured.
It is usually recommended that the probable error should not exceed 20% for the cylindrical volume. The number of plots needing to be measured in order to reach this level of error can be calculated from the formula (see Chapter 3 and Annex III):
n_{o} = (cv^{2} . t^{2}_{, v})/ e^{2}
Once the pilot sampling and the analysis of its results are complete, we can proceed to the final sampling, increasing the number of sample units to reach the number calculated with the above formula. If this number is too large in relation to resources and time available, the acceptable error will have to be reconsidered (as regards both magnitude and level of confidence of the estimation). Accepting a higher level of error or a lower level of confidence significantly reduces the sampling effort. Generally, the number of plots needed for a 20% error with a 95% confidence is in the 20 to 30 range. The data from the additional plots are added to the relevant stratum table and the statistical parameters recalculated to define the mean and final level of error of each variable.
The next step is to construct weight and volume equations for the more important species in each stratum. These will serve to deduce shape factor values which, multiplied by cylindrical volume or basal area, will give the real volume of a tree or cluster of trees. With secondary species, it is advisable to group these and apply the equations of the species most closely related or similar in shape.
In order to construct these equations, it is necessary to measure at least 30 trees of five diameter classes embracing the entire population (6 trees per diameter class), to fell them and crosscut them into regular sections, keeping to local wood production practices (as regards tools, length, felling and lopping procedures). Stems suitable for sawing or use as poles, stakes, etc., are not cut for fuelwood, but are measured and weighed separately. The wood is stacked in keeping with local practice to form “cords”, steres, “metres of wood”, etc., then measured and weighed.
The data for each individual tree are recorded on a separate sheet for each species, indicating:
 tree number;
 diameter at 0.5
and 1.3 m;
 total height and height of bole;
 average diameter, length and weight of logs, poles and
stakes;
 dimension and apparent volume of stacked wood;
 green weight of stacked wood.
From each tree, or at least from two trees in each diameter class, samples are taken to determine specific weight and moisture. These are obtained as transversal slices or disks, 2 or 3 cm thick, cut at random from the stacked wood. They are marked with a crayon, wax pencil or indelible marker, wrapped in a twoply polythene bag, carefully sealed and taken to the laboratory.
To calculate specific weights and moisture see Chapter 2.
With the data obtained from 30 trees, a table of products, weights and volumes is constructed for each species, including:
 diameters (at 0.5 and 1.3 m);

heights;
 basal areas (at 0.5 and 1.3 m);
 cylindrical volume (at 0.5 and 1.3 m);
 green weight of wood;
 stacked
volume;
 moisture content;
 dry
weight of wood;
 volume of boles (whether suitable for
sawing or for poles, stakes, etc.);
 dry weight of
boles;
 total volume;
 total dry
weight.
With this table and Excel functions, regressions between basal area or cylindrical volume can be rapidly calculated as independent variables and as can any other dependent variable of interest, e.g. dry weight of wood, volume of boles, total volume and total dry weight. Graphs can also be plotted and are very useful for understanding relationships between variables. It is important to recall that these relationships are mathematical functions; they indicate that there is a greater or lesser correlation associating two or more variables, but they do not necessarily imply causal relationships. It is generally accepted that, if the coefficient of determination (r^{2}) is greater than 0.8 (because r > 0.9), the function is appropriate and can be used with confidence. More refined analyses of regression quality can be done using the STATISTICA program.
The final outcome will be a table of functions of weight, volume and products for the principal species and groups of secondary species in each of the strata. By incorporating these functions into the respective dendrometric tables, it is possible to calculate a selected variable for each stratum; e.g. stocks of firewood expressed in different forms and units, stocks of roundwood for other uses, stocks of trees producing nonwood forest products, their distribution by species or diameter classes. These unit values are averages per hectare, and for each resource whose area is known it is possible to calculate total values, which is very important for formulating management plans and recommendations.
Summarizing, completion of the first phase of the inventory will have produced a database and the calculation tools needed for estimating with fair approximation the stocks of wood for energy (and other purposes) in areas of priority interest. If the information on land cover and use is periodically updated, this database will help track stock dynamics and make projections for the future.
In the second phase, the principal objective is to make a correct estimate of woodfuel resource productivity (see subchapter 2.2.2). Here, the chief difficulty lies in estimating the age of trees or forest stands. If there are no reliable historical data (which are usually only available for plantations), some form of correlation needs to be found between age, number of growth rings and tree diameter for each species and in each stratum.
Once these relationships have been established, the same data from the sample plots can be used to estimate the age of the respective populations, and dividing stocks by ages will give the annual rate of growth or mean increment for the whole, or disaggregated by species, diameter class or any other grouping of interest. It is also possible to associate these values with, for example, local conditions (soil, climate), present and previous use and management practices.
In order to facilitate this, it is recommended that sample disks be taken at 0.5 m bole height for at least 30 trees of the principal species, something can be done at the same time as the felling and measuring for the weight and volume equations. The duly identified samples (tree number, species, plot, stratum) are sent to a dendrochronological laboratory, where they are sanded and the growth rings counted. The operation is very specialized and needs to be done by a properly qualified and experienced operator.
The information obtained in this way is of immense value. The possession of reliable data on diameter/age ratios permits appreciable savings in time and effort, if we consider the lengthy time and considerable effort required to assess growth rates by means of permanent plots.
More complete information on the subjects discussed above is available in FAO Forestry Papers, Nos. 22/1, 22/2 and 51/1. A detailed description of field procedures is given in the “Plano de Manejo Florestal para a Região do Seridó”.