Forest Inventory and Sampling Design
|Society of American Foresters||International Society of Tropical Foresters|
Forest Inventory and Sampling Design
Inventories are used to make management or policy decisions. For that purpose the most common variable used is the average value of whatever is being measured. Whether this average value is representative and whether it is a reliable estimate of the true value of what has been measured, depends very much on the sampling design used. Complex or specific situations require complex or specific sampling designs. The reader is referred to the specialist literature for these designs. In this section, only some of the basic aspects of forest inventory and sampling design are dealt with.
To assure that the information is representative, it is important to try to include as much as possible of a population’s variety within the sample. This can be achieved by choosing populations as homogeneous as possible, by making sure that the sample units are well distributed over the whole population, by increasing the size of the sample or by adjusting size and shape of the sampling unit.
Optimum sizes and shapes need to be determined for each different type of forests, while statistical calculations help determine the optimum sample size (Wenger 1984, Prodan et al 1997). In even-aged single species forests that are classified according to their site productivity, few sample units of small size in each productivity class may be very representative. In the natural wet tropical forests local diversity in structure and floristic composition may require large sample units, while high diversity between units may require a large number of sample units in order to capture the whole variety of the forest.
The reliability of the data obtained is important to be able to estimate the probability that the averages or totals estimated for the forest are different from the real values in a particular sample of that forest. A company that for example wants to invest in a sawmill will want to know the risk of not being able to harvest the quantity of harvestable timber estimated for its forest. This risk can be expressed in terms of the sampling error, a statistical measure of the reliability of the data. If this error is high, the risk is high, and the company may not want to invest. Thus, it is important to design forest inventories in such a way that the risk, i.e. the sampling error, is reduced to acceptable levels. One can, for example, decrease the variability between sample units by increasing their individual size. This decreases the sampling error. Increasing the number of units has similar effects.
Distribution of Sample Units
Different methods exist to distribute the sample over an area. Statistically, the most reliable distribution is obtained through random selection of the sampling units, where each unit of a population has equal probability to be selected. While such a distribution statistically is the more correct one, it has as a disadvantage that sample units may be grouped in certain areas, with other parts of the forest not being sampled. This may result in statistically reliable but little representative data. This form of distribution also reduces the use of the information for area calculations or for the analysis of forest types or species and tree distribution patterns. The random distribution is useful in some types of research and for exploratory inventories over very large areas, where sample units exists of clusters or groups of plots. In that case, the clusters can be located randomly within the different forest areas (stratification). Random selection, however, is little practical for the distribution of sample plots in most forest inventories.
The most common form of distribution of sample plots is that obtained through systematic sampling: of the sampling frame of N units, the first sample unit to be measured is selected randomly, after which every xth sample unit is selected in such a way that N/x is approximately n, the sample size. As a result one obtains a grid of sample units at equal distances between units within sample lines, and equal distances between the lines. While statistically this form of distribution gives some problems, if applied correctly, it usually results in a better distribution of the sample units and allows better planning and easier access to each unit.
Another form of distribution is that through expert selection: the forester or researcher selects its sample sites as “representative” for the forest area to be sampled. This is acceptable for certain types of research, but in general will give biased information in the case of forest inventories: most people tend to select well-stocked areas as representative, leaving areas with clearings or affected by other natural phenomena out of the sample, thus creating overestimates for the averages of most stand variables (e.g. volume per hectare data).
Permanent Sample Plots
Forest inventories in general are designed to describe a static situation: those aspects of the forest that are of interest in a particular point in time. The number of trees per hectare, volume per hectare, basal area per hectare, relative importance of a species, biodiversity indices are measured at a specific point in time. In order to get an idea of changes over time, be these natural or caused by human activities, it will be necessary to repeat these measurements over time. These should preferably be done within the same (permanent) sampling units. These plots need to be carefully laid out, well-marked in the field and on maps, and should be established in areas where it is likely that they will not be removed by natural or human activities.
See also: Individual, species, or global diversity, Forest Management Plans, Harvest Scheduling/Allowable Cut, Natural Tropical Forests Management, Forest Values, Environmental Impacts and Assessments, Food and Agriculture Organization Statistics, Forest Inventory and Analysis: USA, Measurement of Trees, Forests and Forest Products, National Resources Inventory: USA, Measurement of Wildlife and Other Resources, Photogrammetry and Remote Sensing, Statistical Data Sources: Latin American Forests
Orozco, L., Brumér, C. (eds). Inventarios forestales para bosques latifoliados en América Central. Serie Técnica Manual Técnico no 50. Centro Agronómico Tropical para la Investigación y Educación (CATIE), 2002. Turrialba, Costa Rica.
Prodan, M., Peters, R., Cox, F., Real. Pedro. Mensura Forestal. Serie investigación y educación en desarrollo sostenible no.1. Deutsche Gesellschaft für Technische Zusammenarbeit (GTZ) GmbH, Instituto Interamericano de Cooperación para la Agricultura (IICA), 1997. San José, Costa Rica. 586 p.
Schreuder, H.T., Gregoire, T.G., Wood, G.B. Sampling methods for multiresource forest inventory. John Wiley and Sons Inc. 1993. New York. 446 p.
Wenger, K.F. (ed). Forestry Handbook. Second edition. John Wiley and Sons, Inc. 1984. New York. 1335 p.
Posted: 26 April 2007
Updated 23 August 2007