2. Variables typically assessed in National Forest Assessments

2.1 Types of variables

Observations produce values for variables and are made on measurable objects. The term attribute is frequently used in the same sense as variable, where the term ¿attribute¿ does refer more to the characteristic of the object, and ¿variable¿ has a statistical connotation defining the characteristics as random variables that take on observable values.

For the purpose of NFAs variables may be classified according to different criteria, such as:
  • Classes of variables in a statistical sense, depending on the scale on which they are observed: ¿Measurements¿ may yield a metric value (for metric variables like distance, diameter, height) or a classification into one out of a set of two or more categories (categorical variables like species, forest type, soil type).
  • Distinction between directly observed and derived variables: Some variables are directly measured/observed such as dbh or tree species, and some variables are derived/modeled, like volume and biomass, and most observations of change.
  • Distinction between status and change variables: the majority of measurements give a status value for a given attribute. Directly measurable change attributes are few. Typical examples are increment borings, where the change (increment) of dbh over a certain period can directly be measured, and measurement of length of terminal shoots of coniferous trees.
Number and range of attributes covered in NFAs is wide. Traditionally, biophysical variables are observed in NFAs, but there is a tendency to complement the data on the biophysical resource with stakeholder oriented data on forest use. Also, extension of the scope and goal of forest inventories toward information provision for criteria and indicator (C&I) processes and toward multi-resource inventories/landscape inventories leads to an extension of the types and number of variables included. The number of attributes observed on each plot can be as high as over 250 and is usually not less than about 100, covering concept areas such as ¿land use, forest area, forest type area¿, ¿growing stock¿, ¿carbon balance¿, ¿wood production¿, ¿non-wood forest products¿, ¿biological diversity in production forests¿, ¿soil erosion¿, ¿water conservation in forests¿.

2.2 Data sources typically employed

As NFAs are complex and usually expensive undertakings, an efficient use of various sources of information needs to be made. Field work (sample based observations, be it on field plots or by interviews) and analysis of remote sensing imagery are the central sources of up-to-date data and information. For more details see Overview of options

A multitude of further sources of information are used in the planning, implementation and analysis of an NFA, like maps, previous inventory reports and documents, research studies, expert knowledge, etc.; but rarely are those sources used to retrieve data for analysis.

Photographs of plots and surrounding of the plot are an interesting possibility to document the current situation in a comprehensive manner and allow to qualitatively describing changes at a later point in time. However, this is not frequently applied. It requires an efficient data and image management, and standard analysis procedures are not available.

2.3 Which variables to include in a particular NFA?

The decision which variables to observe in a particular NFA is a strategic one: obviously those variables need to be included that are required to generate the target information. It is a fact that in many forest inventories and NFAs more variables are observed than are eventually analyzed and reported. While this underlines the ¿general documentation character¿ of all forest inventories, and while it can be helpful for unknown posterior information demands, it can also be a waste of resources. (For more information see What information is needed?).

It is advisable to include mainly those variables for which a direct use is known, and on which the inventory will be reporting directly (such as for the variable ¿forest type¿) or which are used as an input for models (such as ¿tree height¿ on which NFAs usually do not report but which is used as input for various derived variables like ¿volume/biomass/carbon¿, ¿forest structure¿ etc.).

In the ideal case, for all variables the analysis methods should be known a priori. By simulating the analysis before starting the data collection, other variables may be identified that should also be included, and possibly other variables are found less useful and are taken out.

It must also be taken into account, that staff (field crews and image interpreters) must be able to make the measurements/observations with reasonable efforts, and that it is also feasible for them to acquire the respective knowledge in training activities with reasonable efforts. Field observation of specific soil and site attributes, for example, will probably be limited to some few variables - unless a soil expert accompanies the field crew. The same holds for more comprehensive inventories of lower plants, of specific non-wood forest products, of wildlife. Also, to expect for inventories in moist tropical forests that all field teams make it to identify all tree species directly in the field, is probably over-ambitious- unless an expert dendrologist is with the field team - which is likely to make the exercise much more expensive.

2.4 Defining the variables: the measurement protocol

In order to guarantee overall interpretable results, variables need to be defined in a comprehensive manner: the inventory staff dealing with the measurements and the data must have the chance to understand what the variables are about.

These definitions are usually written in the inventory manual. Many such manuals exist worldwide, and for a new inventory project one may use those as example or template to develop an own manual.

All variables need to be technically defined in an unambiguous manner, and also the measurement procedure must be defined in detail. A definition without an indication how to carry out the measurement is incomplete, and may lead - without possibly noting it - to inconsistent results. Typical examples are the variables dbh and forest. For the variable ¿dbh¿, a complete definition comprises indications

  • in which height the dbh is to be measured (1.3m),
  • how to proceed in special cases that 1.3m is an impossible height to measure (this part is usually given as a set of drawings),
  • which measurement device to use and what to observe while using it,
  • what measurement unit is to be taken and to what accuracy (e.g. centimeter, to the first decimal).
For the variable ¿forest¿, usually quantitative (minimum crown cover, minimum width, minimum height, ¿) and qualitative (definition of tree, vegetation types, ¿) definition criteria are given. Only in very few cases, however, is there a definition of a measurement procedure responding to the question: how to measure ¿crown cover¿ or ¿stand width¿ in a concrete case.

For categorical variables, each occurring category (class) needs to be considered, either as a separate category, or aggregated with others into one class. For some variables it is recommendable to have the class ¿others¿ for unforeseen cases. The description of the classes must be such that the field crews can make a rapid and clear decision. And the classes should be meaningful for the analysis.