---

Guide to the interpretation of the results

It follows from the method of sampling used in our statistical forest inventory that the results produced for various units of land or for different parameters of the same area (such as hydrology, actual forest association group, etc.) are associated with different reliability values due fundamentally to the varying level of representation of the specific feature in the sample. One may be confronted with that when one performs the confidence interval analysis of even a single table of results (matrix). It is important to decide from both a statistical and a user perspective the level of reliability of data selected for publication. User perspectives dominated during the preparation of the forest inventory website, which is why data with broader confidence intervals are also published. That solution was accepted first of all because the data published in connection with forest resource assessment have no background and intend to close a backlog, which is why the data communicated are considered informative by all means although occasional data uncertainties have always been taken into account. We have also found that aside from a few truly extreme cases where only few measured data existed, confidence intervals started to narrow down as the quantity of data that lend themselves to analysis increased, while estimated values showed no marked change. However, it is extremely important to bear in mind the above during any attempt at interpreting and using the data.

Confidence interval

The bottom and top limits of the confidence interval is displayed at 95% reliability in most cases for any estimates appearing in tables of statistics. That means that the probability of the actual value falling between the top and bottom limits of the confidence interval is 95%.

When "0" is displayed as the bottom limit of a confidence interval for a value in a table of statistics, it means that half of the width of the confidence interval surpasses the calculated value, hence the bottom value is negative, which is in effect displayed as "0".

Certain data with extremely low levels of representativeness have no confidence interval displayed at all to indicate that they have low reliability. A deliberate attempt was made not to apply "cosmetic treatment" to the database when certain data assumed extreme values. Naturally, whenever data error was the reason for such phenomena, such errors have been rectified within permissible limits and it is our intention to do so in the future, yet one must emphasize that the majority of such cases result from the low level of related sampling. Cases of this nature include the number of small trees per hectare in the other coniferous actual forest association type, where the crowds of black cherry regrowth produced extreme data. As the "other coniferous" actual forest association group is represented by a very low number of elements in the complete sample taken, the related effect is also definitive. As the number of elements increase later on, these extreme cases will become less pronounced.

For instance, as regards forest land area, the calculated value of 2,142,000 ha couples with the values 2,091,739 ha and 2,192,261 ha. That means that the probability that the actual value, which does not necessarily equal the calculated value, falls between 2,091,739 ha and 2,192,261 ha is 95%. In this case, the relatively small interval also indicates that our definition of the area can be regarded as sufficiently accurate.

Naturally, that also means that the broader the confidence interval of a data set is, the greater the reservation one should apply in analysing it. Such cases are expected to occur when the reference area unit is reduced (e.g. from country to region or forest region) and when the occurrence of characteristics is rare/less frequent, and hence are represented at low level in the sample (such as rare tree species, rare soil type, etc.).

Obstacles to sampling

Forest land area values include sampling points that are classified as forest land area where sampling failed to occur due to some obstacle. Such hindrances may include, for instance, floods, enclosed military area, etc.

That actually means in territorially based statistics that in case a parameter cannot be sourced from the NFD, than the sampling point in question will be treated as "non-defined" from the point of view of the given parameter. That is to say, for instance, the total value of territorial distribution by Category of naturalness will only yield the total forest land area of the country once "non-defined" values are added in.

A special aspect of this "set of problems" involves statistics, which presents the parameters of sample trees, small trees, stumps and dead wood. For instance, the total wood volume of a tree species group belongs here. This value obviously does not include the wood volumes that could have been measured at the sampling points that were left unrecorded due to an obstacle, which also means that this aggregate value is slightly underestimated at national level. Specific values applicable to a single hectare of the forest land area of the country are also slightly underestimated.

Young stand

This is somewhat similar to the case discussed under the section on sampling points not recorded due to an obstacle, as potential sample trees also exist here, but recording individual sample trees is not possible due to the segmentation of sampling points. The resulting missing "sample tree values", such as wood volume, are also reasonably underestimated.

However, stand level parameters are available, such as Actual forest association, or Canopy closure. Hence, this does not distort statistics, as opposed to the case discussed under sampling points left unrecorded due to an obstacle.

Felling site, understocked, small trees

Parameters concerning tree stand and sample trees are not defined, due to lack of date at these sampling points. They are regarded to be "vacant areas" from a certain respect, to borrow a comparison used in database jargon.

Non-sampling point

This is a special category which is classified as forest land area but where sampling failed to occur as it did not meet the criteria of establishing a sampling point. For instance, the central point is inside a forest land area, but the sampling plot included other areas covering more than 50 m², and not even shifting will improve this condition.

NFB data

Certain data are taken from the NFB. It is our confirmed principle that these data may be adjusted once reliable records and analysis are completed, with particular regard to the case when a sampling plot deviates from the average of the forest land area where it is located. These include management plan identifiers, site data and the age of sample trees.

Other data taken from the Database are also used during statistical processing. For example: primary designation, form of ownership, etc.

NFB statistics

Tables of statistics comparing data provided by Unified Forest Monitoring (UFM) and the NFB are special. These include, for instance, the table on the nation-wide territorial distribution of tree species groups, which compares territorial data calculated during the UFM with territorial data retrieved from the NFB. It must be emphasised that we should try to create common data content/data platforms to support comparisons, provided that it should be evaluated with the remarks made above in mind.

Unique territorial values in the statistics

Due to the above, certain statistics contain columns and/or rows marked non-defined, with total values that keep recurring several times – due to their very nature:

Territorial data

• 311,200 ha: Unstocked area (Understocked forest land area, Felling site, Non-sampling point, Obstacles to sampling, Small trees, Young stand)

• of which 2,400 ha Non-sampling point
• of which 36,400 ha Young Stand
• of which 40,000 ha Obstacles to sampling
  • of which 9,600 ha not subject to management planning or other managed sub-compartment
    • of which 2,800 ha other managed sub-compartment, and 6,800 ha not subject to management planning

• 54,800 ha: not subject to management planning and unstocked area

• 239,200 ha: Not subject to management planning or other managed sub-compartment
  • of which 201,600 ha not subject to management planning