Improving reliability
Improving reliability should be interpreted as increasing the representativeness of sampling. When representativeness increases, data calculated (estimated statistically) for a territorial unit will become less burdened with uncertainty and their confidence interval narrows down. That also brings with it increased accuracy in projecting forest inventory data to territorial units which are smaller than the country as a whole.
Considering that sampling for forest inventory purposes has strong connections with the surveys conducted earlier by the OSG, it is practical to recall a few facts from the past. The program of Observing Tree Stand Growth, which was a predecessor of forest resource assessment, was based on sampling designed to provide reliable data sets at nationwide, (former forest) directorate and potentially at county level. The option to conduct stratified sampling also emerged at that time, since forest cover in individual counties shows wide variations. With that so, one either adopts a strategy of uniform sampling, which will give rise to higher uncertainties of estimates concerning areas where the forest rate is larger, or sampling is adjusted to expected reliability values. If resource management is also considered and the density of sampling is not increased across the whole country, the latter will entail taking samples by applying a grid of higher density in areas with lower forest rates, such as Békés County. When the OSG grid was established, it was decided that a uniform sampling grid should be used and the same decision has remained up to the present day. Yet, the opportunity to increase the density of sampling in these areas and to improve reliability by doing so exists, it is only a question of demand.
Country-wide representativeness is easiest to achieve by increasing the number of samples taken. That is possible in one of two ways. On the one hand, by using a sampling grid of greater density and on the other hand by adding data from other sources to forest resource assessment. To study the details of the latter, see the Chapter on "Forest inventory with multiple sources".
Increasing the density of the sampling grid of the statistical forest inventory (from the current 4 × 4 km mesh) was an automatic consequence at the time the first five-year cycle of the assessment ended. We had to decide between two alternatives: the advantages of returning immediately and increasing the density of the sampling grid to 2.8 × 2.8 km. As forest inventory data will (hopefully) be used in the widest possible circle and at the lowest possible territorial breakdown, improving representativeness was the major factor that motivated the decision.
As a result, the forest inventory to be completed by the end of the second five-year cycle will cover 10 years with representativeness at 200 ha/sampling point (a 2.8 × 2.8 km grid) (the former OSG sampling grid). With that grid density given, data about smaller territorial units, such as forest micro regions or forest management districts will become much more reliable. The sampling launched as part of the second five-year cycle in 2015 was started with the aforementioned considerations in mind. Due to this enlargement, sampling will return to each sampling point once every ten years, which is an acceptable interval given the nature of forests and even allows clearer declaration of changes to certain variables (such as diameter, height, canopy closure, degree of water management). Another advantage lies in the availability of reliable data sets from approximately 10-11 thousand sampling points, and sampling may be repeated in the future. The time series produced that way will also support the detection of changes.
Due to the above, an environment has been created simultaneously that allows the retrieval of data updated for 5 year periods using what are known as moving averages.