0268-B1
Andrius Kuliešis and Albertas Kasperavičius[1]
The design, technology and contents of the national forest inventory (NFI) were based on investigations that have been ongoing in Lithuania since 1966. The main aim of the NFI is to carry out a complex and reliable monitoring of natural resources in the forests. All in all there are about 5 500 permanent sample plots distributed over the forest land. The permanent sample plots are located in square clusters or tracts. Each cluster consists of four sample plots 500 m2 in size. The distance between sample plots in tracts is 250 m. The optimization and estimation of the efficiency of NFI sampling design and improvement of its parameters are the main tasks of our research.
By taking account of volume variation and time spent in measuring, an optimal 500-600 m2 plot size was ascertained. Optimization of plot size and grouping ensures balance between time consumption for driving and walking and for plot measurement as one of the most important features in efficient sampling design.
Maximal representation of forest diversity was ensured by reduction of the number of plots per tract down to four, increasing the distance between plots within tracts up to 250 m, precise fixing of plot location and its partitioning into sectors in respect of stand, site and other conditions.
For many years the main source of information about forests in Lithuania was stand wise forest inventory, which guaranteed the flow of information for a detailed and effective short term planning, organization of forestry on a certain territory. Many quite important characteristics of forests, such as wood increment, mortality, growth balance and allowable cut, tree damages cannot be reliably ascertained by applying previous methods. Lithuanian forest inventory faced many problems after the appearance of private forest sector with very small holdings. Without essential changes in existing assessment methods and technologies used at present, it is impossible to obtain a sufficiently precise information needed to ascertain changes in the state of forests, to perform strategic planning and assessment at the level of the whole country. These circumstances have led to the organization of the National Forest Inventory (NFI) based on sampling method.
The first experimental forest inventory by sampling method was carried out in 1966 - 1968 in Kazlø Rûda and Prienai forests. In 1969 all state forests of Lithuania were inventoried by a statistical method (Fig. 1) (Kuliešis 1971). In order to devise a method of controlling stand yield in 1976 in the Dubrava forest a continuous forest inventory (CFI) was initiated on a large forest scale (Kuliešis 1989). In 1969 temporary plots of varying size clustered into groups of six were used. In Dubrava CFI permanent and temporary plots (also of varying size) clustered into groups of four were combined. The results of performed studies showed the necessity to change plots of varying size into plots of a fixed size (Kuliešis 1994). These earlier and subsequent pilot inventories have laid the foundation for the Lithuanian NFI.
Fig. 1. The history of Lithuanian NFI by sampling method
The Lithuanian NFI, based on the repeatedly measured permanent sample plots, was started in 1998 (Kuliešis 1999). The aim of the inventory is to carry out a complex and reliable monitoring of natural forest resources for an efficient assessment of the main forest parameters in the country or its regions.
The design of inventory is a systematic cluster sampling with a random start. The estimation of the efficiency of NFI sampling design, optimisation and improvement of its parameters are the main objectives of our research. In order to create an optimal sampling design for Lithuanian forest inventory, to carry out the inventory with minimal labour expenditures under a defined accuracy, it is necessary to optimize the size of sample plots, their construction, joining into groups - tracts.
In order to solve the main tasks of the research it was necessary: to estimate the variance of growing stock in the sample plots of different size, to estimate the effectiveness of the NFI sampling design according to the analysis of time input necessary for field operations, to optimise the size of the permanent sample plot, to estimate the influence of clustering of plots on the representation of forest diversity, bias of the mean and the variance of estimated parameters.
National forest inventory is based on the method of continuous, combined, multi-stage systematic sampling with a random start. Three stages of systematic sampling (Fig. 2), combining repeated inventory on permanent plots with the measurements on temporary plots, overground measurements with remote sensing methods enable to monitor the whole territory of the country.
Fig. 2. Three stages of systematic sampling with a random start
The main aim of the inventory is to assess volume increment of trees, mortality and growth balance by direct methods, to control the dynamics of forest area in the country. Temporary plots are aimed at controlling sampling representativeness of permanent plots, at raising growing stock volume estimation efficiency by applying regression estimation methods.
Inventory is carried out each year on the whole territory of Lithuania (Fig. 3). During the first five years the network of permanent sample plots is consistently thickened. It is planned to establish in the forest and measure every year not less than 1000, and totally 5000 permanent sample plots over a five-year period (Fig. 3).
Fig. 3. Allocation of NFI sampling units - tracts in Lithuanian forests
The inventory cycle is 5 years. During the following five year period every year permanent plots will be remeasured and an additional number of temporary plots will be established. One permanent sample plot represents 400 ha.
The principal sample unit is a permanent plot of fixed radius. Sample plots are allocated in tracts, the edges of which are oriented in north - south, east - west directions (Fig 4 a).
Fig. 4. a) distribution and clustering of NFI sample plots; b) design of the permanent sample plot
In order to raise representation of sampling design by minimizing the probability of occurrence of more than one plot on the same site and taking into account the average size of a stand compartment, its configuration, the length of a tract edge in Lithuanian forests was estimated to be 250 m. The area of the main plot in horizontal projection is 500 m2 (Fig 4 b). In the main plot all trees over 14.0 cm in diameter are measured. In the centre of the plot another 100 m2 size plot is singled out, where all trees over 6.0 cm in diameter are measured. In the first quarter of the 100 m2 size plot, i.e. on 25 m2 area, naturally growing trees over 2.0 cm in diameter at 1.3 m height, as well as all planted trees independently of their dimensions are measured and mapped (Fig 4 b).
In order to protect trees from injuries while boring holes in permanent plots, age and radial increment of trees are determined according to analogous trees growing on the angle count plots, allocated nearby the permanent plot in the same stand as the main circular plot. Two angle count plots are allocated at 20 m distance from the centre on both sides of the walking direction (Fig 4 b). Undergrowth and underbrush are inventoried in a 3×20 m strip plot allocated within the main plot along the walking direction 10 m to both sides from the plot centre (Fig 4 b).
Sample plots representing borderlines of different land use categories, stands, linear and other objects, according to the actual situation, are separated into sectors.
The research was based on the data of the NFI permanent sample plots established and measured during 1998 - 2000 in forest land of Lithuania.
In order to estimate the efficiency of NFI sampling design we examined the optimality of sampling design parameters: sample plot size, clustering of plots, size of the cluster. The optimal sample plots size was investigated analyzing the variability of growing stock in the sample plots of various sizes and time expenditures required for the establishment and measurement of such plots. The variability of growing stock was estimated analysing the coefficient of variation of growing stock in the sample plots of various sizes. It was tested according to Fairfield Smith’s empirical “law” (Smith 1938). The regularities of growing stock variation were analysed using the entire scale of variability of natural conditions in Lithuanian forests. The time consumption for fulfilling different operations in the NFI field works was determined using the time photography method. The effectiveness of the sampling design was estimated under the analysis of multiple relations between direct and indirect works fulfilled in the sample plot and tract. The optimal size of a sample plot was ascertained minimising the total cost (total time input) required for carrying out the inventory with a predefined accuracy. The representativeness of the applied sampling design was estimated analysing the mean and the variance of growing stock, comparing the results from clustered plots versus the individually allocated plots.
The estimation of growing stock variance in sample plots is very urgent both for proving optimal sampling designs of forest inventories and for ascertaining spatial forest structure as one of the most important factors of forest sustainability, stability and productivity. The most important and, unfortunately, least investigated are regularities of growing stock variation in stands associations covering a large scale variability of natural conditions.
The variance of growing stock per 1 hectare in all Lithuanian forests is 19 014 (m3/ha)2 on an average. In conventional terms it means a sufficiently high variation, equal to 60 - 65 %. The greatest variance was estimated in the least homogeneous aspen, oak and spruce stands, the lowest being in white alder and pine stands (Kasperavièius et al. 2001a).
The dependence of growing stock variation on forest type, stand age, stocking level, site humidity and fertility level as well as site index was determined. The most stable stands with even spatial structure and the least growing stock variation are those from 40 to 110 years of age, 0.7 - 0.9 stocking level, growing in medium or higher site index conditions (Kasperavièius et al. 2001b). The most even spatial structure is characteristic of pine and white alder stands. The coefficient of growing stock variation (CVq, %) of all stands decreased from 74.2 % to 55.7 % when the sample plot size (q) increased from 100 to 500 m2 (Fig. 5).
Fig. 5. Dependence of tree volume variation on tree species in Lithuanian forests
According to Fairfield Smith’s law, which can be re-expressed as: CVq =kq-b (where: k, b - regression coefficients to be estimated), the average value of the regression coefficient b is equal to 0.18. The higher the regression coefficient b, the more decreases CVq following a corresponding increment of sample plot size. During the study it was ascertained, that the influence of sample plot size changes on the change of tree volume per plot variation coefficient in homogeneous stand communities (b=0.18) is twice less than in separate stands (b=0.33-0.45). These results allow to essentially specify sampling design of inventories in large forest areas and to forecast a predefined assessment accuracy of indices.
The effectiveness of the sampling design was estimated analysing multiple relations between direct and indirect work operations carried out in the sample plot and tract, dependence of time consumption for carrying out different operations on the characteristics of measured objects and their quantities was ascertained.
According to the time input required for carrying out different operations in a tract and sample plot (Fig. 6), the time input structure in the full tract and in the tract with different number of sample plots in each of them was constructed (Fig. 7).
Fig. 6. Time input (min) required to single out and measure permanent sample plot in Lithuanian forests (3 surveyors)
Approximately a half of all time is required by driving and walking to the object and another half - by direct measurements on the plot. Time consumption for direct measurements in the sample plots varies from 34 to 63 % in the Lithuanian NFI sampling design depending on the number of sample plots per tract (Fig. 7). The share of time consumption for direct measurements in the plot increase due to increasing number of plots per tract. The average number of plots per tract for all Lithuanian NFI is 2.8. The equality between the time consumed for direct measurements and time for preparation to measure is mostly desirable and, according to Zeide research results (Zeide 1980), is the sign of optimal forest inventory design.
Fig. 7. Dependence of time input (min) required to measure tract on the number of sample plots in the tract (3 surveyors)
The time input for measuring a sample plot in Lithuanian forests for the crew of three members increases almost by 3.1 min for each 100 m2 with increasing sample plot size from 100 m2 up to 500 m2 (Fig. 8).
Fig. 8. Dependence of time input to measure a plot on the plot size
The range of optimal plot sizes ensuring the lowest inventory cost for different forest types varies from 500 m2 up to 600 m2 depending on the stands structure. The estimated optimal plot size for the inventory of the entire Lithuanian forests is equal to 514 m2 (Fig. 9). The increase of the sample plot size from 100 m2 up to 500 m2 enabled to reduce the total cost for the inventory of Lithuanian forests by about 33 - 39 %.
Fig. 9. Dependence of time input to inventory an object on plot size
Approximately 1.08 of a sample plot or its sector from the same tract fall into the aggregation of stands homogeneous by forest type, age and site index class. The estimation of the growing stock variance in stands homogeneous by forest type, age and site index class has shown that the losses in diversity of the investigated object allocating sample plots in tracts by four are insignificant. The introduced sampling design with clusters of four sample plots reduces the variance of growing stock only by 1 - 2 % for all tree species. Results of this analysis allow to interpret the sampling design of the Lithuanian NFI as the sampling design with individually allocated sample plots.
The increased precision of allocating sample plots using GPS receivers and dividing sample plots into sectors according to actual forest distribution by forest types has improved the objectiveness and reliability of the NFI data. All these measures allowed to increase the efficiency of the sampling design and to reduce the time consumption eliminating less informative data.
1. Tree volume variation in large forest tracts responds to plot size changes 2 - 3 times weaker as compared to the changes in plot volume variation in homogeneous stands. This fact essentially changes optimality parameters of sampling design in large forest tracts as compared to individual stands.
2. By applying peculiarities of the dependence of tree volume variation and time consumption variation on plot size, an optimal 500 - 600 m2 plot size was ascertained, allowing to reduce total labour consumption for all inventory object by 33 - 39 % as compared to 100 m2 size plots.
3. Optimization of plot size and their grouping ensures balance between time consumption for driving - walking and direct operations in the plot as one of the most important features of sampling design optimality.
4. Maximal (even 98 - 99 %) representation of object diversity was ensured by reduction of the number of plots per tract down to 4, increasing of the distance between plots within tract up to 250 m. Almost every sample plot or its sector fall into a stand different by site, prevailing tree species or age. This allows to state, that the Lithuanian NFI sampling design is very close to the sampling design with allocation of individual sample plots.
Kasperavičius, A., Kuliešis, A. 2001a. Influence of cluster sampling design on the representativeness of national forest inventory. Miškininkystė. Vol. 2 (50): 5 - 22. (In Lithuanian).
Kasperavičius, A., Kuliešis, A. 2001b. Analysis of the variation of growing stock in Lithuanian forests depending on environmental conditions and sample plots size. Aplinkos tyrimai, ininerija ir vadyba. Vol. 3 (17): 56 - 66. (In Lithuanian).
Kuliešis, A. 1989. Yield of forest stands and its use. Vilnius, 141 p. (In Lithuanian).
Kuliešis, A. 1994. Estimation of forest stand statistics using sample plots of varying size. Proceedings of the Lithuanian Forest Research Institute. Miškininkystė. Vol. 34: 112 - 128. (In Lithuanian).
Kuliešis, A. 1999. Application of sampling method in Lithuanian national forest inventory. Baltic Forestry. Vol. 5 (1): 50 - 57.
Kuliešis, A., Kasperavičius, A. 1998. Optimization of parameters of sampling units in Lithuanian national forest inventory. Baltic Forestry. Vol. 4 (2): 40 - 50.
Smith, H.F. 1938. An empirical law describing heterogeneity in the yields of agricultural crops. J. Agric. Sci. Vol. 28: 1 - 23.
Zeide, B. 1980. Plot size optimization. Forest science. Vol. 26 (2): 251 - 257.
Кулешиc A. 1971. Experience of growing stock resources estimation in the state forests of Lithuania by sampling method. Kaunas, 183 p. (In Russian).
| [1] Lithuanian State Forest
Inventory and Management Institute, Pramones pr. 11a, LT - 3031 Kaunas,
Lithuania. Tel: +370 37 490222; Email: [email protected] |