This paper discusses the continuous field view of representing geography and its role in forest inventory. Describing all forest attributes at any location or point and storing this information in digital databases, as an alternative to store forest compartments as vector polygons and associated attributes, is considered to be important for development of remote sensing and GIS applications for forest inventories. Main results achieved following this approach during the last decade in Lithuanian university of agriculture are described. Special research polygon near Kaunas was developed. Four auxiliary data sources (Spot Xi, Landsat TM, digital aerial photos and stand-wise inventory material), three estimators (two-phase sampling with stratification, the k-nearest neighbours and regression) as well as different methods for auxiliary data integration were examined to get point-wise forest characteristics. The lowest root mean square errors at a level of virtual sample point using optimal implementation tactics are – for mean diameter 24 %, height 18%, age 28%, basal area 37%, volume per 1 ha 40% and percentage of coniferous trees 29% of the mean value of corresponding forest characteristic. Integrating additional auxiliary information – characteristics of forest compartments, estimated during the conventional stand-wise inventory – and satellite images improved the overall estimation accuracy. Pre-stratification of sample plots using the attributes of compartments improved the estimation accuracy for certain stand groups. A new approach of segmentation, aimed to construct conventional forest compartments – estimation of point-wise forest characteristics for every pixel of satellite imagery and using them instead of original image values – was suggested and investigated.

This paper discusses the usability of non-parametric knn (k-nearest neighbour) method to detect changes in forest areas from satellite images. Spot Xi images acquired 1999, main forest characteristics from field measured sample plots and data of conventional stand-wise forest inventory from the year 1988 were used to estimate the grids of following forest characteristics: mean age of main forest storey, diameter, basal area, height, volume per 1 ha, as well as the percentages of coniferous, soft and hard deciduous tree species. The differences of grids, created using stand-wise forest attributes from the 1988 inventory and estimated using the k-nearest neighbour methods were experimented to detect changes in the forest. 68.7-75.5% of areas, classified as the potential felling areas, were detected to be clear cut areas or young stands less than 15 years according to the data of stand-wise inventory of year 2003. Different settings for the methods investigated are evaluated, too.

This paper deals with the evaluation of the geometrical accuracy of Lithuanian forests compartments geographical data that has been developed using different forest inventory techniques. Geo-reference background database GDB10LT was used as the standard for comparisons. 2500 control points on clearly identifiable places – crossroads, dikes’ intersections, etc. – were selected randomly. The main finding was that the maximal positional root mean square error of clearly identifiable objects in forestry geographic data was 16.47 m (12.37 m and 10.87 m for X and Y coordinates respectively). However, such rather big errors refer to the techniques of GIS database development using paper topographic maps as a background for forest maps and manual digitizing. Enhancement of techniques for GIS database development was found to lead to significant increase in geometrical accuracy of the information.

In forest harvesting process, a logging company and a forest landowner who have various goals concerning forest stands purchase value are involved. The logging company wants to maximize the profit on lower price, but forest landowner - to extract maximum income. Calculating the forest stands value prior to harvesting it is possible to achieve mutually acceptable price. This article describes question statement and primary research of prior calculated and harvested volume difference for further study of forest stands purchase value algorithm optimization. Harvested volume from feed-back data and prior calculated forest stands volume were compared to obtain the difference and understand the economical importance for optimization of harvesting sites maximal value calculation algorithm. For primary study, forest stands prior harvesting calculation model from the logging company was used. Results from prior harvesting calculations were compared with data after harvesting to achieve information about the precision of calculation model. Obtained results showed significant difference between prior calculated and harvested volumes, which in financial matter cause losses to the logging company. Compared to total harvested volume, negative cutting difference was 7% from volume, which in financial terms with annual harvesting of 1 million m**3 make loss of ~2.3 million euro.

In Latvia, almost 50% of all forests are privately owned. Due to hard economic situation in the 1990ies, many private forests have suffered from illegal forestry operations and overexploitation. One of economically most important tree species for private forest owners is Norway spruce. In this study the productivity of state-owned and private spruce forests was compared, assuming that the stand productivity of private forests should be lower compared to those owned by the state. Data gathered in the Forest Resource Inventory in 2004, 2005 and 2006 was used for the analysis. Total standing volume of the dominant stand and the current mean annual volume increment (CMAI) of spruce were used as main productivity indicators. The mean values of the main stand characteristics were found to be rather similar in state and private forests. The only significant differences were discovered in the mean diameter and CMAI of spruce in 60-90 years old stands; in both cases the values were higher in private forests. Site type and mixture degree were tested as the possible influencing factors. Although there were differences in the site type distribution between state and private forests, the influence of this factor on the CMAI of spruce proved to be insignificant. It was found out that the CMAI of spruce depended significantly on the mixture degree. The proportion of mixed stands in the state forests was considerably higher than in the private forests; presumably therefore the value of the CMAI in the state forests was significantly lower.

Researching new remote sensing and data processing methods is very important subject in forestry. The objectives of research are to explore methods to determine single tree characteristics using LIDAR and adapt them for Latvian forest conditions. Different algorithms and mathematical relations for automatic calculation of tree species, coordinates, height and diameter at breast height are described. Within the project four different clustering methods for tree identification were evaluated. The first method's construction is based on reflection point count in certain height range. The second and third methods are searching for global and local maximums on height axis of LIDAR data collection. In the fourth method segmentation of aerial photography is done by using the user selected sample data. Tree tops were s discovered by searching similarly coloured regions. Field measurements were used for the calibration of LIDAR data and analysis. Sample plots were fitted in the study area with different species composition, age and density. The total number of measured trees in sample plots is 1844. Results show that height can be found mainly for I, II, III craft class trees with average error 2.5%. Stem diameter estimation error of pine is 28%, spruce 17%, birch 4.2% and second storey trees 5.4% using linear equations D = 0.6616*H + 4.6969 (for coniferous trees) and D = 0.7756*H + 3.7132 (for deciduous trees). Dividing trees in classes of coniferous and deciduous can be done by using near infra red photography. The total number of first storey trees identified by LIDAR is 91%, by aerial photographic method 94%.

This paper discusses the use of medium resolution Landsat TM satellite images to support conventional approaches of Lithuanian forest inventory practices. Estimation accuracies achieved using just field measured sample plots, Landsat TM satellite images and two non-parametric k-nearest neighbour and most similar neighbour estimators were studied at a level of compartments. 19 mature forest areas, prepared for final felling with GPS measured borders and all trees callipered, were used for validation. Notably higher estimation accuracies were achieved using field sample plots distributed through the whole forest area studied than just ones located on mature forest stands. The root mean square deviations in estimating compartment-wise volume of growing stock per 1 ha was around 27-28% if the best variant of estimation approach was used. Possible influence of the accuracy in locating the borders of validation areas on the estimations is discussed in the paper, too.

This paper introduces the first test results to use laser scanning and high resolution digital colour infrared aerial image data to estimate average tree crown density at a sample plot level. General methodological framework based on two-phase sampling schemes, non-parametric estimators and satellite images as the auxiliary data sets was adopted for the use with airborne data sources. More than 400 circular sample plots were established and measured in a special research forest area near Kaunas, the central part of Lithuania. The tree crown density was visually estimated for every coniferous tree belonging to the 500 square m plot together with other conventional forest parameters. Two variants of digital colour infrared aerial images (ground sampling density 15 and 40 cm), LiDAR point clouds, based on 1 point/square m scanning density and two phase sampling approach with non-parametric k-nearest neighbour and most similar neighbour estimators were used to test the accuracies of tree crown density estimation at a sample plot level. Reliable estimates were found to be possible on pure coniferous stands only. Average tree crown density was estimated with the root mean square error around 17.5-18% at a sample plot level, bearing in mind average crown density around 64% for the whole study area. The estimates were unbiased. Integration of laser scanning based variables with the ones available from digital aerial images resulted in lowest estimation root mean square errors. Laser scanning based variables used as the auxiliary data set independently resulted in better estimation errors than the variables available from digital colour infrared images.

Tree species identification is important for a variety of natural resource management and monitoring activities especially in forest inventory. The objective of research is to identify tree species using digital aerial photography and LIDAR data in Latvian forest conditions. The study outlines a number of tree species identification possibilities: the ability to identify conifers and deciduous trees; the ability to identify pine and spruce; the ability to identify birch, aspen and black alder. The study site is a forest in the middle part of Latvia at Jelgava district (56º39’ N, 23º47’ E). Aerial photography camera (ADS 40) and laser scanner (ALS 50 II) were used to capture the data. LIDAR resolution is 9 points m-2 (500 m altitude). The image data is RGB, NIR and PAN spectrum with 20 cm pixel resolution. During the study a modified region growing algorithm was developed to determine tree canopy and tree species identification using threshold segmentation, Fourier transform, frequency filtering and reverse Fourier transform. Tree species classification of coniferous and deciduous trees is possible in 82% of the cases; the first storey of the trees can be classified correctly in 96% of the cases, but the second storey of the trees only in 49% of the cases. Spruce identification is possible in 81.1% of the cases, for first storey trees in 89.6% of the cases and for the second storey trees in 72.9% of cases. Deciduous tree correct classification is possible in 63% of the cases, birch 75%, black alder 60% and aspen only in 41% of the cases.

The study represents results of theoretical evaluation of forest biomass available for solid biofuel production in pre-commercial thinning in Latvia. The study is based on the National forest inventory (NFI) data; calculations are done for each NFI plot separately. The calculation is done in three steps – selection of the NFI sample plots, which fulfils criteria for the pre-commercial thinning, development of the diameter distribution table, setting the criteria of the thinning intensity, calculation of extractable biomass. Thinning from below (removal of the smallest trees) is considered in calculation. Two types of biomass are accounted – full tree (aboveground biomass) and stem-wood (stem biomass). The study demonstrates that pre-commercial thinning could become an important source of forest biomass in Latvia (15400 GWh of primary energy according to current situation in forests); however, dimensions of trees and harvesting conditions might be challenging for production. The most of the potential biofuel resources are located in stands with average tree higher than 8 m; therefore, it is reasonably to develop and introduce technologies applicable for production of partially delimbed trees.