Combining sampling-based forest inventory data and remote sensing to improve volume estimates of mature forest stands

In Lithuania, sampling-based surveys of mature forests started in 2008–2009. These surveys focused on statistically sound and NFI-compatible estimates at the forest enterprise level. They are based on measurements of large number of sample plots, thus making forest inventory labor-intensive and costly. The results of such inventories are hardly applicable at local level. This study tested several methodological approaches to improve forest enterprise and local level stand volume estimates for mature stands utilizing the Landsat 5 Thematic Mapper (TM) images and non-parametric k-nearest neighbor estimation technique. As the study object we used information available from the surveys of mature forest stands conducted in 4 state forest enterprises in Northern part of Lithuania (total area around 6000 sq. km), with over 4500 precisely geo-referenced circular field plots and major forest characteristics measured in the field. Information on National Forest Inventory (NFI) plots for the area of the satellite image scenes as well as the stand-wise forest inventory data was available for the study at the State Forest Survey, too. Different approaches to reduce the number of field plots to be measured during the surveys of mature stands were tested considering the estimate achieved with the full set of sample plots as a true value: using just the NFI plots, using subsets of sample plots from the survey of mature stands, using stratified subsets of sample plots utilizing the stand-wise forest inventory data. Simultaneously, the potential of providing local level stand volume estimates utilizing the no cost satellite images, surveyed sample plots and information borrowing technique was discussed, tooVytauto Didžiojo universitetasŽemės ūkio akademij