Predicting tree diameter distributions from airborne laser scanning, SPOT 5 satellite, and field sample data in the Perm Region, Russia

A tree list is a list of trees in the area of interest containing, for example, the species, diameter, height, and stem volume of each tree. Tree lists can be used to derive various characteristics of the growing stock, and are therefore versatile and informative sources of data for several forest management purposes. Especially in heterogonous and unmanaged forest structures with multiple species, tree list estimates imputed from local reference field data can provide an alternative to mean value estimates of growing stock (e.g., basal area, total stem volume, mean tree diameter, mean tree height, and number of trees). In this study, reference field plots, airborne laser scanning (ALS) data, and SPOT 5 satellite (Satellite Pour l'Observation de la Terre) imagery were used for tree list imputation applying the k most similar neighbors (k-MSN) estimation method in the West Ural taiga region of the Russian Federation for diameter distribution estimation. In k-MSN, weighted average of k field reference plots with highest similarity between field reference plot and target (forest grid cell, or field plot) based on ALS and SPOT 5 features were used to predict the mean values of growing stock and tree lists for the target object simultaneously. Diameter distributions were then constructed from the predicted tree lists. The prediction of mean values and diameter distributions was tested in 18 independent validation plots of 0.25-0.5 ha in size, whose species specific diameter distributions were measured in the field and grouped into three functional groups (Pines, Spruce/Fir, Broadleaf Group), each containing several species. In terms of root mean squared error relative to mean of validation plots, the accuracy of estimation was 0.14 and 0.17 for basal area and total stem volume, respectively. Reynolds error index values and visual inspection showed encouraging results in evaluating the goodness-of-fit statistics of the estimated diameter distributions. Although estimation accuracy was worse for functional group mean values and diameter distributions, the results indicate that it is possible to predict diameter distributions in forests of the test area with the tested methodology and materials.