Predicting Some Stand Attributes of Scots Pine Stands Using Landsat 8 OLI Satellite Image

Abstract

 Monitoring forest resources using satellite images has been employed for different forest inventory purposes. This study used remote sensing data to derive regression models for estimating some stand attributes, including mean diameter, stand basal area, stand volume, number of trees, and stand density of pure Scots pine (Pinus sylvestris L.) stands. Field measurements were conducted within the 135 sample plots to obtain the above-mentioned stand attributes data. Reflectance values, vegetation indices, and texture values of each sample plot were generated from Landsat 8 OLI satellite images. The data obtained from sample plots were randomly selected and divided into two groups, consisting of 101 sample plots (75% of total data) for derivation of models, and 34 sample plots (25% of total data) for validation of the derived models. The prediction strengths of seven independent variable groups (i.e., reflectance values, vegetation indices, texture values, reflectance values and vegetation indices, reflectance values and texture values, vegetation indices and texture values, reflectance values, vegetation indices, and texture values) were also compared. A multiple linear regression analysis was utilized to fit stand at-tributes based on the derived independent variable groups. Three good-ness-of-fit statistics (R2, RMSE and MAE) were used to compare the different prediction models. Results revealed a moderate success of the derived regression models. Best models for mean diameter, stand basal area and number of trees were achieved with vegetation indices and tex-ture values as independent variable group, with R2 values of 0.492, 0.338 and 0.534, respectively.