Estimation of forest stand volumes by Landsat TM imagery and stand-level field-inventory data

Abstract

Strategic decisions concerning the timing and location of forest operations require accurate information about the state of forests. In regional forest management, it is essential that the inventory data cover all the forests of the area. A field inventory is expensive, and due to cost-effectiveness the forests not under active management are omitted, which tends to create gaps. Some inventory data for these gaps can potentially be obtained by remote sensing. In this study, forest stand parameters were estimated by means of Landsat TM images and existing stand-level forest data of neighbouring area. The non-parametric k-nearest-neighbour ( kNN) estimator was employed in the estimation of stand volumes by tree species. Different spectral features were tested in the estimation, and the average values of the pixels within a stand core proved to give the best volume estimates. When the kNN estimator and the average of the stand core pixels were employed in the estimation, the estimation error (RMSE) of the total volume was 48%. The errors for other volume estimates were higher; for spruce the RMSE was 81%, and in the case of the volumes of pine and broad-leaved trees, the RMSEs were over 100%. The estimation results obtained are not accurate enough for forest management purposes at the forest stand level. However, the approximate volume estimates derived by the method can be useful in areas where no other forest information is available.