Evaluation of remote sensing techniques for estimation of forest variables at stand level

摘要: There is a continuous need for accurate forest description. Forest data at stand level required in forestry planning, particular when scheduling treatments within the next few years. Collection of typically acquired with subjective surveying methods field. However, field work labor intensive and therefore an expensive method. An alternative to use remotely sensed by, e.g. optical, radar, laser sensors. In present thesis, different remote sensing techniques estimation variables have been evaluated. All studies were performed hemi-boreal coniferous dominated test site southern Sweden (lat. 58°30’N, long. 13°40’E), enabling detailed comparison. Remotely related using regression analysis. The root mean square error (RMSE) average stem volume airborne scanning data, CARABAS-II radar aerial photo-interpretation, multi-spectral optical satellite was found be 13%, 19%, 21-24%, 24-32%, respectively. analyses clearly demonstrate that single technique those investigated gives most tree height estimates level. from are better than commonly achieved accuracy variable photo-interpretation Z/I DMC images agreement results conventional film-based panchromatic photos. Hence, indicate digital photogrammetric workstation could replace photos analog or analytical stereoplotters without loss accuracy. Combining two more sources which complementary possibility improving accuracies. Here, particularly successful approaches found. A combination improved RMSE 11-12%. Using about 15%. application sector restricted costs availability data. Sweden, might supplied by several companies operating pure commercial market. National Land Survey has mission photo mapping on regular basis, but also companies. For governmental policy effects practice what offered users.