Doctoral thesis, 2012
Estimation of canopy structure and individual trees from laser scanning dataLindberg, Eva
AbstractDuring the last fifteen years, laser scanning has emerged as a data source for forest inventory. Airborne laser scanning (ALS) provides 3D data, which may be used in an automated analysis chain to estimate vegetation properties for large areas. Terrestrial laser scanning (TLS) data are highly accurate 3D ground-based measurements, which may be used for detailed 3D modeling of vegetation elements. The objective of this thesis is to further develop methods to estimate forest information from laser scanning data. The aims are to estimate lists of individual trees from ALS data with accuracy comparable to area-based methods, to collect detailed field reference data using TLS, and to estimate canopy structure from ALS data. The studies were carried out in boreal and hemi-boreal forests in Sweden. Tree crowns were delineated in three dimensions with a model-based clustering approach. The model-based clustering identified more trees than delineation of a surface model, especially for small trees below the dominant tree layer. However, it also resulted in more erroneously delineated tree crowns. Individual trees were estimated with statistical methods from ALS data based on field-measured trees to obtain unbiased results at area level. The accuracy of the estimates was similar for delineation of a surface model (stem density root mean square error or RMSE 32.0%, bias 1.9%; stem volume RMSE 29.7%, bias 3.8%) as for model-based clustering (stem density RMSE 33.3%, bias 1.1%; stem volume RMSE 22.0%, bias 2.5%). Tree positions and stem diameters were estimated from TLS data with an automated method. Stem attributes were then estimated from ALS data trained with trees found from TLS data. The accuracy (diameter at breast height or DBH RMSE 15.4%; stem volume RMSE 34.0%) was almost the same as when trees from a manual field inventory were used as training data (DBH RMSE 15.1%; stem volume RMSE 34.5%). Canopy structure was estimated from discrete return and waveform ALS data. New models were developed based on the Beer-Lambert law to relate canopy volume to the fraction of laser light reaching the ground. Waveform ALS data (canopy volume RMSE 27.6%) described canopy structure better than discrete return ALS data (canopy volume RMSE 36.5%). The methods may be used to estimate canopy structure for large areas.
Keywordsforest inventory; individual trees; canopy structure; laser scanning; LiDAR; ALS; TLS
Published inActa Universitatis Agriculturae Sueciae
2012, number: 2012:33
Publisher: Institutionen för skoglig resurshushållning, Sveriges lantbruksuniversitet