Comparing different methods for prediction of mean crown height in Norway spruce stands using airborne laser scanner data

In this study, different methods were used to predict mean crown height of Norway spruce-dominated stands by means of low pulse density airborne laser scanning (ALS) data. The methods were based on statistical modelling, properties of the laser point clouds or combinations of them. Separate modelling data were used for model calibration and two different validation datasets were used to assess the accuracy of the results. The results obtained were partly contradictory, showing varying performance of different methods using different datasets. However, there were also notable differences between the methods used to obtain crown height by field measurements. The root mean square error figures of crown height predictions were at minimum between 1.0 and 1.5 m. This study showed that statistical modelling based on ALS height metrics was a good approach if the relationship between mean crown height and the ALS information was corresponding in the modelling data and in the application phase. A method based on the alpha shape technique was also an accurate alternative. Methods that rely directly on the laser point cloud to predict mean crown height without any calibration were good alternatives to get relatively accurate results but there are still drawbacks (area of calculation unit) in their applicability.