As an active remote sensing system airborne laser scanning (ALS) is well suited to achieve normalized digital surface models (nDSMs) by subtracting digital terrain models (DTMs) from digital surface models (DSMs). The nDSM represents object heights and is an important data source for the derivation of various forest parameters such as tree height, stem volume or biomass. The validation of the derived results as well as the comparison of the results from different study areas is often a challenging task due to different sampling designs and accuracies of forest inventory data being used as ground truth data. In this study we use 17 fully callipered samples, covering an area of 4.55 ha in total, to assess the accuracies of stem volume and biomass maps for different Austrian test sites. For the callipered samples all trees with a diameter at breast height ≥10.5 cm were measured. For the calibration of the stem volume and biomass models available national forest inventory (NFI) as well as local forest inventory (LFI) data are used, which are both based on angle count sampling plots. This verification approach guarantees firstly the independency of calibration and validation data and secondly it allows accuracy analyses for different reference units. For the study area Montafon the relative differences of stem volume and biomass range between -20.0% and 57.4% and between 16.3% and 56.2% respectively for twelve coniferous dominated sample areas with ~0.25 ha each. For a reference unit with an area of ~3.0 ha the relative differences decrease to 15.7% and 19.3% for stem volume and biomass respectively. For the study area Tyrol deciduous and coniferous models were applied. The calculated relative differences of stem volume and biomass vary between -25.8% and -10.3% and -18.5% and 3.1% respectively for the two coniferous dominated sample areas with an area of ~0.38 ha each. For the two deciduous dominated sample areas with an area of ~0.38 ha, both the relative difference of stem volume and biomass vary between -10.0% and 0.6% and -3.3% and 1.1% respectively. The average relative differences for all sample areas of the Tyrol study area with a total area of ~1.5 ha is -1.2% and -1.9% for the stem volume and the biomass, respectively. As the estimations of the stem volume and biomass maps are based on federal state wide data sets (ALS and NFI) the findings of this study are of high practical relevance for integrating ALS derived forest parameters into operational forest inventories. EARSeL eProceedings 11, 1/2012 75

In this study a semi-empirical model that was originally developed for stem volume estimation is used for aboveground biomass (AGB) estimation. The semi-empirical model is based on the relative heights of first echo LiDAR point cloud data and assumes a linear relationship between AGB and canopy volume. However, the usage of point cloud data leads to a computationally demanding task when processing large point cloud datasets for the generation of area-wide AGB maps. In the presented study the effects of using rasterized LiDAR data as input for the AGB model are investigated in order to speed up processing and to make use of the model on large spatial datasets. The canopy volumes are calculated from a Canopy Height Model (CHM). The optimum resolution of the CHM is determined by analyzing the effects of varying cell sizes (1.0 m, 1.5 m, 2.0 m, 3.0 m) on the achievable accuracies. Calibrating the model with rasterized input data having a spatial resolution of 2.0 m instead of using first echo point cloud data leads to a slight increase of the coefficient of determination (R = 0.70 to R = 0.72) and a slight decrease of the standard deviation of the prediction errors. For calibrating the model reference AGB is calculated per sample plot from local forest inventory data by means of averaged weighted (according to tree species and age class composition) extension factors. The influence of using rasterized LiDAR input data on the achievable accuracy of the assessed AGB is investigated for a coniferous dominated study area in Vorarlberg, Austria.

a) Institute of Photogrammetry and Remote Sensing, Vienna University of Technology Gushausstrase 27-29, 1040 Wien, Austria b) Department of Forest Inventory at the Federal Research and Training Center for Forests, Natural Hazards and Landscape, Seckendorff-Gudent-Weg, 1130 Vienna, Austria c) alpS-Center for Natural Hazard Management, Grabenweg 3, 6020 Innsbruck, Austria d) LASERDATA GmbH Management and Analysis of Laserscanning Data, Technikerstr. 21a, 6020 Innsbruck, Austria e) Stand Montafon Forstfonds, Montafonerstrase 21, 6780 Schruns, Austria

The overall goal of this study was to describe a novel area-based semiempirical model for estimating growing stock from small-footprint light detection and ranging (lidar) data. The model assumes a linear relationship between growing stock and lidar-derived canopy volume that is stratified according to several canopy height classes to account for height dependent differences in canopy structure and nonlinear tree size-shape relationships. It was applied over a 128 km2 alpine area in Austria where operational forest inventory data and lidar data acquired in winter and summer were available. The analysis showed that the semiempirical model was quite robust against changes in laser point density and acquisition time. Further, it was found that the model performed as well as a widely used iterative regression method based on a multiplicative model. Both models reached a high coefficient of determination (R2 = 0.76–0.86) and a standard deviation of the residuals in the order of 20.4%–29.1%. Although it is less flexible than the multiplicative model, the advantages of the semiempirical model are its simplicity and the fact that its coefficients can be physically interpreted. These traits can be expected to enhance the applicability of the model in regions where high-quality inventory data are lacking.

This paper evaluates the performance of a recently developed approach for wide-area stem volume estimations based on airborne laser scanning (ALS) and national forest inventory (NFI) data in the case where data recorded under operational conditions are used as input. This entails that neither ALS data nor NFI samples were collected and optimized for the current study. The approach was tested for the Austrian state of Vorarlberg, which covers an area of 2601 km2 and encloses about 970 km2 of forest land. ALS data with point densities varying between 1 and 4 points m−2 were acquired in the framework of a commercial state-wide terrain mapping project during several winter- and summer-flight campaigns. The stem volume model was calibrated with all NFI data available for Vorarlberg, whereas additional local forest inventory data were used for independent validation. Moreover, several relevant operational issues were addressed in this study, such as the determination of the optimum area used to calculate the reference laser metrics input to the model, the effect of gridding point cloud data to speed up processing, and the stratification of input data into coniferous and deciduous sample plots. Without tree species stratification and based on the 3D laser heights model, calibration provided a maximum R2 of 0.79 and a standard deviation (SD) of residuals derived from cross-validation of 107.4 m3 ha−1 (31.5%). Calibrating the model only with coniferous samples increased the achieved R2 to 0.81 and decreased SD to 104.8 m3 ha−1 (29.7%). As only eight NFI sample plots were available for deciduous forest a robust calibration of a separate model could not be obtained. Calibrating the model with a rasterized canopy height model (CHM) instead of using the 3D laser heights just led to a slight decrease in accuracy (R2 = 0.75, SD = 120.9 m3 ha−1 (35.5%) without forest-type stratification and R2 = 0.78 and SD = 117.2 m3 ha−1 (33.1%) for the coniferous stem volume model). Finally, the stem volume model calibrated with CHM data was adopted to generate a stem volume map of the entire State of Vorarlberg. Validation of this map with the additional local forest inventory data confirmed the accuracies (R2 = 0.75; SD = 135.6 m3 ha−1 (32.3%)) that were derived during calibration of the stem volume model based on the NFI data. The models and methods presented in this study are used operationally for forest and environment policy purposes and practical applications in Austria.

Airborne laser scanning (ALS) is an active remote sensing technique that uses the time-of-flight measurement principle to capture the three-dimensional structure of the earth’s surface with pulsed lasers that transmit nanosecond-long laser pulses with a high pulse repetition frequency. Over forested areas most of the laser pulses are reflected by the leaves and branches of the trees, but a certain fraction of the laser pulses reaches the forest floor through small gaps in the canopy. Thus it is possible to reconstruct both the three-dimensional structure of the forest canopy and the terrain surface. For the retrieval of quantitative forest parameters such as stem volume or biomass it is necessary to use models that combine ALS with inventory data. One approach is to use multiplicative regression models that are trained with local inventory data. This method has been widely applied over boreal forest regions, but so far little experience exists with applying this method for mapping alpine forest. In this study the transferability of this approach to a 128 km2 large mountainous region in Vorarlberg, Austria, was evaluated. For the calibration of the model, inventory data as operationally collected by Austrian foresters were used. Despite these inventory data are based on variable sample plot sizes, they could be used for mapping stem volume for the entire alpine study area. The coefficient of determination R2 was 0.85 and the root mean square error (RMSE) 90.9 m3ha-1 (relative error of 21.4%) which is comparable to results of ALS studies conducted over topographically less complex environments. Due to the increasing availability, ALS data could become an operational part of Austrian’s forest inventories.