Marius Hauglin

Research Scientist

(+47) 994 59 839
marius.hauglin@nibio.no

Place
Ås H8

Visiting address
Høgskoleveien 8, 1433 Ås

Abstract

Background The age of forest stands is critical information for forest management and conservation, for example for growth modelling, timing of management activities and harvesting, or decisions about protection areas. However, area-wide information about forest stand age often does not exist. In this study, we developed regression models for large-scale area-wide prediction of age in Norwegian forests. For model development we used more than 4800 plots of the Norwegian National Forest Inventory (NFI) distributed over Norway between latitudes 58° and 65° N in an 18.2 Mha study area. Predictor variables were based on airborne laser scanning (ALS), Sentinel-2, and existing public map data. We performed model validation on an independent data set consisting of 63 spruce stands with known age. Results The best modelling strategy was to fit independent linear regression models to each observed site index (SI) level and using a SI prediction map in the application of the models. The most important predictor variable was an upper percentile of the ALS heights, and root mean squared errors (RMSEs) ranged between 3 and 31 years (6% to 26%) for SI-specific models, and 21 years (25%) on average. Mean deviance (MD) ranged between − 1 and 3 years. The models improved with increasing SI and the RMSEs were largest for low SI stands older than 100 years. Using a mapped SI, which is required for practical applications, RMSE and MD on plot level ranged from 19 to 56 years (29% to 53%), and 5 to 37 years (5% to 31%), respectively. For the validation stands, the RMSE and MD were 12 (22%) and 2 years (3%), respectively. Conclusions Tree height estimated from airborne laser scanning and predicted site index were the most important variables in the models describing age. Overall, we obtained good results, especially for stands with high SI. The models could be considered for practical applications, although we see considerable potential for improvements if better SI maps were available.

To document

Abstract

Boreal forests constitute a large portion of the global forest area, yet they are undersampled through field surveys, and only a few remotely sensed data sources provide structural information wall-to-wall throughout the boreal domain. ArcticDEM is a collection of high-resolution (2 m) space-borne stereogrammetric digital surface models (DSM) covering the entire land area north of 60° of latitude. The free-availability of ArcticDEM data offers new possibilities for aboveground biomass mapping (AGB) across boreal forests, and thus it is necessary to evaluate the potential for these data to map AGB over alternative open-data sources (i.e., Sentinel-2). This study was performed over the entire land area of Norway north of 60° of latitude, and the Norwegian national forest inventory (NFI) was used as a source of field data composed of accurately geolocated field plots (n=7710) systematically distributed across the study area. Separate random forest models were fitted using NFI data, and corresponding remotely sensed data consisting of either: i) a canopy height model (ArcticCHM) obtained by subtracting a high-quality digital terrain model (DTM) from the ArcticDEM DSM height values, ii) Sentinel-2 (S2), or iii) a combination of the two (ArcticCHM+S2). Furthermore, we assessed the effect of the forest- and terrain-specific factors on the models’ predictive accuracy. The best model (,i.e., ArcticCHM+S2) explained nearly 60% of the variance of the training set, which translated in the largest accuracy in terms of root mean square error (RMSE=41.4 t ha−1 ). This result highlights the synergy between 3D and multispectral data in AGB modelling. Furthermore, this study showed that despite the importance of ArcticCHM variables, the S2 model performed slightly better than ArcticCHM model. This finding highlights some of the limitations of ArcticDEM, which, despite the unprecedented spatial resolution, is highly heterogeneous due to the blending of multiple acquisitions across different years and seasons. We found that both forest- and terrain-specific characteristics affected the uncertainty of the ArcticCHM+S2 model and concluded that the combined use of ArcticCHM and Sentinel-2 represents a viable solution for AGB mapping across boreal forests. The synergy between the two data sources allowed for a reduction of the saturation effects typical of multispectral data while ensuring the spatial consistency in the output predictions due to the removal of artifacts and data voids present in ArcticCHM data. While the main contribution of this study is to provide the first evidence of the best-case-scenario (i.e., availability of accurate terrain models) that ArcticDEM data can provide for large-scale AGB modelling, it remains critically important for other studies to investigate how ArcticDEM may be used in areas where no DTMs are available as is the case for large portions of the boreal zone.

To document

Abstract

Accurately positioned single-tree data obtained from a cut-to-length harvester were used as training harvester plot data for k-nearest neighbor (k-nn) stem diameter distribution modelling applying airborne laser scanning (ALS) information as predictor variables. Part of the same harvester data were also used for stand-level validation where the validation units were stands including all the harvester plots on a systematic grid located within each individual stand. In the validation all harvester plots within a stand and also the neighboring stands located closer than 200 m were excluded from the training data when predicting for plots of a particular stand. We further compared different training harvester plot sizes, namely 200 m2, 400 m2, 900 m2 and 1600 m2. Due to this setup the number of considered stands and the areas within the stands varied between the different harvester plot sizes. Our data were from final fellings in Akershus County in Norway and consisted of altogether 47 stands dominated by Norway spruce. We also had ALS data from the area. We concentrated on estimating characteristics of Norway spruce but due to the k-nn approach, species-wise estimates and stand totals as a sum over species were considered as well. The results showed that in the most accurate cases stand-level merchantable total volume could be estimated with RMSE values smaller than 9% of the mean. This value can be considered as highly accurate. Also the fit of the stem diameter distribution assessed by a variant of Reynold’s error index showed values smaller than 0.2 which are superior to those found in the previous studies. The differences between harvester plot sizes were generally small, showing most accurate results for the training harvester plot sizes 200 m2 and 400 m2.

To document

Abstract

Many remote sensing-based methods estimating forest biomass rely on allometric biomass models for field reference data. Terrestrial laser scanning (TLS) has emerged as a tool for detailed data collection in forestry applications, and the methods have been proposed to derive, e.g. tree position, diameter-at-breast-height, and stem volume from TLS data. In this study, TLS-derived features were related to destructively sampled branch biomass of Norway spruce at the single-tree level, and the results were compared to conventional allometric models with field measured diameter and height. TLS features were derived following two approaches: one voxel-based approach with a detailed analysis of the interaction between individual voxels and each laser beam. The features were derived using voxels of size 0.1, 0.2, and 0.4 m, and the effect of the voxel size was assessed. The voxel-derived features were compared to features derived from crown dimension measurements in the unified TLS point cloud data. TLS-derived variables were used in regression models, and prediction accuracies were assessed through a Monte Carlo cross-validation procedure. The model based on 0.4 m voxel data yielded the best prediction accuracy, with a root mean square error (RMSE) of 32%. The accuracy was found to decrease with an increase in voxel size, i.e. the model based on the 0.1 m voxel yielded the lowest accuracy. The model based on crown measurements had an RMSE of 34%. The accuracies of the predictions from the TLS-based models were found to be higher than from conventional allometric models, but the improvement was relatively small.

To document

Abstract

The use of forest biomass for bioenergy purposes, directly or through refinement processes, has increased in the last decade. One example of such use is the utilization of logging residues. Branch biomass constitutes typically a considerable part of the logging residues, and should be quantified and included in future forest inventories. Airborne laser scanning (ALS) is widely used when collecting data for forest inventories, and even methods to derive information at the single-tree level has been described. Procedures for estimation of single-tree branch biomass of Norway spruce using features derived from ALS data are proposed in the present study. As field reference data the dry weight branch biomass of 50 trees were obtained through destructive sampling. Variables were further derived from the ALS echoes from each tree, including crown volume calculated from an interpolated crown surface constructed with a radial basis function. Spatial information derived from the pulse vectors were also incorporated when calculating the crown volume. Regression models with branch biomass as response variable were fit to the data, and the prediction accuracy assessed through a cross-validation procedure. Random forest regression models were compared to stepwise and simple linear least squares models. In the present study branch biomass was estimated with a higher accuracy by the best ALS-based models than by existing allometric biomass equations based on field measurements. An improved prediction accuracy was observed when incorporating information from the laser pulse vectors into the calculation of the crown volume variable, and a linear model with the crown volume as a single predictor gave the best overall results with a root mean square error of 35% in the validation.

To document

Abstract

Airborne laser scanning data and corresponding field data were acquired from boreal forests in Norway and Sweden, coniferous and broadleaved forests in Germany and tropical pulpwood plantations in Brazil. Treetop positions were extracted using six different algorithms developed in Finland, Germany, Norway and Sweden, and the accuracy of tree detection and height estimation was assessed. Furthermore, the weaknesses and strengths of the methods under different types of forest were analyzed. The results showed that forest structure strongly affected the performance of all algorithms. Particularly, the success of tree detection was found to be dependent on tree density and clustering. The differences in performance between methods were more pronounced for tree detection than for height estimation. The algorithms showed a slightly better performance in the conditions for which they were developed, while some could be adapted by different parameterization according to training with local data. The results of this study may help guiding the choice of method under different forest types and may be of great value for future refinement of the single-tree detection algorithms.

Abstract

The aim of this study was to validate and compare single-tree detection algorithms under different forest conditions. Field data and corresponding airborne laser scanning (ALS) data were acquired from boreal forests in Norway and Sweden, coniferous and broadleaved forests in Germany, and pulpwood plantations in Brazil. The data represented a variety of forest types from pure Eucalyptus stands with known ages and planting densities to conifer-dominated Scandinavian forests and more complex deciduous canopies in Central Europe. ALS data were acquired using different sensors with pulse densities varying between the data sets. Field data in varying extent were associated with each ALS data set for training purposes. Treetop positions were extracted using altogether six different algorithms developed in Finland, Germany, Norway and Sweden, and the accuracy of tree detection and height estimation was assessed. Furthermore, the weaknesses and strengths of the methods under different forest conditions were analyzed. The results showed that forest structure and density strongly affected the performance of all algorithms. The differences in performance between methods were more pronounced for tree detection than for height estimation. The algorithms showed a slightly better performance in the conditions for which they were developed, while some could be adapted by different parameterization according to training with local data. The results of this study may help guiding the choice of method under different conditions and may be of great value for future refinement of the single-tree detection algorithms.