prilagođeno pretraživanje po punom tekstu

ŠUMARSKI LIST 11-12/2021 str. 72     <-- 72 -->        PDF

White, J.C., M. A. Wulder, M. Vastaranta, N. C. Coops, D. Pitt, M. Woods, 2013: The utility of image-based point clouds for forest inventory: A comparison with airborne laser scanning. Forests, 4, 518–536.
White, J.C., C. Stepper, P. Tompalski, N.C. Coops, M.A. Wulder, 2015: Comparing ALS and Image-Based Point Cloud Metrics and Modelled Forest Inventory Attributes in a Complex Coastal Forest Environment. Forests, 6 (10): 3704-3732.
White, J.C., N. C. Coops, M. A. Wulder, M. Vastaranta, T. Hilker, P. Tompalski, 2016: Remote Sensing Technologies for Enhancing Forest Inventories: A Review. Canadian Journal of Remote Sensing 42 (5): 619-641.
World reference base for soil resources (WRB), 2006: A framework for international classification, correlation and communication. World Soil Resources Reports No. 103, FAO, Rome, Italy: 128.
The application of digital aerial photogrammetry in forest inventory has been in the focus of a number of studies during the last decade (White et al. 2016, Goodbody et al. 2019). The results were tested and evaluated mostly on plot-level, and less often on stand-level (Bohlin et al. 2012, Rahlf et al. 2014, Gobakken et al. 2015, Pitt et al. 2015, Stepper et al. 2015, Puliti et al. 2016, Balenović et al. 2017, Iqbal et al. 2019). In almost all studies, a classic Area-Based Approach (ABA) which includes modelling at plot-level and ‘wall-to-wall’ mapping to estimate forest variables at stand-level were applied. A somewhat different ABA approach which implies direct modelling and estimation at stand-level were proposed by Balenović et al. (2017). This new approach, hereinafter referred to as Stand-Based Approach (SBA), is based on existing and easily available photogrammetric materials and data (aerial images from regular national topographic surveys, official national digital terrain data) as well as on data from existing forest management plans. The main precondition for the SBA method application is the approximate time coincidence between the time of aerial images acquisition and the time of the field data collection for the forest management plan generation. Similar to classical ABA, SBA also requires models (equations) for forest variable estimations. However, in SBA case, models are developed on the certain number of stands (subcompartments) of similar structural characteristics (e.g. forest management classes). In SBA, independent variables are metric stand-level data obtained from aerial images and its product (normalized point clouds or canopy height models), while reference (modelling or validation) data are obtained from regular forest management plans. Compared to classical ABA, SBA does not require additional field measurements, and therefore presents a fast and cost-effective alternative to ABA. An additional assumption is that models developed for the certain area can be applied for other forest areas with similar forest characteristics.
This work presents the continuation of previous study (Balenović i dr. 2017) with the aim to additionally test the effectiveness and accuracy of SBA method. More precisely, SBA method and existing models of stand volume estimation originally developed for lowland pedunculate oak (Quercus robur L.) of Spačva basin forest complex (Eastern Croatia) were tested in pedunculated oak forest of Pokupsko basin forest complex (Central Croatia).
A total of 87 even-aged pedunculate oak forest stands of Jastrebarski lugovi management unit were included in this study (Figure 1, Table 1). Photogrammetric data (aerial images, digital terrain data) were provided by the Croatian State Geodetic Administration, and were used to generate Digital Surface Model (DSM) and Digital Terrain Model (DTM). A raster Canopy Height Model (CHM) of 5 m resolution was generated by subtracting DTM from DSM (Figure 2). Metrics extracted from CHM for each stand and used for stand-level volume estimation are presented in Table 2. Equations (1) and (2) present photogrammetric models for stand-level volume estimation. A more detailed description of the models can be found in Table 3. SB-1 and SB-2 models consist of independent variables and parameters (regression constant and coefficients) originally developed for the Spačva basin area (Table 4). PB-1 and PB-2 models consisted of the same variables as SB-1 and SB-2 models, but their parameters were developed for the present study area (Pokupsko basin) (Table 4). All models were validated using the reference stand volume from the forest management plan. SB-1 and SB-2 models were validated using the entire dataset (87 stands), whereas PB-1 and PB-2 models were validated using the randomly selected 29 stands (other 58 stands were used for parameters estimation).
According to validation results (Table 2, Figure 3), PB models showed considerably greater performance than SB-models. Compared to SB-1 model, PB-1 model achieved 11% higher R2adj values, for 3,92% MD% values (absolute), and for 6,44% higher RMSE% values. Also, the results showed that the inclusion of stand age (SA) as an additional predictor in SB-2 and PB-2 models did not contribute