Day 4

Detailed paper information

Back to list

Paper title Hemiboreal forest growing stock estimation by airborne and satellite stereo imagery
Authors
  1. Grigorijs Goldbergs Institute of Electronics and Computer Science Speaker
Form of presentation Poster
Topics
  • A3. Biosphere
    • A3.06 Biomass monitoring
Abstract text Understanding the hemiboreal forestland's role in the continental carbon cycle requires reliable quantification of their growing stock (forest biomass) at the regional scale. Remote sensing complements traditional field methods, enabling indirect fine-scale estimation of forest 3D structure parameters (primarily tree height) from high-density 3D point clouds by avoiding destructive sampling. In addition, carbon accounting programs and research efforts on climate-vegetation interactions have increased the demand for canopy height information, an essential parameter for predicting regional forest biomass [1]. Unfortunately, relatively high acquisition costs prevent airborne laser scanning (ALS), the most efficient and precise tool, from regularly mapping forest growing stock and dynamics. Therefore, in the last decade, there has been increasing interest in using very high resolution (ground sample distance (GSD) < 0.5 m) satellite-derived stereo imagery (VHRSI) to generate canopy height models (CHM) analogous to LiDAR point clouds to support forest inventory and monitoring. Despite the large offer of VHRSI sensors on the market (GeoEye, WorldView etc.), image-derived CHM performance for retrieving the forest inventory data in various geographical regions is still not fully understood [2]. However, while the ALS can penetrate the forest canopy and characterise the vertical distribution of vegetation, the VHRSI image-based point clouds only represent the non-transparent outer “canopy blanket” cover of dominant trees.

Thus, the present study assesses the potential of VHRSI sensors for an area-based prediction of growing stock (m3 ha-1) by deriving the main forest canopy height metrics from image-based point clouds and validating against the Latvian National Forest Inventory (NFI) data. The study area represents a typical hemiboreal forestland pattern across the eastern part of Latvia with predominantly mature, dense, closed-canopy evergreen pine, spruce and deciduous birch, black alder tree species.

The study workflow was divided into two stages. During the first stage, the study: (1) evaluated and compared the vertical accuracy and completeness of CHMs derived from airborne and VHRSI stereo imagery to reference LiDAR data; (2) analysed the differences in the CHM height estimates associated with different tree species; (3) examined the effect of sensor-to-target geometry (specifically base-to-height ratio) on matching performance and canopy height estimation accuracy [3]. As a result, the study confirmed the tendency for canopy height underestimation for all satellite-based models. The image-based CHMs of forests with dominated broadleaf species (e.g., birch and black alder) showed higher efficiency and accuracy in canopy height estimation and completeness than trees with a conical crown shape (e.g., pine and spruce). Furthermore, this research has shown that determining the optimum base-to-height (B/H) ratio is critical for canopy height estimation efficiency and completeness using image-based CHMs. This study found that stereo imagery with a B/H ratio of 0.2–0.3 (or convergence angle range 10°–15°) is optimal for image-based CHMs in closed-canopy hemiboreal forest areas.

At the second stage (currently being implemented), the study: (1) establish allometric relationships between field-derived (harvester data) individual tree volume and tree height; (2) use estimations from individual tree LiDAR measurements as training/reference data of growing stock for study area plots; (3) utilise a two-phase analysis that integrates both individual tree detection and area-based approaches (ABA) for precise forest growing stock estimation by using CHMs derived from airborne and VHRSI stereo imagery; (4) assesses the effect of ABA plot size on image-based CHM models performance and accuracy. The main goal of this study stage is to demonstrate that where field-plot (NFI) data are spatially limited, it is possible to use a hierarchical integration approach based on upscale forest growing stock estimates from individual trees to broader landscapes [4]. As for practical application and as an auxiliary tool for planning and managing forestry, the proposed method of mapping forest growing stock based on image-derived canopy height metrics will also be of great importance. However, compared to LiDAR, it is vital to remember that optical sensors are strongly influenced by solar illumination, sun-to-sensor and sensor-to-target geometry. The insufficient sunlight during the winter season, and summer season clouds, sometimes restrict the use of satellite sensors, making image-based vegetation monitoring problematic. The positive results of this study will facilitate Latvian regional forest growing stock inventories, monitoring and mapping by using VHRSI sensors as an adequate low-cost alternative to LiDAR data.



1. Fang, J.; Brown, S.; Tang, Y.; Nabuurs, G.-J.; Wang, X.; Shen, H. Overestimated Biomass Carbon Pools of the Northern mid- and High Latitude Forests. Clim. Change 2006, 74, 355–368, doi:10.1007/s10584-005-9028-8.
2. Fassnacht, F.E.; Mangold, D.; Schäfer, J.; Immitzer, M.; Kattenborn, T.; Koch, B.; Latifi, H. Estimating stand density, biomass and tree species from very high resolution stereo-imagery-towards an all-in-one sensor for forestry applications? Forestry 2017, 90, 613–631, doi:10.1093/forestry/cpx014.
3. Goldbergs, G. Impact of Base-to-Height Ratio on Canopy Height Estimation Accuracy of Hemiboreal Forest Tree Species by Using Satellite and Airborne Stereo Imagery. Remote Sens. 2021, 13, 2941, doi:10.3390/rs13152941.
4. Goldbergs, G.; Levick, S.R.; Lawes, M.; Edwards, A. Hierarchical integration of individual tree and area-based approaches for savanna biomass uncertainty estimation from airborne LiDAR. Remote Sens. Environ. 2018, 205, 141–150, doi:10.1016/j.rse.2017.11.010.