Forest structure and disturbance dynamics detected with high-resolution airborne LIDAR

概要

A significant proportion of the global land is covered by forests, which are critically important for the carbon uptake by land. But, the carbon uptake function of forests in Japan and other mid-latitude regions is threatened by increases in frequency and strength of major storm events under the changing climate regime. Hence, it is necessary to describe forest structure at the regional scale, to set the baseline, and to monitor forest responses to major storm events. To do so, it is important to make use of the recent advances in remote sensing techniques for accurate evaluations of forest structures with sufficient details at the level comparable to conventional field-based approaches, but also with a sufficiently large spatial coverage. This thesis reports a new methodology to use airborne laser scanning (ALS) to achieve these goals, demonstrating its feasibility by analyzing the 230 km2 area of forests north of Kyoto City. This scale is much larger than the spatial scales to which ALS had been previously applied. The relevant background information, as summarized above, is explained in

Chapter 1 through literature review, before describing the overall objectives of the thesis.

Chapter 2,“Forest canopy height variations in relation to topography and forest types in central Japan with LiDAR,”reports variations of forest canopy structure in terms of the height of individual trees with a 1-m resolution digital surface model (DSM) in combination with 1-m resolution terrain model (DTM) to calculate the canopy height (CHM), from which the height of the individual trees was estimated as the maximum canopy height within a 2-m radius. This resulted in the identification of 14,466,340 canopy trees, from which 480,000 trees were randomly chosen to analyze the relationship of tree height to elevation, local topography (slope angle and direction, convexity), forest type (planted and natural stands of evergreen conifers, deciduous broadleaf stands, evergreen broadleaf stands), and distance from the nearest gaps in the canopy. In addition, for the area within the national forest boundary, the effect of stand age on tree height was also examined. The multivariate statistical analyses show that a Random Forest model, a type of machine-learning technic, is the best to describe how these variations influence the height of individual trees.

Chapter 3 “Impact of typhoons on forest landscape: estimating biomass loss using LiDAR-derived tree height and satellite optical data” evaluates how the study area was impacted by a major typhoon that affected the area in 2018 (‘Jebi’). The technic developed in Chapter 2 was used to estimate individual tree height, which was important in estimating the aboveground biomass at the stand level. Combining the LiDAR-based data with satellite optical data from the study region from before and after the typhoon, it was estimated that 1.07% of the study area exhibited canopy height loss. The results also show that conifer plantations (which occupied 45.3% of the area) were particularly vulnerable (85% of the damaged area). The biomass loss was estimated to be 1.44% of the pre-typhoon aboveground biomass of the study area. Conifer plantations, in which tree height tended to be taller, lost 2.44% of the pre-typhoon biomass. These estimates matched but were likely to be underestimated, in comparison to the results of more-accurate ground- and drone-based estimations for a smaller subset of the study area.

Chapter 4 “Assessing forest vulnerability to typhoon damage using high-resolution airborne LiDAR and mesoscale meteorological data” reports how wind characteristics of the typhoon‘Jebi’ (which traveled over Kyoto on September 4, 2018) explains the vulnerability of the forest in combination with the characteristics of the forest stands and the underlying topography. The analysis with machine learning techniques found that combinations of the slope aspect, canopy height, and wind speed explain the likelihood of windthrow. Treefalls were more likely in valleys than on ridges, because in valleys that opened to the strongest winds coming from SW and SE, apparently funneling resulted in powerful winds to cause treefalls. In addition, the vulnerability of trees to typhoon wind increased consistently with the increase of canopy height and wind speed. Chapter 5 summarizes how the techniques and findings reported in Chapters 2-4 might advance the field, and how it may be informative for forest management in central Japan.