研究目的
The objective of this study was to assess the utility of bitemporal ALS data for classi?cation of Hdom change, AGB change, forest disturbances and forestry activities. We distinguished between the following change classes: (i) increasing versus decreasing Hdom, (ii) increasing versus decreasing AGB, (iii) undisturbed versus disturbed forest and (iv) forestry activities namely untouched, partial harvest and clearcut.
研究成果
The results show that bitemporal data acquired as part of repeated ALS-based forest inventories can be used to classify various changes in forest structure reliably. Changes in dominant height and aboveground biomass can be expected to be classified with overall accuracies >90%, and forest disturbances and forestry activities with overall accuracies of 89% and 88%, respectively. The availability of bitemporal field and ALS data is expected to increase, enabling the operational application of the methods proposed.
研究不足
The study acknowledges challenges such as measurement errors, co-location errors between field and ALS data, and allometric errors. Additionally, minor disturbances may not always be captured due to the temporal resolution of 11–15 years.
1：Experimental Design and Method Selection:
The study used the k-nearest neighbor (kNN) method for classification of forest changes based on ALS data. The methodology involved developing statistical relationships between forest attributes and ALS metrics at the level of sample plots.
2：Sample Selection and Data Sources:
Data were obtained from 558 field plots and four repeated ALS-based forest inventories in southeastern Norway, with temporal resolutions ranging from 11 to 15 years.
3：List of Experimental Equipment and Materials:
Airborne laser scanning (ALS) data were acquired using Optech ALTM 1233, ALTM 3100, and Riegl LMS Q-1560 scanners. Field data were collected using differential global navigation satellite systems (dGNSS) with Javad Legacy 20-channel receivers and a Topcon HiPer SR.
4：Experimental Procedures and Operational Workflow:
Field plots were revisited and remeasured unless a final harvest had taken place. ALS data were processed to generate terrain surface models and compute heights relative to the ground.
5：Data Analysis Methods:
The kNN method was applied for classification, with performance assessed according to overall accuracy, user's accuracy, and kappa from leave-one-out cross-validation.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
